Citation
A CMOS wireless interconnect system for multigigahertz clock distribution

Material Information

Title:
A CMOS wireless interconnect system for multigigahertz clock distribution
Creator:
Floyd, Brian A., 1974-
Publication Date:
Language:
English
Physical Description:
ix, 278 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Antennas ( jstor )
Capacitance ( jstor )
Electric potential ( jstor )
Frequency dividers ( jstor )
Inductors ( jstor )
Receivers ( jstor )
Signals ( jstor )
Transistors ( jstor )
Transmitters ( jstor )
Vibration ( jstor )
City of Gainesville ( local )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2001.
Bibliography:
Includes bibliographical references (leaves 268-277).
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Brian A. Floyd.

Record Information

Source Institution:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
47260541 ( OCLC )
ocm47260541

Downloads

This item has the following downloads:


Full Text










A CMOS WIRELESS INTERCONNECT SYSTEM FOR MULTIGIGAHERTZ CLOCK DISTRIBUTION














By


BRIAN A. FLOYD










A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2001














ACKNOWLEDGMENTS

I would like to begin by thanking my advisor, Professor Kenneth O, whose constant encouragement, guidance, and friendship have been invaluable. I also would like to thank Professors Fox, Fossum, and Frank for their interest in this work and their time commitment in serving on my committee.

Much appreciation goes to the Semiconductor Research Corporation (SRC) for funding this work, and to the SRC and Intersil for sponsoring my fellowship. Thanks also go to SRC mentor Dr. Scott List of Intel and to fellowship liaison Dr. Ken Ports of Intersil. Furthermore, I would like to thank the sponsors of the Copper Design Challenge--the SRC, Novellus, SpeedFam-IPEC, and UMC-- as well as IBM and the Navy for providing access to advanced CMOS technologies.

I would like to thank my colleagues Chih-Ming Hung and Kihong Kim, whose beneficial discussions, advice, and friendship have contributed immensely to this work. Also, I would like to recognize former SRC group members K. Kim, J. Mehta, H. Yoon, D. Bravo, and the next generation, J. Caserta, X. Guo, W. Bomstad, N. Trichy, T. Dickson, and R. Li, for their teamwork on this project.

Finally, I am grateful to my wife, Caroline, to whom this work is dedicated. Her love, prayers, and support have meant more than I can say. Also, I would like to thank my parents for their love and encouragement throughout the years. Most importantly, I would like to thank God for sustaining me each and every day and for being my ultimate source of strength, renewal, and hope.

ii














TABLE OF CONTENTS


page

ACKNOW LEDGMENTS .................................................. ii

AB STRA CT ........................................................... viii

CHAPTERS

1 INTRODUCTION ....................................................1

1.1 Global Interconnect Challenges ..................................1
1.2 Proposed Interconnect System ...................................4
1.2.1 Potential Solutions ......................................4
1.2.2 Description of Wireless Interconnect System .................. 5
1.2.3 Clock Receiver and Transmitter Architectures ................. 6
1.2.4 Potential Benefits.......................................7
1.2.5 Areas of Research.......................................8
1.3 Overview of Dissertation........................................9

2 CMOS LOW NOISE AMPLIFIERS......................................12
2.1 Overview ....................................................12
2.1.1 ScopeofLNAResearch.................................12
2.1.2 Performance Metrics of LNAs............................14
2.2 Possible LNA Topologies ......................................15
2.2.1 Common-Gate CMOS LNA..............................15
2.2.2 Source-Degenerated CMOS LNA .......................... 17
2.3 Input Matching for Source-Degenerated LNA ...................... 18
2.3.1 InputImpedance.......................................18
2.3.2 Input Match to Withstand Component Variations .............. 19
2.4 Output Matching for Source-Degenerated LNA ..................... 22
2.5 Gain of Source-Degenerated LNA ...............................25
2.5.1 Gain Driving Resistive Load .............................25
2.5.2 Methods to Maximize Gain ..............................26
2.5.3 Gain Driving Capacitive Load ............................28
2.6 Noise Parameters .............................................29
2.6.1 Noise Sources.........................................29
2.6.2 Noise Parameters of Single Transistor ....................... 30
2.6.3 Optimum Qgs for Minimum Noise.........................33


iii









2.6.4 Optimum Width ofM2..................................36
2.7 Design Methodologies for Source-Degenerated LNA ................ 38
2.7.1 Derivation-Based Methodology ............................ 38
2.7.2 Alternative Constant Available Gain Methodology ............. 39
2.8 Sum m ary ..... ...............................................41

3 CMOS LNA IMPLEMENTATION AND MEASUREMENTS................. 42

3.1 Overview ..... ...............................................42
3.2 Passive Components...........................................42
3.2.1 Inductor Design Techniques..............................42
3.2.2 Capacitor implementation................................44
3.3 A 900-MHz, 0.8-gm CMOS LNA................................45
3.3.1 Circuit Implementation..................................45
3.3.2 M easured Results......................................47
3.3.3 Summary for900-MHz LNA.............................51
3.4 An 8-GHz, 0.25-gm CMOS LNA................................52
3.4.1 Circuit Implementation..................................52
3.4.2 Inductor Characteristics .................................54
3.4.3 M easured Results......................................55
3.5 A 14-GHz, 0.18-im CMOS LNA................................57
3.5.1 Circuit Implementation..................................57
3.5.2 Inductor Characteristics .................................58
3.5.3 Transistor Characterization ................................62
3.5.4 M easured Results......................................64
3.6 A 23.8-GHz, 0.1-gjm SOI CMOS Tuned Amplifier .................. 67
3.6.1 Circuit Implementation..................................67
3.6.2 Source-Follower.......................................69
3.6.3 Gain of 24-GHz LNA...................................70
3.6.4 M easured Results......................................72
3.7 Sum m ary ....................................................75

4 CMOS FREQUENCY DIVIDERS......................................77
4.1 Overview ..... ...............................................77
4.2 Frequency Divider Using Source-Coupled Logic .................... 78
4.2.1 Circuit Description.....................................78
4.2.2 Latched Operating Mode ................................80
4.2.3 Quasi-Dynamic Operation ...............................81
4.3 Injection Locking of SCL Frequency Divider ....................... 82
4.3.1 Oscillation of SCL Divider...............................83
4.3.2 Description of Injection Locking ........................... 85
4.3.3 Simulation of Injection-Locked Divider...................... 87
4.3.4 Implications of Injection Locking for Clock Distribution ........ 89
4.4 A 10-GHz, 0.25-gm CMOS SCL Divider .........................89
4.4.1 Circuit implementation..................................89
4.4.2 M easured Results......................................92


iv








4.5 A 15.8-GHz, 0.18-gm CMOS SCL Divider........................95
4.5.1 Circuit Implementation..................................95
4.5.2 M easured Results......................................97
4.6 Programmable Frequency Divider Using SCL ...................... 98
4.6.1 Motivation and General Concept ........................... 98
4.6.2 System Start-Up Methodology ............................ 102
4.6.3 Circuitry for Programmable SCL Divider ................... 104
4.6.4 Simulated Results.....................................109
4.6.5 Testing Methodology ..................................114
4.6.6 Measured Results.....................................116
4.7 A 0.1-gm CMOS [DP]2 Divider on SOI and Bulk Substrates......... 118
4.7.1 Circuit Implementation.................................118
4.7.2 Effect of SOI on Circuit Performance ...................... 120
4.7.3 Measured Results.....................................123
4.8 Sum m ary ...................................................125

5 SYSTEM REQUIREMENTS FOR WIRELESS CLOCK DISTRIBUTION...... 127
5.1 Overview ..... ..............................................127
5.2 Power Transfer from Clock Source to Local Clock System .......... 128
5.2.1 Transmitter Power ....................................129
5.2.2 Antenna Specifications.................................130
5.2.3 Receiver Gain........................................131
5.2.4 Matching Between Antennas and Circuits ................... 132
5.3 Definition of Clock Skew and Jitter.............................135
5.4 Clock Skew Versus Amplitude Mismatch ......................... 136
5.4.1 LatchedM ode........................................137
5.4.2 Injection-Locked Mode ................................140
5.5 Clock Jitter Versus Signal-to-Noise Ratio......................... 142
5.5.1 Phase Noise of Frequency Dividers ........................ 143
5.5.2 Conversion From Additive Noise to Phase Noise ............. 144
5.5.3 Output Phase Noise of Clock Receiver ..................... 146
5.5.4 Jitter in Clock Receiver ................................147
5.6 Sensitivity and Noise Requirements.............................. 151
5.6.1 Receiver Sensitivity ...................................151
5.6.2 Receiver Noise Requirements ............................ 152
5.7 Estimation of Total Noise, SNR, and Jitter for 0.25-gm Receiver ..... 154 5.8 Linearity Specification........................................155
5.9 Sum m ary ...................................................159

6 WIRELESS INTERCONNECTS FOR CLOCK DISTRIBUTION ............. 162
6.1 Overview ...................................................162
6.2 On-Chip Antennas ...........................................163
6.2.1 Antenna Fundamentals.................................163
6.2.2 Types of Antennas .....................................166
6.2.3 Propagation Paths for On-Chip Antennas ................... 168


v








6.3 Antenna Characteristics in 0.25-p.m CMOS ....................... 170
6.3.1 Measured Characteristics for Antennas ..................... 171
6.3.2 Integrated Antenna with LNA ............................ 174
6.4 Wireless Interconnect in 0.25-gm CMOS ......................... 175
6.5 Antenna Characteristics in 0.18-gm CMOS ....................... 178
6.5.1 Chip Implementation ..................................178
6.5.2 Antenna Descriptions..................................179
6.5.3 Measured Antenna Characteristics.............. ........... 181
6.6 Wireless Interconnects in 0.18-gm CMOS ........................ 186
6.6.1 Clock Transmitter.....................................186
6.6.2 Voltage Controlled Oscillator............................. 187
6.6.3 Power Amplifier......................................191
6.6.4 Clock Transmitter with Integrated Antenna.................. 194
6.6.5 Single Clock Receiver with Integrated Antenna .............. 197
6.6.6 Simultaneous Transmitter and Receiver Operation ............ 199
6.7 Double-Receiver Wireless Interconnect........................... 200
6.7.1 Measurement Setup ...................................201
6.7.2 Demonstration of Double-Receiver Interconnect.............. 203
6.7.3 Measured Skew Between Two Clock Receivers .............. 206
6.7.4 Measured Jitter of Clock Receivers ........................ 208
6.8 Sum m ary ...................................................210

7 FEASIBILITY OF WIRELESS CLOCK DISTRIBUTION SYSTEM .......... 212
7.1 Overview ...................................................212
7.2 Power Consumption Analysis..................................212
7.2.1 Power Consumption Comparison Methodology .............. 213
7.2.2 Clock Distribution Systems ..............................216
7.2.3 Results and Conclusions for Power Consumption ............. 218
7.3 Process Variation ............................................219
7.3.1 Simulation Methodology ...............................219
7.3.2 LNA and Frequency Divider Variation ..................... 221
7.3.3 Clock Receiver Variation ................................ 222
7.3.4 Conclusions for Process Variation ......................... 225
7.4 Synchronization .............................................226
7.4.1 Clock Skew..........................................227
7.4.2 Clock Jitter ..........................................230
7.4.3 Conclusions for Synchronization .......................... 233
7.5 Latency of 0.18-gim Wireless Interconnect ........................ 233
7.6 Intangibles..................................................235
7.7 Conclusions and Future Work..................................236
7.7.1 Feasibility Summary...................................236
7.7.2 Conclusions for Wireless Clock Distribution Systems .......... 237
7.7.3 Broader Applicability..................................238
7.7.4 Suggested Future Work ................................239



vi









APPENDICES

A THEORY FOR COMMON-GATE LOW NOISE AMPLIFIER ............... 241
A.1 Input Impedance .............................................241
A .2 G ain ..... ..................................................241
A.3 Noise Factor ................................................242

B OUTPUT MATCHING USING CAPACITIVE TRANSFORMER ............ 244
B.1 Two-Element Matching Technique............................... 244
B.2 Application to Capacitive Transformer in LNA .................... 245

C DERIVATION OF NOISE PARAMETERS .............................247
C. 1 Equivalent Input Noise Generators............................... 247
C.2 Noise Parameters in Impedance Form ............................ 248
C.3 Alternative Noise Parameters..................................250

D NOISE PARAMETERS OF MOSFET..................................252
D. 1 Transistor Model and Equivalent Input Noise Generators............ 252
D.2 Noise Parameters for Complete Model ........................... 254
D.3 Case Studies: Noise Parameters for Second-Order Effects ........... 255
D.3.1 Case 1-Effect of Cgd..................................255
D.3.2 Case 2 - Effect of GIN .................................257
D.3.3 Case 3 - Effect of Rb, Excluding Cgb......................258

E INJECTION LOCKING OF OSCILLATORS............................260
E. 1 Overview ...................................................260
E.2 Theory for Injection Locking ..................................261
E.2.1 Basic M odel.........................................261
E.2.2 Differential Locking Equation ............................ 262
E.2.3 Locking Range and Locking Signal Level................... 263
E.2.4 Steady-State Phase Error ...............................263
E.3 Phase Noise of Injection-Locked Oscillators ...................... 264

F QUALITY FACTOR OF RING OSCILLATOR ........................... 265

G RELATIONSHIP BETWEEN JITTER AND PHASE NOISE ................ 267

LIST OF REFERENCES................................................268

BIOGRAPHICAL SKETCH.............................................278






vii














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy A CMOS WIRELESS INTERCONNECT SYSTEM
FOR MULTIGIGAHERTZ CLOCK DISTRIBUTION By

Brian A. Floyd

May 2001

Chair: Kenneth K. O
Major Department: Electrical and Computer Engineering

As the clock frequency and chip size of high-performance microprocessors increase, it becomes increasingly difficult to distribute signals across the chip, due to increasing propagation delays and decreasing allowable clock skew. This dissertation presents the design, implementation, and feasibility of a wireless interconnect system for clock distribution. The system consists of transmitters and receivers with integrated antennas communicating via electromagnetic waves at the speed of light. A global clock signal is generated and broadcast by the transmitting antenna. Clock receivers distributed throughout the chip detect the signal using integrated antennas, amplify and divide it down to a local clock frequency, and buffer and distribute these signals to adjacent circuitry.

First, the design and implementation of CMOS receiver circuitry used for wireless interconnects is presented. A design methodology is developed for CMOS low noise amplifiers and demonstrated with a 0.8-tim, 900-MHz amplifier achieving a 1.2-dB noise figure and a 14.5-dB gain. Amplifiers are also demonstrated at 7.4, 14.4, and 23.8 GHz, viii








using 0.25-, 0.18-, and 0.10-gm technologies, respectively. A design methodology based on injection locking is developed for CMOS frequency dividers, and a programmable divider which limits clock skew is presented. Dividers operating up to 10, 15.8, and 18.8 GHz are demonstrated, implemented in 0.25-, 0.18-, and 0.10-gm technologies, respectively.

Results for the overall wireless interconnect system are then presented. System requirements (gain, matching, noise, linearity) for wireless clock distribution are derived, including specifications for signal-to-noise ratio versus clock jitter, and amplitude mismatch versus clock skew. Wireless interconnect systems are demonstrated for the first time using on-chip antenna pairs, clock receivers, and clock transmitters. The interconnects operate across 3.3 mm at 7.4 GHz, using a 0.25-grm technology, and across 6.8 mm at 15 GHz, using a 0.18-pm technology. Using the 6.8-mm, 15-GHz interconnect, a 25-ps clock skew and 6.6-ps peak jitter have been measured at 1.875 GHz for two receivers separated by ~3 mm. Finally, the wireless interconnect system is analyzed in terms of power dissipation, synchronization, process variation, latency, and area. These results indicate the feasibility of an intra-chip wireless interconnect system using integrated antennas.


















ix














CHAPTER 1
INTRODUCTION


1.1 Global Interconnect Challenges

According to the 1999 International Technology Roadmap for Semiconductors (ITRS) [SIA99], at the 0.10 and 0.05-gjm technology generations, chip areas for high-performance microprocessors are projected to be approximately 620 and 820 mm2, respectively. On-chip global clock frequencies are projected to be 2 and 3 GHz, while local clock frequencies are projected to be 3.5 and 10 GHz, respectively. These trends for high-performance microprocessors are shown in Table 1-1. In such integrated circuits (ICs), the delay associated with global interconnects--those which connect functional units across the IC--has become much larger than the delay for a single logic gate (herein termed gate-delay). This is shown in Figure 1-1, which plots the global interconnect delay and gate-delay versus minimum feature size [SIA99, Boh95]. As feature size decreases, gate-delay decreases, illustrating the well-known fact that transistor scaling improves device performance and chip density simultaneously. However, the propagation delay of a voltage wave is approximately equal to 0.35RC12 [Wes92], where 1 is the length of a wire, and R and C are its resistance- and capacitance-per-unit-length. As the CMOS technology is scaled to smaller feature sizes, both the RC time-constant [Boh95, Tau98] and the chip area (12) increase. Therefore, the global interconnect delay increases and quickly begins to dominate the overall system delay.




1






2

Table 1-1 Technology Trends of Semiconductor Industry for Microprocessors'

Year 1999 2002 2005 2008 2011
Technology Node 180nm 130nm 100nm 70nm 50nm
Microprocessor Gate 140 85-90 65 45 30-32
Length (nm)
Microprocessor Chip 450 -508 622 713 817
Size (mm2)
Linear Dimension of 21.2 22.5 24.9 26.7 28.6
Chip (mm)
Local CLK (MHz) 1,250 2,100 3,500 6,000 10,000 Global CLK (MHz) 1,200 1,600 2,000 2,500 3,000
Metal Layers 7 8 9 9 10 Power Dissipation (W) 90 130 160 170 174 10% Global Skew 83 62 50 40 33
Requirement (ps)
'Source: 1999 International Technology Roadmap for Semiconductors [SIA99]

100 1
Gate-delay
SGlobal delay, Al and K=4
80 * Global delay, Cu and c=2


" 60

_ Global
40 AI, xcconv

Global
20 Gate Cu, low-K


500 350 250 180 130 100 70 50
Technology Generation (nm)

Figure 1-1 Global interconnect delay and gate-delay versus technology generation for
aluminum and copper metallization and conventional and low-K dielectrics.


To offset this problem, copper (Cu) interconnects and low-K dielectrics have been

introduced, decreasing the global propagation delay by a ratio of approximately PCuilow/

PAlKconv, where p is the resistivity of copper (aluminum) and K is the relative dielectric






3

constant for low-Kc (conventional) dielectrics. This corresponds to a downward shift in the global delay plotted in Figure 1-1. Unfortunately, since the delay scales with chip area, despite the addition of copper and low-K dielectrics, global interconnect delay will continue to increase with succeeding generations of microprocessors. Copper and low-K dielectrics only extend the lifetime of conventional (i.e., traditional conductor) interconnect systems a few technology generations. In particular, the global delay is still much larger than the gate-delay for the 0.10-gm technology node and beyond.

The global interconnect delay is especially detrimental to global clock signal distribution. Global clock signals need to be distributed across the microprocessor with skews of less than ten percent of the global clock period. With each succeeding generation of microprocessor, the clock frequency increases, decreasing the clock period and thus the skew requirement in absolute time. This is in contrast to chip area and propagation delay, which are both increasing. Hence, techniques are required to equalize the increasingly large delays of each distributed clock signal to even greater accuracies or lower absolute clock skews. Another serious issue with global clock distribution is the dispersion associated with interconnect resistance. A non-zero interconnect resistance causes the harmonics of the clock signal to travel at different velocities through the interconnect, resulting in an increase (i.e., slowing) of the rise and fall times of the signal. As the interconnect length is increased, the rise and fall times increase, and they can ultimately limit the maximum frequency of the signal [Deu98].

Both of the previously stated problems have been somewhat circumvented in the ITRS for 0.13-gm generations and beyond by distributing a lower frequency global clock and allowing functional units to operate off of a higher frequency local clock. However,






4

referring to Table 1-1, there is an increasing gap between the projected local and global clock signal frequencies, underscoring the shortcomings of current global clock distribution systems. Clearly, advanced interconnect systems capable of distributing high frequency signals with short propagation delays and minimal power dissipation are needed to address these concerns.


1.2 Proposed Interconnect System

1.2.1 Potential Solutions

Potential solutions to address the limitations of conventional global interconnect systems can currently be categorized as follows: (1) further modifying the properties of conductors, such as cooled metal [All00] or superconductive metal, (2) shortening the distance of global interconnects by using three-dimensional structures, or (3) shifting away from synchronous computer architectures towards asynchronous architectures. A final category of solutions requires thinking even more "outside of the box" and entails research of a more fundamental nature as follows: (4) using alternative mediums to distribute signals, such as optical [Mil97] or organic mediums. However, all of the examples just listed require significant development and/or changes to either the semiconductor materials, manufacturing process, or circuit design process. An alternative global interconnect system of the fourth category is to distribute signals at the speed of light using microwaves and antennas, employing a conventional CMOS technology. This system is termed a wireless or radio-frequency (RF) interconnect system [097, 099, Flo00a].






5

IC edge Receiving Antennas (PC Board/MCM)

RX RX RX RX Integrated Circuits

RX RX RX RX fp f rut
RXI RXI RX RX transmitted TX clock signal
RX RX RX RX RX RX RX RX
Transmitting
Antenna (with
RX=Receiver parabolic reflector) TX=Transmitter Cn
N
(a) (b)
Figure 1-2 Conceptual system illustrations of (a) intra-chip and (b) inter-chip wireless
interconnect systems for clock signal distribution.


1.2.2 Description of Wireless Interconnect System

The wireless interconnect system consists of integrated receivers and transmitters with on-chip antennas which communicate across a single chip or between multiple chips via electromagnetic waves. Wireless interconnects can be used for both data and clock signals. However, for wireless data, a modulation scheme is required, while for a wireless clock, only a single tone is required. Therefore, wireless clock distribution is a natural first step for evaluating the potential of wireless interconnects in general as well as for developing the key components of a wireless interconnect system. For these reasons, wireless clock distribution is studied in this work.

A conceptual illustration of a single-chip or intra-chip wireless interconnect system for clock distribution is shown in Figure 1-2(a). An approximately 20-GHz signal is generated on-chip and applied to an integrated transmitting antenna which is located at one part of the IC. Clock receivers distributed throughout the IC detect the transmitted






6

20-GHz signal using integrated antennas, and then amplify and synchronously divide it down to a -2.5-GHz local clock frequency. These local clock signals are then buffered and distributed to adjacent circuitry. Figure 1-2(b) shows an illustration of a multi-chip or inter-chip wireless clock distribution system. Here the transmitter is located off-chip, utilizing an external antenna, potentially with a reflector. Integrated circuits located on either a board or a multi-chip module each have integrated receivers which detect the transmitted global clock signal and generate synchronized local clock signals.


1.2.3 Clock Receiver and Transmitter Architectures

Figure 1-3(a) shows a simplified block diagram for an integrated clock transmitter. The -20-GHz signal is generated using a voltage-controlled oscillator (VCO). The signal




LPa -o k ...... *............

Loop - Transmitting fREF PF Filter Antenna (-20 GHz)

: - Freq. Divider -.


(a)



ReceivingL LNA Frequency YLocal Clock Antenna Divider Output (-20 GHz) (~2.5 GHz) Analog Output
Buffers Buffers
(b)
Figure 1-3 Block diagrams of (a) integrated clock transmitter and (b) integrated clock
receiver.






7

from the VCO is then amplified using a power amplifier (PA), and fed to the transmitting antenna. The VCO is phase-locked to an external reference using a phase-locked loop (PLL), providing frequency stability. The PLL consists of a phase-frequency detector (PFD), a loop filter, the VCO, and a frequency divider.

Figure 1-3(b) shows a block diagram for an integrated clock receiver. The global clock signal is detected with a receiving antenna, amplified using a low noise amplifier (LNA), and divided down to the local clock frequency. These local clock signals are then buffered and distributed to adjacent circuits. The amplifier is tuned to the clock transmission frequency to reduce interference and noise. Since the microprocessor is extremely noisy at the local clock frequency and its harmonics, transmitting the global clock at a frequency higher than the local clock frequency provides a level of noise immunity for the system [Meh98]. Also, operating at a higher frequency decreases the required antenna size. The receiver is implemented in a fully differential architecture, which rejects common-mode noise (e.g., substrate noise) [Meh98, Brav00a], obviates the need for a balanced-to-unbalanced conversion between the antenna and the LNA, and provides dual-phase clock signals to the frequency divider.


1.2.4 Potential Benefits

The wireless clock distribution system would address the interconnect needs of the semiconductor industry in providing high-frequency clock signals with short propagation delays. These needs would be met while providing multiple benefits. First, signal propagation occurs at the speed of light, shortening the global interconnect delay without requiring integrated optical components. Second, the global interconnect wires used in conventional clock distribution systems are eliminated, freeing up these metal layers for






8

other applications. Third, referring to Figure 1-2(b), the inter-chip clock distribution system can provide global clock signals with a small skew to an area much greater than the projected IC size. This is an additional benefit, possibly allowing synchronization of an entire PC board or a multi-chip module (MCM). Fourth, in the wireless system, dispersive effects are minimized since a monotone global clock signal is transmitted. Fifth, another benefit is a more uniformly distributed power load equalizing temperature gradients across the chip. Sixth, by adjusting the division ratio in the receiver, higher frequency local clock signals [SIA99] can be obtained, while maintaining synchronization with a lower frequency system clock. Seventh, an intangible benefit of wireless interconnect systems is the effect they could have on microprocessor or system implementations, potentially allowing paradigm shifts such as drastically increased chip size. Finally, compared to other potential breakthrough interconnect techniques, such as optical, superconductive, or organic, a wireless approach based on silicon seems to be a potential solution which is compatible with the technology trends of the semiconductor industry.


1.2.5 Areas of Research

The main areas of research for wireless clock distribution are as follows: integrating compact power-efficient antenna structures, identifying noise-coupling mechanisms for the wireless clock distribution system and estimating the signal-to-noise ratio that can be achieved on a working microprocessor, implementing the required 20-GHz circuits in a CMOS process consistent with the ITRS, and characterizing a wireless clock distribution system in terms of skew and power consumption and estimating the overall feasibility of the system. The first two items have been discussed in detail in separate research






9

dissertations [Meh98, Brav00a, Kim00a] and are still being actively pursued, while the last two items are the subject of this dissertation.


1.3 Overview of Dissertation

This dissertation focuses on the implementation of an intra-chip wireless clock distribution system and evaluation of its feasibility, serving as a natural first step for evaluating the potential of both intra- and inter-chip wireless interconnects. This work will emphasize the design and implementation of RF CMOS receiver circuitry, and will evaluate the system feasibility by implementing both single-link (one transmitter/one receiver) and multi-link (one transmitter/multiple receivers) wireless interconnects.

The low noise amplifier (LNA) is the first component in the clock receiver, and through its low noise figure and moderate gain, it approximately sets the signal-to-noise ratio of the entire receiver. Generally, CMOS LNAs have had inferior performance compared to silicon bipolar or gallium-arsenide (GaAs) LNAs. Also, there are very few examples of CMOS LNAs operating above 5.8 GHz. Chapter 2 presents a new design methodology for source-degenerated CMOS LNAs, examining input and output matching, gain, and noise parameters. The effects of substrate resistance, gate-induced noise, and input-matching variations are considered. These methodologies are demonstrated in Chapter 3, where the implementation and measured results of source-degenerated CMOS LNAs operating at 0.9, 8, and 14 GHz, and implemented in 0.8-, 0.25-, and 0.18-jm CMOS technologies are presented. Also, a 23.8-GHz tuned amplifier with a common-gate input implemented in a partially-scaled silicon-on-insulator (SOI) 0.1-pim CMOS technology is presented. These LNAs are intended for use in clock receivers; however, the results are also applicable to standard CMOS receivers.






10

In the clock receiver, the frequency divider translates the ~20-GHz global clock signal to a -2.5-GHz local clock signal. The divider must be capable of high frequency of operation and low input-sensitivity while consuming minimal power. Also the dividers should be synchronized between receivers to reduce the clock skew. The design and implementation of high-frequency CMOS frequency dividers are presented in Chapter 4. This section will present a design methodology based on injection locking for dividers implemented with source-coupled logic (SCL). These dividers oscillate in the absence of an input signal; hence, this oscillation can be injection-locked to provide frequency division from a low input-swing voltage signal. Measured results of SCL dividers operating up to 10 and 15.8 GHz, at 2.5 and 1.5 V, using 0.25- and 0.18-gm CMOS technologies are presented. Also, a dual-phase dynamic pseudo-NMOS divider [Yan99a] implemented in a partially-scaled 0.1-[tm CMOS technology is presented, which operates up to 18.75 and 15.4 GHz on SOI and bulk substrates at 1.5 V, respectively. Finally, an initialization and start-up methodology to synchronize the dividers in each clock receiver is presented, which allows for decreasing the systematic skew.

In Chapter 5, the system requirements for the wireless clock distribution system are developed, translating the clock metrics of skew and jitter into standard RF metrics. To maximize the power transfer from the clock source to the local clock system, matching, gain, and antenna requirements are discussed. Clock skew and jitter are then used to set requirements for the power-level at the input of the frequency divider, the signal-to-noise ratio (SNR) at the input of the frequency divider, the noise figure for the LNA with source-follower buffers, and an I1P3 for the LNA and source-follower buffers. The overall system requirements are summarized and tabulated.






11

The major goal of this research is to demonstrate the operation of a wireless clock distribution system and evaluate its feasibility. Chapter 6 tackles the first of these two goals by demonstrating the operation and plausibility of a wireless clock distribution system. First, an overview of antenna fundamentals is presented along with measured on-chip antenna characteristics for test-chips implemented in 0.25- and 0.18-gm CMOS technologies. Antenna transmission gain, phase, and impedance results for linear dipole, zigzag dipole, loop antennas, and antennas in direct contact with the substrate are presented. Two single-receiver interconnects and one single-transmitter interconnect are then presented, demonstrating wireless interconnects for the first time. These interconnects operate at 7.4 GHz across 3.3 mm using the 0.25-grm technology and at 15 GHz across 5.6 and 6.8 mm using the 0.18-gm technology. Also, a double-receiver wireless interconnect with a programmable divider is demonstrated, and the clock skew and jitter are obtained.

Chapter 7 evaluates the feasibility of a wireless clock distribution system in terms of power consumption, process variation, synchronization, latency, area, and design verification. The worst-case clock skew and jitter are estimated and compared to the measured skew and jitter presented in Chapter 6. This chapter will show that comparable power consumption, skew, and jitter can be obtained with a wireless clock distribution system, as compared to conventional systems, with potential costs of added area and more difficult design verification. Finally, this chapter contains conclusions on the overall feasibility of a wireless clock distribution system and on this work in general, and suggests future work.














CHAPTER 2
CMOS LOW NOISE AMPLIFIERS


2.1 Overview

2.1.1 Scope of LNA Research

A key building block for the clock receiver, as well as for the front-end of superheterodyne and direct conversion receivers, is the low noise amplifier (LNA). Due to its moderate to high gain as well as its low noise figure, the LNA approximately sets the overall signal-to-noise ratio of the receiver by reducing the impact of noise from subsequent stages. This is equivalent to saying that when the LNA is properly designed, the total receiver noise figure is roughly that of the LNA. For a cascaded system, the total noise factor (F) and noise figure (NF) are given by the following formulas [Gon97]: (S/N)nut F2- 1 F3- 1
F = =iput F, + + ..., and (2.1) (S/N)output G1 G1G2

NF = 10 - log(F), (2.2) where (S/N) is a signal-to-noise ratio, and Fi and Gi are the noise factors and available power gains, respectively, of individual stages in the receiver. As can be seen, the noise contributions from latter stages are divided by the total gain preceding them--a process known as "input-referring." Thus, to minimize the total noise figure, the first stage in the receiver should amplify the input signal while adding minimal noise.

For the wireless interconnect application, a CMOS technology is required to be consistent with the ITRS. However, RF CMOS circuitry has only recently been under


12






13

investigation by researchers, with most of the research occurring at frequencies below 6 GHz. The limited work in CMOS is due to the stringent performance requirements of conventional wireless standards (e.g., Global System for Mobile (GSM) or Global Positioning Systems (GPS)), which has necessitated the use of GaAs or silicon bipolar technologies. However, CMOS technologies can potentially reduce the overall cost of the transceiver by allowing increased levels of integration, with the single-chip radio being an ultimate goal. For CMOS technology to compete with silicon bipolar or GaAs technologies for wireless applications, it must at least deliver the minimum necessary performance at a reduced system-level cost. From a performance standpoint, to be competitive with bipolar or GaAs LNAs, CMOS LNAs must equal or surpass their low power consumptions of approximately 10 mW and their low noise figures of approximately 2 dB [Sha97]. From a cost standpoint, the LNA should be implemented in a standard digital CMOS technology without any specialty passive components, and the LNA should require a minimal number of external components.

While recent works have demonstrated the potential of CMOS LNAs for -l-GHz applications, they generally have difficulty in attaining both low noise figure and low power consumption simultaneously [Sha97, Stu98, Hua98]. Alternately, if they do meet the low noise and low power, it is made possible by using either modified CMOS technologies or external matching networks [Gra00, Hay98]. Also, at this time, other than work that the author has participated in or originated, CMOS LNAs operating above -5.8 GHz have not been reported. The following two chapters present the design and implementation of CMOS LNAs operating between 0.9 and 24 GHz. An explicitly defined design methodology is presented in this chapter which considers gain, noise figure, input






14

matching, and output matching. Implementation results for four different LNAs, operating at 0.9, 8, 14, and 23.8 GHz, respectively, are then presented in the next chapter.


2.1.2 Performance Metrics of LNAs

As mentioned above, to minimize the receiver's noise figure, the LNA is required to have moderate to high gain and low noise figure. Examining (2.1), it would seem that the LNA's gain can be increased arbitrarily, thereby minimizing the noise figure and improving the sensitivity of the receiver. However, this is not the case due to the effect of LNA gain on receiver linearity. Linearity is typically measured in terms of a third-order intercept point (IP3) which is usually input-referred (IIP3). The IP3 is the output power at which the third-order intermodulation products are equal to the desired linear component. The expression for the total IIP3 for a cascaded system can be expressed as follows:

1 _ 1 G G1G2
1+ - + + ... (2.3) IIP3T IIP31 IIP32 IIP33

where IP3i and Gi are the input-referred IP3 (in watts) and the power gain (in watts/watt) of each individual stage of the receiver. For gains greater than one, the linearity of latter stages will dominate the total receiver linearity. Therefore, to maximize the receiver's IIP3, the latter stages' linearity should be maximized while reducing or limiting the total preceding gain. Thus, the gain of the LNA (i.e., G1) cannot be increased arbitrarily, since that would degrade the receiver's 11P3. Therefore, (2.1) and (2.3) imply an acceptable range of LNA gains which will meet both the system's noise figure and linearity requirements. Also from (2.3), it can be seen that the requirement for the LNA's IIP3 is not as stringent as that for subsequent stages (e.g., mixer). Typically, the LNA is designed to have a power gain of approximately 15 dB, a noise figure of less than 2 dB, and an IIP3 of -5 dBm, all the while consuming less than ~10 mW of power.






15

In superheterodyne and direct-conversion receivers, the LNA is preceded by the antenna, a duplexer filter or transmit/receive switch, and an optional pre-select filter. Since these components are not typically integrated, the input to the LNA is driven through a 50-92 transmission line. Therefore, the LNA's input should be matched to 50 2 For the wireless clock receiver application, the input of the LNA should be conjugately matched to the antenna impedance, since transmission lines are not required. The output matching of the LNA depends on whether the LNA drives an off-chip component, such as an image-reject filter for superheterodyne architectures, or an on-chip component, such as a mixer for direct-conversion architectures or a frequency divider for the clock receiver application. When driving an off-chip component, the LNA's output should be matched to 50 U The input and output matching criteria for a 50-92 match are specified in terms S11 and S22, where both should be less than -10 dB.


2.2 Possible LNA Topologies

Designing an LNA consists of meeting the gain, noise, matching, and linearity performance metrics while minimizing power consumption and cost (where all of the aforementioned tend to trade-off to a certain extent with one another). Towards this end, there are two main circuit topologies for CMOS that will be discussed--common-gate and common-source with inductive degeneration.


2.2.1 Common-Gate CMOS LNA

The first possible topology employs a common-gate amplifier with source inductance, shown in Figure 2-1(a). Appendix A contains derivations for the input impedance, gain, and noise figure for this common-gate topology. Here, the input is a parallel resonant






16

L g I I I I I I


M1 VgI In M
In

Ls Ls


(a) (b)

Figure 2-1 Potential CMOS LNA topologies: (a) common-gate or (b) common-source with
inductive degeneration.

network composed essentially of the source inductance (Ls), the gate-to-source capacitance (Cgs), and one over the transconductance (1/gm). The input impedance is designed to be -50 2; however, to maximize the gain and minimize the noise figure, the input impedance can be set to -35 , while still achieving an S11 of -15 dB. Thus, gm is set to approximately 1/35 i-1, while Ls is chosen to parallel-resonate with Cgs at the operating frequency. In addition to easily providing the input match, the common-gate topology exhibits good linearity, due to the source degeneration of the transistor provided by Rs.

A drawback of this topology is its higher noise figure. As shown in Appendix A, the noise factor for this topology is approximately the following: F = + - (gR 1 ' (2.4) a (gimRS)

where a is the ratio between the device transconductance (gm) and the short-circuit drain conductance (gd0). In the long-channel limit, y and a are 2/3 and 1, respectively, while y can be significantly larger than 2/3 for short channel lengths, due to hot-electron effects [Jin85]. Thus, in the long-channel limit and with gm=1/35 -1, the noise factor is 1.47, yielding a noise figure1 of 1.66 dB. However the noise figure can be significantly larger in






17

the short-channel regime and when taking into account other sources of noise (e.g., substrate resistance, inductor parasitic resistance, and gate-induced noise [Zie86, Sha97, Tsi99]), with typical experimental values exceeding 3 dB.

The high noise figure for the common-gate topology is a major disadvantage. Fundamentally, the high noise figure is due to gm being constrained by the input matching condition, which thereby constrains the noise figure. This is analogous to saying that the input matching conditions for optimal power match and noise match are not coincident. Therefore, a topology which decouples gm from the input matching condition would add an additional degree of freedom to the design, and hopefully superimpose the power and noise matching conditions. A topology which provides this decoupling is shown in Figure 2-1(b), which consists of a common-source amplifier with inductive degeneration. To obtain the extra degree of freedom in the design, an inductor (Lg) is added in series with the gate. As will be shown in the next section, this topology allows the power and noise matching conditions to be met simultaneously, while exhibiting sufficient linearity and allowing for low power consumption.


2.2.2 Source-Degenerated CMOS LNA

Figure 2-2 shows a simplified schematic of a CMOS LNA with inductive source degeneration. This LNA is matched to 50 Q at both the input and the output. A single-stage topology is used to minimize the power dissipation and to improve 1-dB compression point (PldB) and IP3 performance. The circuit gain is provided by a cascode




1. If gm = 1/50 K-1, then the noise figure would be approximately 2.2 dB for the
long-channel limit.






18


Vdd


Ld
C Out
ZLeq





Zin Vg 2 M2 1 R






Figure 2-2 Schematic of the source-degenerated CMOS LNA amplifier, which has a reduced Miller effect. Also, a cascode exhibits high reverse isolation, which simplifies the design procedure by decoupling the input and output matching conditions. While this topology is similar to other reported CMOS LNAs, fundamental differences include the omission of a 50-2 output buffer, the use of shielding structures, and the use of a capacitive transformer at the output.


2.3 Input Matching for Source-Degenerated LNA

2.3.1 Input Impedance

It can readily be shown that the input impedance of the source-degenerated LNA, neglecting gate-to-drain capacitance (Cgdl), is as follows: Zi=" jo)(Lg +Ls)+. + -..lL . (2.5) RjoSCgs Cgsbt L
*Lb
/\f VS * t LS Vgy

Figure 2-2 Schematic of the source-degenerated CMOS LNA amplifier, which has a reduced Miller effect. Also, a cascode exhibits high reverse isolation, which simplifies the design procedure by decoupling the input and output matching conditions. While this topology is similar to other reported CMOS LNAs, fundamental differences include the omission of a 50-92 output buffer, the use of shielding structures, and the use of a capacitive transformer at the output.


2.3 Input Matching for Source-Degenerated LNA

2.3.1 Input Impedance

It can readily be shown that the input impedance of the source-degenerated LNA, neglecting gate-to-drain capacitance (Cgdl), is as follows: Zin =- jco(Lg + Ls) + - + 91Ls. (2.5)






19

A salient feature of this topology is that source degeneration provides a real term to the input impedance, which can then be used to match to 50 91 This real term is not a resistance per se (hence it will not generate thermal noise), but rather when a current is applied to the input node, the voltage that develops at that node has components both in phase (real term) and �90 degrees out of phase (imaginary terms due to Lg, Ls, and Cgsl). The in-phase component results from the series feedback in the source of M1.

As can be seen from (2.5), this network takes the form of a series resonant circuit, with resonant frequency
1
o = [(Lg + Ls)Cgsl] 2 (2.6) Thus, at series resonance, the input impedance becomes Zin = mI Ls = oTLs, (2.7) gsl

which is a function of the bias condition, the channel length of M1, and Ls. The quality factor of this network, including the source resistance, Rs, is

1
Qi. = I). (2.8) n =ooCgs I (Rs + 0oTLs) (2.8) Thus, to design the input matching network for a given CgsI to achieve an SIl <-10 dB and series resonance, the following relations are used: 26 96
< L < - (2.9) COT s (OT

1
Lg - 1 - Ls. (2.10) 0oCgsl

2.3.2 Input Match to Withstand Component Variations

A methodology is then needed to choose Cgsl (i.e., the width of MI) and yield Ls and Lg. One way is to choose Cgsl to meet the input matching condition over an entire






20

operating frequency band, while withstanding component variations in Lg, Ls, and CgstQualitatively, the input impedance is more sensitive to component variations for high-Q networks. Thus, there is a maximum value for the quality factor which ensures matching for a given frequency band and for a given set of component tolerances. An alternative input quality factor independent of Ls can be defined as follows:

1
Qgs = oCgsRs (2.11) gs mCgsl RS,

Assuming that inductors have a tolerance of TL, Cgs1 has a tolerance of Tc, and Ot has a tolerance of TT, equation (2.5) becomes the following:

1
Zin = COTLs(1 � TL)(1 T) + jo(Lg + Ls)(1 TL) + . (2.12)
_ _ _( � joCl(1 �Tc) = COTLs(1 � TL)(1 � TT) + JQgsRsO (1 � TL) - o 1 ) (COO co (I �TO) Note that the variation in coT (which has a IVsatI/L dependence in the velocity-saturated regime [Tsi99], where Vsat is the saturation velocity) is partially correlated to the variation in Cgs (which has a WLCox dependence), through variations in the length of the device. The impedance in (2.12) has to satisfy the input matching condition at both the upper and
B
lower ends of the frequency band (i.e., co = Co B , where B is the total bandwidth). The input impedances which satisfy S 1 < -10 dB, can be visualized on a Smith chart as those impedances located within a circle centered at the origin with radius 0.32 (1010/20), as shown in Figure 2-3. If the input impedance is normalized and written as zin=(r)+j(i), the required value for i to result in a specific Sl1 for a given value of r is as follows:


S (r + 1)2 (2.13))
i 1-i (2.13)






21


j0.5 -2 Impedances satisfying
S11<-10 dB



0


-jO.2 -j5


-j0.5 -j2
-j
Figure 2-3 Smith chart showing impedances which satisfy Sl < -10 dB. The worst case matching condition occurs either at the lower end of the frequency band when the component tolerances cause the resonant frequency to increase or at the upper end of the frequency band when the component tolerances cause the resonant frequency to decrease. Both conditions yield approximately the same value for Qgs. Looking at the first
B
condition (i.e., L=L(1-TL), Cgs1=Cgsl(1-Tc), o = o- 2, and coT=oT(1-TT)), the required value of Qgs can be solved by normalizing (2.12) and then applying (2.13), as follows: o(1- Tc) Sid2(r+1)2-(r-1)2 Qgl = , (2.14) (1 - 2 (1- T,)(1 - TC)) 1 - ISzlz2 where
0TLs
r = L (1 - T)(1 - TT), (2.15)
RS


0 B 2QB(2 2Q


QB = - (2.17)






22

Table 2-1 Required Value of Qgs for Component Variations with S11 < -10 dB GSM GSM GSM ISM ISM ISM Band 935-960 935-960 935-960 2.4-2.5 2.4-2.5 2.4-2.5
MHz MHz MHz GHz GHz GHz QB 37.9 37.9 37.9 24.5 24.5 24.5
ITLI, ITc, ITTI 5% 5% 10% 5% 5% 10%

orLs 35 50 50 35 50 50 Qgs 2.9 4.8 2.4 2.6 4.3 2.3


Table 2-1 shows the required value of Qgs for the GSM and ISM (Industrial, Scientific, and Medical) bands, for varying component tolerances and values for tLs. It can be seen that as QB (the quality factor implied by the operating frequency band) decreases, so too does Qgs, as expected. Also, as the component tolerance becomes larger, the required value of Qgs decreases, indicating that high-Q networks are more sensitive to component variations, as originally asserted. Finally, as tLs becomes closer to 50 , Qgs increases, indicating that the network can withstand larger component variations. Note that choosing a smaller value of Qgs will result in additional margin for input matching variations. The results presented in Table 2-1 show that choosing Qgs between 2.3 and 4.8 will result in S11 < -10 dB over the entire band, while withstanding between 5 to 10% of component variations. Once Qgs is specified, Cgsl and hence W1 can be determined, allowing the inductor values to be chosen as given by (2.9) and (2.10).


2.4 Output Matching for Source-Degenerated LNA

Driving a 50-4 load while providing sufficient power gain requires either an output matching network or a 50- buffer. Since a buffer typically would dissipate additional power while also degrading the circuit linearity, a single-stage design with an output






23

Vdd
(a) ZM2 = RP2 11 1
(a) j0P
H7-Matching Out
Network
Vg2 M2 RL= 50 Q
' VS


(b)
IC2
RP~IIR~ IId! 02C

RPLdilRp CLd+CP2 I L 01 RL= 50 I I
I I

fl-Matching Network

Figure 2-4 Output matching equivalent circuits for CMOS LNA



matching network is preferable. As shown in Figure 2-2 and Figure 2-4(a), the output impedance of M2 is transformed to 50 Q using a three-element matching network consisting of a capacitive transformer (C1 and C2) and a shunt inductor (Ld). This three-element network is also known as a fI matching network. Compared to a two-element matching network (or "L" network), a II network allows multiple sets of component values for Ld, C1, and C2. Each set corresponds to a different loaded quality factor for the 1I network. The two-element network, on the other hand, allows only one set of component values for Ld and C2 (C1 is no longer present), with the quality factor of the loaded network fixed by the impedance transformation ratio [Bow82]. Therefore, the Hf network provides added flexibility to the designer, allowing multiple values for the drain inductor.






24

Figure 2-4(b) shows an equivalent circuit for the output matching network, including the parasitic series resistance (rd) and shunt capacitance (CLd) associated with Ld. The output impedance of the cascode is represented as a resistance RP2 (on the order of kilo-ohms), in parallel with a capacitance Cp2, where Cp2 consists of Cgd2 and Cdb2. For optimal power transfer, the matching network would transform Rp2 to 50 1 However, this implies a quality factor of Q =_ Rp2/50 - 1 for the matching network, which is on the order of 10 for Rp2 _ 5 kil Since the matching network includes an on-chip spiral inductor, the quality factor of the network is limited by that of the inductor, QLd. To include the finite Q of the spiral inductor, rEd can be transformed to an equivalent resistance shunting the inductor, as follows:


RPLd= (QLd2 + 1) rLd QLd2 rLd. (2.18) Since RpLd is typically much less than Rp2, the matching network transforms a complex source, consisting of Rpd in parallel with the capacitance Cp2+Cd, into the 50- load.

A dilemma now appears, in that the impedance to be transformed (Rpd) is a function of the network providing the impedance transformation (Ld). Therefore, an iterative approach can be taken using gain and output matching as criteria, as follows: (1) an initial value is chosen for Ld, (2) RpLd and Cd are then estimated along with ZM2 in Figure 2-4(a), (3) values for C1 and C2 are designed to achieve a 50-4 match, and (4) S21 and S22 are checked against their specifications and Ld is updated if necessary. The initial value for Ld is chosen based on the desired gain, using formulas derived in the next section. The equations for C1 and C2 can be readily derived [Ho00], and are given in Appendix B. Thus, any standard two-element matching technique can be used to choose C1 and C2 for completing the match of ZM2 to 50 Q (e.g., [Bow82]).






25

2.5 Gain of Source-Degenerated LNA

For an amplifier driven by a source VS, with source impedance Rs, and loaded at the output by RL via transmission lines with characteristic impedances of Rs and RL, respectively, the forward transducer power gain is given by [Gon97] LGT Load 2 4Vut 2 RS 2 Rs G PAVS -S21 = 4 VS L = 4IAj R-L (2.19) where PLoad is the power delivered to the load, and PAVS is the power available from the source. For Rs=RL=50 0, the transducer power gain is then GT = IS21 2 = 12 - Av 2. (2.20)


2.5.1 Gain Driving Resistive Load

It can be shown that the voltage gain from the source to the load, A, assuming that the input is at series resonance and again neglecting Cgdl, is as follows:


IAI (Vgsl( VdI (Vd2X Vout (2.21)
= = Vs Vgs I VdI Vd2

mi 1 o +C2RL (ain)(gm2 + o dj (m2 ZLeq(O0o)[) I+ j00l( C 2)RL


where Cdl is the total capacitance at the node connecting the drain of M1 to the source of M2, and Qin is the quality factor of the input series resonant circuit, as given by equation (2.8). Here, ZLeq is the total equivalent impedance at the drain node of M2, as shown in Figure 2-2. Since the output matching network essentially matches the equivalent parallel resistance of inductor Ld to the load resistance, RL, then Zuq is approximately equal to the following:
12 1
ZLeq(0o) =- 2QLdrLd = QLdcooLd. (2.22)






26

The voltage gain can then be written as

1 _(_o) IAvl=o = I . Qin mi . (QLd'o) -o. Ld) n(o) , (2.23) where

8(0@) = gm2 (2.24) ginm2 + jcoCdl

1 + jco(C1 + C2)RL (2.25) n (0) = (2.25) jIoC2RL

Here, 8(o) is a low-pass filter response, where the low-pass filter is composed of Cdl and 1/gm2; n(co) is a complex capacitive transformer ratio. At sufficiently high frequencies,
Cl
n(0) reduces to 1+ - ; however, this approximation must be justified and is rarely approC2,
priate for typical values of C2. By inspection, the ratio 18/nI is always less than 1. Further simplification can be applied to (2.23) to obtain a more useful equation. Substituting (2.8) and noting that o- gml/Cgsl, it can be shown that S2S1 = 2.(AC, = . QLd Ld o) (2.26) IS21 = RS + O)TLs n(0)


2.5.2 Methods to Maximize Gain

To maximize the gain (S21), four steps can be taken. First, examining the effect of the drain inductor on S21 requires examining the term QLdLd/n(co), where n(wo) depends on QLd and Ld via C1 and C2. However, since n(coo) is approximately a capacitance ratio, its dependence on QLd and Ld is weak. Therefore, S21 increases with increasing QL and increasing Ld. This can be seen in Figure 2-5, where S21 is plotted (a) versus Ld for constant QI and (b) versus QLd for constant Ld. For each data point, the output is matched to 50 U1 Therefore, the LNA should be designed such that Ld and its quality






27

factor is maximized2. Second, to maximize the gain, or should be maximized. This can be achieved in two ways, as follows: (1) bias the transistor for maximum 4r, which can increase the power consumption, or (2) use a more advanced (i.e., costly) technology (shorter channel length) for a given power consumption. Thus, cost and power consumption should be balanced to achieve the correct trade-off for o4. Third, looking in the denominator of (2.26), it can be seen that the gain and the input matching criteria trade-off via the term OcTLs. Since series feedback is used in the source of MI, the gain is reduced while the input series resistance (Rs) is increased by a factor of one plus the loop gain (l+WirLs/RS). Returning to (2.9), the input matching condition can still be met by choosing orLs =_ 35 0, as opposed to 50 L. Thus, by allowing a small amount of mismatch at the input (i.e., S11 = -15 dB), the gain is increased by approximately 18%, or 1.4 dB. Finally, the term 18/nI should be maximized, where 8 can be maximized by reducing Cd1 through judicious layout, and n can be minimized by decreasing or even eliminating C1.


17 16

16 --. ----------- ---. 15 -----------15 --.- ..- . ...-- .-. ..----------- 14 ---. --. -...---... .---...-.-.



11
. . . . . . . . .. 1 3 - - .-.. . . .. . . . . . . . . . .
13 12.. . .. ... ... ... . . . . .. .-

12 ----- -. ----.-.--- . -------. ---11 . . . . .10
4 5 6 7 8 9 10 11 12 13 14 3 4 5 6 7 8 9 10 Ld (nH) QLd
(a) (b) Figure 2-5 S21 at resonance versus (a) Ld for constant QL and (b) QId for constant Ld. The
output is always matched to 50 L.


2. Note that if QL becomes too large, the LNA can become intolerant to process variations.






28

Looking at these four options, QLd and oT have the most significant impact on gain. Intuitively, high-quality factor inductors result in LNAs which consume lower power. This is the case for [Hay98], where external matching networks are used to obtain high gain and low noise at a very small power. However, for a fully integrated LNA, external matching networks are taboo, since the extra components increase the cost and since their high Q's are not very repeatable. Thus, to obtain the desired gain using on-chip inductors with limited quality factors, the power consumption has to be increased. The quality factor of Ld is limited by the number of metal layers present, the oxide thickness between the top-level metal and the substrate, the substrate resistance, and the type of material used for the metal interconnect (aluminum or copper).


2.5.3 Gain Driving Capacitive Load

When the LNA does not drive 50 , and instead drives a capacitive load (i.e., mixer or frequency divider), the gain is significantly increased. Thus, when the capacitive transformer (C1 and C2) is removed, ZLeq consists only of the drain inductor with its parasitics in parallel with an equivalent load capacitance composed of CLd, CM2, and the capacitance of the subsequent stage. Thus, ZLeq(o) becomes twice that given by equation (2.22). Also, with the output matching network removed, n(w) is no longer required. The voltage gain then becomes the following:


IAv, = = 2. (T'QL d I(o). (2.27) = **Rs + OTLs

With the elimination of n(co) and the additional factor of two, the gain can be significantly larger. This means that the desired gain of -15 dB can be achieved at a reduced power consumption (i.e., mr). Also, since C1 and C2 are removed, the total capacitance resonating with Ld is small. Thus, Ld can be increased, further increasing the gain.






29

2.6 Noise Parameters

The noise factor of any two-port network can be classified in terms of four noise parameters as follows:
G2
F = Fmin+ [ Zs- Zopt2], (2.28) where Fmin is the minimum obtainable noise factor, Gn is the noise conductance, Zopt Ropt + jXopt is the optimum source impedance resulting in F = Fmin, and ZS = Rs + jXs is the actual source impedance. Note that these noise parameters are more convenient than the four traditional noise parameters (Fmin, Rn, Gopt Bopt) [Gon97] for the source-degenerated LNA, due to the source being in an impedance form rather than an admittance form. A review of this set of noise parameters in impedance form is contained in Appendix C.


2.6.1 Noise Sources

The primary noise source within the CMOS LNA is thermal noise generated by the distributed channel of the MOS transistor. A secondary noise source is thermal noise generated by the series resistance of spiral inductors. Channel noise can be lumped into two current-noise sources--one at the drain and one at the gate. The drain noise-current has the following power spectral density:

.2
'd 4kTygdo, (2.29) Af

where y is a bias-dependent parameter (0.67 for long-channel devices), k is Boltzmann's constant (1.38 x 10-23 J/K), T is absolute temperature, and gdo is the short-circuit drain conductance. The gate noise-current, a second-order effect known as gate-induced noise (GIN) [Zie86, Sha97, Tsi99], has the following power spectral density:






30

.2 ~2 2
S- 4kT Cgs (2.30) Af 5gdo

where 8 is another bias-dependent parameter (1.33 for long-channel devices). Since the origin of both the drain and gate noise currents is the channel, these noise sources are correlated, having a correlation coefficient (c) of j0.395 for long-channel devices. Due to the gate-to-channel capacitance providing a 900 phase shift, c is complex.

Aside from GIN, there are three other second-order noise sources within the LNA, as follows: channel noise of M2, thermal noise associated with gate resistance, and thermal noise generated in the substrate [Kis99]. To understand the relative importance of these second-order effects, the noise parameters for a single transistor will be derived in the following section.


2.6.2 Noise Parameters of Single Transistor

Figure 2-6 shows the high-frequency, quasi-static model for the intrinsic MOSFET [Tsi99], including noise sources. Deriving the noise parameters for this single transistor including drain and gate-induced noise, the body transconductance (gmb), substrate


Cgd
g +d j d


SC gs Vgs 9mVgs gmbVbs
rCgb - S dbS Vbs
SCsb b s



RFigure 2-6 High-frequency, quasi-static model of intrinsic MOSFET.

Figure 2-6 High-freauency, auasi-static model of intrinsic MOSFET.






31

resistance (Rsub), and all of the capacitances (Cgs, Cgd, Cgb, Cdb and Csb) is an arduous (albeit doable) task, yielding cumbersome equations. These equations can be found in Appendix D, both for the general case considering all second-order effects, as well as case-studies examining individual second-order effects. Rather than show these equations here, the resultant parameters are evaluated numerically using Matlab. Figures 2-7(a)-(d) show the effects of Qgs, Rsub, each of the capacitances--gate-to-drain (Cgd), gate-to-body (Cgb), drain-to-body (Cdb), and source-to-body (Csb)--and gmb on derived noise parameters for a single MOS transistor with GIN turned on and off. Each noise parameter is plotted versus the following normalized parameters: Qgs ranging from 0 to 10 with GIN both on and off (a first-order effect including width dependence of drain noise), Rsub ranging from 0 to 1000, and gmb/gm, CgbCgs, Cgd/Cgs, and Cdb,sb/Cgs each ranging from 0 to 1. The following are assumed: y = 1.2, 1 , and 8 = 2y. For each individual effect examined, the other parameters are held constant according to the assumptions listed in Figure 2-7.

The results for Fmin, plotted in Figure 2-7(a) show a strong dependence on each of the effects considered. First, to reduce Fmin, Rsub should be either very small (short) or very large (open). Substrate resistance can be minimized by surrounding the transistor with large-area substrate contacts very close to the channel. Second, to reduce Fmin, the mechanisms which couple substrate noise into the transistor (i.e., gmb, Cgb, and Cdb) should in general be minimized, where Cgb and Cdb are layout-dependent. Third, Fmin decreases with increasing Cgd. This is due to feedback through Cgd decreasing the uncorrelated noise resistance, Ru, caused by GIN. Finally, since Qgs is inversely proportional to device width, the channel thermal noise and the coupled substrate noise are reduced as Qgs







32

1.4 0.03

1.3 Fmin Gn

- 1.2 -0.02
V0
E
m > o

EC
". 1.0 - rGo 0.01
0.9 -o0 GIN 0

0.8 1 ' ' 1 0.00
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Normalized Parameter Normalized Parameter
(a) (b)
80 p400 , , , 4
Ropt Xopt
60 .-300

r40 L 200
0 0

20 - 100

0 " 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Normalized Parameter Normalized Parameter
(c) (d)
LEGEND -49m/m ASSUMPTIONS gmb = 0.25gm
e-eQgs/10, GIN on >-Cdb/Cgs y= 1.2 Icl = 0.395 Cgb = 0.3Cgs SQgs/10, GIN off x Cgb/Cgs (o)r = 4o00o Rsub=10 Cgd = 0.2Cgs
,--+ Rsub (k) ---Cgd/Cgs 8 = 2y Qgs = 3.0 Cdb = Cgs


Figure 2-7 Noise parameters (a) Fin, (b) Gn, (c) Ropt, and (d) Xopt, for the following
normalized parameters: Qgs with GIN on and off, Rsub, gmb, Cdb, Cgb, and Cgd.
For each parameter studied, all other parameters are held constant using the values listed in the assumption table, and GIN is on (except, obviously, for Qgs,
GIN off case).

increases. Therefore, Fmin decreases accordingly. For very low values of Qgs, Fmin

decreases due to the transistor and its parasitic capacitances being so large. Specifically,

Csb begins to shunt the substrate noise to ground, decreasing its contribution to Fin. The

value of Qgs for which Fin is maximum depends on the value of Rsub.






33

2.6.3 Optimum Qg, for Minimum Noise

While Fmin is strongly dependent on multiple effects (GIN, Rsub, gmb, Cgd, and Cgb), Gn and Xopt are only strongly dependent on Qgs with GIN on or off, and Ropt is only strongly dependent on Qgs with GIN on. Therefore, only GIN and Qgs need to be considered for Gn and Zopt. With this simplification, a design procedure for minimum noise can be developed. First, Fmin is minimized with respect to Rsub and each of the capacitances. Second, the actual noise factor (F) is minimized with respect to Qgs, including the effects of GIN, through choosing the noise matching condition to approximately coincide with the power matching condition. Figure 2-8 shows Fmin and F versus Qgs with GIN on. Here, Sl1 is constrained to be -10 or -15 dB, with orLs set to 50 1 Constraining S1I prevents Zs from being equal to Zopt, resulting in a higher noise figure. As can be seen from Figure 2-8, optimizing F with respect to Qgs is much more valuable than optimizing Fmin. Also, since Fmin is relatively flat compared to F, only the quantity F-Fmin has to be


4.0

3.5

S3.0

" 2.5
_-0 Fmin
"o 2.0' F, Sp=-10 dB .--"- F, S11=-15 dB
0
z 1.5

1.0

0.5 I I I
0 2 4 6 8 10 Qgs
Figure 2-8 Minimum noise figure and actual noise figure when the input matching is
constrained by S 11. The difference between F and Fmin indicates that constraining
S 11 causes a noise mismatch.






34
optimized. From the results shown in Figure 2-7, it can be seen that F-Fmin (a function of Gn, Zopt, and Zs) can be optimized considering GIN and Qgs only.

The noise parameters of a cascode (M1 and M2) with and without source degeneration are shown in Table 2-2, including GIN for M1. Here, Cgdl,2, gate and substrate


Table 2-2 Noise Parameters in Impedance Format

Cascode (M1 and M2) Cascode with Ls
( o2 o __1 Gn' = Gn (2.32)
Gn = T 1dol b2 = (2.31)
n ~ 1d, .' 2 a 2((T - gsR,

Rop A I2 A -Ic2 gsR (2.33) Ropt = Ropt (2.34)
O )oCgslb2 QgSRs (2.33)

Sb = QgsRs (2.35) Xopt = Xopt - oLs (2.36)
Xopt oCgs 2 gss (2.35)


Fmin = 1+ 2 - -- (2.37) Fmin'= Fmin (2.38)



resistance, Cgbl,2, and gate-induced noise of M2 have all been neglected. The noise parameters for a single transistor under the same assumptions are derived in Appendix D, section D.3.2. The following variables have been introduced: (2gdo2(OC 2 (2.39)
(W2) =1 + ?l- l\ m __ 2.9 Y1gdol 9m2

2
A = , (2.40) b = + Alc|, (2.41) b2 = + 2AIcl + A2, (2.42)






35

a = m (2.43) gdo

From (2.31),(2.33), and (2.35), it can be seen that Gn is inversely proportional to Qgs, while Ropt and Xopt are directly proportional to Qgs. This agrees very well with the noise parameters plotted in Figure 2-7.

Source degeneration modifies the noise parameters [Har73]. For the specific case corresponding to neglecting Cgd and transistor output impedance (the model used to calculate the original noise parameters), all but Xopt remain unchanged (a benefit of using impedance noise parameters). These modified noise parameters are also shown in Table 2-2. Through setting C = 1 and 81 = 0, the noise parameters for a single transistor, excluding GIN and Rsub, can be obtained. Examining (2.28),(2.31), and (2.37), it can be seen that as with the gain analysis, or should be maximized to reduce noise figure.

These noise parameters for the source-degenerated cascode are substituted into (2.28), along with Zs = Rs + jooLg, yielding the following:


FFmin o b gA bIc ()] (24)
-Fmin ~l~r T gs Qgs b2 L gs-'- T ,(2.44) where
Omo(Lg + Ls)
QL= - O(L+L) (2.45) RS

By inspection of (2.44), setting QL = Qgs (corresponding to series resonance at the input) does not result in minimum noise; thus, there is a trade-off between the noise matching and the power matching conditions. For a given value of S11 and assuming orLs = Rs, the following required value of QL for minimum noise can be obtained through the aid of (2.13):






36
4 S,2
QL = Qgs - 2. (2.4
1 - 1S11 2.6

Thus, the input should be designed such that the series resonance occurs at a frequency higher than the operating frequency. Note that for S11 = 0, QL = Qgs, as expected. Substituting (2.46) into (2.44), taking the partial derivative with respect to Qgs, and setting the result to zero yields the optimum Qgs for minimum F-Fmin as follows:

/2

QPt = 1 + 4 S . (2.47) A 2 1-S1112

For the assumptions listed in Figure 2-7 with Sll = -10 dB, and setting (=1, then QOpt equals 2.62 and QL is 1.95. For S11 = -15 dB, QOpt is 2.32 and QL is 1.96, while for S = -= dB, corresponding to a perfect input match, QOpt L = 2.18. At these optimal Qgs values, F-Fmin is 0.02, 0.07, and 0.17 for S1 = -10, -15, and -cc dB, respectively3, when considering GIN only (i.e., (2.44)). When considering all of the second-order effects (i.e., as in Figure 2-7), F-Fmin is 0.24, 0.38, and 0.65, respectively4. Therefore, matching the input for perfect series resonance results in a higher NF, as is the case for [Sha97].


2.6.4 Optimum Width of M2

The optimum sizing of M2 can be obtained by minimizing ((W2). Since gm2 and gdo2 are directly proportional to W2, while Cdl has components directly proportional to W1 and to W2, then (W2) has components proportional to and inversely proportional to





3. The impact in decibels depends on the particular value of Fin. For Fin = 1 dB, the corresponding noise figures are 1.1, 1.2, and 1.6 dB, respectively.

4. For Fmin = 1 dB, the corresponding noise figures are 1.8, 2.2, and 3.8 dB, respectively.






37

W2. Thus, an optimum value of W2 exists which minimizes (. The capacitance at the drain of M1 (Cdl) can be written as follows, neglecting interconnect capacitance: Cdl Cdll W + Cdl2W2, (2.48) where Cdl 1W1 consists of Cdbl plus a fraction of Cgdl, and Cdl2W2 consists of Csb2 and Cgs2. Differentiating (2.39) with respect to W2 assuming constant drain current and constant Vds1 yields the following:

Cd11
W2 = Cd--W1. (2.49) dl2WI

Thus, the capacitance at the drain-node of M1 should be equally divided between M1 and

M2. Figure 2-9 shows a plot of simulated noise factor versus W2 for a 0.25-gm technology. For this simulation the default noise model for BSIM3v3 is being used. The optimum width of M2 is 60 gm, when W1 is set to 90 gm. Calculations show that for this same technology, the ratio Cdl 1/Cdl2 is equal to 0.6, yielding a calculated optimal W2 of 53 im.

1.50
1.49 1.48 1.47
0 1.46
0
ci 1.45
LL
a 1.44
0
z 1.43
1.42 W2pt=60m

1.41
I
1.420 30 40 50 60 70 80 90 100 110 Width M2 (im) Figure 2-9 Noise factor versus width of cascoded transistor (M2). The optimum width occurs
when the capacitance at the intermediate node between M1 and M2 is equally
balanced between M1 and M2.






38

Thus, the simulated and calculated optimal values for W2 agree reasonably well. Also, as can be seen from Figure 2-9, choosing W2/W1 ranging from 0.55 to 0.75 will result in minimal noise contribution from M2. Therefore, the noise parameters yield the optimum values for W1 and W2, through equations (2.47) and (2.49), respectively, as well as the optimum gate inductance for a given input reflection coefficient through equation (2.46).


2.7 Design Methodologies for Source-Degenerated LNA

2.7.1 Derivation-Based Methodology

Now, a complete design methodology based on the previous derivations has been developed for the source-degenerated CMOS LNA, assuming a given technology. First, choose Qgs to be nominally 2.6. This value should be the minimum of those obtained from (2.14) and (2.47), yielding minimum noise and robustness against component variations. Once Qgs is defined, WI can be determined based on Cox. Second, using the gain and noise equations along with the desired power consumption, estimate the required co'r, yielding an approximate operating point for MI. Third, determine W2 using (2.49). Fourth, set Ls nominally to 50/or, where lower values can be used to slightly boost the gain while sacrificing input matching robustness. Fifth, determine Lg using (2.45) and (2.46) based on the desired value for S 1. Sixth, using gain equation (2.26), estimate the value of Ld needed to achieve a desired value for S21 (note, 18/ni can be estimated to be

-0.2). Seventh, extract parasitics of Ld and cascode, and match the output to 50 Q using any standard ideal matching technique, determining values for C1 and C2.






39


RS
Input Output
(a) r\j Vs Matching Matching RL Network Network Fs IFL


Ld
Rs Vg2 L .IRS Input 0- F Output


(b) f\j V Matching FL Matching RL Network Is Ls Network



Figure 2-10 Illustration of constant available gain methods, where for the gain stage either, (a)
a single transistor is used or (b) a source-degenerated cascode with inductive load
is used.


2.7.2 Alternative Constant Available Gain Methodology

The above design methodology is sensitive to the derived noise parameters. As SPICE intrinsic noise models continue to improve, it is easier and "more" accurate to use CAD tools to obtain the noise parameters of an intrinsic model with additional extrinsic parasitics also modeled. A methodology for choosing Lg, C1, and C2 well-suited to CAD, is the constant available gain method. Traditionally, the constant available gain method is used to generate ideal matching networks, when given the S-parameters and noise parameters of a single transistor [Gon97], as illustrated in Figure 2-10(a). However, since lossy inductors are used at both the drain and the source, a composite circuit including Ls, Ld, and the cascode is used in place of the active device, as illustrated in Figure 2-10(b). Values for W1, W2, Ls and Ld are chosen using the previous methodology. The S-parameters and noise parameters of the composite circuit including the inductors are extracted using






40

GA Circles
s plane I j2
*! 0.7 d
j0. . * 15'dB ** 5

* . . ,NF Circles



-j0.2 -j5



-J
Figure 2-11 Constant available gain and NF circles for cascode with Ls and Ld. SPICE. Using these parameters, constant available gain (GA) circles and constant noise circles are plotted on the Smith chart, along with input and output stability circles. Using these circles, the reflection coefficient of the source and the load, Fs and 1L, can be determined. Once Fs and PL are determined, the input and output matching networks can readily be designed.

Figure 2-11 shows a plot of simulated constant gain and noise circles on the Smith chart. Assuming Qgs is chosen using (2.47), Ropt should be close to the 50- circle. The value of Lg closest to lopt, while still meeting the desired S 11 specification, can then be read off the Smith chart. The output is then conjugately matched (i.e., FL = rout*), using the following formula:

S21S12Fs
Fou= L* = S22 + 1- lS (2.50) This determines the output matching network, the simplest of which is a capacitive transformer, C1 and C2. To obtain C1, a constant conductance circle is traversed from the center of the Smith chart, while to obtain C2, a constant resistance circle is traversed from the endpoint of the previous path to "L=r=out*, as shown in Figure 2-12.






41









O o _CRL

-jo. C2 L -J5

-jO. -j2

-J

Figure 2-12 Output matching network design to match 50 Q to Fout*.


2.8 Summary

Common-gate and source-degenerated LNAs have been presented. An explicit design methodology has been developed for a source-degenerated CMOS LNA, considering gain, NF, input matching, and output matching. In particular, the optimal sizing of the transistors has been presented for robustness against input matching variations and for minimal noise figure. The noise parameters (in impedance form) for a MOSFET have been derived to understand the effect of various second-order effects on noise performance. Gate-induced noise and substrate resistance both affect the noise parameters significantly. Using these results, a design methodology for minimum noise while constraining the input matching condition has been presented, which yields an optimal value for the size of the primary transistor in the cascode. An optimum sizing for the cascoded device was obtained through minimization of noise figure. Finally, an alternative design methodology based on the constant available gain method has been presented.














CHAPTER 3
CMOS LNA IMPLEMENTATION AND MEASUREMENTS

3.1 Overview

The previous chapter presented common-gate and source-degenerated LNAs which can achieve high gain, low noise figure, and low power consumption. Also, a detailed design methodology for source-degenerated LNAs was presented, considering gain, NF, input matching, and output matching. Methodologies, however, are only as good as their demonstrations; therefore, this chapter presents the implementation of CMOS LNAs, including design of passive components. Low noise amplifiers operating at 0.9, 8, 14, and 24 GHz are presented. The LNAs are implemented in 0.8-, 0.25-, 0.18-, and

0.10-igm CMOS technologies, respectively.


3.2 Passive Components

3.2.1 Inductor Design Techniques

The quality factor of the on-chip spiral inductors directly affects both LNA gain and noise figure. As already discussed, QLd should be maximized to improve gain while decreasing power consumption. Although the series resistance of Lg was not included in (2.44), it can easily be shown that the series resistance of Lg (rLg) increases F by the quantity rLg/Rs. For these reasons, high quality-factor inductors are required for Lg and Ld. Depending on the inductance and the operating frequency, it may not be practical to integrate Lg, which is the case for a 900-MHz LNA.



42






43





rL
s L--'gnd

Cox Cox - - shield
R __L C Cubstrated
sub Csub sub sub ed


(a) (b) Substrate Figure 3-1 (a) General two-port lumped-element model for spiral inductor. (b) Illustration of a spiral inductor above a patterned-ground shield. The shield terminates most of
E-field while allowing B-field to pass through, increasing Q.


A two-port lumped-element model for a general spiral inductor is shown in Figure 3-1(a). Using this model, the quality factor (Q) for an inductor with one port shorted to ground can be written as follows [Lon97], where substrate capacitance (Csub) and port-to-port capacitance (Cs) have been neglected:


(Q (oL)1 +04L2 C2xRsub/rs-1 (3.1) Q-'s 1 +- 02C'---2 231 ox1 + sxRub )

This accounts for loss in both the series resistance of the spiral and in the conductive silicon substrate. Note that the interaction of the substrate with the magnetic field is not captured with this equation (i.e., eddy-current effects). To maximize Q, the series resistance

(rs) should be minimized, oxide capacitance (Cox) should be minimized, and the substrate resistance (Rsub) should either be very small or very large. This trend for Rsub is the same as that seen in the noise-figure analysis. In fact, this is a general trend which can be applied to any RF circuit or component [FloOb].






44

One way to minimize the substrate resistance of the inductor is to use patterned ground shields [Yue98, Che98]. Referring to Figure 3-1(b), ground shields provide a low-resistance path to ground for capacitively coupled currents through termination of the E-field. This reduces Rsub to approximately 5-10 a and virtually eliminates Csub. The shield is patterned to prevent eddy currents from flowing in the shield by not terminating the B-field, which would decrease L and increase rs. Instead, the eddy currents are forced into the substrate, with the magnitude of these currents being inversely proportional to the substrate resistivity [Flo00b]. In the commonly used substrate resistivity range, the substrate loss increases with decreasing resistivity. This effect shows up as an increase in the series resistance (rs) in the inductor model. This effect is reduced for smaller-area inductors, since the magnetic field does not penetrate as deeply into the substrate, thereby reducing the eddy currents. The net result is an increase in inductor quality factor at the cost of increased parasitic capacitance or decreased self-resonant frequency.


3.2.2 Capacitor implementation

Capacitors C1 and C2 in Figure 2-2, as well as the bypass capacitors for bias voltages are implemented using accumulation-mode MOS capacitors [Hun98]. These capacitors achieve high Q-factors, and are compatible with standard digital CMOS technologies. The output dc voltage across C1 is 0 V, consistent with driving an image-rejection filter. A concern with using this capacitor structure is the voltage-dependent capacitance and its effect on PldB and IP3 performance. However, measurements show that PldB and IP3 vary by less than 0.5 dB for the output de voltage ranging from ground to Vdd-






45

Vdd=2.7V

Cpd Lbw=2.2nH Cbyp=500pF 100fF rs=0.550

Ld=8.1nH
rs=8.52 C2=1.7pF
Coxi,2=0.8pF rs=0.6. Lbw=2nH Out = PAD Vg2 2 =2.1pF Cpd
Cpd W=450u rs=0.6 100TfF Bias Tee 100fF rs=10 rg=1.50
Bias Tee Cd Cy=OF L08
r -- -- I M 7
In I Cbt L=16nH W=80u
s rg=0.5 Lgnd=0.5nH
Lbt Cpd "Ls=1.4nH
I I 100fF rs=0.5Q Cpd
L - J 100fF
Son-Cip off-chip
Vg- gn gn
Figure 3-2 Schematic of 900-MHz source-degenerated CMOS LNA including on-chip and
package parasitics.


3.3 A 900-MHz, 0.8-11m CMOS LNA

3.3.1 Circuit Implementation

To verify the design methodology for CMOS source-degenerated LNAs, a 900-MHz version was implemented in a 0.8-grm CMOS technology with 3 metal layers. This technology was obtained through MOSIS. Figure 3-2 shows the schematic of the LNA. The important on-chip and package parasitics are shown, including series resistances for inductors and capacitors (rs), pad capacitance (Cpd), gate resistance (rg), inductor parasitic capacitance (cox), and bondwire/ground inductance (LbwLgnd). Both bias voltages, Vg1 and Vg2, are generated off chip for flexibility in testing. Figure 3-3(a) shows a die photograph of the 900-MHz LNA [Flo99a]. The circuit area is 720x720 um2. This






46




UII











(a) (b) Figure 3-3 Die photographs of (a) 0.8-gm, 900-MHz single-ended CMOS LNA and (b)
0.25-gm, 8-GHz differential CMOS LNA.

LNA was packaged, therefore gold bond-wires are used to implement Ls as well as to improve the quality factor of Ld. The on-chip portion of Ld consisted of a ground-shielded, 4.25-turn, 8-nil spiral inductor with a quality factor of 4.5 at 900 MHz. It was implemented using metals 2 and 3 shunted together. The bondwire and pin added an additional 2.2 nH to Ld. For this operating frequency and technology, Ls is -1.4 nil. An off-chip inductor is used for Lg, while an off-chip bypassing capacitor is required for Vdd. Finally, ground-shielded pads [Col98] are used to minimize the substrate resistance, improve the noise figure and reverse isolation, and control the pad impedance.

The 900-MHz LNA was packaged in an SOIC-like test package with low ground inductance, and mounted on a multilayer board. A parasitic inductance therefore exists between the on-chip ground and the board-level ground. If this inductance is too large, the reverse isolation is greatly reduced, due to feedback [Col98]. To minimize the ground inductance, 11 ground pads are included on chip and down-bonds from the on-chip ground to the paddle of the package are used, reducing the inductance to less than 0.4 nil. The paddle is then directly soldered to the board ground.






47

3.3.2 Measured Results

The LNA was characterized at four different operating points to study the trade-off between power consumption and NF. Figure 3-4(a) shows a plot of measured gain and NF versus bias current. Operating points 1 to 4 correspond to minimum to maximum NF and maximum to minimum power, respectively. At 10 mA of current (operating point 1), the LNA achieves a NF of 1.2 dB, which is competitive with most GaAs and bipolar LNAs. However, the power consumption is 30 mW, which is quite large. As the power consumption is decreased to 8.1 mW (operating point 3), the NF only increases to 1.78 dB. This result meets both of the desired performance metrics for CMOS LNAs (NF < 2 dB, and power < 10 mW). Measurements show that to achieve a NF of 2 dB, only 6.3 mW is required. Operating points 3 and 4 have comparable NF's to [Stu98] and [Hua98], but at only a fraction of their power. Much of this power savings comes from the ability to drive



Operating Point
25 43 2 ' 16 6.5 . 15

NF 5.5
2.0- -14 1
4.5


Z a)Iu1.0- n -10 1.5 3 Gan1.5 . 3 + F **Op. Point 11n
0.5 1 . I I ' 8 0.5 I . P 0 2 4 6 8 10 12 0.6 0.7 0.8 0.9 1.0 1.1 1.2
Bias Current (mA) Frequency (GHz)
(a) (b)
Figure 3-4 (a) Measured gain (S21) and noise figure (NF) versus bias current. Four different operating points are shown, numbered 1 to 4, corresponding to minimum to maximum NF and maximum to minimum current, respectively. (b) Gain and NF
versus frequency for operating points 1 & 3.






48

50 92 without an additional output buffer. Figure 3-4(b) shows a plot of the transducer gain (S21) and NF versus frequency for the LNA, for operating points 1 and 3. At 900 MHz, the measured transducer gains (S21) are 14.5 and 10.5 dB at 30 and 8.1 mW, respectively.

Figures 3-5 (a) and (b) show the input and output reflection coefficients (S and S22) for operating points 1 and 3. The relatively high S1l's are due to the bond-wire inductance being too small, generating only -15 Q of real impedance. To improve the input match to under -10 dB, Ls should be slightly increased while Lg can be decreased to maintain the noise match. S22 is below -10 dB over a 137-MHz span. The very low S22 corresponds to a conjugate match at the output.

To measure Fmin, the measurement setup shown in Figure 3-6(a) was used. External tuners are inserted at the input and output of the packaged LNA, and the impedances of these tuners are adjusted until the optimum noise matching condition (i.e., Fopt or Zopt) occurs, corresponding to Fmin. Figure 3-6(b) shows the optimum source-reflection coefficient (ropt) versus frequency for operating points 1 and 3, plotted on Smith charts. Figure

-1 , -1
**Op. Point 1
m-mOp. Point 3
-2 -5

S-3 - -9

cD -4 c/ -13

-5- -17

- -21 1
0.6 0.8 1.0 1.2 0.6 0.8 1.0 1.2
Frequency (GHz) Frequency (GHz)
(a) (b)
Figure 3-5 (a) Input (SII) and (b) Output (S22) reflection coefficients versus frequency,
operating points 1 and 3.






49

Noise Parameter Measurement Setup

(a) Noise Bias DUT Bias Noise Figure
Source Tee Tuner (W Lg) uner Tee Meas. System DC Tuner DC Control
Fopt, Operating Point 1 Fopt, Operating Point 3



(b)


-*Op. Point 1
D-Op. Points
2.5
S2.0
1.5 0.
L. 1.0 T 0
0.5 A - ' .
(c) 0.80 0.85 0.90 0.95 1.00 Frequency (GHz)
Figure 3-6 (a) Measurement system used to obtain Fin and Fopt.
(b) Optimum source reflection coefficient for minimum noise (Fopt).
(c) Minimum noise figure versus frequency.

Table 3-1 Measured noise parameters at 900 MHz for 0.8-jm CMOS LNA Operating Current Fmin F(S11 = -oodB) F(S 11=-10dB) Point (mA) (dB) opt (dB) (dB)

1 10 1.08 0.355L 1660 1.46 1.12 2 5 1.32 0.395Z 1670 1.79 1.38 3 3 1.51 0.421Z 1650 2.02 1.56 4 2.35 1.66 0.434L 1630 2.18 1.72



3-6(c) shows the measured Fmin versus frequency for operating points 1 and 3, along with a curve fit to the measured data. Table 3-1 shows the measured noise parameters (excluding Rn) for each operating point. At 900 MHz, Fmin is 1.08 and 1.51 dB, for operating






50

points 1 and 3, respectively. To match the input to obtain Fmin, Lg should be increased and an off-chip capacitor should be added. This transforms the 50- source impedance to Zopt (equivalent to reflection coefficient Fopt). However, FTopt for the LNA is not equal to S 11", therefore the noise matching condition and the power matching condition do not coincide. Table 3-1 also shows the NF corresponding to perfect input power matching (i.e., S11 =

- -dB). The noise parameters can therefore be used to derive the optimal input matching network, such that minimum noise figure is obtained while S11 is constrained to be -10 dB. As shown in Table 3-1, these noise figures are 1.12 and 1.56 dB for operating points 1 and 3, which are very close to their respective Fmin. To obtain these NF at S1l=-10 dB, Lg should be increased by 4 nH and a 4-pF shunt capacitor should be added to the LNA input.

The reverse isolation for each operating point is very good with more than 42 dB of isolation for each bias condition. As mentioned earlier, this is due to the use of a package with low ground-inductance, ground-shielded pad structures, a ground-shielded inductor, and a cascode amplifier. The measured PjdB and IP3 data for each operating point are shown in Figure 3-7. As can be seen, the power of the intermodulation product (P2fl-2) is unchanged by the current level in the LNA. However, the output power of the fundamental (Pfl) depends on the operating power gain at the specific current level. Therefore, the IIP3 relatively scales in accordance with the gain. The output power level at which the LNA compresses is a function of the supply voltage, which is 2.7 V for operating points 2-4 and 3 V for operating point 1. Therefore, the PldB (output) is approximately the same for each operating point, while PldB (input) scales with the gain. The measured IIP3's at 30 mW and 6.3 mW are -1 dBm and -3.8 dBm. Due to the use of a single-stage topology, the amplifier is very linear, and is suitable for GSM applications.






51



20







E -60 GPS-40 Ill Ill


-80 2fi iI
-35 -30 -25 -20 15 -10 -5 0 Pin (dBm)
Figure 3-7 Measured PldB and IIP3, operating points 1-4.


Table 3-2 Summary of measured characteristics for 900-MHz CMOS LNA fo = 900 MHz Op. Point 1 Op. Point 2 Op. Point 3 Op. Point 4

Power (mW) 30 13.5 8.1 6.3
Gain (dB) 14.5 12.4 10.5 9.4
50- Noise Figure (dB) 1.2 1.5 1.8 2.0
Fmin (dB) 1.08 1.32 1.51 1.66
F (S11=-10 dB) (dB) 1.12 1.38 1.56 1.73

VDD (V) 3.0 2.7 2.7 2.7 S11 (dB) -4.7 -4.1 -3.9 -3.8 S22 (dB) -15.9 -16.7 -16.9 -17.0 S12 (dB) < -42 < -42 < -42 < -42
IIP3 (dBm) -1 -1.9 -3.3 -3.8
PldB (input) (dBm) -12.4 -10.4 -6.3 -3.8



3.3.3 Summary for 900-MHz LNA

A summary of the measured characteristics for each operating point is shown in Table 3-2. Operating point 1 has an F.in which is 0.08 dB higher than a 900-MHz CMOS






52

LNA implemented in a 6-level-metal 0.35-gm BiCMOS process [Gra00], which is currently the lowest NF for any CMOS LNA. However, considering that [Gra00] employs technology two generations beyond 0.8-g.m, with an oT approximately five times as large, and that the back-end metal process and substrate in [Gra00] allow for inductor quality factors approximately 3 times as large as the 0.8-gm process, the result achieved in this work is excellent. Also, the performance goals of NF < 2 dB and power < 10 mW have been achieved, demonstrating the competitiveness of CMOS for wireless applications. Finally, the LNA exhibits excellent linearity and is suitable for wireless applications.


3.4 An 8-GHz, 0.25-pm CMOS LNA

3.4.1 Circuit Implementation

Having demonstrated the design methodology for source-degenerated CMOS LNAs at 900 MHz, both a single-ended and differential version of the source-degenerated LNA was implemented in a standard 0.25-ptm CMOS technology, operating at -8 GHz. This LNA was also implemented with an entire clock receiver, as will be presented in Chapter 6. The technology provides 5 metal layers, and both p- (-8 O-cm) and p+ with epitaxial layer (~0.01 Q-cm) substrates, allowing the effects of substrate resistivity on LNA performance to be studied. This technology was obtained once again through MOSIS.

Figure 3-3(b) shows a die photograph of the 8-GHz differential LNA, while Figure 3-8 shows a schematic of the differential LNA, including important parasitics (pad capacitances are implied for each labelled input). The die size is 730x510 p.m2. All three inductors, Ld, Lg, and Ls are integrated, as well as 20-pF bypass capacitors. Symmetry was maintained top-to-bottom for the differential layout to avoid any systematic errors from layout mismatch. Again, ground-shielded pads are used throughout the layout.






53

L=2.2nH
Vdd=2.5 V rs=70
cox=46f F
bypl Cd rLd2 A- Rsub=150 20pF Ld L CLd Fd
V1I 2 C4=102fF
Out- C2 Ld1 Id2 rs=0.42 Out+ C Jw C1 M3 M4 134fF 33fF C,_ c,,,2 W=49.5u rs
Inductors & 20pF L =0.25u 0.42
Parasitics
In+ rL 1 gi 14 W=91.2u m g2 rL q2 InS= PAD _M_ L=0.3nH cox=67fF
Cpd=39fF Vgc W=91.2u rs=1.22 Rsub=150 Lb0p3u cox=21fF
M, L =0.25uu



L20pF YRsub=15

Figure 3-8 Schematic of 8-GHz fully differential LNA showing parasitics.


This LNA is a fully differential version of that given in Figure 2-2. As mentioned in the first chapter, a differential topology is required for the clock receiver application. Transistor M5 provides tail current to the differential amplifier, where Vbias is externally generated. An important difference between this LNA and the 900-MHz LNA, is that inductors Lgl,2 and Lsl,2 are integrated. This LNA was originally designed to provide 12.6 dB of gain while driving 50 0, and the anticipated NF was approximately 3 dB. The higher anticipated NF, compared to the 900-MHz result, is due to short-channel effects increasing y [Jin85], as well as the integration of Lg. According to simulations, the integration of Lg increases the NF by approximately 0.6 dB. First, the increase is due to thermal noise generated by the series resistance of Lg (rLgi), whose Q factor is 10 (5) for p- (p+) substrates. Second, the parasitic capacitances at the input and output ports of Lg (CLgi) move the input match away from the optimal noise match. Therefore, the integration of Lg






54

imposes severe restrictions on the achievable noise figure. For standard wireless applications at this frequency range, Lg is approximately 2 nH, which is suitable for bondwire implementation. This would increase the Q dramatically and help the noise figure. However, for the wireless interconnect application, since the antenna is integrated, Lg has to be integrated as well.


3.4.2 Inductor Characteristics

To demonstrate the effect of substrate resistance on inductor quality factor, 1.8-nH spiral inductors have been implemented in the 5-level-metal, 0.25-gm CMOS process on both p- and p+ (with epitaxial layer) substrates. These inductors were used for Lg in the LNA. The inductor is implemented using shunted metals 3, 4, and 5 layers with 2.5 turns and an area of 10x 100 m2. The quality factor is estimated using both the half-bandwidth method [098] and the conventional method (Q1l=-imag(yll)/real(yl1)). Figure 3-9 shows

12
S10
.2- Qbw P
a 8
L
Z, 6
6 - abw P+

2

Frequency (G z)
2.25
S2.00 S1.75
CO)
3 1.50
1.25 ,
3 4 5 7 8
Frequency (GRz) Figure 3-9 Quality factors and inductance of Lg versus frequency for different substrate
resistivities (p+ - 0.01 a-cm, p- with epi ~ 8 -cm). The inductor is implemented
in a 5-metal-layer 0.25-im CMOS process.






55

the measured Q and inductance of these inductors for both substrate types. The quality factor of the inductor on p+ substrates (Q=5) is -50% of that on p- substrates (Q=10). For p+ substrates, the inductance slightly decreases, while the series resistance nearly doubles, due to the induced eddy currents. Measurements also show that for a 1-nH inductor with an area of 90x90 jim2, the Q reduction is only -30% [Flo00b]. This signifies that smaller area inductors are beneficial for low-resistivity substrates.


3.4.3 Measured Results

Figure 3-10(a) shows the measured S-parameters for the differential LNA. To obtain this balanced measurement using an unbalanced network analyzer, 1800 couplers were placed at the input and output of the LNA, acting as baluns. Driving 100 Q differentially, the LNA provides 10 dB of gain, with input and output reflection coefficients of less than -15 dB, consuming 21 mW from a 2.5-V supply. This gain is - 3 dB lower than that expected, due to a discrepancy between the simulated transistor characteristics and the measured transistor characteristics. In particular, Cgs was 80% larger than that simulated; thus, o)r was lower and the gain was decreased, as given by (2.26).

Figure 3-10(b) shows the measured gain and NF of the single-ended LNA for both p- and p+ substrates. Each LNA consumes 10 mW from a 2.5-V supply. The power gain of the LNA on a p- substrate is ~7.3 dB, while that for the LNA on a p+ substrate is -1 dB. For the p- substrate, there is a 2.7-dB difference between the gain for the single-ended and differential LNAs, whose origin is unknown. Theoretically, the gains should be the same when each half-circuit of the differential LNA is operated at the same power consumption as the single-ended LNA. Referring to Figure 3-9, the spiral inductor Q for p+ substrate is






56

10
10 - % -, 14


0 5 S2 12 S22 ' 10
-10- . O< S21' P+ '
m -n C 0 8
-20. $11 S (.9"-5 - '
-30 - -N F, p'
-04

-40 * * *' -10 L
5 6 7 8 9 10 3 4 5 6 7 8 9 1V
Frequency (GHz) Frequency (GHz)
(a) (b) Figure 3-10 (a) Measured S-parameters of differential LNA. (b) Measured gain and NF for
single-ended NF for both p+ and p- substrates.


-50% of that for p- substrate, and the inductance is smaller. Therefore, the resonant frequency of the p LNA is about 25% higher due to the decreased inductance. Simulation results show that -4 dB of the gain reduction can be attributed to the 50% reduction in the Q for Ld. Another -1 dB comes from the additional loss due to the increase in rLg.

The NF of the single-ended p LNA is -7.4 dB, which is 2.5 dB higher than the noise figure for the p- LNA. The increased NF is partly attributed to the increased thermal noise generated by the series resistance of the inductors. However, a large part of the increased NF is due to the decreased gain, resulting in higher input-referred noise from output noise components. Although the substrate resistance is higher for p-, the use of multiple-fingered transistors with many substrate contacts helps to mitigate its effect on NF. These results reveal the benefits of high-resistivity substrates for RF applications.






57

Table 3-3 Summary of measured results for 8-GHz, 0.25-gm LNAs.

Differential LNA (p- substrate) Single-Ended LNAs

Vdd 2.5 V Vdd 2.5 V
Power 21 mW Power 10 mW
Consumption Consumption
Resonant 7.4 GHz Res. Frequency 6 GHz
Frequency (p- substrate)
Gain 10 dB Gain (p- substrate) 7.3 dB Sil -26 dB NF (p- substrate) 4.9 dB S22 -15.5 dB Res. Frequency 7.5 GHz (p+ substrate)
Reverse 32 dB Gain (p+ substrate) 1 dB
Isolation
NF (p+ substrate) 7.4 dB


3.5 A 14-GHz, 0.18-tm CMOS LNA

3.5.1 Circuit Implementation

A final implementation of a source-degenerated LNA was done in a 0.18-gm CMOS technology with 6 layers of copper interconnects [Flo01a]. This technology was obtained from UMC as part of the SRC Copper Design Challenge, sponsored by the Semiconductor Research Corporation (SRC), Novellus, SpeedFam-IPEC, and UMC. The substrate resistivity is 15-25 2-cm. The circuits were designed to operate at -21 GHz, while measurements reveal ~ 14 GHz operation.

For this technology, both a single-stage differential LNA and an LNA with source-follower buffers were implemented. The operation of source-followers will be presented in section 3.6. The schematic of the single-stage LNA is identical to the 8-GHz version shown in Figure 3-8, where component values and sizes were modified, as reflected in Figure 3-11. Figure 3-12 shows a schematic of the fully differential LNA with






58

L=0.61 nH
Vdd=1.5 V rs=1.20 Cbyp-- cox=26f F
Cbypl / Rsub=100
20pF rLdl Ldl CLd rLd2 I
O--V"L / C4=56fF Out- 2 1 d2 rs=0.2G Out+


NW r l 1g M L-=-Obp218 L018uT Z
Vg2 C 3 CNW CN C M M4L 140fF 10fF Oyp2 W=24u rs
Inductors & T 20pF L=0.18u 0.22 aParasitics



LgL Li CLs1 CLs2 Ls2CLg3 T OLg4 rLs1 Ls2 L-0.61nH Irs=2.42
* = PAD V i M5 L=0.12nH cox=15fF
Cpd=39fF gC W=192u rs=0.33Q Rsub=10Q b L=0.18u cox=13fF 20pF 01 Rsub=100 Figure 3-11 Schematic of 14-GHz fully differential LNA showing parasitics.


source-follower buffers. This circuit is used in the clock receiver, driving the frequency

divider. Two different input matching conditions were implemented. The input matching

condition was 100 Q (differentially) when the LNA was measured using 50-2 test equipment, and 125-j55 Q when the LNA was driven by the on-chip antenna. Therefore, the

reactance of the gate inductors for the latter case was increased by 27 1 (each side or 55 Q

differentially). The single-stage differential LNA was differentially matched at the output

to 100 Q, using capacitive transformers. Figure 3-13(a) shows a die photograph of the

0.18-gm differential LNA. The die size is 745x410 jm2.


3.5.2 Inductor Characteristics

The gate and drain inductors for the 14-GHz LNA have the following design

parameters: inductance = 0.6 nH, turns = 2.5, area = 67x67 im2, width = 5-tm, and turn






59

Vdd

W=67.2u Ldl Ld2 W=67.2u L =0.18u L =0.18u


M3 M4t


Vgcd M5 M M2 M6 Vg
W=67.2u Lgl Ls1 Ls2 Lg2 W=67.2u L =0.18u L =0.18u In+ InSource Source Follower Vbias M5 Follower
Buffer Buffer Low Noise Amplifier

Figure 3-12 Schematic of differential 14-GHz LNA with source-followers.



















(a) (b)

Figure 3-13 Die photographs of (a) 0.1 8-jim, 14-GHz differential CMOS LNA and
(b) 0.1-uim, 23.8-GHz SOI CMOS tuned amplifier.






60

0.9 , . , 20




- 0.6



0.4 0
C0.3 .0II 111

0 5 10 15 20 0 5 10 15 20

Frequency (GHz) Frequency (GHz)
Figure 3-14 Measured (a) inductance and (b) quality factor (Qbw) for Lg, and Ld. spacing = 1.25-gm. Patterned-ground shields were again used to decrease the substrate resistance. Since this technology has a high-resistivity substrate, the increase in series resistance, and hence degradation in Q, caused by eddy currents in the substrate is small, as was shown in section 3.4.2. Ld was implemented using metals 5 and 6 shunted together, decreasing the series resistance and maximizing Q. Simulations show that for Lg, minimizing the parasitic capacitance has a greater effect on NF than minimizing the series resistance (provided the Q is > 20). This is due to the capacitance moving the input matching condition away from the optimum noise match. Therefore, due to the availability of copper, only metal 6 was used for Lg.

A key benefit of copper metallization is its reduction of interconnect series resistance. For inductors, the Q will be almost doubled as compared with the same-thickness aluminum (equal to the ratio of the Al and Cu resistivities). Both metals 5 and 6 were to have a nominal thickness of 1 gm. However, during fabrication these two metal layers had to be thinned from 1 pm to approximately 0.5 im. Therefore, rs was increased and Q was decreased. Figure 3-14 shows the measured inductance and quality factor (Qbw [098])






61


Table 3-4 Comparison Between Expected and Measured Inductor Characteristics

Inductor (15 GHz) L (nH) rs () Cox (fF) Rsub (9) Q

Lgate Expected 0.61 2.41 15.2 10 21.4
Measured 0.73 6 26 -10 10.9
Ldrain Expected 0.61 1.2 25.9 10 28.6
Measured 0.69 3.5 30 -10 15


versus frequency for Lg and Ld. The best Q is approximately 10 at 10 GHz, and increases to approximately 20 at 20 GHz. Table 3-4 shows a comparison between the expected and measured inductor characteristics. First, the measured inductance is ~0.1 nH larger than that expected, resulting in decreased circuit resonant frequency. Second, the measured series resistance is more than twice that expected, due to increased metal sheet resistance, while the measured Q is about half that expected.

Although the Q's are not as high as expected, they are certainly adequate for implementing RF circuitry, provided that the impact of lower Q is accounted for during the design phase. These results illustrate the dramatic effect that decreasing the top-level metal(s) sheet resistance has on inductor characteristics. Nevertheless, for constant metal thickness, copper metallization greatly improves inductor characteristics over aluminum metallization. Thus, these results are approximately twice as good as an all-aluminum process with the same metal thicknesses, or about comparable to an aluminum process with twice the metal thickness. These measurement results suggest that greatly improved inductor characteristics can be attained by thickening metal 5 and 6 layers.






62

3.5.3 Transistor Characterization

The 0.18-9m LNA and the entire wireless interconnect system were designed to operate at -21 GHz. However, as will be shown in the subsequent sections, the operating frequency for these circuits has shifted down to -14 GHz. Referring to Table 3-4, it can be seen that the measured inductance and its parasitic capacitance were slightly higher than expected (within 0.1 nH). At -20 GHz, these discrepancies can account for at most a 3-GHz frequency shift. To understand the remaining frequency shift, the designs have been reviewed to make sure that all of the metal parasitics have been properly accounted, and that the transistors have been specified correctly including the area and perimeter of drains and sources. The remaining frequency shift is therefore believed to be primarily due to the differences between the CAD intrinsic transistor model used during the design process and the actual fabricated transistor characteristics. Note that the models used to design these RF circuits were the standard BSIM models, to which extrinsic parasitic components important at radio frequencies are added (e.g., Rgate, Rsub, and capacitances associated with interconnecting the multiple fingers of the transistor). This technique has been used very effectively in the past to yield simulated results which agree very closely with measured results.

The S-parameters of individual NMOS and PMOS transistors have been measured to quantify the difference in capacitance between the model and the actual transistors. Both the NMOS and PMOS transistors have W=18.8 gm and L=0.18 pm, and are implemented with 0.94-gm long fingers (total fingers = 20). The S-parameters of the transistors are then transformed into y-parameters. Parasitics associated with the pads are de-embedded through y-parameter subtraction of an open test structure. With the transis-






63

0.81.1
-e Measured y221.0 -eMeasured Y22
0.7 - Simulated y22 0.9 - Simulated y22
0.6 NMOS 0.8 PMOS
0.5 VgsOV, Vds=.2V 0.7 Vgs=0V, Vds=-0.2V
0.4 Cdd=31.4 fF U 0.6 C0dd=43.4 fF E 0.4 E 0.5
0.3 Cdd=7.2 fF 0.4
0.27 0.3 Cdd= 19.2 f F
0..
0.32d
0.2 0.2
1 1.5 2 2.5 3 3.5 4 1 1.5 2 2.5 3 3.5 4
Frequency (GHz) requency (GHz)
1.0 ,,,0.9 ,,
0.9- Measured y, e 0.8 IMeasured Yll
0.9 - Simulated y1 1 Cgg=36.4 fF 0.8 Simulated y 1 f
0.8 NMOS 0.7 PMOS 0.7 Vgs=O.9V, Vds--o0.1V 0.6 Vgs=-.1V, Vds=-O.1V
0.6
,0.6 c/) 0.5 E 0.5 E C0.4 gg=28.9 fF
0.4 C=26.9fF 0.4
0.3 0.3
0.2 0.2
0.1 . 0.1
1 1.5 2 2.5 3 3.5 4 1 1.5 2 2.5 3 3.5 4
Frequency (GHz) Frequency (GHz)
Figure 3-15 Measured versus simulated Y ll (with transistor in linear region) and Y22 (with
transistor off) for NMOS and PMOS transistors. The slope of these lines over 27r
is equal to the capacitances Cgg and Cdd for YlII and Y22, respectively.


tor biased in the linear region, Yl 1 yields the total gate capacitance (Cgg = Cgs + Cgd + Cgb

= WLCox + 2WCov). With the transistor turned off, the channel conductance is zero, and

Y22 yields the total drain capacitance (Cdd = Cdb + Cgd), while Y12 yields Cgd. Although

there are more accurate ways to obtain these capacitances, this measurement provides a

quick and easy way to observe any capacitance trends. Figure 3-15 shows the measured

versus simulated y I and y22 for the NMOS and PMOS transistors. The slope of each of

these lines is proportional to capacitance, therefore the steeper the line, the higher the

capacitance. As can be seen, the measured capacitances are all higher than the simulated

capacitances (note--because the open pad structure used for de-embedding does not

include the interconnects to the transistor, the measurements overestimate the capacitances






64

by a few fF). For NMOS, the gate capacitance is -1.3 times larger, while for PMOS it is 1.2 times larger. For both NMOS and PMOS, the total drain capacitance is approximately twice that which is simulated, with the PMOS difference being more severe. This can result in as much as a 1.4 times (F) reduction in resonant frequency for a tuned circuit.


3.5.4 Measured Results

Figure 3-16(a) shows the measured gain (S21) and noise figure (NF) for the single-stage LNA matched to 100 Q at the input and output, while Figure 3-16(b) shows the measured input and output reflection coefficients (S11 and S22). Driving 100 Q differentially, the LNA provides 8 dB of gain at 13.3 GHz, with S11 and S22 equal to -8 and -15 dB, respectively. The LNA consumes 12 mW from a 1.5-V supply. The reverse isolation is better than 27 dB. Finally, the 100-4 noise figure is 7.7 dB at resonance. The results are summarized in Table 3-5.


10 I I I I 12 0 ii

8 11
-4
6 S21 - 10

C14 ,9U- 81


2 8
-12
0 NF 7
S22
-2 6 -16 10 11 12 13 14 15 16 10 11 12 13 14 15 16
Frequency (GHz) Frequency (GHz)
(a) (b) Figure 3-16 (a) Measured gain and noise figure for 0.18-jm LNA.
(b) Measured input and output reflection coefficients for 0.18-utm LNA.






65

Table 3-5 Summary of measured results for 14-GHz, 0.18-gm LNAs with copper.

Single-Stage Differential LNA Differential LNA with Source-Followers

Center Frequency 13.3 GHz Center Frequency 14.4 GHz
Gain 8 dB Gain at Vgc=0.5V 21 dB S11 -8 dB Gain at Vgc=0.9V 25 dB S22 -15 dB Sll -5 dB
Reverse Isolation 27 dB Reverse Isolation 32 dB
Noise Figure 7.7 dB Noise Figure 8 dB

Vdd 1.5 V Vdd 1.5 V
Power Consump- 12 mW Power Consump- 28.2 mW
tion tion (Vgc=0.5V)


Simulations predicted a 14-dB gain, a 3-dB NE, and a 21-GHz resonant frequency. The reduced resonant frequency has already been accounted for, due to the increased inductance and the larger transistor capacitances. Referring to equation (2.26), a 50% reduction in Q results in a 6-dB reduction in IS21 12. This accounts for the drop in S21 from the expected value of 14 dB to the measured value of 8 dB. The understanding of the higher noise figure is not as straight-forward, though. First, these noise measurements were obtained through on-wafer probing using external baluns, which yields higher noise figures (on the order of 1-2 dB) as compared to packaged chips, due to variations in the contact resistance of the probe. Second, generally speaking, as the gain is decreased, the noise figure increases, due to higher input-referred noise. Third, decreased inductor Q results in more thermal noise generated by the resistive loss of the inductors. This is particularly severe for Lg and Ls, which are located at the input of the LNA. Fourth, the modeling of the transistor's thermal noise coefficient (y, which was assumed to be 1.17 in simulations based on 0.25-um measurement results, is expected to increase with






66


20
S21
10

0-
o Si
-0 -10-20 - S22

-30

-40,_ _ A
10 11 12 13 14 15 16 Frequency (GHz)
Figure 3-17 Measured S-parameters for differential LNA with source-follower buffers.


decreasing channel length due to increased hot-carrier-induced noise) and the body resistance can greatly influence the noise figure. According to SPICE simulations, increasing y to 2, and Rsub from 10 to 20 f, increases the noise figure by 2 dB. Finally, BSIM3v3's modeling of the transistor's noise parameters is not very accurate; therefore, BSIM3v3's optimal noise matching condition versus the actual optimum can be very different. This can result in operating at a suboptimal input noise-matching condition. All of these effects taken together can yield the 4.7-dB increase in noise figure.

Figure 3-17 shows the measured S-parameters for the differential LNA with source-follower buffers (driving a 100-0 load). For Vgc-0.5 V, the gain of the LNA/buffer circuit increases to 21 dB, while consuming a total of 28.2 mW from a 1.5-V supply. Table 3-5 summarizes the measured results. The resonant frequency has increased to 14.4 GHz. The gain is increased compared to the single-stage LNA result due to removal of the capacitive transformer and the presence of negative conductance at the output nodes of the LNA. Too much negative conductance, however, can result in unstable operation by






67

causing the total resistance at the output node of the LNA to be negative. By increasing Vgc to 0.9 V, the gain of the LNA/buffer circuit could be increased to 25 dB; however, S22 becomes greater than 1, indicating unstable operation. Therefore, Vgc is chosen such that stability is ensured (i.e., S22 < 0 dB).


3.6 A 23.8-GHz, 0.1-11m SOI CMOS Tuned Amplifier

3.6.1 Circuit Implementation

Since the wireless clock distribution system requires RF circuits operating above 15 GHz, it is necessary to evaluate the potential of deep submicron CMOS technologies for this frequency range. As CMOS technologies approach the sub-tenth-micron regime, the cutoff frequencies will exceed 100 GHz. Also, multiple interconnect layers in these processes will facilitate on-chip passive components with acceptable quality factors. These trends point to CMOS being a viable option for RF applications at 18 GHz and above. Recent results which attest to this fact include a 0.18-gm distributed amplifier used as a 16.6-GHz oscillator [Kle99], a 0.18-gm, 23-GHz amplifier using a simple cascode and on-chip transmission lines for matching [Yan99b], and a 0.1-gm, 25.9-GHz voltage-controlled oscillator [HunOOb]. From these, it can be seen that CMOS transistors can be used to implement circuits operating above 20 GHz.

To further investigate the RF potential of deep submicron CMOS technology including noise and linearity performance, and to evaluate the utility of lumped passive components in 20-GHz amplifiers, a tuned amplifier operating at 23.8 GHz has been implemented [Flo01b] in a partially-scaled 0.1-gpm silicon-on-insulator (SOI) CMOS technology from IBM. The process uses a 0.35-pm design rule set for all dimensions except for the 0.1-urm effective gate length (0.15-pm drawn gate length) and 3-nm gate






68

Vdd=1.5 V L=0.55nH
L=0.55nH Cpd 1 Cbyp cox=25f
rs=4.30 Ysf 150fF 4.8pF Ld2 Rsub=3000
cox=25fF Ldl C uboo
Rsub=3000 C2 Out 75fF
Cpd C=75fF C
V 150fF I M, W=60u | L rs=5 C1
g1.' W=120u L =0.15u "M C-175fF
Cbyp L=0.15u W=60u
4.8pF . pd L =0.15u

SVg 150fF M4
Cpd 2 W=60u
75fF L=0.55nH L byp W=60u L =0.15u
rs=4.30 s CbypF W=60u
cox=25fF 4.8pF L=0.15u=PAD
Rsub=3000

Figure 3-18 Schematic of 23.8-GHz amplifier in a 0.1-jtm SOI CMOS technology.


oxide thickness. Two metal layers (0.7 gm each) are supported and the substrate resistivity is 1-2 i-cm. Partially-depleted, floating-body SOI transistors are used for all circuitry.

Figure 3-18 shows a schematic of the fully-integrated tuned amplifier including, important parasitics. A three-stage topology is used to boost the circuit gain and reverse isolation. As can be seen, this topology is very different than the single-stage source-degenerated LNAs previously presented, providing a look at the high-frequency performance of other circuit topologies. For ease of input matching, a common-gate input stage with a shunt inductor is used. An overview of the common-gate topology is contained in section 2.2 and Appendix A. At resonance, the input impedance to the amplifier is 1/gml. Therefore, gml is designed to be 0.02 f-1 to provide a 50-0 match. Following the input stage is a source-follower buffer which then drives a common-source cascode stage with a tuned load. The output of this stage is matched to 50 Q using a capacitive transformer. This stage is a non-degenerated version of the LNAs discussed earlier. Figure






69

3-13(b) shows a die photograph of the 0.1-gm LNA. The die size is 0.44x0.51 mm2. Ground-shielded diamond-shaped pads are used at the input and output to reduce the parasitic capacitance to the substrate. Three on-chip spiral inductors of 0.55 nH are used in the circuit. Finally, on-chip accumulation-mode MOS capacitors of 4.8 pF are used to bypass Vdd, Vg&, and Vgc.


3.6.2 Source-Follower

The source-follower is used to shift the dc bias level from the output of the first stage (at Vdd) to the input of the third stage. The bias point of the source-follower is controlled through input Vge, which controls the current through M2 and sets the Vgs of M3. A source-follower with a capacitive load can have a negative input conductance. Since inductive source-degeneration provides a positive, real input impedance for the transistor, by reason it follows that capacitive source-degeneration provides a negative, real input impedance for the transistor.

Referring to Figure 3-18, it can be shown that the input admittance looking into M3 is as follows:
2 2
Re(Ysf) = Gsf t [gTCgs3-gm3Cgs3CLT]



mB Cgs3(gm3T +T + (Cgs3CLT+ LT)) Im(Yf) = BSf = (gm3 + gT)2 + 2(Cgs3 + CLT)2 +Cgd3


where gT is the total output conductance at the source-follower output node (equal to gds3+gds2) and CLT is the total capacitance at the same node (equal to Cgs4+Csb3+Cdb2+Cgd2). The voltage gain through the source-follower buffer is as follows: A f gin3 + JoCgs3 (3.4) A gm3 + T + jo(CgAs3 + CLT)






70

which has a magnitude less than one, a pole at O)=(gm3+gT)/(Cgs3+CLT) < o,r and a zero at o--r. Negative conductance indicates that energy is being generated. When a source-follower is used after an amplifier containing an inductive load, this energy actually replenishes some of the energy lost in the inductor, causing the quality factor to increase. Hence, the negative conductance of the source-follower increases the gain of the LNA through increasing the load quality factor.

Figure 3-19 shows the measured input conductance versus control voltage (Vgc), versus frequency for a source-follower implemented in the 0.25-g.m CMOS test chip. This buffer is driving an open pad structure having a capacitance of - 40 fF. The conductance becomes more negative with increasing frequency in agreement with (3.2). Also, the conductance becomes more negative as Vgc increases to 1.2-1.4 V, and then begins to become less negative for further increases in Vgc. This is due to gm3 increasing for Vgc between 0 to 1.2 V, causing Gsf to decrease. However, the transistors in the source-follower eventually enter the linear regime, causing gT to increase and gm3 to decrease; thus, Gsf becomes less negative.


3.6.3 Gain of 24-GHz LNA

The voltage gain of the three-stage LNA can be readily derived using equations (A.3), (3.4), and (2.23). Assuming Gsf < 0, the total voltage gain is as follows:

IAvL o=, = Av, cg Av, sf AV, cs (3.5)


= gmIQLdi oLdI IA s1 QLd2 oLd2 (o)
S+ gmlRS ( m4 n(oO
eff '1 + L I OC (io)h (gmlQdoLdAl gm3 gs3 Q2m4UQLd2 o dn2 )
1+ gl R Jigm3 + T+ j (C3s3 +LT






71

x 10-4
1.5

X 10-4 1
2
1 0.5
- 0 0

" -11
--0.5
-2
-3 -1
-1.5
4
1 -2 Freq. (GHz) .5
Vgc
82
Figure 3-19 Input conductance to source-follower buffer driving a capacitive load versus
frequency and control voltage.


where the factors 8 and n have been defined in (2.24) and (2.25), and Avsf is defined in (3.4). An effective quality factor has been introduced for Ldl, to account for the negative input conductance of the source-follower buffer, as follows:

Qeff QLd (3.6) Ld I - ( Gf - ZL(O)) (3.6)

Impedance ZLd is the conductance of the drain inductor of the first stage at (tank) resonance. Such a substitution for the quality factor can be used whenever a tuned gain stage is followed by a source-follower buffer. With the input impedance designed to be Rs=50 Q (i.e., gml = 0.02 11) and with Ldl=Ld2, the total forward transducer gain becomes

1 gm4 eff 2 2
S21 = 2 .-IAv I = Q L Ld2 o2Ld1 Av, sf . (3.7)


It can be seen that the denominator of (3.6) can become zero, indicating that "infinite" gain is obtained. Thus, a finite output signal can be obtained when no input is present to the amplifier, meaning that the circuit will oscillate. Obviously, this condition is to be






72

avoided. Thus, the conductance of the source-follower is adjusted such that the gain is improved while stability is maintained.


3.6.4 Measured Results

The measured S-parameters of the SOI amplifier are shown in Figure 3-20(a). At 23.8 GHz the amplifier is perfectly matched to 50 Q at the input, with an S11 of -45 dB, and well-matched at the output, with an S22 of -9.4 dB. The transducer gain (S21) is 7.3 dB while the reverse isolation is 27 dB, which are both excellent for a CMOS circuit at this frequency range. Also, the gain is greater than 0 dB beyond 26 GHz. The total supply current is 53 mA for a 1.5-V supply, which is high due to the use of multiple amplifier stages.

The measured quality factor (Q) of the on-chip spiral inductors is -2 at 24 GHz. This value is lower than the expected value of 4.8, which resulted in the gain being ~ 15 dB (201og{4.8/2}2) lower than its simulated value of 22 dB. The reason for this Q degradation, however, is not fully understood.

10

O21 8. Gas 14

.0 4
-10o

S- 10E
-20 0
50-K, NF 0

-30 2 8 S12)
-40 0 6
20 21 22 23 24 25 20 21 22 23 24 25 26
(a) Frequency (GHz) (b) Frequency (GHz)
Figure 3-20 (a) Measured S-parameters and (b) measured operating power gain (Gp) and
50- noise figure (NF) vs. frequency.






73

Examining these results, it can be concluded that the gain performance of this 23.8-GHz amplifier is limited primarily by the inductor quality factor. Since this design was completed before the effects of substrate resistance on inductor Q [Flo00b] were fully understood, patterned ground shields were not used even though the substrate resistivity of 1-2 G2-cm is in the range where ground shields would have enhanced the inductor Q. The substrate resistance for these inductors is on the order of 300 " while the parasitic capacitance to the substrate is 25 fF (refer to inductor model in Figure 3-1(a)). According to equation (3.1), the inductor Q is greatly reduced at 24 GHz, due to the substrate resistance. In fact, had patterned ground shields been used, then the substrate resistance would have been reduced to -10-15 Q, while the parasitic capacitance would have increased. This would have approximately doubled the inductor Q, potentially increasing the gain by 20log{2}2=12 dB (for the case when the source-follower is adjusted for close to zero input conductance). Further improvements could be obtained by improving the back-end process by increasing the top-level metal thickness, increasing the distance between the top-level metal and the substrate, and using copper rather than aluminum for the metal (as will be the case for fully-scaled 0.1-gm technology). Such improvements should easily lead to inductor Q > 30 [Bur98] at 20 GHz, as has been shown in this chapter. This would dramatically increase the amplifier gain and decrease the power consumption.

Using simulated values for all quantities except QLd in (3.7), and then inserting the extracted QL value of 2, results in S21 = 9.2 dB, which is close to the measured S21. The discrepancy is attributed to differences between the transistor model and the actual device. The estimated input conductance to the source-follower is -2 mS, resulting in a Qeff of 2.8. Therefore, the negative conductance of the source-follower improves S21 by a factor of






74

-1.4 or 3 dB. However, IA,sls is -0.7; therefore, the net gain in S21 due to the source-follower is 1.4*0.7 _ 1 (0 dB). Thus, for this amplifier, while the source-follower is providing a needed de level shift between stages 1 and 3, the gain is unaffected. A potential better use for the source-follower would be to use it purely as a negative conductance generator, without passing a signal through the stage. The power consumption could be reduced by reusing the current in the gain stage with the source-follower.

Multistage amplifiers can exhibit poor linearity, and negative conductance can potentially affect this linearity even further. The measured PldB is -10 dBm (-16.2 dBm referred to the input), while the IIP3 is -7.8 dBm. These linearity data are considerably better than those obtained in [Ho98], an encouraging result, indicating that acceptable linearity can be achieved with amplifiers utilizing source-followers.

Figure 3-20(b) shows the measured noise figure (NF) and associated gain (G.) versus frequency. The gain is 8.1 dB at 23.8 GHz, while the noise figure is very high at 10 dB. The high noise figure is caused by the reduced first-stage gain, due to the low inductor Q. Also, the noise associated with the input inductor, LS, increases NF as well. The low gain allows noise from stages two and three to refer to the input of the circuit. While such a noise figure certainly is too high for typical wireless communication systems, it can be decreased through circuit and inductor optimization. A final clarification on the amplifier's performance is that although the NF is larger than the gain, it does not mean that the amplifier is unusable (i.e., that the output signal-to-noise ratio (SNR) will always be < 1). Noise figure is a relative quantity, representing degradation of SNR. Therefore, a high NF degrades the dynamic range of a circuit through raising the minimum detectable signal.






75

Table 3-6 Summary of measured results for 23.8-GHz SOI CMOS tuned amplifier.

Parameter Result Parameter Result Resonant Frequency 23. 8 GHz NF 10 dB S21 7.3 dB PIdB(in) -16.2 dBm S11 -45 dB IIP3 -7.8 dBm
S22 -9.4 dB Vdd 1.5 V S12 -27 dB Supply Current 53 mA


A summary of the results for the tuned amplifier is shown in Table 3-6. Excellent gain and isolation is obtained for CMOS circuits at this frequency. While only moderate noise and linearity performance are obtained, once a fully-scaled CMOS technology with a more advanced back-end process is used, the results should dramatically improve. These improvements will be due to increased inductor Q, decreased parasitic capacitance which allows the inductance and hence gain to increase, and finally further circuit optimization. Finally, these results show that a 0.1-gm CMOS technology should be to able support RF applications above 20 GHz, and suggest even higher operating frequencies for further scaled CMOS technologies.


3.7 Summary

The validity of the LNA design methodologies presented in the previous chapter was demonstrated through implementation of a 0.8-gm, 900-MHz CMOS LNA achieving a 1.08-dB minimum noise figure at 30 mW, and a 1.5-dB minimum noise figure at 8 mW. These results meet all of the required performance metrics for LNAs, demonstrating the competitiveness of CMOS LNAs as compared to GaAs or silicon bipolar technologies. Using a 0.25-um CMOS process, 8-GHz single-ended and differential LNAs have been






76

implemented on both p+ and p- substrates. The differential LNA on p- substrate provides 10-dB of gain, with a 4.9-dB noise figure, consuming 21 mW from a 2.5-V supply. Using a p+ substrate degrades the gain by - 6 dB and the noise figure by ~ 2.5 dB, due primarily to reduced inductor Q.

A 13.3-GHz differential source-degenerated LNA with 8 dB of gain has been implemented in a 0.18-gim CMOS technology. The circuit was designed to operate at 21 GHz; however, larger than expected inductance and transistor capacitance decreased the resonant frequency. This shows that as the operating frequency of the circuit increases above 10 GHz, modeling of the transistors and passive components becomes increasingly important. Currently, the accuracy is not scaling with frequency, indicating that larger deviations between simulated and measured performances will occur at higher frequencies. Although copper interconnects have increased the inductor Q in the 0.18-gim technology, an unplanned thinning of the metal layer during fabrication has mitigated the overall improvement. Even still, inductor Q's of -28 at 15 GHz have been obtained. A test-circuit containing the differential source-degenerated LNA plus a pair of source-followers has been implemented, showing a gain of 21 dB at 14.4 GHz.

Finally, using a 0.1-gjm partially-scaled SOI CMOS technology, a multistage tuned amplifier with a common-gate input and source-followers has been implemented. Source-followers generate negative conductance and provide a dc level-shift. The input admittance and gain of this stage were presented. This amplifier achieves 7.3 dB of gain at 23.8 GHz, with positive gains beyond 26 GHz. This result is comparable to the highest operating frequency to date [Yan99b] for CMOS amplifiers.













CHAPTER 4
CMOS FREQUENCY DIVIDERS


4.1 Overview

In the clock receiver, the frequency divider translates the global clock signal to the local clock signal. The divider should have a maximum operating frequency above 15 GHz and should dissipate a minimal amount of power. In addition, the divider should be capable of locking to very low-level input signals--specified in terms of input sensitivity--which will improve the minimum detectable signal (MDS) of the clock receiver. Finally, the signals reaching each of the frequency dividers within the clock receivers will potentially have different phases and amplitudes, both resulting in clock skew, as well as random noise, resulting in clock jitter. Clock skew and clock jitter as a result of these mechanisms will be discussed in Chapter 5. However, to account for the phase difference and, hence, reduce the clock skew, the dividers should be synchronized between receivers.

Using silicon and SiGe bipolar technologies, frequency dividers operating up to 82 GHz have been reported [Was00a, Was00b, Kna00, Wur00], and using CMOS technologies, dividers operating up to 26.5 GHz have been reported [WanOO, Wet00, Raz94, Kur97]. Clearly, silicon is capable of supporting divider circuits operating above 20 GHz. In this chapter, the design and implementation of high-frequency CMOS frequency dividers operating up to 18.75 GHz and employing either source-coupled logic (SCL) or dual-phase dynamic pseudo-NMOS [DP]2 logic [Biy96, Yan99a] will be presented. The SCL dividers are implemented in 0.25- and 0.18-jim bulk CMOS technologies, operating 77






78

up to 10 and 15.8 GHz, respectively. The [DP]2 divider is implemented in a partially-scaled, 0.1-gm CMOS technology with bulk and SOI substrates. A design methodology for SCL dividers based on injection locking which allows for maximizing the operating frequency and minimizing the required voltage swing of the divider's input signal is discussed. Also, using SCL, a new programmable divider is developed which allows the start-up state of the divider to be controlled. Such programming will decrease the systematic clock skew, provided that the dividers are properly initialized and synchronously released from the start-up state. The initialization and start-up methodology for the dividers are also presented in this chapter.


4.2 Frequency Divider Using Source-Coupled Logic

4.2.1 Circuit Description

Figure 4-1(a) shows a block diagram of a divide-by-eight (8:1) frequency divider employing SCL. The divider consists of three cascaded 2:1 dividers with dual-phase inputs and outputs. Each 2:1 divider, shown in Figures 4-1 (b) and (c), consists of two SCL D-latches in a master-slave configuration [War89, Flo00a, Hun 01]. The outputs are tied to the inputs with inverted phase to perform a toggle operation.

Source-coupled logic (also known as MOS current-mode logic (MCML)) has some distinct advantages for the wireless interconnect application. First, a small-level input voltage can be detected on the clock input. This is due to the differential structure of the latch as well as the mechanism of division, which will be discussed shortly. Therefore, the input sensitivity for the divider is high, improving the MDS of the receiver. Second, the latch can be designed such that the outputs do not swing rail-to-rail, decreasing the






79

(a)
8:1 Divider CLK Q CLK Q CLK Q
Inputs 1 (2:1) I (2:1) (2:1) Outputs








--- -s
& CLKb QbR CLKb Qb 2CLKb b

I

aMaster Slave

(b) D Q O- D Q Qb
2:1 Divider Db Qb i Db
(Toggle F/F) d b V b b



CLK CLKb Vdd Vdd









SVbias
MM7 8





CLK M1 CL M2














Figure 4-1 Block diagrams of (a) 8:1 divider consisting of three cascaded 2:1 dividers, and
(b) 2:1 divider consisting of two D-latches in master/slave. (c) Schematic of 2:1
divider implemented using source-coupled logic.






80

power consumption for a given operating frequency. Third, the use of a differential structure reduces the switching noise of the latch, by keeping the supply current approximately constant. Thus, the noise added by the divider is reduced, improving the output jitter.

A conventional SCL latch has a current source between the sources of M1,2 and M11,12 and ground, resulting in constant tail current. This current source is omitted for low voltage operation due to the limited supply head-room, causing the total current flowing through the evaluate and hold stages to not be constant. However, for fast transitions, the 2:1 divider always has both an evaluate and a hold stage on, thus the supply current remains relatively constant (reducing the switching noise).

The voltage Vbias is grounded to increase the maximum operating frequency by operating M7,8 in the linear region, which lowers the RC time constant for nodes Q and Qb. Furthermore, when the voltage of Q is increasing (during a low to high transition), the drain-to-source voltage and the output resistance of M9 (in linear region) are decreasing. This nonlinearity of M9 helps to pull up Q to its logic high, which in turn helps to increase the maximum operating frequency at the same power consumption. Similarly, when Q is decreasing, the nonlinearity tends to push down Q to its logic low. In addition to the advantage of high speed operation, this topology has reduced power consumption compared to the topology in [Cra95], due to the elimination of the folded diode-connected transistors from Q and Qb to ground.


4.2.2 Latched Operating Mode

The 2:1 divider implemented with SCL can operate in one of two modes. The first mode is the standard digital operation of latching, while the second is an injection-locked mode which will be presented in the next section. The latching mode of operation occurs






81

when the input clock signals have a large voltage swing. Here, M3,4 act as an evaluation stage when the latch is transparent, and regenerative pair M5,6 act as a hold stage when the latch is opaque. Thus, for CLK = high, the master is transparent and the slave is opaque, with Qi and Qbi taking the value of D and Db. For CLK = low, the master is opaque and the slave is transparent, and nodes Qi and Qbi are passed to the output. As is evident from this description, cascading two level-sensitive D-latches in a master-slave configuration results in an edge-triggered D-flip-flop (DFF). It can easily be verified that connecting the outputs of a DFF back to its inputs with inverted phase results in a toggle flip-flop. Thus, for every clock pulse, the output is toggled, resulting in a divide-by-two operation.


4.2.3 Quasi-Dynamic Operation

The question arises as to whether or not this SCL latch is static or dynamic. In other words, is there a lower frequency limit for this latch? The answer to this question depends on the sizing of the transistors in the latch [Mur95], and because of this, the latch topology is quasi-dynamic. To illustrate this, the PMOS load transistors are first modeled as resistors, with a resistor RpM. Consider the case when the output of the master D-latch, Qbi is first evaluated low and then held low. Since there is no tail-current source, the currents through M1 and M2 are not necessarily equal. The current values depend on the sizing of M1 and M2 and the input clock signals. The current pulling Qbi low during evaluation for the master DFF is then modeled as fl(I,), where f, depends on the sizing of M3 and M4, the voltages Q and Qb, and the evaluation current, I,. Thus, Qbi has a voltage during evaluation equal to

LOWEva,,l = Vdd - f (I1)RPM7. (4.1)






82

Likewise, during hold operation of the master D-latch, the current pulling Qbi low is modeled as f2(2), where f2 depends on the sizing of M5 and M6, the voltages Qi and Qbi, and the hold current, 12. Thus, Qbi has a voltage during hold equal to LOWHold = Vdd-f 2(2)RM7g. (4.2) If f2(12) > f (I), the output voltage swing of the latch is decreased, and the voltages Qi and Qbi might not be large enough to drive the next master-slave flip-flop.

For a given fl(Ii) to f2(12) ratio, a lower frequency limit for the 2:1 divider exists, set by how fast the nodes can change from LOWEval to LOWHold. As long as the input frequency is high enough, then LOWEval cannot fully charge to LOWHold, and the divider functions properly. This lower-frequency limit was observed in simulations, where the minimum frequency was approximately one-eighth of the maximum frequency for the chosen transistor widths. To make the quasi-dynamic latch fully static, first, tail current sources should be added and W1 should be set equal to W2. This then makes I1 equal to 12. Second, the widths of M3-6 should all be equal [Mur95], forcing fl(II) to be equal to f2(I2).


4.3 Injection Locking of SCL Frequency Divider

The second mode of operation for the 2:1 SCL divider is injection locking. Injection locking is the process of synchronizing an oscillator with an incident signal. Oscillators can be injection-locked to the fundamental [Adl46, Pac65, Uzu85], subharmonics [Zha92], and superharmonics [Sch71, Rat99] of the input signal. This section will examine the oscillation of the SCL divider and discuss how the divider can be injection-locked. The actual theory behind injection locking is presented in Appendix E. This appendix reviews the basic operation of injection-locked oscillators (ILO's), and presents the






83

locking range, its dependence on signal level, the steady-state phase error between input and output signals, and the phase noise of LO's.


4.3.1 Oscillation of SCL Divider

To understand the phenomenon of injection locking, the oscillation of the 2:1 divider with no input present must first be understood. When the input clock signal swings are very small, then CLK and CLKb are approximately equal at their common-mode value. Thus, both the master and slave latches are semi-transparent, allowing signals to propagate through both latches. Referring to Figure 4-1(b), if the delay from the gate to the drain of M3 is equal to tpd, then the total delay around the loop is equal to 4tpd. Thus,
1
the divider oscillates at a frequency equal to 4pd and the signal at the drain of M3 will 4trpd
lag the signal at the gate of M3 by 900. The divider circuit can thus be likened to a two-stage ring oscillator with an inversion in the loop'.

Additional understanding of the oscillator can be gained by solving for the small-signal loop gain of the divider. First, assume the two latches are identical and that each latch is symmetric (i.e., W3=W4, W5=W6, and W7=W8). Consider the case when CLK and CLKb are held at a constant and equal value. For this case, M1 and M2 become current sources. Cross-coupled transistors M5,6 can be represented as negative resistances with values of -1/gee, where gce is the transconductance of M5. The PMOS load can be modeled as a conductance, gp Finally, the total capacitance at the output nodes of the latch can be represented by CL. Figure 4-1(c) can then be simplified to Figure 4-2, by





1. An actual expression for the propagation delay through each latch was not derived, since
this delay can readily and more accurately be obtained with SPICE.






84


0(gP'gcc)'l (gPgcc)l

++ ++ b

SCL - bi -





Figure 4-2 Representation of source-coupled latch as a fully-differential amplifier and the 2:1
divider as an oscillator.


representing each source-coupled latch as a fully-differential amplifier. The small-signal loop gain (T) of the 2:1 divider can be shown to be as follows:


( g m3 2(4.3) P p- gCC + joCL)

For oscillations to begin, the phase of T should be 3600 and ITI > 1. Choosing ITI > 1 ensures start-up and causes the amplitude of oscillation to increase until non-linearities within the amplifier limit the amplitude. This amplitude condition requires gm3 > (OCLFor the phase of T to be 3600, the phase-shift through each differential amplifier should be 900. This corresponds to gp = gcc. This Barkhausen criterion for oscillators applies to small-signal sinusoidal oscillations. However, with a loop gain greater than one, the oscillator will become nonlinear and its amplitude and frequency will be determined by the large-signal characteristics of the circuit. The divider then behaves as a relaxation oscillator (also known as an astable multivibrator) [Sed91].

Relaxation oscillators are hysteretic comparators which charge and discharge capacitances located in a feedback loop. The oscillation frequency of a relaxation oscillator depends on the capacitance being charged, as well as the current drive of the






85

comparator. It is interesting to note that any ring oscillator whose propagation delay is set by the charging of RC circuits can be considered to be a relaxation oscillator [Ega99]; thus, the two models used to describe the divider's oscillation are equivalent.


4.3.2 Description of Injection Locking

Since the 2:1 divider can self-oscillate, it is possible to injection-lock the divider to an input signal whose frequency is close to either the fundamental, subharmonic, or superharmonic of the oscillation frequency. Thus, for a 2:1 injection-locked frequency divider (ILFD), the input signal is close to the second harmonic of the oscillation frequency, herein referred to as fiso. The locking mechanism of the oscillator, for an input signal at twice the natural frequency (a superharmonic), is created by nonlinear properties in the oscillator which generate intermodulation products close to the natural frequency [Rat99]. Referring to Figure 4-1(c), the drain node of M1 oscillates at twice the natural frequency; hence, this node is naturally suited to injecting a signal at this double frequency. An alternative and equivalent understanding of the circuit is that injection locking is being used to synchronize this "double-frequency" oscillator (drain of M1) at a frequency near fiso. This understanding is more straightforward and avoids the complications surrounding superharmonic injection locking.

The range of input frequencies over which the ILFD can be locked or "pulled" is a function of the input voltage swing and the frequency-dependent energy-loss in the oscillator. Appendix E shows that the voltage swing of the locking signal (VL) is related to the locking frequency range (Ao) and the oscillator output voltage swing (Vo), as follows: VL
- > AOA . (4.4) V0






86

Here, Vo is the voltage swing of the double-frequency oscillator (drain of M1), and A is equal to the derivative of phase-shift around the oscillator with respect to frequency. For a tuned oscillator, A is equal to 2Q/oo. Appendix F shows that the equivalent quality factor of a ring-oscillator is approximately equal to i/2. Thus, an injection-locked ring oscillator has the following input-swing to locking-range relationship (also known as the input sensitivity of the divider):

VL (ACo . (4.5) VIT0 (0

A voltage conversion gain can be defined for the frequency divider, equal to the ratio of the output swing to the input swing. Since the gain occurs at two different frequencies, it is referred to as conversion gain. This gain can be related to (4.5) by relating the output swing of the divider to the swing of the double-frequency oscillator. Thus, the circuit has "infinite" gain at fiso. Also, (4.5) shows that the further away the input frequency is from fiso, the larger the input voltage-swing has to be.

To control the oscillation frequency and conversion gain of the divider, the transistor sizes, input common-mode voltage, and supply voltage can all be adjusted. Referring to Figure 4-1(c), the speed of switching node Qi,bi HIGH depends on the size and IVgsI of M7,8 (note Vdd dependence), while the speed of switching node Qi,bi LOW depends on the sizes of M3,4 and MI,2 as well as the common-mode voltage on the clock inputs. For both cases, the switching speed is decreased by the presence of regenerative pair M5,6, since the positive feedback has to be overcome by the charging or discharging mechanism. Therefore, all of the transistor sizes, the supply voltage, and the clock common-mode voltage determine the self-oscillation frequency for the 2:1 ILFD. Note that the transistor sizes are






87

used to adjust the oscillation frequency during design, while the voltages can be used to

adjust the oscillation frequency during circuit testing.


4.3.3 Simulation of Injection-Locked Divider

This section will present simulation results for an ILFD, showing the fiso and

demonstrating the dependence of the input swing on the locking range. To extract fiso, the

8:1 divider is simulated using a transient analysis, with both CLK and CLKb held at

approximately mid-supply. As with other oscillator simulations, initial conditions are

required, where nodes Qi and Q are initialized to the supply and Qbi and Qb are initialized

to ground. Finally, since this is an analog simulation, the SPICE tolerances should be

decreased as compared to a standard digital simulation.

Figure 4-3 shows the simulated oscillation of the 8:1 ILFD, for a 0.25-gm CMOS

process. Since Vdd = 2.5 V, CLK and CLKb are set to ~1.3 V. The output of the first 2:1

divider is oscillating at 6.1 GHz, while the following 2:1 dividers are locked at 3.05 and

2 .4 --...! - ...... .......... i.... ... .... -.. ..
2.4 4
2.2 - -. --- - . div. 2 - -- - - -
6.1 GHz - ------2.0 -
1.8 --- .
1.6
0)m1.4 75 1.2
>1.0

0.8-------- ---- ---- -
0.6 - -0.4 -.-.----0div. 8
0.2 ------------ .S GH z ~. .
0 1 1 1 1
3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
Time (ns)
Figure 4-3 Simulation of self-oscillation of 8:1 source-coupled divider. The clock inputs
are held at 1.3 V, causing the first 2:1 divider to oscillate at 6.1 GHz for
Vdd=2.5V (fiso = 12.2 GHz).






88

10' 2.6
Divider stops locking
2.2

100 locking range 1.8 co
01. 10"1 1.0 fiso 0.6
Input, fin=8 GHz.
10-2 . .
7 8 9 10 11 12 13 14 0.2 0 1 2 3 4 5
Frequency (GHz) Time (ns)
(a) (b)
Figure 4-4 (a) Simulated input sensitivity of 0.25-gm CMOS 8:1 divider (b) Exponentially
decreasing input signal at 8 GHz and output signal of 8:1 divider, illustrating
divider's dependence on input voltage swing.


1.5 GHz, respectively. Thus, fiso is equal to 12.2 GHz. This simulation can be iterated through various device sizes to set fiso to the desired operating frequency. Additionally, this simulation can be run for the second and third 2:1 dividers to set their respective self-oscillation frequencies to one-fourth and one-eighth of the desired operating frequency. Note that for an 8:1 asynchronous divider, any of the three 2:1 dividers can oscillate provided that the input signal level to each divider is small. However, for the typical case where the first 2:1 divider oscillates with a large output signal, the following 2:1 dividers operate in a latched mode.

The input-swing versus locking-frequency-range relationship is shown in Figure 4-4(a), which plots the simulated input peak-to-peak voltage swing versus input frequency. To obtain these input swings using a single simulation, a damped sinusoid was injected into the divider, as shown in Figure 4-4(b). Eventually, the divider cannot lock to the input signal, at which point the input voltage swing is determined. This simulation is






89

repeated for every desired input frequency. As can be seen from Figure 4-4(a), the required input swing dips at the self-oscillation frequency (12.2 GHz), and then rapidly increases for higher frequencies2. The locking frequency range for a given input swing is equal to the distance between the left and right portions of the curve. This simulation verifies that larger input voltage swings result in larger locking ranges.


4.3.4 Implications of Injection Locking for Clock Distribution

The frequency divider will operate in the injection-locked mode for small-level input signals, and in the latched mode for large input signals. Therefore, to minimize the minimum detectable signal (MDS) of the receiver, the divider should be injection-locked, with its input-referred self-oscillation frequency near the operating frequency of the receiver. This provides an alternative interpretation for the wireless clock distribution system. Using this interpretation, each receiver is initially a free-running oscillator. The transmitted global clock signal is then used to injection-lock each oscillator and therefore synchronize the system. Noise and interference can corrupt the global reference signal as well as the phase of the oscillators themselves, resulting in phase noise or clock jitter.


4.4 A 10-GHz, 0.25-11m CMOS SCL Divider

4.4.1 Circuit implementation

A 10-GHz 128:1 source-coupled frequency divider has been implemented using a standard 0.25-ptm CMOS technology. A block diagram of the divider is shown in Figure





2. Note that for wireless interconnects, the region corresponding to input frequencies above
fiso should be avoided since the receiver's MDS would be severely degraded.






90


CLK Q - CLK Q - CLK Q
Inputs (2:1) (2:1) (2:1)
-C CLKb b - CLKb Qb > CLKb b SCL Divide-by-8


clk Q clk Q clk Q clk Q Output

D Q QbD Qb D Qb TSPC Divide-by-16

Figure 4-5 Block diagram of 128:1 divider.


Vdd Vdd Vdd Vdd

M3 clk~fM
Ms cl M6 9 Mg10

D cIkH M2 M5 clk M8 b M cilk M4 M7 M1



Figure 4-6 Schematic of true single-phase clocked (TSPC) latch.


4-5. The first 8:1 divider is implemented with SCL, while the lower frequency 16:1 asynchronous divider is implemented using true-single-phase-clocked (TSPC) logic [Yua89]. Figure 4-6 contains a schematic of the TSPC latch. Simulations indicate that this latch can operate up to - 2 GHz for the chosen device sizes. All of the devices for the SCL and TSPC dividers shown in Figure 4-1 and Figure 4-6 are implemented with a channel length of 0.25 gim, whereas the channel widths are given in Table 4-1.






91

Table 4-1 Transistor sizing for 0.25-Jim 128:1 frequency divider SCL 8:1 Divider
Transistors First 2:1 Second 2:1 Third 2:1
W1, W2, WIl, W12 12 gm 3 gm 3 gm W3, W4, W13, W14 6 gm 6 gm 6 gm W5, W6, W15, W16 2.1 gm 2.1 gm 4.2 Jm W7, W8, W9, W10 5.1 gm 2.55 gm 2.55 Jm TSPC 16:1 Divider
Sizes Are Same for each 2:1 TSPC Divider W2, W3, W4, W5, W6, W7, W8, W10 6 gm

W1, W11 3 Jm W9 12 gm
/12Jim














Figure 4-7 Die photograph of 0.25-gm, 128:1, 10-GHz source-coupled divider.


Figure 4-7 shows a die photograph of the 128:1 divider [Flo00a]. The die size is 730x510 gm2. On-chip 20-pF bypass capacitors are included between Vdd and ground for the 8:1 SCL divider, 16:1 TSPC divider, and output buffer. Separate supplies are used for each circuit to prevent any low-frequency supply noise from the latter stages of the divider from corrupting the high-frequency operation of the first 8:1 divider. In particular, when




Full Text
62
3.5.3 Transistor Characterization
The 0.18-|im LNA and the entire wireless interconnect system were designed to
operate at ~21 GHz. However, as will be shown in the subsequent sections, the operating
frequency for these circuits has shifted down to ~14 GHz. Referring to Table 3-4, it can be
seen that the measured inductance and its parasitic capacitance were slightly higher than
expected (within 0.1 nH). At ~20 GHz, these discrepancies can account for at most a
3-GHz frequency shift. To understand the remaining frequency shift, the designs have
been reviewed to make sure that all of the metal parasitics have been properly accounted,
and that the transistors have been specified correctly including the area and perimeter of
drains and sources. The remaining frequency shift is therefore believed to be primarily due
to the differences between the CAD intrinsic transistor model used during the design pro
cess and the actual fabricated transistor characteristics. Note that the models used to
design these RF circuits were the standard BSIM models, to which extrinsic parasitic com
ponents important at radio frequencies are added (e.g., Rgate, Rsub and capacitances asso
ciated with interconnecting the multiple fingers of the transistor). This technique has been
used very effectively in the past to yield simulated results which agree very closely with
measured results.
The S-parameters of individual NMOS and PMOS transistors have been measured
to quantify the difference in capacitance between the model and the actual transistors.
Both the NMOS and PMOS transistors have W=18.8 |im and L=0.18 |im, and are imple
mented with 0.94-jxm long fingers (total fingers = 20). The S-parameters of the transistors
are then transformed into y-parameters. Parasitics associated with the pads are
de-embedded through y-parameter subtraction of an open test structure. With the transis-


151
Figure 5-8 Simulation results of RMS jitter versus input SNR from 0.25-|im frequency
divider after a division by 2 and division by 4. The simulated jitter (points) is
compared to the predicted jitter (line) from (5.35).
(corresponding to integrating the power spectral density to lower frequencies), the differ
ence between the expected and measured jitter is anticipated for short observation times.
5.6 Sensitivity and Noise Requirements
5.6.1 Receiver Sensitivity
Having defined the SNR required at the input of the divider, the receiver sensitivity
can now be defined. Sensitivity is defined as the minimum signal level which can be
detected for the required SNR. At the divider input, this signal level is
Psig = SNR NdivT = SNR kTfQ F(f) G(f)df (5.37)
= SNR kT F(fa) G(f0) BWN
where (5.27) and (5.28) have been used. The temperature in this equation refers to the
equivalent temperature of the antenna, representing the total noise at the input of the


CHAPTER 1
INTRODUCTION
1.1 Global Interconnect Challenges
According to the 1999 International Technology Roadmap for Semiconductors
(ITRS) [SIA99], at the 0.10 and 0.05-|im technology generations, chip areas for high-per
formance microprocessors are projected to be approximately 620 and 820 mm2, respec
tively. On-chip global clock frequencies are projected to be 2 and 3 GHz, while local clock
frequencies are projected to be 3.5 and 10 GHz, respectively. These trends for high-perfor
mance microprocessors are shown in Table 1-1. In such integrated circuits (ICs), the delay
associated with global interconnectsthose which connect functional units across the
IChas become much larger than the delay for a single logic gate (herein termed
gate-delay). This is shown in Figure 1-1, which plots the global interconnect delay and
gate-delay versus minimum feature size [SIA99, Boh95]. As feature size decreases,
gate-delay decreases, illustrating the well-known fact that transistor scaling improves
device performance and chip density simultaneously. However, the propagation delay of a
voltage wave is approximately equal to 0.35RC/2 [Wes92], where / is the length of a wire,
and R and C are its resistance- and capacitance-per-unit-length. As the CMOS technology
is scaled to smaller feature sizes, both the RC time-constant [Boh95, Tau98] and the chip
area (l2) increase. Therefore, the global interconnect delay increases and quickly begins to
dominate the overall system delay.
1


APPENDICES
A THEORY FOR COMMON-GATE LOW NOISE AMPLIFIER 241
A.l Input Impedance 241
A.2 Gain 241
A.3 Noise Factor 242
B OUTPUT MATCHING USING CAPACITIVE TRANSFORMER 244
B. 1 Two-Element Matching Technique 244
B.2 Application to Capacitive Transformer in LNA 245
C DERIVATION OF NOISE PARAMETERS 247
C.l Equivalent Input Noise Generators 247
C.2 Noise Parameters in Impedance Form 248
C.3 Alternative Noise Parameters 250
D NOISE PARAMETERS OF MOSFET 252
D.l Transistor Model and Equivalent Input Noise Generators 252
D.2 Noise Parameters for Complete Model 254
D.3 Case Studies: Noise Parameters for Second-Order Effects 255
D.3.1 Case 1 Effect of Cgd 255
D.3.2 Case 2 Effect of GIN 257
D.3.3 Case 3 Effect of Rb, Excluding Cgb 258
E INJECTION LOCKING OF OSCILLATORS 260
E.l Overview 260
E.2 Theory for Injection Locking 261
E.2.1 Basic Model 261
E.2.2 Differential Locking Equation 262
E.2.3 Locking Range and Locking Signal Level 263
E.2.4 Steady-State Phase Error 263
E.3 Phase Noise of Injection-Locked Oscillators 264
F QUALITY FACTOR OF RING OSCILLATOR 265
G RELATIONSHIP BETWEEN JITTER AND PHASE NOISE 267
LIST OF REFERENCES 268
BIOGRAPHICAL SKETCH 278
vii


164
non-directive or isotropic antenna. Hence, the directivity is maximum in the direction of
maximum radiation. The higher the directivity, the stronger the antennas radiated energy
is for that direction. All practical antennas are directional (D>1), in that the directivity is
not constant for all directions. A subcategory of directional antennas is the omnidirec
tional antenna, whose directivity is essentially constant across a certain plane.
The gain between a transmitting/receiving antenna pair can be determined using
Friis transmission equation [Bal97], which describes the power received to the power
transmitted between two antennas as follows:
(6.1)
where e¡ is an efficiency representing loss in the conductors and dielectrics (equal to the
radiation resistance divided by the sum of the radiation resistance and the loss resistance
[Bal97]), r¡ is the reflection coefficient at the antenna terminals, D¡ is the directivity, and A
and r are the wavelength and separation distance, respectively. When characterizing an
antenna pair using S-parameters, this Friis transmission equation is equal to IS2il2
[KimOOa], which is equal to the transducer gain (GT), when the antennas are measured
with 50-Q transmission lines and termination impedances [Gon97]. To characterize the
antenna performance under matched conditions, a transmission gain (Ga) is defined as
G
(6.2)
(l-|Sn|2)(l-|S22|2)
Referring to (6.1), Ga is equal to the quantity eterDtDr(A/47tr)2. This gain is equal to the
power available at the output divided by the power delivered to the input, representing the
best possible gain for the antennas, where both antennas are conjugately matched.


246
the output of the cascode, Zse, including the drain inductor should first be calculated, as
follows:
z=(^W¡+MC+Cp2)-i-J=R+iX (R4)
where Rse and Xse are the real and imaginary parts of Zse, respectively. Setting Rp = RL =
50 2, and then using (B.l) yields the quality factor, Q, of the L matching network. Equa
tions (B.2) and (B.3) are then used to find C2 and Cj, respectively, as follows:
C2
1
n(X-QRie)
(B.5)
Cl
Q
50 to'
(B.6)
If either (or both) of these values is negative, it signifies that an inductor should be used in
place of the capacitor. If this is unacceptable for the LNA design, then the value of Ld (and
its associated parasitics) should be modified.


80
power consumption for a given operating frequency. Third, the use of a differential struc
ture reduces the switching noise of the latch, by keeping the supply current approximately
constant. Thus, the noise added by the divider is reduced, improving the output jitter.
A conventional SCL latch has a current source between the sources of Mj 2 and
Mu 12 and ground, resulting in constant tail current. This current source is omitted for low
voltage operation due to the limited supply head-room, causing the total current flowing
through the evaluate and hold stages to not be constant. However, for fast transitions, the
2:1 divider always has both an evaluate and a hold stage on, thus the supply current
remains relatively constant (reducing the switching noise).
The voltage Vbias is grounded to increase the maximum operating frequency by
operating M7 8 in the linear region, which lowers the RC time constant for nodes Q and
Qb. Furthermore, when the voltage of Q is increasing (during a low to high transition), the
drain-to-source voltage and the output resistance of M9 (in linear region) are decreasing.
This nonlinearity of M9 helps to pull up Q to its logic high, which in turn helps to increase
the maximum operating frequency at the same power consumption. Similarly, when Q is
decreasing, the nonlinearity tends to push down Q to its logic low. In addition to the
advantage of high speed operation, this topology has reduced power consumption com
pared to the topology in [Cra95], due to the elimination of the folded diode-connected
transistors from Q and Qb to ground.
4.2.2 Latched Operating Mode
The 2:1 divider implemented with SCL can operate in one of two modes. The first
mode is the standard digital operation of latching, while the second is an injection-locked
mode which will be presented in the next section. The latching mode of operation occurs


110
Figure 4-21 Output waveforms of 8:1 divider for each start-up condition
This effect can be demonstrated and analyzed using SPICE. Figure 4-22 shows the
simulated signals for two programmable 8:1 dividers operating at an input frequency of 30
GHz. In this simulation, the clock signals are always present, and the INI signal is being
switched repeatedly from high to low, as indicated in the plots. One divider is always ini
tialized to CNT = #000, while the other divider is always initialized to CNT = #111. Ide
ally, this should correspond to a phase offset of 45 between the two dividers each time the
divider is initialized. As can be seen from Figure 4-22, after the first two initializations, the
output waveforms take the desired 45 difference. However, after the third initialization,
the output phase difference is 90. The failure of the circuit can be seen by examining the
output waveforms from the first, second, and third 2:1 dividers (referred to as the 2:1,4:1,
and 8:1 outputs). Since one divider is initialized to state #000 and the other is initialized to
state #111, each time INI = 1, the 2:1, 4:1, and 8:1 outputs should be all zeros or all ones,
for the #000-divider and #111-divider, respectively. From the waveforms in Figure 4-22,


92
Signal
Generator
Figure 4-8
Bias-T A
Bias-T 2
A
Gnd-Sig-
180
Coupler
Sig-Gnd
I
t
Probe
Signal-
Ground
Probe
Oscope
50Q. Semi-Rigid Cables
Measurement setup used to characterize dual-phase frequency divider.
the output buffer switches its 50-2 load, a large current spike results, which can then
cause the supply voltages to ring with a long time constant. In addition to the bypass
capacitors, 50-Q resistors are included between the two inputs of the 128:1 divider to
ground. Since the input impedance of the divider is capacitive, and the source is 50 Q, a
reflection coefficient with magnitude of 1 results. This causes the input voltage to be twice
as large as that applied by the source, potentially damaging the input transistors.
4.4.2 Measured Results
A measurement setup has been developed to characterize the divider on-wafer, and
is shown in Figure 4-8. This setup consists of an external signal generator followed by a
180 coupler, which acts as a balun. Thus, the unbalanced signal from the sweeper is con
verted to differential input signals for the divider whose power is (ideally) 3 dB less than
the power at the input of the balun. To set the common-mode voltage for the input signals,
bias-Ts are placed on the differential and common-mode inputs for the 180 coupler. This
is preferable to placing the bias-Ts on the output of the coupler since phase mismatch in
the bias-Ts will be eliminated. Semi-rigid cables are then used to connect the balun to the
ground-signal-signal-ground probe. Finally, the divider is probed and the output is mea
sured using an oscilloscope. A key requirement for this measurement setup is that the


23
Vdd
Figure 2-4 Output matching equivalent circuits for CMOS LNA
matching network is preferable. As shown in Figure 2-2 and Figure 2-4(a), the output
impedance of M2 is transformed to 50 Q using a three-element matching network
consisting of a capacitive transformer (Cj and C2) and a shunt inductor (Ld). This
three-element network is also known as a n matching network. Compared to a two-ele
ment matching network (or L network), a II network allows multiple sets of component
values for Ld, Cj, and C2. Each set corresponds to a different loaded quality factor for the
II network. The two-element network, on the other hand, allows only one set of compo
nent values for Ld and C2 (Cj is no longer present), with the quality factor of the loaded
network fixed by the impedance transformation ratio [Bow82]. Therefore, the II network
provides added flexibility to the designer, allowing multiple values for the drain inductor.


229
with respect to the output signal. Finally, mismatches in the load capacitance of the clock
receivers will generate clock skew. These capacitances are assumed to be matched to
within 1%; thus, the skew due to this mismatch is also 1%.
The total worst-case estimated skew can now be obtained. Since each of the ran
dom skew components are independent, the individual variances can be summed. Also,
since the numbers listed in the previous paragraph represent 3-g variations, the total 3-g
I 2 2 2
variation is equal to a/3.8 + 3 + 1 = 4.9%. Therefore, the worst-case uncorrected skew
for a wireless clock distribution system is estimated to be the mean skew (6.3%) plus the
3-g random skew (4.9%), resulting in 11.2% skew, with respect to the output signal. This
skew estimation assumes that there is no AGC in the receiver. With AGC, the skew due to
amplitude mismatch would be eliminated, and the skew due to process variation would be
reduced. Assuming that AGC is used, the worst-case skew would be less than 10.2%.
Finally, the use of more sophisticated biasing techniques should reduce the pro-
cess-induced skew further.
As was shown in Chapter 4, programmable dividers can be used to start-up the
receiver in a different state and to reduce clock skew. The types of clock skew that can be
corrected for include both systematic and random skew, since both are time-invariant. For
an 8:1 divider, the skew will be limited to under ~~, or 6.25%. Thus, the worst-case
16
skew using a programmable divider is 6.25%, which meets the system specification of less
than 5-10% of skew. Note, however, that to correct for the skew using programmable
dividers requires a start-up methodology for the clock receivers which would estimate the
uncorrected skew using a phase detector, modify the states of the receiver accordingly, and
then synchronously release the clock receivers from the initialized state (refer to section


56
Figure 3-10 (a) Measured S-parameters of differential LNA. (b) Measured gain and NF for
single-ended NF for both p+ and p" substrates.
~50% of that for p' substrate, and the inductance is smaller. Therefore, the resonant fre
quency of the p+ LNA is about 25% higher due to the decreased inductance. Simulation
results show that ~4 dB of the gain reduction can be attributed to the 50% reduction in the
Q for Ld. Another ~1 dB comes from the additional loss due to the increase in rLg.
The NF of the single-ended p+ LNA is ~7.4 dB, which is 2.5 dB higher than the
noise figure for the p" LNA. The increased NF is partly attributed to the increased thermal
noise generated by the series resistance of the inductors. However, a large part of the
increased NF is due to the decreased gain, resulting in higher input-referred noise from
output noise components. Although the substrate resistance is higher for p\ the use of
multiple-fingered transistors with many substrate contacts helps to mitigate its effect on
NF. These results reveal the benefits of high-resistivity substrates for RF applications.
NF (dB)


69
3-13(b) shows a die photograph of the 0.1 -|im LNA. The die size is 0.44x0.51 mm2.
Ground-shielded diamond-shaped pads are used at the input and output to reduce the para
sitic capacitance to the substrate. Three on-chip spiral inductors of 0.55 nH are used in the
circuit. Finally, on-chip accumulation-mode MOS capacitors of 4.8 pF are used to bypass
Vdd> Vgi and Vgc.
3.6.2 Source-Follower
The source-follower is used to shift the dc bias level from the output of the first
stage (at Vdd) to the input of the third stage. The bias point of the source-follower is con
trolled through input Vgc, which controls the current through M2 and sets the Vgs of M3. A
source-follower with a capacitive load can have a negative input conductance. Since
inductive source-degeneration provides a positive, real input impedance for the transistor,
by reason it follows that capacitive source-degeneration provides a negative, real input
impedance for the transistor.
Referring to Figure 3-18, it can be shown that the input admittance looking into M3
is as follows:
Re{Ysf) = Gsf =
03 [8TCgs3~ Sm3Cgs3CLT]
(&3 + Sr)2 + 2(Css3 + C'ir)2
(3.2)
Im(Ylf) = Blf = <*3 + ij0 +< (Css3 + CLTf
where gT is the total output conductance at the source-follower output node (equal to
gds3+gds2) and CLT is the total capacitance at the same node (equal to
Cgs4+Csb3+Cdb2+Cgd2) The voltage gain through the source-follower buffer is as follows:
^ ^/w3 +
V'Sf 8m3 + St + j((Cgs3 + CL.t)
(3.4)


250
Xopt = -Xc. (C.ll)
The minimum noise figure (Fmin) can now be obtained by plugging in (C.10) and (C.l 1)
into (C.9), as follows:
^>,= 1+2 Gn(Ropt + Rc). (C.12)
Thus, the total noise factor can be represented as
? = F,,n + Y\.\zs-zop£], (C-i3)
s
where the noise parameters in impedance form are Fmin, Gn, and Zopt.
C.3 Alternative Noise Parameters
The noise performance of any two-port network can be represented by any four
noise parameters (e.g., [Fmin, Gn, Ropt, Xopt], [Fmin, Rn, Gopt, Bopt], etc.). Surveying the
results in the previous section, an alternative set of parameters which come directly from
the equivalent input noise generators can be defined, from which any other noise parame
ter set can be easily obtained. These alternative parameters are Rn, Gn, and Pn, where Gn is
given by (C.6), while
R
n
4kT
(C.14)
P
n
4kT '
(C.15)
Here, Pn is a dimensionless correlated power ratio. These parameters naturally are
needed when trying to obtain any of the traditional noise parameters; thus, they are conve
nient for numerical techniques. Examining these new parameters, it can easily be shown
that


195
Output to
Spectrum
Analyzer
(a)
-50
1 1 1
1 1 1
i
Ol
o
1 1 1 r~
1 1 1
. d = 3 mm
d = 5.6 mm
.
o
CD
1
-60 dBm
-60
-
-
n-70
n -70
-69 dBm
X
X
E
E
m
CQ
-80
-80
m

-90
-90
.100' 1 1 >1 1 1 1 -loot 1 1 1__1 1 i 1
15.08 15.09 15.10 15.11 15.12 15.095 15.105 15.115 15.125 15.135
Frequency (GHz)
(b)
Frequency (GHz)
(c)
Figure 6-24 (a) Measurement setup used to characterize wireless clock transmitter.
(b) & (c) Output spectra from receiving antenna detected from clock transmitter
across a 3- & 5.6-mm distance.
using receiving antennas. Figure 6-24(a) shows a block diagram of the measurement setup
used to test the clock transmitter. The DC control and supply voltages were supplied to a
transmitter driving a zigzag dipole antenna. The output spectrum was then obtained by
probing receiving antennas located at 3-mm and 5.6-mm separations from the transmitting
antenna. Figures 6-24(b) and (c) show the resultant output spectra measured at 3-mm and
5.6-mm distances, respectively, having peak output power levels of -60 and -69 dBm at
~15 GHz. The transmitter consumes ~48 mW of power. Thus, an on-chip clock transmitter
with an integrated antenna is demonstrated for the first time.


223
Figure 7-6 Simulated process variation of (a) voltage gain and (b) phase-shift of receiver
from antenna input to input of frequency divider.
variation in the resonant frequency can be large. The maximum gain at 8.5 GHz varies
from 27 to 36 dB, with the resonant frequency varying by -120 MHz. As can be seen, the
gain variation is primarily due to the resonant frequency shifting, rather than the peak gain


266
Taking the derivative of (F.l), and applying the result to (F.3) results in [Ega99]
Q =
n
2'
(F.4)
Thus, ring oscillators have very low Qs. As a result, they can be injection locked over a
very large frequency, since the locking frequency range is inversely dependent on Q, as
shown in Appendix E. Also, due to the low Q, ring-oscillators typically have poor phase-
noise performance, resulting in higher jitter.


271
[GraOO]
G. Gramegna, A. Magazz, C. Sclafani and M. Paparo, Ultra-wide
dynamic range 1.75dB noise-figure, 900 MHz CMOS LNA, ISSCC Digest
Technical Papers, pp. 380-381, Feb. 2000.
[GuoOl]
X. Guo, R. Li, and K. 0, Impact of a propagating layer underneath a sili
con wafer on the performance of integrated antennas, to be published,
2001.
[GutOO]
V. Gutnik and A. P. Chandrakasan, Active GHz clock network using dis
tributed PLLs, IEEE J. Solid-State Circuits, vol. 35, no. 11, pp. 1553-
1560, Nov. 2000.
[Haj99]
A. Jahimiri, S. Limotyrakis, and T. H. Lee, Jitter and phase noise in ring
oscillators, IEEE J. Solid-State Circuits, vol. 34, no. 6, pp. 790-804, June
1999.
[Har73]
K. Hartmann and M. Strutt, Changes of the four noise parameters due to
general changes of linear two-port circuits, IEEE Trans. Electron Devices,
vol. 20, pp. 874-877, Oct. 1973.
[Hay98]
G. Hayashi, H. Kimura, H. Simomura, and A. Matsuzawa, A 9mW
900MHz CMOS LNA with mesh arrayed MOSFETs, Symp. VLSI Circuits
Dig. Tech. Papers, pp. 84-85, June 1998.
[Ho98]
Y.-C. Ho, K. Kim, B. Floyd, C. Wann, Y. Taur, I. Lagnado, and K. O., 4-
GHz and 13-GHz tuned amplifiers implemented in a 0.1 -|im CMOS tech
nology on SOI, SOS, and bulk substrates, IEEE J. Solid-State Circuits,
vol. 33, no. 12, pp. 2066-2073, Dec. 1998.
[HoOO]
Y.-C. Ho, Implementation and improvement for RF low noise amplifiers
in conventional CMOS technologies, Ph. D. Dissertation, University of
Florida, Gainesville, FL, 2000.
[Hua98]
Q. Huang, P. Orsatti, and F. Piazza, Broadband, 0.25-|im CMOS LNAs
with sub-2dB NF for GSM applications, Proc. Custom Integrated Circuits
Conference, pp. 67-70, May 1998.
[Hun98]
C.-M. Hung, Y.-C. Ho, and K. O, High-Q capacitors implemented in a
CMOS process for low-power wireless applications, IEEE Trans. Micro-
wave Theory and Techniques, vol. 46, pp. 505-511, May 1998.
[HunOOa]
C.-M. Hung, Investigation of a multi-GHz single-chip CMOS PLL fre
quency synthesizer for wireless applications, Ph. D. Dissertation, Univer
sity of Florida, Gainesville, FL, 2000.


109
Xf
State Initialization
Xi
X,
Inputs
S1 SO
elk Q
2:1 Prog.
Divider
elkb Qb
INI
3O
S1 SO
elk Q
2:1 Prog.
Divider
elkb Qb
INI
S1 SO
elk Q
2:1 Prog.
Divider
Olclkb Qb
INI
Outputs
Figure 4-20 Block diagram of programmable 8:1 frequency divider
(CNT) of the divider (i.e., decimal 0-7). This CNT will come from the 3-bit A/D converter
after the phase detector.
4.6.4 Simulated Results
Basic functionality of the initialization circuit is demonstrated in Figure 4-21,
which shows the simulation results for an 8:1 divider for X2XjX0 ranging from binary
#000 to #111. As can be seen, the output phase can be varied in increments of 7t/4 (2rc/M)
or 45. This simulation was performed under identical conditions for each state. Thus, the
divider was initialized at time zero, and then after a certain delay the INI signal was
released along with the input clock signals.
There is an inherent flaw, however, in the implementation of this programmable
divider which subsequent measurement results revealed. The failure surfaced when
attempting to repeatedly initialize the divider to the same CNT. When the divider had
already been running and then INI was invoked and then released, the divider would ini
tialize to any of the eight possible states. This means that the actual initialized value for a
given CNT depends on the current state of the divider when INI transitions high.


208
point of zero input skew. From this intersection, the skew added by the measurement sys
tem can be determined to be -35 ps. This then allows the on-chip skew to be determined by
subtracting -35 ps from the measured off-chip skew, resulting in the second x-axis in Fig
ure 6-30. Thus, the minimum on-chip skew is 25 ps, which is 4.7% of the local clock
period. By adjusting the start-up states of the dividers, the skew can be reduced to under
5%, which is the requirement set for global clock distribution systems. This same skew
extraction methodology has been applied to the case where Zgc is used as the transmitting
antenna, showing 20 ps of skew, or 3.75%. This is significant, since the measurement case
has at least 3.6-4.4% of systematic skew due to time-of-flight delay mismatch and ampli
tude mismatch, in addition to process variation.
Unfortunately, the uncorrected or unprogrammed skew can not be determined
from this measurement. This is due to the state-dependent initialization failure of the
dividers. If the dividers initialize properly, then the uncorrected skew would be the skew
measured when both dividers are initialized to the same state. However, since each divider
currently initializes to a random state, it is impossible to know if the dividers are initial
ized to the same state, which would yield the uncorrected skew. The results do, however,
show that programming the dividers can tune out systematic skew.
6.7.4 Measured Jitter of Clock Receivers
The jitter of these receivers has been measured using the setup already shown in
Figure 6-27. To measure the peak-to-peak jitter, the triggered waveforms can be displayed
with an infinite persistence. The jitter is measured off of the input 15-GHz signal rather
than one of the 1.875-GHz output signals. This is because the trigger signal is also a
1.875-GHz clock signal which has its own jitter. Thus, measuring the jitter of the 15-GHz


108
S1 S1 S1 S1 SO SO SO so
Figure 4-19 Schematic of initialization circuitry implemented with complementary
pass-transistor logic.
restore the voltage levels. A benefit of CPL is that only NMOS transistors are used for the
logic array, allowing for very compact layout.
Finally, to realize the 8:1 programmable divider, three 2:1 programmable dividers
are cascaded. Since the state-control inputs S1 and SO are the desired input and output val
ues of each 2:1 divider, then SO for the first 2:1 divider is equal to SI for the second 2:1
divider, and SO for the second 2:1 divider is equal to SI for the third 2:1 divider. The
resultant 8:1 divider block diagram is shown in Figure 4-20. Input SI for the first 2:1
divider is set to ground, since simulations show that a total of only eight states rather than
sixteen are possible for this divider (i.e., SI = 0 yields 8 states and SI = 1 yields the same
8 states). State control inputs X2XjXo form a binary number equal to the desired count


133
M (i-|rcl2)(i-|raJ)
11 r r |2
A cA ant\
(5.3)
where rc is the reflection coefficient of the circuit and rant is the reflection coefficient of
the antenna, both with respect to 100 £2 (differentially). In the wireless clock distribution
system, there are two mismatch factorsone at the transmitter and one at the receiver;
thus, these factors should be maximized. The specification for the total mismatch loss
(Lmm) in the system is 1 dB, given by the following:
101og,.L xJ-
(5.4)
10v1*1 anil lrl ant
Typically, impedance mismatch is specified as either a voltage-standing-wave ratio
(VSWR) or a reflection coefficient. These parameters are not straightforward in this sys
tem, due to the complex impedances of the source (circuit) and the load (antenna). As was
done in [Gon97], equation (5.2) can be rewritten as follows:
PM = Pays = Pays 0 |r/). (5-5)
where req is an equivalent reflection coefficient, defined by
Ir,| = (5-6)
Now a VSWR can be defined as
VSWR =
1+lrJ
Hrl
(5.7)
For a 1-dB total mismatch loss, the equivalent specifications for Teq and VSWR at each
antenna are -9.6 dB and 2.0, respectively.
Given an antenna impedance (or reflection coefficient), the range of allowable cir
cuit impedances (or reflection coefficients) is required. T^ is expressed in terms of rc and
rant by plugging (5.3) into (5.6), resulting in


214
Figure 7-1
size factor: 1 a
stage number: n=1 n=2
'Load
Figure 7-2 Schematic of tapered buffer used in global clock distribution. Each stage is a
times as large as the previous buffer stage, resulting in minimum delay.
The total global capacitance can be allocated among three components, as follows:
Co, Cw, and CL [Flo99b]. These components are illustrated in Figure 7-1. CG is the equiv
alent capacitance of the highest level network which delivers the clock from its source to
various sectors or spine locations distributed throughout the chip. CG includes the total
capacitance of the final driver stage (herein termed the sector buffer) plus any buffers lead
ing up to the sector buffer. The multi-stage buffers, shown in Figure 7-2, are assumed to be
tapered, where each buffer stage is a times as large as the previous buffer stage. It can be
shown that this results in minimum delay through the buffer chain [Mea80]. The total
capacitance of an N-stage buffer driving a capacitive load, Cj^^, can be estimated as


161
Table 5-1 System specifications for wireless clock distribution at 15 GHz.
Performance Metric
Specification
Global clock frequency
15 GHz
Local Clock Frequency
1.875 GHz
Output Clock Skew
< 5% (of local clock period)
Output Clock Jitter
Peak-to-peak cycle-to-cycle <3%
RMS cycle-to-cycle <0.5%
Transmitter Power
10 dBm
Antenna-to-Antenna
Gain with Interference+
-47 dB at 0.5-cm separation
-63 dB at 1.5-cm separation
(both including 7-dB interference loss)
Receiver Gain+
63 dB
(Gain = antenna-pair loss)
(LNA resonant freq. = fiSO of divider)
Matching Between
Antenna and Circuit
Total mismatch loss < 1 dB;
req < -9.6 dB; VSWR < 2.0
Receiver Sensitivity+
-54 dBm
Noise Figure+
14 dB
IIP3+
-12 dBm
Amplitude Matching
at Divider Input
(for <0.5% clock skew)
within 0.3 dB
(worst-case: at least 1 divider at MDS)
within 2.9 dB
(all dividers at least > MDS + 3 dB)
Signal-to-Noise Ratio
at Divider Input
12 dB
(for < 0.5% RMS jitter)
Receiver Power Consumption
40 mW per receiver at 1.5 V^d
+Note: These specifications are dependent on global clock frequency.


99
propagation delays, of the clock signals has to be under a certain tolerance (typically less
than 5-10% of a clock cycle). The difference has a static or time-invariant component
resulting in clock skew and a dynamic or time-variant component resulting in clock jitter.
For the wireless clock distribution system, the static skew has both random and
systematic components. Random clock skew is that which is not known a priori, due to
gain and phase differences amongst receivers. These differences will be caused primarily
by process variation and temperature gradients. Also, currently lumped into the random
category are differences in the antenna gain, phase, and impedance, due to metal structure
interference effects [YooOO, KimOOb], Note, however, that as both the antenna and system
electromagnetic modeling become more robust and accurate, this skew will shift from ran
dom to systematic skew.
Systematic clock skew is caused by differences in flight times of the electromag
netic wave between the transmitter and receivers. Referring to Figure l-2(a) for a
intra-chip wireless clock distribution system, the difference in path lengths between the
receivers nearest and farthest from the antennas can be a significant portion of the chips
linear dimension. For example, for a 25 x 25-mm2 chip with an evenly distributed grid of
16 receivers, the maximum path-length difference is ~9 mm. At 15 GHz, this is close to
the global clock signals wavelength in silicon-dioxide (10.1 mm) and about 1.5 times the
wavelength in silicon (5.8 mm), causing a significant phase mismatch. For an inter-chip
distribution system, the path lengths will be much more equalized, which is a benefit over
the intra-chip distribution system.
One method to remove the systematic skew is to ensure that the global clock sig
nals have the same phase at each receiving antenna. This can be accomplished by forcing


24
Figure 2-4(b) shows an equivalent circuit for the output matching network, includ
ing the parasitic series resistance (rLd) and shunt capacitance (CLd) associated with Ld.
The output impedance of the cascode is represented as a resistance RP2 (on the order of
kilo-ohms), in parallel with a capacitance CP2, where CP2 consists of Cgd2 and Cdb2. For
optimal power transfer, the matching network would transform Rp^ to 50 1 However, this
implies a quality factor of Q = jRp2/50 1 for the matching network, which is on the
order of 10 for R^ = 5 kil Since the matching network includes an on-chip spiral induc
tor, the quality factor of the network is limited by that of the inductor, QLd. To include the
finite Q of the spiral inductor, r^ can be transformed to an equivalent resistance shunting
the inductor, as follows:
RpLd = (QLd + ^)rLd=QLd rLd (2.18)
Since RPLd is typically much less than RP2, the matching network transforms a complex
source, consisting of RPLd in parallel with the capacitance Cpj+Cj^j, into the 50-2 load.
A dilemma now appears, in that the impedance to be transformed (RPLd) is a func
tion of the network providing the impedance transformation (Ld). Therefore, an iterative
approach can be taken using gain and output matching as criteria, as follows: (1) an initial
value is chosen for Ld, (2) RPLd and are then estimated along with ZM2 in Figure
2-4(a), (3) values for Cj and C2 are designed to achieve a 50-Q match, and (4) S21 and S22
are checked against their specifications and Ld is updated if necessary. The initial value for
Ld is chosen based on the desired gain, using formulas derived in the next section. The
equations for Q and C2 can be readily derived [HoOO], and are given in Appendix B.
Thus, any standard two-element matching technique can be used to choose Cj and C2 for
completing the match of ZM2 to 50 Q (e.g., [Bow82]).


236
potentially be decreased. Thus, cost-functions for both the active area and the metal area
should be developed to evaluate the total area cost of a wireless interconnect system.
Another intangible feasibility item is the ability to accurately verify if the system
will function properly, known as design verification. With any global clock distribution
system, it is imperative to meet the clock requirements with first silicon. For conven
tional interconnect systems, sophisticated computer aided design (CAD) tools have been
developed to model the transmission-line effects, extract the parasitics associated with a
line, and model current return-paths. For a wireless clock distribution system, CAD tools
would also have to be developed, or at least improved, to model the antenna characteris
tics, the effects of metal interference structures between the antennas, the digital switching
noise, and the RF performance of the circuits (requiring accurate inductor and transistor
modelling). While this hurdle is not insurmountable, it is quite large, pointing out the need
for additional research.
7.7 Conclusions and Future Work
7.7.1 Feasibility Summary
The feasibility of a wireless clock distribution system has been evaluated, yielding
the following results. First, the power consumption of a wireless clock distribution system
is similar to that of conventional systems, where the clock receivers and transmitter will
consume less than 2% of the system power. Second, the clock skew of the wireless clock
distribution system will be less than 6.25%, while the peak-to-peak clock jitter should be
less than 3%, and both of these numbers have been verified with measurements. Thus, a
wireless system should be able to meet the skew and jitter requirements for clock distribu
tion systems. Third, process variation has been shown to affect both the skew and the basic


243
Neglecting all other sources of noise except the thermal noise in the MOS transistor
(lumped only at the drain), the total short-circuit output current can be shown to be
*sc ^mVRs i p
1 + 8mKS
(A.6)
where Gm is defined in (A.2) and id is the channel thermal noise lumped at the drain. This
drain thermal noise has a PSD of
.2
= 4 kTygdo,
(A.7)
where y is a bias-dependent parameter (0.67 for long-channel devices), k is Boltzmanns
constant (1.38 x 10'23 J/K), T is absolute temperature, and gdo is the short-circuit drain
conductance. Since these two noise sources are uncorrelated, the output PSD is
.2 ~2
lsc =
~ ,ld
vRs +
8m
(A.8)
where (A.2) has been substituted. Therefore, the noise factor is
ld 4kTY8do
F = l + ^= = 1+ ,
2 2 2
8mVRs
= 1+1.
(A.9)
g^(4kTRs) a
where a is the ratio between the device transconductance (gm) and the short-circuit drain
conductance (gd0)-


76
implemented on both p+ and p' substrates. The differential LNA on p" substrate provides
10-dB of gain, with a 4.9-dB noise figure, consuming 21 mW from a 2.5-V supply. Using
a p+ substrate degrades the gain by ~ 6 dB and the noise figure by ~ 2.5 dB, due primarily
to reduced inductor Q.
A 13.3-GHz differential source-degenerated LNA with 8 dB of gain has been
implemented in a 0.18-|im CMOS technology. The circuit was designed to operate at 21
GHz; however, larger than expected inductance and transistor capacitance decreased the
resonant frequency. This shows that as the operating frequency of the circuit increases
above 10 GHz, modeling of the transistors and passive components becomes increasingly
important. Currently, the accuracy is not scaling with frequency, indicating that larger
deviations between simulated and measured performances will occur at higher frequen
cies. Although copper interconnects have increased the inductor Q in the 0.18-pm technol
ogy, an unplanned thinning of the metal layer during fabrication has mitigated the overall
improvement. Even still, inductor Qs of ~28 at 15 GHz have been obtained. A test-circuit
containing the differential source-degenerated LNA plus a pair of source-followers has
been implemented, showing a gain of 21 dB at 14.4 GHz.
Finally, using a 0.1-|i.m partially-scaled SOI CMOS technology, a multistage tuned
amplifier with a common-gate input and source-followers has been implemented.
Source-followers generate negative conductance and provide a dc level-shift. The input
admittance and gain of this stage were presented. This amplifier achieves 7.3 dB of gain at
23.8 GHz, with positive gains beyond 26 GHz. This result is comparable to the highest
operating frequency to date [Yan99b] for CMOS amplifiers.


75
Table 3-6 Summary of measured results for 23.8-GHz SOI CMOS tuned amplifier.
Parameter
Result
Parameter
Result
Resonant Frequency
23. 8 GHz
NF
10 dB
S2i
7.3 dB
PidB(in)
-16.2 dBm
Sn
-45 dB
nP3
-7.8 dBm
S22
-9.4 dB
Vdd
1.5 V
S12
-27 dB
Supply Current
53 mA
A summary of the results for the tuned amplifier is shown in Table 3-6. Excellent
gain and isolation is obtained for CMOS circuits at this frequency. While only moderate
noise and linearity performance are obtained, once a fully-scaled CMOS technology with
a more advanced back-end process is used, the results should dramatically improve. These
improvements will be due to increased inductor Q, decreased parasitic capacitance which
allows the inductance and hence gain to increase, and finally further circuit optimization.
Finally, these results show that a 0.1 -|im CMOS technology should be to able support RF
applications above 20 GHz, and suggest even higher operating frequencies for further
scaled CMOS technologies.
3.7 Summary
The validity of the LNA design methodologies presented in the previous chapter
was demonstrated through implementation of a 0.8-|im, 900-MHz CMOS LNA achieving
a 1.08-dB minimum noise figure at 30 mW, and a 1.5-dB minimum noise figure at 8 mW.
These results meet all of the required performance metrics for LNAs, demonstrating the
competitiveness of CMOS LNAs as compared to GaAs or silicon bipolar technologies.
Using a 0.25-(xm CMOS process, 8-GHz single-ended and differential LNAs have been


200
respectively. Thus, there is a 25.7-dB and 34.7-dB deficit in system gain for achieving a
fully integrated half- and full-chip wireless clock distribution system, respectively. A pri
mary reason for this deficit is the reduced thickness of metals 5 and 6. As has been shown,
the effect of lower than expected inductor Q and reduced operation frequency has reduced
the overall system gain. First, the 50% Q degradation reduced the PA gain by at least 14.5
dB (Q and gm effect) and has reduced the LNA gain by approximately 6 dB. Second, refer
ring to Figure 6-13(a), operating the antennas at 15 GHz, as opposed to at 21 GHz, results
in approximately 8 dB less transmission gain for the full-chip transmission distance. Also,
the mismatch in the VCO frequency range and peak gain frequency for the PA led to
another 8.5-dB degradation of the system gain (the PA appears to currently be a sensitive
spot in the system). Therefore, a 37-dB difference can be accounted for, indicating that the
fully-integrated system would have worked if the copper thickness was doubled and the
operating frequency had been above 18 GHz. As it is, by just doubling the metal thickness
of metal 5 and 6 layers, the system gain will increase by approximately 20.5 dB, which
should be enough to allow operation of the fully-integrated half-chip distribution system.
As mentioned earlier, these designs were implemented as part of the SRC Copper
Design Challenge. Due to circuit fabrication problems at UMC and the thinning of the
top-layer metals, the masks are being regenerated and another run is forthcoming. With
this new run, it is expected that all of the circuit performances will dramatically improve
and that simultaneous operation of a clock receiver and transmitter will be achieved.
6.7 Double-Receiver Wireless Interconnect
A critical test for the wireless clock distribution system is to demonstrate two
clock receivers operating simultaneously and synchronously from the same transmitted


117
Figure 4-27 Measured outputs of phase detector used to characterize initialization circuitry.
state-dependent initialization failure had not yet been identified. One divider was always
initialized to state #000, while the other dividers state was programmable. Measurements
of the two-divider test structure reveal that the output phase difference between the two
dividers can take on one of eight discrete values, indicating that the start-up phase of the
dividers was being adjusted. These eight values are shown in Figure 4-27 versus the ideal
programmed state. However, for each programmed state, the output would not take on the
same phase value for multiple repetitions of the experiment (i.e., INI switching on and
off). For example, when the divider was programmed to state #100, this should correspond
to a 180 phase difference and the output of the phase detector should be at its minimum
point. However, the phase detector output would take on any of the eight possible values,
since the initialized value depends on the current state of the divider. Thus, every program
mable state yields any of the eight possible phase values, indicated by dotted lines in Fig
ure 4-27. This shows that the initialization process is not working correctly, which led to
the development of the new DFFs to remove this state-dependence. The fact that the


176
Dipole Antenna
Low Noise Amplifier
(with source-followers)
Figure 6-8
Frequency Divider
and Output Buffers
Die photograph of 0.25-|xm clock receiver with linear dipole antenna.
mismatch. Additionally, multiple substrate connections should be included for both the
LNA and divider. In particular, to minimize any substrate crosstalk, a guard-ring structure
should separate the LNA from the divider. Finally, the LNA and buffer should have a sep
arate supply voltage from the frequency divider.
Figure 6-9(a) shows the measurement setup used to demonstrate a single-receiver
wireless interconnect. A 7.4-GHz global clock signal is externally generated and then
amplified to 21 dBm to overcome the low antenna transducer gain. This signal is con
verted to a balanced signal and injected to the on-chip transmitting antenna. The global
clock signal propagates 3.3 mm across the chip to the receiver, whose output is measured
using an oscilloscope. Figure 6-9(b) shows plots of the input voltage (before the external
amplifier) to the transmitting antenna and the output voltage for the wireless clock
receiver. As can be seen, a 925-MHz (7.4 GHz divided by eight) local clock signal is gen
erated, demonstrating the operation of the single-receiver on-chip wireless interconnect
for clock distribution. The reduced voltage swing at the output is from driving a 50-2
load. As inferred from the measured antenna-pair transducer gain of -64 dB, the received
signal level is -43 dBm at the LNA input. Since the receiver would not lock to the


A CMOS WIRELESS INTERCONNECT SYSTEM
FOR MULTIGIGAHERTZ CLOCK DISTRIBUTION
By
BRIAN A. FLOYD
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
2001


276
[War89]
K. Ware, H.-S. Lee, and C. G. Sodini, A 200-MHz CMOS phase-locked
loop with dual phase detectors, IEEE J. Solid-State Circuits, vol. 24, no. 6,
pp. 1560-1568, Dec. 1989.
[Was 00a]
K. Washio, E. Ohue, K. Oda, R. Hayami, M. Tanabe, H. Shimamoto, T.
Harada, M. Kondo, 82 GHz dynamic frequency divider in 5.5 ps ECL
SiGE HBTs, ISSCC Digest Technical Papers, pp. 210-211, Feb. 2000.
[WasOOb]
K. Washio, R. Hayami, E. Ohue, K. Oda, M. Tanabe, H. Shimamoto, and
M. Kondo, 67-GHz static frequency divider using 0.2-|im self-aligned
SiGe HBTs, IEEE Radio Frequency Integrated Circuits Symp., pp. 31-34,
June 2000.
[Web97]
C. F. Webb, C. J. Anderson, L. Sigal, K. L. Shepard, J. S. Liptay, J. D. War-
nock, B. Curran, B. W. Krumm, M. D. Mayo, R J. Camporese, E. M.
Schwarz, M. S. Farrell, R J. Restle, R. M. Averill III, T. J. Siegel, W. V.
Houtt, Y. H. Chan, B. Wile, Nguye, A 400-MHz S/390 microprocessor,
IEEE J. Solid-State Circuits, vol. 32, no. 11, pp. 1665-1675, Nov. 1997.
[Wes92]
N. Weste and K. Eshraghian, Principles of CMOS VLSI Design, A Systems
Perspective, 2nd Edition, Reading, MA: Addison Wesley, 1992.
[WetOO]
M. Wetzel, L. Shi. K. Jenkins, R. del la Houssaye, Y. Taur, P. Asbeck, and
I. Lagnado, A 26.5 GHz silicon MOSFET 2:1 dynamic frequency
divider, IEEE Microwave and Guided Wave Letters, vol. 10, no. 10, pp.
421-423, Oct. 2000.
[WurOO]
M. Wurzer, T. Meister, H. Knapp, K. Aufinger, R. Schreiter. S. Boguth, L.
Treitinger, 53 GHz static frequency divider in a Si/SiGe bipolar technol
ogy, ISSCC Digest Technical Papers, pp. 206-207, Feb. 2000.
[Yan99a]
H. Yan, M. Biyani, and K. K. O. A high-speed CMOS dual-phase
dynamic-pseudo NMOS ([DP]2) latch and its application in a dual-modu
lus prescaler, IEEE J. Solid State Circuits, vol.34, no. 10, pp. 1400-1404,
Oct. 1999.
[Yan97]
H. C. Yang, L. K. Lee, R. S. Co, A low jitter 0.3-165 MHz CMOS PLL
frequency synthesizer for 3 V/5 V operation, IEEE J. Solid State Circuits,
vol.32, no. 4, pp. 582-586, Apr. 1997.
[Yan99b]
H. Yano, Y. Nakahara, T. Hirayama, N. Matsuno, Y. Suzuki, and A.
Furukawa, Ku-band Si MOSFET monolithic amplifiers with low-loss on-
chip matching networks, IEEE Radio Frequency Integrated Circuits
Symp.,pp. 127-130, June 1999.


258
resonance should occur at a higher frequency, as indicated in [Sha97]. Second, an optimal
source resistance is now present, causing Fmin to be greater than one. This is due to the
drain and gate noises not being fully correlated. Third, Fmin is proportional to co/Op, mean
ing that as technologies scale to shorter channel lengths, lower noise figures will occur for
a given operating frequency. Fourth, the optimal source impedance is directly proportional
to Qgs, where Qgs is defined in (2.11). This means that as the device becomes smaller (or
frequency decreases), Qgs and Zopt both increase.
D.3.3 Case 3 Effect of Rb. Excluding
Substrate resistance generates thermal noise which is coupled into the device
through bulk transconductance and bulk capacitances (Cgb, Cdb, Csb). The inclusion of
Cgb complicates the equations considerably; therefore, the equations for the full model
should be used to examine the effect of Cgb on noise. Neglecting Cgb and Cgd, then y2] =
gm, A = 0, and CGT = Cgs. The bulk admittance, Yb, is defined in (D.9), where Cgb = 0.
This results in the following input noise generators:
(D.36)
(D.37)
The noise parameters can then be derived, and are given in the following table, D-4, where
Ob accounts for the effect of bulk capacitance on noise.
Examining these noise parameters, it can first be seen that substrate resistance
changes Ropt, while leaving Xopt unmodified. Second, Fmin has a non-monotonic relation
ship with Rb, peaking at a certain value of Rb. For Rb=0, OBRb=0, and hence Fmin = 1. On


55
the measured Q and inductance of these inductors for both substrate types. The quality
factor of the inductor on p+ substrates (Q=5) is -50% of that on p' substrates (Q=10). For
p+ substrates, the inductance slightly decreases, while the series resistance nearly doubles,
due to the induced eddy currents. Measurements also show that for a 1-nH inductor with
an area of 90x90 |im2, the Q reduction is only -30% [FloOOb]. This signifies that smaller
area inductors are beneficial for low-resistivity substrates.
3.4.3 Measured Results
Figure 3-10(a) shows the measured S-parameters for the differential LNA. To
obtain this balanced measurement using an unbalanced network analyzer, 180 couplers
were placed at the input and output of the LNA, acting as baluns. Driving 100 Q. differen
tially, the LNA provides 10 dB of gain, with input and output reflection coefficients of less
than -15 dB, consuming 21 mW from a 2.5-V supply. This gain is 3 dB lower than that
expected, due to a discrepancy between the simulated transistor characteristics and the
measured transistor characteristics. In particular, Cgs was 80% larger than that simulated;
thus, (Oj was lower and the gain was decreased, as given by (2.26).
Figure 3-10(b) shows the measured gain and NF of the single-ended LNA for both
p' and p+ substrates. Each LNA consumes 10 mW from a 2.5-V supply. The power gain of
the LNA on a p" substrate is -7.3 dB, while that for the LNA on a p+ substrate is -1 dB.
For the p' substrate, there is a 2.7-dB difference between the gain for the single-ended and
differential LNAs, whose origin is unknown. Theoretically, the gains should be the same
when each half-circuit of the differential LNA is operated at the same power consumption
as the single-ended LNA. Referring to Figure 3-9, the spiral inductor Q for p+ substrate is


APPENDIX A
THEORY FOR COMMON-GATE LOW NOISE AMPLIFIER
A. 1 Input Impedance
Figure A-l shows the small-signal model of a common-gate low noise amplifier
(LNA). By inspection, the input admittance for the common-gate amplifier is
1
Y- = S- + JaC^ja>L
(A.1)
Thus, to achieve an input impedance of ~50 £2, gm is set between 1/70 and 1/35 O'1, while
Ls is chosen to parallel-resonate with Cgs at the operating frequency.
A.2 Gain
To calculate the gain of the common-gate amplifier, the circuit transconductance is
first calculated. Assuming that the source inductor parallel resonates with Cgs], the circuit
transconductance is as follows:
241


202
Figure 6-27 Diagram of measurement setup developed for double-receiver interconnect.
Measuring the clock skew between two clock receivers using wafer probing would
require matching the delay through the external measurement system, such that the mea
sured clock skew is due only to the skew in the circuits themselves. Since this is difficult at
~2 GHz, the clock skew is detected on chip using a phase detector (PD). The PD converts
on-chip skew to voltage, as presented in Chapter 4, section 4.6.5. The PD is symmetric
with respect to both input signals, preventing a systematic skew or phase difference from
arising in the PD itself. As shown in Figure 4-26(b), the output voltage response from the
PD is approximately linear versus input skew, and is symmetric with respect to the y-axis.
A customized probe was required to measure the two receivers and the phase
detector output simultaneously. Since each clock receiver has probes on three sides, the
outputs from two sides of each receiver were routed such that one probe could be used in


180
zigzag antenna is identical to the antenna used for the receiver, while the loop antenna has
a diameter of 200 |im and a trace-width of 10 |lm, implemented in metal 6. The gain from
a loop antenna to zigzag antenna will be analyzed for different distances and angles.
This test chip also includes zigzag antennas in direct contact with the substrate,
where the zigzags have the same dimensions as presented above. This idea is motivated by
the fact that there will be many on- and off-chip metal interference structures between the
antennas. As shown in Figure 6-2 and discussed in section 6.2.3, the interference struc
tures can severely attenuate the direct path between two metal antennas. A more suitable
medium for wireless interconnects is the substrate. Figure 6-12 shows a cross-section of a
chip containing substrate antennas. If the substrate and back-side dielectric are being
exploited, then the antennas do not need to be implemented in top-level metal. Instead, the
antennas are implemented with n-plus in p-substrate and with p-plus in n-well. This can
potentially alleviate metal routing constraints around the antenna.
Transmitting
Antenna
p- substrate
Metal or Dielectric Medium
Receiving
Antenna

0
o
w
o
o
c
O)
c
1
Figure 6-12
Cross-sectional illustration of antennas in direct contact with substrate.


193
Frequency (GHz)
(a)
Frequency (GHz)
(b)
Figure 6-23 (a) S-parameters of power amplifier (driving 50-2 differentially).
(b) Operating power gain of the PA when connected to a zigzag antenna.
for the PA. The frequency range of the baluns is limited to 18 GHz; thus, the measurement
r\
is also limited to this frequency range. The PA demonstrates an IS2il of 5.4 dB at 18 GHz,
while consuming 41 mW from a 1.3-V supply. The S-parameters are obtained when driv
ing a 100-Q load. However, since the antenna impedance is not equal to 100 Q and has a
reactive component, the gain and resonant frequency for the PA driving the antenna are
different than that from the S-parameter measurements. The operating power gain, plotted
in Figure 6-23(b), is the ratio between the power delivered to the antenna and the power
delivered to the input of the PA. The results show that the PA has a 6-dB operating power
gain at 18 GHz. Because the peak gain frequency of the PA is de-tuned, the operating
power gain at 15 GHz is less than one. Since the clock distribution system operates at 15
GHz, the PA is actually attenuating the signal from the VCO. If the power consumption of
the PA is increased to 75 mW, the operating power gain at 15-GHz can be increased to 0
dB. Finally, the power delivered to the transmitting antenna is obtained from the


187
Figure 6-18 Die photograph of 0.18-gm clock transmitter with zigzag dipole antenna.
reader can refer to [HunOOa]. Instead, the discussion will focus on the measured character
istics and their implication for wireless interconnect design.
Figure 6-18 shows a die photograph of the clock transmitter with a zigzag antenna.
The size of the transmitter, including the unused portions on either side of the circuit, is
0.64x2 mm2, while the active areas (excluding pads) is 0.4x0.29 mm2. The layout is sym
metric left-to-right. Multiple substrate connections are included throughout the circuit.
6.6.2 Voltage Controlled Oscillator
Figure 6-19 shows a schematic of the VCO and PA. Cross-coupled transistors,
and M2, form a positive feedback, providing negative resistances to the LC tanks. Instead
of a conventional tail-current source, a PMOS transistor, M7, is placed at the bottom to
perform a comparable function. This allows the VCO core output to be dc coupled to the
PA. Simulation shows that when using this configuration, the phase noise contributed from
the noise present at the gate of M7 is lower as compared to that of a conventional current
source [HunOOc]. To generate the same negative resistance with a given current, the
required width of a PMOS transistor is larger than that of an NMOS transistor, due to


9
dissertations [Meh98, BravOOa, KimOOa] and are still being actively pursued, while the
last two items are the subject of this dissertation.
1.3 Overview of Dissertation
This dissertation focuses on the implementation of an intra-chip wireless clock dis
tribution system and evaluation of its feasibility, serving as a natural first step for evaluat
ing the potential of both intra- and inter-chip wireless interconnects. This work will
emphasize the design and implementation of RF CMOS receiver circuitry, and will evalu
ate the system feasibility by implementing both single-link (one transmitter/one receiver)
and multi-link (one transmitter/multiple receivers) wireless interconnects.
The low noise amplifier (LNA) is the first component in the clock receiver, and
through its low noise figure and moderate gain, it approximately sets the signal-to-noise
ratio of the entire receiver. Generally, CMOS LNAs have had inferior performance com
pared to silicon bipolar or gallium-arsenide (GaAs) LNAs. Also, there are very few exam
ples of CMOS LNAs operating above 5.8 GHz. Chapter 2 presents a new design
methodology for source-degenerated CMOS LNAs, examining input and output matching,
gain, and noise parameters. The effects of substrate resistance, gate-induced noise, and
input-matching variations are considered. These methodologies are demonstrated in Chap
ter 3, where the implementation and measured results of source-degenerated CMOS LNAs
operating at 0.9, 8, and 14 GHz, and implemented in 0.8-, 0.25-, and 0.18-p.m CMOS
technologies are presented. Also, a 23.8-GHz tuned amplifier with a common-gate input
implemented in a partially-scaled silicon-on-insulator (SOI) 0.1 -|im CMOS technology is
presented. These LNAs are intended for use in clock receivers; however, the results are
also applicable to standard CMOS receivers.


61
Table 3-4 Comparison Between Expected and Measured Inductor Characteristics
Inductor (15 GHz)
L (nH)
rs(Q)
cox m
Rsub (^)
Q
Cgate
Expected
0.61
2.41
15.2
10
21.4
Measured
0.73
6
26
~10
10.9
Ldrain
Expected
0.61
1.2
25.9
10
28.6
Measured
0.69
3.5
30
~10
15
versus frequency for Lg and Ld. The best Q is approximately 10 at 10 GHz, and increases
to approximately 20 at 20 GHz. Table 3-4 shows a comparison between the expected and
measured inductor characteristics. First, the measured inductance is ~0.1 nH larger than
that expected, resulting in decreased circuit resonant frequency. Second, the measured
series resistance is more than twice that expected, due to increased metal sheet resistance,
while the measured Q is about half that expected.
Although the Qs are not as high as expected, they are certainly adequate for
implementing RF circuitry, provided that the impact of lower Q is accounted for during
the design phase. These results illustrate the dramatic effect that decreasing the top-level
metal(s) sheet resistance has on inductor characteristics. Nevertheless, for constant metal
thickness, copper metallization greatly improves inductor characteristics over aluminum
metallization. Thus, these results are approximately twice as good as an all-aluminum pro
cess with the same metal thicknesses, or about comparable to an aluminum process with
twice the metal thickness. These measurement results suggest that greatly improved
inductor characteristics can be attained by thickening metal 5 and 6 layers.


102
Figure 4-16 Illustrations of concept of divider programming to reduce systematic skew. Two
receiver start with a skew of 20% and by selecting the correct state of each, the
skew can be reduced to 3% (for this example).
Finally, the initialization circuitry should not load down the divider, which would lower
the maximum frequency and change the conversion gain versus frequency.
4.6.2 System Start-Up Methodology
It has been shown conceptually how programmable frequency dividers can be used
to reduce and potentially eliminate systematic clock skew. These programmable dividers
will rely on some initialization signal (INI) which will denote whether the divider is to be
initialized or to operate freely. Therefore, a start-up methodology is needed for the system
to initialize all of the receivers to the correct state or count (CNT), and then to release
these receivers from the initialized state to the free-running state.
For programming the dividers, there are two basic methodsdynamic and static.
One is to correct the system dynamically with feedback. Here, the phase-offset between
each clock receiver and its nearest neighbors can be measured with a phase detector, and
then the CNT of the dividers can be adjusted accordingly. The system is basically


225
7-7(b)~the majority of the output clock waveforms are at 1.0625 GHz, with similar skew
to that just mentioned. However, there are a few iterations where the output waveforms are
at a different frequency, showing that the divider is self-oscillating. This means that the
input signal to the divider is not large enough to lock the divider. Finally, at a -31-dBm
input powerFigure 7-7(c)virtual chaos is evident, with most of the clock waveforms at
different frequencies. At this power level, the system has failed.
7.3.4 Conclusions for Process Variation
Table 7-3 summarizes the process variation results. Most significant among these
results are the phase-shifts due to process variation, resulting in output clock skew, and the
large variation in fiSO- Also, these results show that as the input signal level decreases, the
receiver becomes more susceptible to process variation and eventually fails at low input
levels. Largely, this problem is due to the input signal level to the divider not being large
enough to lock the divider. In other words, this means that the overall receiver gain, con
sisting of the amplifier and divider conversion gain, has its peak at a different frequency.
These results first motivate the need for some type of automatic gain control
(AGC) in the receiver. The AGC circuits could tune the resonant frequency of the ampli
fier and/or adjust the peak gain level. In so doing, the variation in phase-shift through the
amplifier would be controlled as well. Most importantly, the fiSO of the injection-locked
frequency divider should be tunable, such that the input signal will always be able to lock
the divider. This tunability could be obtained by designing the first 2:1 injection-locked
divider as a differential voltage-controlled oscillator (e.g., [Rat99]). Second, these results
have been obtained with very primitive biasing techniques. Using more advanced biasing
techniques should improve the circuits immunity to process variation.


38
Thus, the simulated and calculated optimal values for W2 agree reasonably well. Also, as
can be seen from Figure 2-9, choosing W2/Wj ranging from 0.55 to 0.75 will result in
minimal noise contribution from M2. Therefore, the noise parameters yield the optimum
values for Wj and W2, through equations (2.47) and (2.49), respectively, as well as the
optimum gate inductance for a given input reflection coefficient through equation (2.46).
2.1 Design Methodologies for Source-Degenerated LNA
2.7.1 Derivation-Based Methodology
Now, a complete design methodology based on the previous derivations has been
developed for the source-degenerated CMOS LNA, assuming a given technology. First,
choose Qgs to be nominally 2.6. This value should be the minimum of those obtained from
(2.14) and (2.47), yielding minimum noise and robustness against component variations.
Once Qgs is defined, Wj can be determined based on Cox. Second, using the gain and
noise equations along with the desired power consumption, estimate the required coy,
yielding an approximate operating point for Mj. Third, determine W2 using (2.49).
Fourth, set Ls nominally to 50/tOx, where lower values can be used to slightly boost the
gain while sacrificing input matching robustness. Fifth, determine Lg using (2.45) and
(2.46) based on the desired value for Sjj. Sixth, using gain equation (2.26), estimate the
value of Ld needed to achieve a desired value for S21 (note, 15/nl can be estimated to be
~0.2). Seventh, extract parasitics of Ld and cascode, and match the output to 50 £2 using
any standard ideal matching technique, determining values for Cj and C2.


34
optimized. From the results shown in Figure 2-7, it can be seen that F-Fmin (a function of
Gn, Zopt, and Zs) can be optimized considering GIN and Qgs only.
The noise parameters of a cascode (Mj and M2) with and without source degener
ation are shown in Table 2-2, including GEN for Mj. Here, Cgdl>2, gate and substrate
Table 2-2 Noise Parameters in Impedance Format
Cascode (Mj and M2)
Cascode with Ls
G (aA2b _fTi VY 1 r3n
G; = Gn (2.32)
" Tl^0lUrJ bl Ui 2JUJ QgsRs (~' }
Rop; = Kpt (2-34)
R _AJi-M2_AJ-\c
Kpt m c h ~ h ^ssKS
oLgs\b2 y b2 )
<2-35)
Xop¡ = XoPt (doLs (2-36)
'--i+2U)Vcwl (2-37>
Fmin = R min (2-38)
resistance, Cgbl2, and gate-induced noise of M2 have all been neglected. The noise param
eters for a single transistor under the same assumptions are derived in Appendix D, section
D.3.2. The following variables have been introduced:
(2.39)
A =
V 5y
(2.40)
bx = £ + A|c|,
(2.41)
b2 = C + 2A|c|+A2,
(2.42)


CHAPTER 6
WIRELESS INTERCONNECTS FOR CLOCK DISTRIBUTION
6.1 Overview
The clock receiver, shown in Figure l-3(b), contains an integrated receiving
antenna, a fully-differential LNA, a pair of source-follower buffers, a frequency divider
which translates the global clock frequency to the local clock frequency, and output buff
ers which drive the local-clock capacitive load. The design and implementation of CMOS
LNAs and frequency dividers has been presented in chapters 2-4, while chapter 5 pre
sented the system requirements for wireless clock distribution. This chapter uses these
concepts and circuits to implement wireless interconnect systems and to demonstrate sys
tem operation. A critical component of a wireless interconnect system is the antenna,
which acts as a transducer between the electromagnetic wave and the current and voltage
within the circuitry. This chapter presents a brief review of antenna fundamentals. Then,
the measured characteristics of on-chip antenna pairs and antennas connected to an LNA
are presented. These results are from a 0.25-|im CMOS test chip. Using this chip, a
7.4-GHz single-receiver wireless interconnect across a 3.3-mm distance is presentedthe
first on-chip CMOS wireless interconnect ever demonstrated. A second test chip was
implemented in a 0.18-|im CMOS process with copper interconnects. In this chip, multi
ple antenna test structures are included and characterized. Also, in the 0.18-pm test chip,
single-receiver and double-receiver wireless interconnects are demonstrated across 5.6-
and 6.8-mm interconnection distances at 15 GHz. These receivers have initialization
162


51
Table 3-2 Summary of measured characteristics for 900-MHz CMOS LNA
f0 = 900 MHz
Op. Point 1
Op. Point 2
Op. Point 3
Op. Point 4
Power (mW)
30
13.5
8.1
6.3
Gain (dB)
14.5
12.4
10.5
9.4
50-0 Noise Figure (dB)
1.2
1.5
1.8
2.0
Fmin (dB)
1.08
1.32
1.51
1.66
F(Sn=-10dB) (dB)
1.12
1.38
1.56
1.73
VDD(V)
3.0
2.7
2.7
2.7
Sn (dB)
-4.7
-4.1
-3.9
-3.8
S22 (dB)
-15.9
-16.7
-16.9
-17.0
$12 (dB)
<-42
<-42
<-42
<-42
IIP3 (dBm)
-1
-1.9
-3.3
-3.8
pldB (input) (dBm)
-12.4
-10.4
-6.3
-3.8
3.3.3 Summary for 900-MHz LNA
A summary of the measured characteristics for each operating point is shown in
Table 3-2. Operating point 1 has an Fmin which is 0.08 dB higher than a 900-MHz CMOS


232
distribution system should be independent of the digital supplies. Even still, noise will
couple to these supplies from both the digital circuits and the switching of the receivers.
Second, the amount of switching noise will most likely be larger than the -74 dBm mea
sured in [BravOOa, BravOOb] (which was for only one noise generating circuit). If the
amount of noise is quadrupledcorresponding to noise generators on all four sides of the
receiverthen the power would increase to -68 dBm, the SIR would decrease to 14 dB, the
peak jitter due to interference would increase to 2.4%, and the total peak jitter would
increase to 2.7%. This would only leave 1.3% of peak-jitter margin for the noisy supplies.
In Chapter 6, the measured peak-to-peak jitter, with a quiet supply and without any
digital switching noise, was 1.24%. This jitter is due to thermal noise, corresponding to a
19.7 dB SNR. As can be seen, this agrees very well with the estimated peak jitter from
thermal noise of 1%, validating both the simulated SNR and the SNR-jitter relationship.
Therefore, these results suggest that the peak-to-peak jitter due to thermal noise, interfer
ence, and source jitter should be approximately 2.4%, meeting the system specification of
3% peak-to-peak jitter. However, more work is required to quantify how much jitter is
introduced by noisy supplies. Also, the jitter from a PLL at 15 GHz should be quantified
and checked against the assumed value. If the jitter specification is exceeded by the noisy
supplies and the PLL, then the receiver circuitry could be optimized to decrease the jitter
due to thermal noise and interference. One potential improvement would be to use a differ
ential LC-oscillator as the first 2:1 injection-locked divider [Rat99]. Due to the higher Q of
this circuit as compared to the low Q of a ring oscillator, the output phase noise and,
hence, jitter will be decreased. This decrease is due to the filtering provided by the
injection-locking, where the output phase can only follow the input phase over a limited
locking range, as shown in Appendix E.


48
50 Q without an additional output buffer. Figure 3-4(b) shows a plot of the transducer gain
(S21) and NF versus frequency for the LNA, for operating points 1 and 3. At 900 MHz, the
measured transducer gains (S21) are 14.5 and 10.5 dB at 30 and 8.1 mW, respectively.
Figures 3-5 (a) and (b) show the input and output reflection coefficients (Sn and
S22) for operating points 1 and 3. The relatively high S^s are due to the bond-wire induc
tance being too small, generating only ~15 Q. of real impedance. To improve the input
match to under -10 dB, Ls should be slightly increased while Lg can be decreased to main
tain the noise match. S22 is below -10 dB over a 137-MHz span. The very low S22 corre
sponds to a conjugate match at the output.
To measure Fmin, the measurement setup shown in Figure 3-6(a) was used. Exter
nal tuners are inserted at the input and output of the packaged LNA, and the impedances of
these tuners are adjusted until the optimum noise matching condition (i.e., ropt or Zopt)
occurs, corresponding to Fmin. Figure 3-6(b) shows the optimum source-reflection coeffi
cient (ropt) versus frequency for operating points 1 and 3, plotted on Smith charts. Figure
Figure 3-5 (a) Input (Su) and (b) Output (S22) reflection coefficients versus frequency,
operating points 1 and 3.


169
p- substrate
Dielectric Medium
Transmitting
Antenna
/////////////////////////////
Metal (Heatsink)
Path A
Receiving
Antenna
Figure 6-2 Cross-sectional illustration of chip for metal antennas with interference structures.
Three possible signal paths between antennaspath A is direct path, path B is
substrate surface wave, and path C is substrate reflection.
Beneath the oxide is the substrate. This substrate then is connected to a heatsink,
which is used to remove the heat from the microprocessor. Experimental results, though,
show that this situation is not very conducive to on-chip signal transmission for substrate
resistivity less than ~ 20 Q-cm. Instead, if a dielectric layer is placed between the substrate
and heatsink, the antenna gain can increase by as much as 8 dB [KimOOa, GuoOl],
depending on the dielectric material and its thickness. Since heat still needs to be removed
from the chip, this dielectric layer has to be thermally conductive. A possible material
which meets both criteria is aluminum nitride (AIN) [GuoOl]. Aluminum nitride has a
thermal conductivity approximately the same as that of silicon and about 10 times that of
glass, and its dielectric constant is 8.5.
Due to the interference structures, the direct path from one antenna to another,
labeled path A in Figure 6-2, will be very attenuated for on-chip wireless interconnection.
This direct path is also referred to as a surface wave. The gain of this path will depend on


123
O
-20
E -40
CQ
o -60
-80
-100
0
-20
E -40
CQ
o -60
-80
-100.(
Figure 4-
higher than this, then the bipolar leakage should not degrade performance. However, the
lower frequency cutoff for an SOI divider is increased, which can potentially reduce the
overall operating frequency range. Another implication of this is that for an SOI [DP]
divider with a high division-ratio, a different DFF topology is needed for the latter stages.
4.7.3 Measured Results
A 4:1 SOI divider was measured on-wafer using high-frequency probes. A spec
trum analyzer was required to examine the output signal due to frequency limitations of
the oscilloscopes trigger (trigger frequency limited to 2 GHz). A 180 coupler was used
as a balun to generate the clock and elk signals. Measurements show that the SOI divider
can operate between frequencies of 9.1 and 18.75 GHz for a 1.5-V supply. Figure 4-32
-40 -20 0 20 40
Frequency Offset from 4.6875 GHz (kHz)
32 Output spectrum of SOI 4:1 frequency divider for fin = 18.75-GHz and Vdd=1.5 V.


93
(a) (b)
Figure 4-9 Measured results for 0.25-|im divider: (a) Maximum input frequency and power
consumption versus supply voltage, (b) Input sensitivity versus frequency versus
supply voltage.
phase difference between the two input signals has to be as close to 180 as possible, else
the divider will not operate up to its maximum operating frequency.
The measured maximum input frequency versus supply voltage for the 128:1
divider is shown in Figure 4-9(a). Also shown is the measured power consumption versus
supply voltage for the 8:1 divider core. At a 2.5-V supply, the 8:1 divider operates up to
9.98 GHz, while consuming 18.4 mW. The measured output waveform for a 9.98-GHz
input signal is shown in Figure 4-10. The reduced output swing is due to driving a 50-Q
output load. At 1.5 V, the divider can operate up to ~6.6 GHz while consuming 3.3 mW.
These are excellent speed and power trade-offs for a 0.25-|im technology.
To measure the input sensitivity of the divider, the minimum input power level for
each input frequency was recorded for multiple supply voltages. Figure 4-9(b) shows the
resultant plots. For a 2.5-V supply, a 2.2-Vplc_pk input signal is required to operate at the


81
when the input clock signals have a large voltage swing. Here, M3 4 act as an evaluation
stage when the latch is transparent, and regenerative pair M5 6 act as a hold stage when the
latch is opaque. Thus, for CLK = high, the master is transparent and the slave is opaque,
with Q¡ and Qbi taking the value of D and Db. For CLK = low, the master is opaque and the
slave is transparent, and nodes Q¡ and Qbi are passed to the output. As is evident from this
description, cascading two level-sensitive D-latches in a master-slave configuration results
in an edge-triggered D-flip-flop (DFF). It can easily be verified that connecting the outputs
of a DFF back to its inputs with inverted phase results in a toggle flip-flop. Thus, for every
clock pulse, the output is toggled, resulting in a divide-by-two operation.
4.2.3 Ouasi-Dvnamic Operation
The question arises as to whether or not this SCL latch is static or dynamic. In
other words, is there a lower frequency limit for this latch? The answer to this question
depends on the sizing of the transistors in the latch [Mur95], and because of this, the latch
topology is quasi-dynamic. To illustrate this, the PMOS load transistors are first modeled
as resistors, with a resistor RPM. Consider the case when the output of the master D-latch,
Qbi is first evaluated low and then held low. Since there is no tail-current source, the cur
rents through Mj and M2 are not necessarily equal. The current values depend on the siz
ing of Mj and M2 and the input clock signals. The current pulling Qb¡ low during
evaluation for the master DFF is then modeled as fj(Ii), where fj depends on the sizing of
M3 and M4, the voltages Q and Qb, and the evaluation current, Ij. Thus, Qbi has a voltage
during evaluation equal to
LOWEval = Vm-WORpm.
(4.1)


120
Figure 4-29 Schematic of 4:1 [DP] synchronous divider.
period. This delay gives an indication of the speed of standard digital latches used in pipe
lined circuitry for this process.
Using [DP]2 logic, 4:1 and 128:1 frequency dividers have been implemented
[FloOlc] in a partially-scaled 0.1-p.m CMOS technology on silicon-on-insulator (SOI) and
bulk substrates from IBM. The process uses a 0.35-p.m design rule set for all dimensions
except for the 0.1-|im gate length and 2.9-nm gate oxide thickness, while providing two
metal layers. For SOI, partially-depleted, floating-body transistors are used for all cir
cuitry. A die photograph of the 4:1 divider is shown in Figure 4-30. The die size is
0.51x0.55 mm2.
4,7.2 Effect of SOI on Circuit Performance
A benefit of using floating body SOI transistors is that the threshold voltage is
reduced due to Vbs > 0. This Vbs is caused by charging of the body node through regener
ation current [Suh94]. With a lower threshold, the gate-overdrive increases, resulting in
more channel current. Hence, the switching speed is increased, resulting in higher maxi
mum operating frequencies for the SOI divider as compared to bulk.


245
r n
I Xs=QRse |
I % 1
i _L i
¡ xp=rp/q ¡
¡ T;
L J
L-Matching Network
Figure B-1 L network matching design.
into the matching network. Hence, the total reactances in the matching network and the
source/load should be equal to the values Xs or Xp, as given by (B.2) and (B.3).
B.2 Application to Capacitive Transformer in LNA
Figure B-2 shows the output matching network for the LNA, including the capaci
tive transformer. Here, the output impedance of the cascode has been modeled by
Rp2llCp2, while the drain inductor is modeled as a parallel network composed of Ld, RPLd,
and Cj^j. As can be seen, C] is the shunt reactance, while C2 plus the equivalent series
reactance from Ld and CLd+CP2 is the total series reactance. Hence, the total impedance at
R|_= 50 Q
Z
se
L _ J V
L-Matching 2
Network F
Figure B-2 Small-signal model of common-gate low noise amplifier.


153
NF(fa) + lOlog 10(5WN) < 107dB. (5.40)
The difficulty with this equation is that BWN is defined through simulations. As defined,
BWn is the bandwidth from which most of the thermal noise is generated. Since the LNA
is a tuned circuit, both the input thermal noise and most of the noise generated in the LNA
will be shaped by the tuned response. The response of the LNA can be modeled by the
response of a parallel RLC network driven by a current source. Simulations show that the
BWn for a parallel RLC network with quality factor, QLd, and resonant frequency,/G, is
BWn
(5.41)
Referring to the results in Chapter 3, is approximately 10 at 10 GHz (for the 0.25-pm
LNA) and 15 at 15 GHz for the 0.18-|im LNA. Thus, BWN = 2 GHz and 101ogj0(BWN) =
93 dB. With this assumption, the maximum noise figure can be specified as 14 dB, with
the explicit equation being
fkT \
NF(fo)max = Psens\dBm-^log[^ySNRmin\dB-lOloglo(BWN). (5.42)
Thus, the system specification for maximum noise figure of the receiver, including the
antenna noise figure, atfQ is set to 14 dB.
This noise figure specification is for a -54-dBm sensitivity. However, as mentioned
earlier, if the input interference to the antenna is significant, then the NF in (5.42) is the
total noise figure of the antenna plus the receiver (i.e., F.^ + Frcvr 1). As the total noise
and interference at the receiver input is quantified, the 14-dB specification for NF may
become infeasible. As such, the sensitivity of the receiver will have to be modified to a
larger power to maintain the 12-dB SNR at the divider input. In other words, currently, the
sensitivity specification is driving the noise figure specification. However, with the total


196
While this first on-chip wireless transmitter demonstration is significant, the
received power level is very low. To understand this, the power delivered to the zigzag
transmitting antenna by the PA is obtained using the measured Ga for the 6.7-mm antenna
separation and calculating the mismatch loss in the system. At 15 GHz, the mismatch loss
at the PA/antenna interface is 2.5 dB and the mismatch loss at the antenna/spectrum ana
lyzer interface is 0.25 dB. These numbers are obtained using (5.3). Referring to Figure
6-13(a), the antenna transmission gain is -53 dB for a 6.7-mm separation.While this is not
the measured gain for a 5.6-mm separation, it is within the deviation expected from vary
ing metal interference structures, which is 5-10 dB, as demonstrated in Figure 6-14(b).
Therefore, with an antenna transmission gain of -53 dB, the extracted power available
from the PA is -13.25 dBm (-69 + 0.25 + 53 + 2.5), while the power delivered to the zigzag
antenna is -15.75 dBm (-13.25 2.5).
The PA was originally designed to deliver +12 dBm to the antenna; thus, there is a
28-dB difference between the expected and measured results. Referring back to section
6.6.3, the reduction in inductor Q due to thin metals 5 and 6 accounts for 7 dB of the PA
gain reduction. Second, operating the PA at 15 GHz, as opposed to at its resonant fre
quency (18 GHz), accounts for another 8.5 dB. Third, the dc-bias level provided by the
VCO at the input of the PA is less than that used to characterize the stand-alone PA. This
reduced bias point is due to the VCO core current being increased to overcome the Q
reduction. Thus, the transconductance of the PA is decreased, resulting in another ~7.5-dB
degradation. Taking all of these effects into account, the expected power delivered by the
PA should be -11 dBm. This is close to the measured PAs available power of -13.25 dBm,
which assumes that the PA and the antenna are conjugately matched, as was the case in the


242
r ldl V1 f l/Sm )_ Sm
" VS v/ HRj+1/J 1+Sms
(A.2)
'S vs
As can be seen, Rs degenerates the transistor. Therefore, the common-gate LNA is typi
cally more linear than an inductively-degenerated common-source LNA.
For a tuned load output, the S2i is equal to
N = 2-KUco0 =
... 2Sm\ZLeq((0o)\
+ 8 mRS
(A.3)
where ZLeq(co0) is equal to the total load impedance at the amplifier output at resonance.
Assuming that the output matching network is the same as that used in the source-degen
erated LNA (i.e., a matching network containing a drain inductor and a capacitive trans
former), then
i 8mQLd^oRd
1
1 + 8mRS
(A.4)
where and Ld are the quality factor and inductance of the drain inductor, and h(G0o) is
a complex capacitive-transformer ratio defined in (2.25).
A.3 Noise Factor
To derive the noise factor of this amplifier, the short-circuit output current is calcu
lated for the entire circuit, including the source resistance. Converting this to a power
spectral density (PSD) and taking the ratio of the total PSD to the PSD from the source
resistance yields the noise factor.
The thermal noise generated by the source resistance is represented by a voltage
source with a PSD given by the following:
v\ = AkT Rs.
(A.5)


40
rs P
NF Circles
Figure 2-11 Constant available gain and NF circles for cascode with Ls and Ld.
SPICE. Using these parameters, constant available gain (GA) circles and constant noise
circles are plotted on the Smith chart, along with input and output stability circles. Using
these circles, the reflection coefficient of the source and the load, Ts and TL, can be
determined. Once Ts and TL are determined, the input and output matching networks can
readily be designed.
Figure 2-11 shows a plot of simulated constant gain and noise circles on the Smith
chart. Assuming Qgs is chosen using (2.47), Ropt should be close to the 50-Q circle. The
value of Lg closest to ropt, while still meeting the desired Sn specification, can then be
read off the Smith chart. The output is then conjugately matched (i.e., Tl Tout ), using
the following formula:
1 Silt's'
(2.50)
This determines the output matching network, the simplest of which is a capacitive trans
former, Cj and C2. To obtain Cj, a constant conductance circle is traversed from the center
of the Smith chart, while to obtain C2, a constant resistance circle is traversed from the
endpoint of the previous path to ^L^out as shown in Figure 2-12.


249
To account for correlation between vn and in, vn can be divided into components which are
correlated (vc) and uncorrelated (vu) with in, as follows:
Vn = Vu + Vc = Vu + Zcin-
(C.4)
The correlation impedance, Zc, is
7 *n*Vn
(C.5)
n
The power spectral densities of each of these noise sources (in, vu) can be repre
sented by either a resistance (for vu) or a conductance (for in), as follows:
.2
G =
n 4kTAf
(C.6)
R.. =
4kTAf'
where Ru can be obtained using the equivalent input noise generators, as follows:
(C.7)
R.. = ^-.-\Zc\2Gn = R-|Z/G.
4kTAf
Substituting (C.2), (C.4), (C.6), and (C.7) into (C.3) yields
(C.8)
, |v. + ',(Zi + Zo)f .,,2
' Rs,Zs + Zel '
(C.9)
Taking the partial derivatives of (C.9) with respect to the source resistance and reactance,
and setting the results equal to zero gives the optimum source impedance resulting in min
imum noise factor, as follows:
(C.10)


CHAPTER 4
CMOS FREQUENCY DIVIDERS
4.1 Overview
In the clock receiver, the frequency divider translates the global clock signal to the
local clock signal. The divider should have a maximum operating frequency above 15
GHz and should dissipate a minimal amount of power. In addition, the divider should be
capable of locking to very low-level input signalsspecified in terms of input sensitiv
itywhich will improve the minimum detectable signal (MDS) of the clock receiver.
Finally, the signals reaching each of the frequency dividers within the clock receivers will
potentially have different phases and amplitudes, both resulting in clock skew, as well as
random noise, resulting in clock jitter. Clock skew and clock jitter as a result of these
mechanisms will be discussed in Chapter 5. However, to account for the phase difference
and, hence, reduce the clock skew, the dividers should be synchronized between receivers.
Using silicon and SiGe bipolar technologies, frequency dividers operating up to 82
GHz have been reported [WasOOa, WasOOb, KnaOO, WurOO], and using CMOS technolo
gies, dividers operating up to 26.5 GHz have been reported [WanOO, WetOO, Raz94,
Kur97]. Clearly, silicon is capable of supporting divider circuits operating above 20 GHz.
In this chapter, the design and implementation of high-frequency CMOS frequency divid
ers operating up to 18.75 GHz and employing either source-coupled logic (SCL) or
dual-phase dynamic pseudo-NMOS [DP]2 logic [Biy96, Yan99a] will be presented. The
SCL dividers are implemented in 0.25- and 0.18-p.m bulk CMOS technologies, operating
77


259
Table D-4 Noise parameters considering substrate resistance and excluding Cgb
Alternative Noise Parameters
Impedance-Based Noise Parameters
fi3 =
Sdo
^,3=l+2
G3 = (D'40)
P.3 = (D.41)
do
Z-3 (D42)
the other hand, for very large Rb, Fmin approaches 1 as well (the limit of 0BRb as Rb
approaches infinity is zero, since <1>B approaches zero faster than Rb approaches infinity).
Fmin is larger than 1 for Rb values in between 0 and infinity. This shows that short- or
open-circuiting the substrate is valuable for minimizing noise figure. This is the case for
most other RF performance metrics as a function of substrate resistance. For Cdb ~ Cgs, r|
= 0.25, and co/Op ~ 0.25, Fmin peaks at around 500 2 For this case, the substrate resis
tance range between 40 and 10 k£2 should be avoided if possible to minimize Fmin.
Care must be taken in trying to obtain the noise parameters as a function of only Rb
and drain noise (i.e., neglecting all bulk capacitances). On first glance, it seems that this
can be achieved by setting The characteristic of decreasing Fmin for very large values of Rb will not be obtained,
since Ob should approach zero as Rb approaches infinity. This means that for large values
of Rb, the bulk capacitances (Cdb and Csb) cannot be neglected.


97
Figure 4-13 Measured results of 0.18-fim frequency divider
(a) Maximum input frequency and self-oscillation frequency versus Vdd.
(b) Input sensitivity versus frequency for Vdd=1.5 and 2.1 V.
SCL divider, 8:1 TSPC dividers, and output buffer. Again, separate supplies are used for
the SCL and TSPC dividers to prevent any low-frequency supply noise from the latter
stages of the divider from corrupting the high-frequency operation of the first 8:1 divider.
Also, 50-2 resistors are included between the two inputs of the 64:1 divider to ground.
4.5.2 Measured Results
Using the measurement setup in Figure 4-8, the maximum input frequency and
power consumption (for 8:1 divider core) versus supply voltage plots for the 64:1 divider
were obtained, and are shown in Figure 4-13(a). For a 1.5-V Vdd, the maximum input fre
quency is 15.8 GHz, and the power consumption is 4.5 mW. The maximum input fre
quency increases to 20.4 GHz when Vdd is increased to 2.1 V, while the power
consumption increases to 12.2 mW. These are excellent power/speed trade-offs. Figure
4-14 shows the output waveforms for fin = 15.8 and 20.4 GHz, at Vdd = 1.5 and 2.1 V,


58
L=0.61nH
Figure 3-11 Schematic of 14-GHz fully differential LNA showing parasitics.
source-follower buffers. This circuit is used in the clock receiver, driving the frequency
divider. Two different input matching conditions were implemented. The input matching
condition was 100 2 (differentially) when the LNA was measured using 50-2 test equip
ment, and 125-j55 2 when the LNA was driven by the on-chip antenna. Therefore, the
reactance of the gate inductors for the latter case was increased by 27 Q. (each side or 55 Q.
differentially). The single-stage differential LNA was differentially matched at the output
to 100 Q, using capacitive transformers. Figure 3-13(a) shows a die photograph of the
0.18-(im differential LNA. The die size is 745x410 |im2.
3.5.2 Inductor Characteristics
The gate and drain inductors for the 14-GHz LNA have the following design
parameters: inductance = 0.6 nH, turns = 2.5, area = 67x67 |J,m2, width = 5-pm, and turn


126
2.1V has been implemented. Again, this divider can detect very low-level input signals
close to the input-referred self oscillation frequency. Both of these results demonstrate that
CMOS frequency dividers can operate at the frequency required by the clock receiver and
that injection locking can be used to enhance the system gain.
Using the SCL divider topology, a method to offset built-in phase delay differences
has been developed. This methodology allows the dividers to be initialized to one of M
states, where M represents the division ratio. By doing this, the systematic skew in the sys
tem can be reduced to under percent. Using the 0.18-|lm CMOS technology, a
divider with this initialization circuity has been implemented. The results show that the
circuit is able to initialize to one of 8 different states. However, for a given desired input
state, the output will take any of the 8 possible states, rather than the desired state each
time. This failure mechanism was identified and a new latch topology has been proposed.
Finally, using a partially-scaled 0.1-(im SOI and bulk CMOS process, a dual-phase
dynamic pseudo-NMOS [DP]2 divider has been implemented. These dividers operate up
to 18.75 and 15.4 GHz on SOI and bulk substrates, respectively at 1.5 V. This result is the
highest known operating frequency to date for a full-swing frequency divider. Also,
extracted [DP]2 latch delays are 13.4 and 16.2 ps, respectively, for the SOI and bulk divid
ers. These results show that 0.1 -[im CMOS can support digital circuits operating above 15
GHz, and suggest even higher frequencies for fully-scaled 0.1 -|i,m technology and beyond.


165
Therefore, Ga is used to characterize on-chip antenna performance [KimOOa]. Note, that in
actual system implementations, the antennas would only be matched for a limited fre
quency range; thus, Ga would only be equal to the actual gain over those frequencies.
Friis transmission formula assumes that the transmitting and receiving antennas
are far enough apart such that they are in the far-field region. In the far-field region, only
radiation terms (1/r) of the electromagnetic field dominate (power density ~ 1/r2), and the
angular distribution of these terms is independent of distance [Wan86]. A near-field, region
also exists, where the reactive terms of the electromagnetic field dominate (i.e., the power
density is complex). This reactive power refers to energy storage in the electric and mag
netic fields; thus, wave propagation is not occurring. Finally, an intermediate-field region
exists where radiation terms dominate; however, their angular distribution is a function of
distance.
Multiple definitions abound for the boundaries between these regions. For an elec
trically short dipole antenna, whose length (L) is less than 0.33A,, the far-field criterion is
[Bal97, KimOOa] as follows:
These electrically short antennas are used in wireless interconnects, due to restricted die
area for the antennas. If the symbol is interpreted as a single order of magnitude, the
boundary between the intermediate- and far-field regions becomes as follows:
r>1.6A,, (6.4)
which agrees exactly with [Ban99]. Note that this criterion change as the antenna becomes
electrically longer (i.e., L goes up, or frequency goes up), with the boundary shifting to 5L
for 0.33X < L < 2.5\ and 2L2k for L > 2.5A, [Ban99].


264
a frequency divider, this phase difference will be with respect to the input frequency; thus,
the output steady-state phase error is that given in (E.8) divided by N.
E.3 Phase Noise of Injection-Locked Oscillators
The phase transfer-function of an ELO with a noisy signal, with amplitude Vn,
added at the input can be shown to be [Llo98, Rat99, Kur68]
O
out
cose
AV0 ss vn
in ^ Vl >l'
+ Tv¡cos6-*
(E.9)
The last term, Vn/VL, represents the input phase noise. The division by the input injected
power level is due to the conversion from input additive noise to phase noise [Ega90]. Tak
ing the magnitude squared of (E.9), the output phase noise of the ILO is
$,<*> = ^7^-2
(E.10)
where Sout and S¡n are the output and input power spectral densities (PSDs) of the phase,
and B can be shown to be:
s2 = (j^cos6-)2 = (a^)2-(a)2- (E11)
If the ILO is being used as a frequency divider, the output phase PSD in (E.10) is divided
by N2, where N is the division ratio [Rat99]. Examining (E.10), it can be seen that the ILO
behaves like a first-order phase-locked loop with respect to input phase noise [Rat99], fil
tering the noise with a low-pass response. Here, the bandwidth of the equivalent low-pass
filter is a function of the input injection level.


254
In these equations, A is the voltage division ratio between vgs and vbs, Yb is the admittance
at the body node, rj is the ratio between gmb and gm, a is the ratio between gm and gdo, and
CGT is the total gate capacitance. Expressions for these variables are as follows:
, _vb, jcSb
n =
8mb
8 m
a =
8m
8 do
-1
[RbiCgb + Csb + Qfc)] >
CGT ~ Cgs+Cgd+Cgb
(D.8)
yb = 4- + mCsb + Clt + CJb) = J-[ 1 + j(o>/a>B)], (D.9)
(D.10)
(D.l 1)
(D.12)
(D.13)
D.2 Noise Parameters for Complete Model
Using (D.5) and (D.6), noise parameters for the MOSFET including all second-
order effects can be derived, as outlined in Appendix C. The alternative noise parameters
(Rn, Gn, and Pn), defined in (C.6), (C.14), and (C.15), are given in Table D-l, where
=
i + [coc^/o^j
(D.14)
1, 1 + (O)/COg)
As can be seen by examining (D.15), (D.16), and (D.17), the gate-to-body capacitance
greatly complicates all of the equations.
These alternative noise parameters can then be used to find more conventional
noise parameters, using (C.16) for Zc, (C.8) for Ru, (C.10) for Ropt, (C.ll) for Xopt, and


175
(a) (b)
Figure 6-7 (a) Measured transducer gain between antenna pair and antenna pair with LNA.
(b) Extracted and measured gain of LNA.
measured gain of the LNA from section 3.4. The antenna and antenna-with-LNA gain
measurements are noisy due, in part, to insufficient averaging on the network analyzer
and, in part, to the interference structures located between the antennas. To indicate the
qualitative trend of the antenna with an LNA measurement, a curve fit is included. These
results show that the LNA is amplifying the signal received by the antenna.
6.4 Wireless Interconnect in 0.25-|xm CMOS
Figure 6-8 contains a die photograph of the clock receiver with an integrated
dipole antenna. The size of the receiver including the antenna and unused portion on
either side of the receiver is 0.7 x 2 mm2, while the size of the receiver alone (excluding
pad area) is 0.4 x 0.5 mm2. Since the clock receiver is fully differential, the receiver layout
should be symmetric for each half-circuit of the receiver, including all components, supply
voltage lines, and pads. This avoids any systematic errors being introduced from layout


Master
Slave
Q1
o
o
T
1
o
o
T-
T-
o
T-
o
T
o
1
o
T
0)
CD
0
0
0
0
£
0
to
03
03
03
CO
03
00
to
CO
CO
CO
CO
CO
CO
CO
CO
C0|
Figure 4-17 (a) Block diagram of 2:1 frequency divider using SCL and (b) ideal input and
output waveforms for divider and associated states.
employing SCL. A schematic of the SCL DFFs is shown in Figure 4-l(c). Figure 4-17(b)
shows ideal waveforms for the input clock signal as well as for all four of the drain
nodes in the divider, labelled Ql, Ql, Q2, and Q2. Examining these waveforms, it can be
seen that the divider has four possible states. State variables S1 and SO are introduced to
identify these states, where SI is the desired value of the divider input (i.e., clock signal)
and SO is desired value of the divider output, defined to be Q2.
The initialization circuitry should set all of the drain nodes for each of the four
possible states. This results in the truth table shown in Table 4-4, which is obtained with
the help of Figure 4-17(b). The drain nodes can then be defined logically by the state vari
ables, as shown in the table. Therefore, during initialization, drain nodes Q2 and Q2 are
set to SO and SO, respectively, while drain nodes Ql and Ql are set to the exclusive-nor


192
second stage acts as a class-E amplifier without a bandpass filter (traditionally used to
select the fundamental), and is tuned together with the antenna impedance. Hence, as the
amplifier switches, both the fundamental and higher-order harmonics are transmitted. The
size of the transistors in stage 2 is 3 times larger than that of the first stage. Due to the thin
gate oxide and, thus, low gate-oxide breakdown voltage, the maximum output power is
designed to be ~12 dBm (the voltage amplitude of the differential output is 2 V and the
real part of the antenna impedance is -125 2). The power efficiency is -60%. The output
power can be simply controlled by the supply voltage. Figure 6-22 shows a simulated dif
ferential output waveform across the antenna. The output waveform is not sinusoidal at the
zero-crossing point due to the passing of higher-order harmonics in the PA, introducing
distortion to the signal.
Figure 6-23(a) shows the measured small-signal S-parameters for a differential PA
(referenced to 100 Q). The input of this stand-alone PA had 50-Q. resistors to ground at
each input node. To obtain the measurement, baluns are placed at the input and output of
the PA to transform the unbalanced signals from the network analyzer to balanced signals
Figure 6-22 A simulated output waveform across the antenna from the PA.


APPENDIX B
OUTPUT MATCHING USING CAPACITIVE TRANSFORMER
B. 1 Two-Element Matching Technique
Multiple methods, both graphical and analytical, exist to quickly match one
impedance to another using two reactive elements. An analytical method found in
[Bow82] which is both easy to understand, use, and code (i.e., into Matlab), is summa
rized here for convenience. The simplest matching scenario is matching two resistances
using an L or two-element network. In this case, the larger resistance, termed Rp,
requires a shunt reactive component (decreasing the impedance), while the smaller resis
tance, termed Rse, requires a series reactive component. The sign of these reactances (i.e.,
inductive or capacitive) are opposite of one another for the net reactance to be zero.
The formulas for this method are as follows:
e = Jfe-1'
= e*.
R
(B.l)
(B.2)
(B.3)
where Xs and Xp are the required series and shunt reactances of the L network. An illustra
tion of this procedure is given in Figure B-l. The value of the inductance and capacitance
of each component can easily be obtained from the reactance for a given angular fre
quency. To handle complex sources and loads, the reactance of the source/load is absorbed
244


37
W2. Thus, an optimum value of W2 exists which minimizes C,. The capacitance at the drain
of M] (Cdl) can be written as follows, neglecting interconnect capacitance:
Cd\ = Cd\\W ^ +Cd\2^2' (2-48)
where CdllW] consists of Cdbl plus a fraction of Cgdl, and Cdl2W2 consists of Csb2 and
Cgs2- Differentiating (2.39) with respect to W2 assuming constant drain current and con
stant Vdsl yields the following:
W2 = ^Vi. (2.49)
Thus, the capacitance at the drain-node of M] should be equally divided between Mj and
M2. Figure 2-9 shows a plot of simulated noise factor versus W2 for a 0.25-|im technol
ogy. For this simulation the default noise model for BSIM3v3 is being used. The optimum
width of M2 is 60 |im, when W j is set to 90 |lm. Calculations show that for this same tech
nology, the ratio Cdll/Cdi2 is equal to 0.6, yielding a calculated optimal W2 of 53 (im.
Figure 2-9 Noise factor versus width of cascoded transistor (M2). The optimum width occurs
when the capacitance at the intermediate node between M¡ and M2 is equally
balanced between M¡ and M2.


273
[Kur97] M. Kurisu, M. Nishikawa, H. Asazawa, A. Tanabe, M. Togo, and A.
Furukawa, An 11.8-GHz 31-mW CMOS frequency divider, Symp. VLSI
Circuits Dig. Tech. Papers, pp. 73-74, June 1997.
[Lee66] D. B. Leeson, A simple model of feedback oscillator noise spectrum,
Proc. IEEE, vol. 54, pp. 329-330, Feb. 1966.
[Lee98] T. H. Lee, The Design of CMOS Radio Frequency Integrated Circuits,
Cambridge, UK: Cambridge University Press, 1998.
[Llo98] O. Llopis, M. Regis, S. Desgrea, and J. Graffeuil, Phase noise perfor
mance of microwave analog frequency dividers application to the charac
terization of oscillators up to the mm-wave range, IEEE Int. Frequency
Control Symp., pp. 550-554, May 1998.
[Lon97] J. R. Long and M. A. Copeland, The modeling, characterization, and
design of monolithic inductors for silicon RF ICs, IEEE J. Solid-State
Circuits, vol. 32, no. 3, pp.357-369, Mar. 1997.
[Mai97] C. A. Maier, J. A. Markevitch, C. S. Brashears, T. Sippel, E. T. Cohen, H.
Blomgren, J. G. Ballard, J. Pattin, V. Moldehnauer, J. A. Thomas, and G.
Taylor, A 533-MHz BiCMOS superscalar RISC microprocessor, IEEE J.
Solid-State Circuits, vol. 32, no. 11, pp. 1625-1634, Nov. 1997
[Man98] T. Manku, Microwave CMOSdevices and circuits. Proc. Custom Inte
grated Circuits Conference, pp. 59-66, May 1998.
[Mea80] C. Mead and L. Conway, Introduction to VLSI Systems, Reading, MA:
Addison-Wesley, 1980, pp. 12-15.
[Meh98] J. Mehta, An investigation of background noise in ICs and its impact on
wireless clock distribution, Masters Thesis, University of Florida,
Gainesville, 1998.
[Mil97] D. A. B Miller, Optical interconnect technologies for Si ULSI, IEDM
Technical Digest, pp. 342-347, Dec. 1997.
[Mur95] K. Murata, T. Otsuji, E. Sano, M. Ohhata, M. Togashi, and M. Suzuki, A
novel high-speed latching operation flip-flop (HLO-FF) circuit and its
application to a 19-Gb/s decision circuit using a 0.2-p.m GaAs MESFET,
IEEE J. Solid-State Circuits, vol. 30, no. 10, pp. 1101-1108, Oct. 1995.
[097] K. K. O, K. Kim, B. Floyd, and J. Mehta, Inter and intra-chip clock signal
distribution using microwaves, 1997 IEEE Solid State Circuits and Tech
nology Committee Workshop on Clock Distribution, Oct. 1997, Atlanta,
GA.


154
noise and interference at the input of the receiver fully quantified, this noise will drive the
sensitivity specification.
5.7 Estimation of Total Noise. SNR, and Jitter for 0.25-pm Receiver
The concepts in the past two sections have been applied to the 0.25-|im LNA and
source-follower circuits. Figure 5-9(a) shows the simulated gain and NF for the LNA. The
maximum gain is 22.8 dB and the minimum NF is 6.4 dB at 8 GHz. Figure 5-9(b) shows
the power spectral density of the noise at the input of the frequency divider, which was
given in (5.26). Integrating this PSD over all frequencies yields the total thermal noise
powerNdivT from (5.27)-which is -53 dBm. To perform an integral in SPICE over all
frequencies, the frequency limits are increased until the total integral does not change (~40
GHz for this case). Using the values 22.8 and 6.4 dB for G(f0) and NF(f0), respectively,
and NdivX of -53 dBm, BWN is calculated to be 1.5 GHz. This is shown in Figure 5-9(b),
Figure 5-9 (a) Simulated gain and noise figure for 0.25-|lm LNA with source-followers, (b)
Simulated power spectral density of thermal noise at input of frequency divider,
along with equivalent noise bandwidth (BWN).


125
Table 4-5 Performance summary for 0.1-p.m dividers.
SOI Divider
Bulk Divider
Division Ratio
4:1
128:1
vdd
1.5 V
1.5 V
Max. Input Freq.
18.75 GHz
15.4 GHz
Output Freq.
4.6875 GHz
120 MHz
Min. Input Freq.
9.1 GHz
<3.0 GHz
Power (of 4:1 core)
13.5 mW
9.8 mW
[DP]^ latch delay
13.4 ps
16.2 ps
The results are summarized in Table 4-5. This result is the highest operating fre
quency for a full-swing frequency divider to date. Also, these results show that a 0.1 -(lm
CMOS technology can support digital circuits operating above 15 GHz, and suggest
higher operating frequencies for fully-scaled 0.1 -|im technology and beyond.
4.8 Summary
An 8:1 divider topology based on a new source-coupled logic latch has been devel
oped. Two different modes of operation have been identified, including a standard digital
latching mode and an injection-locked mode. Injection locking is used to synchronize an
oscillator to a low-level input signal whose frequency is a superharmonic of the oscillation
frequency. Using injection locking allows the divider to lock to very low-level input sig
nals, resulting in conversion gain for the divider.
Using a standard 0.25-|im CMOS process, a 128:1 divider operating up to ~10
GHz for Vdd = 2.5 V has been implemented. For this supply voltage, the input-referred
self-oscillation frequency is ~7.3 GHz. Thus, the divider can detect very low-level input
signals (~40 mVpk_pk) close to this frequency. Using a 0.18-|im CMOS technology with
copper interconnects, a 64:1 divider operating up to 15.8 GHz at 1.5 V and 20.4 GHz at


231
circuit. Finally, jitter in the source or PLL will correspond to jitter on the output signal,
where the jitter as a percentage of the period is divided by 8 through the clock receiver.
Estimations for each of these jitter components will now be presented. First, the
jitter due to thermal noise has been simulated with SPICE. As was shown in section 5.7,
the simulated SNR due to thermal noise was 21.8 dB, resulting in -1% of peak jitter. This
simulation assumed an input thermal noise of kT. Second, to estimate the amount of inter
ference at the receiver input, digital test circuits or noise generators have been placed very
close to antennas [BravOOa, BravOOb], allowing the power spectral density to be obtained.
Measurements and simulations in [BravOOa, BravOOb] show that the total power due to
digital switching noise is -74 dBm. If the input signal level is at the desired receiver sensi
tivity of -54 dBm, then the signal-to-interference ratio (SIR) is 20 dB, resulting in 1.2% of
peak jitter. Third, since a CMOS PLL has not been built at 15 GHz, the peak jitter of the
PLL at 15 GHz is assumed to be 5%. Compared to results found in literature [Yan97,
Boe99, TamOO, Kae98, Mai97], this jitter specification should be attainable. With respect
to the output divide-by-8 signal, this results in 0.625% of peak jitter. Therefore, the total
peak jitter due to thermal noise, interference, and source (PLL) jitter is
n 2 2
V1 + 1.2 +0.625 = 1.68%. Since the peak-jitter specification is 3%, this leaves a
J32 1.682 = 2.5% margin.
The above jitter estimate is optimistic for two main reasons. First, the estimations
were made with noiseless or quiet supplies. When the microprocessor is operating, there
will be significant noise on the power and ground lines, even with power-supply decou
pling. This will generate jitter in both the transmitter and receiver circuits, consuming part
of the 2.5% margin. To help offset this jitter, the analog supplies for the wireless clock


LIST OF REFERENCES
[Adl46]
R. Adler, A study of locking phenomena in oscillators, Proc. IRE, vol.
34, pp. 351-357, June 1946.
[A1100]
I. Aller, K. Bernstein, U. Ghoshal, H. Schettler, S. Schuster, Y. Taur, and 0.
Torreiter, CMOS circuit technology for sub-ambient temperature opera
tion, ISSCC Digest Technical Papers, pp. 214-215, Feb. 2000.
[Bal97]
C. A. Balanis, Antenna Theory: Analysis and Design, 2nd edition, New
York: Wiley, 1997, pp. 133-164.
[Ban99]
R. Bansal, The far-field: how far is far enough? Applied Microwave and
Wireless, pp. 58-59, Nov. 1999.
[Bel95]
A. Bellaouar and M. I. Elmasry, Low-Power Digital VLSI Design, Circuits
and Systems, Boston, MA: Kluwer Academic Publishers, 1995, pp. 203-
207.
[Biy96]
M. Biyani, A divide-by-128/129 dual modulus prescaler using a 0.8-fim
CMOS process, M. S. dissertation, University of Florida, Gainesville, FL,
1996.
[Boa82]
A. D. Boardman, Electromagnetic Surface Modes, New York: John Wiley
& Sons, 1982.
[Boe99]
D. W. Boerstler, A low-jitter PLL clock generator for microprocessors
with lock range of 340-612 GHz, IEEE J. Solid-State Circuits, vol. 34, no.
4, pp. 513-519, Apr. 1999.
[Boh95]
M. T. Bohr, Interconnect scalingthe real limiter to high performance
ULSI, IEDM Technical Digest, pp. 241-244, Dec. 1995.
[Bow82]
C. Bowick, RF Circuit Design, Indianapolis, IN: H.W. Sams Publishing,
1982.
[BravOOa]
D. Bravo, Investigation of background noise in integrated circuits relating
to the design of an on-chip wireless interconnect system, M. S. disserta
tion, University of Florida, Gainesville, FL, 2000.
268


261
input and output signals. This theory is a summary of the relevant results presented in
[Adl46] and [Pac65]. Finally, the phase noise of the ILO will be presented.
E.2 Theory for Injection Locking
E.2.1 Basic Model
Figure E-l(a) shows the basic model for an ILO. Here, the oscillator is modeled
simply as a feedback network with a forward path of H(s). This model can therefore be
applied to either LC oscillators or ring oscillators. In the absence of an input signal, the
circuit will oscillate at its natural frequency, to0. At this resonant frequency and only at this
frequency, the output of the circuit, V0, and the error signal, E, are in phase. When a lock
ing signal with amplitude VL is injected into the loop at a frequency coL = co0 + Ago, a dif
ferent error signal is produced which is phase-shifted (and magnitude-shifted) with respect
to V0. The loop will then oscillate at a frequency where the phase-shift through H(s) is the
opposite of the phase-shift induced between Vo and E.
Phasor representations can be used to illustrate the circuit operation, where E is the
vector sum of the phasors VL and V0, as shown in Figure E-l(b). Here, VL is assumed to
de/dt
Figure E-l (a) Basic feedback model of an injection-locked oscillator,
(b) Phasor representation of the system.


APPENDIX D
NOISE PARAMETERS OF MOSFET
D. 1 Transistor Model and Equivalent Input Noise Generators
In this appendix the noise parameters of a single intrinsic MOSFET are derived,
including the effects of thermal noise in the channel, lumped both at the gate (gate-
induced noiseGIN) and the drain, substrate resistance, and all of the parasitic capaci
tances. Figure D-l shows a high-frequency, quasi-static model of the MOSFET [Tsi99]
including noise sources. The noise sources have the following power spectral densities:
.2
'-£f = UTySdo,
(D.l)
-A = 4m
A/
(D.2)
iftb AkT
2
A/ Rb
(D.3)
Vbs
+
'Rb
Figure D-l High-frequency, quasi-static model of intrinsic MOSFET [Tsi99].
252


CHAPTER 7
FEASIBILITY OF WIRELESS CLOCK DISTRIBUTION SYSTEM
7.1 Overview
While wireless interconnects have been demonstrated, overall system feasibility
has yet to be determined. One important feasibility issue is quantifying how much power a
wireless clock distribution system will consume and evaluating if this consumption is
within the budget allowed by the microprocessor. Another feasibility issue is evaluating
the effects of process variation on operation of the wireless interconnect. Perhaps the most
important feasibility issue is the clock skew and jitter of the system, since a clock distribu
tion system inherently has to be synchronized, else the whole function of a global clock
signal is compromised. This chapter estimates the worst-case clock skew and jitter that
will be obtained and compares these estimations to the measured skew and jitter presented
in Chapter 6. Also, the latency of the wireless interconnect is presented, along with how
this latency should scale with technology generations. Two intangible feasibility issues are
the area consumed by the transmitter and receivers, which has to be weighed against the
top-level metallization being conserved, and the design verification of the system. Using
these results, the current feasibility of the system is discussed. This is then followed by
overall conclusions for wireless clock distribution and suggested future directions.
7.2 Power Consumption Analysis
Whenever circuits are optimized for the highest possible operating frequency, it is
usually accompanied by an increase in the power dissipation. Since the wireless system
212


274
[098] K. K. 0, Estimation methods for quality factors of inductors fabricated in
silicon integrated circuit process technologies, IEEE J. Solid-State Cir
cuits, vol. 33, no. 8, pp. 1249-1252, Aug. 1998.
[099] K. K. O, K. Kim, B. A. Floyd, J. Mehta, and H. Yoon, Inter and intra-chip
wireless clock signal distribution using microwaves: a status of a feasibility
study, 1999 Govt. Microcircuit Applications Conf. Dig. Papers, pp. 306-
309, Mar. 1999.
[Pac65] L. J. Paciorek, Injection locking of oscillators, Proc. IEEE, vol. 53, no.
11, pp. 1723-1727, Nov. 1965.
[Pel95] M. M. Pelella, J. G. Fossum, Dongwook Suh, S. Krishnan, K. A. Jenkins,
Low-voltage transient bipolar effect induced by dynamic floating-body
charging in PD/SOI MOSFETS, Proc. IEEE Int. SOI Conf., pp. 8-9, Oct.
1995.
[Pra95] G. A. Pratt and J. Nguyen, Distributed synchronous clocking, IEEE
Trans. Parallel and Distributed Systems, vol. 6, no. 3, pp. 314-328, Mar.
1995.
[Rat99] H. R. Rategh and T. H. Lee, Superharmonic injection-locked frequency
dividers, IEEE J. Solid-State Circuits, vol. 34, no. 6, pp. 813-821, June
1999.
[Raz94] B. Razavi, K. F. Lee, and R. H. Yan, A 13.4-GHz CMOS frequency
divider, ISSCC Digest Technical Papers, pp. 176-177, Feb. 1994.
[Raz95] B. Razavi, K. F. Lee, and R. H. Yan, Design of high-speed, low-power fre
quency dividers and phase-locked loops in deep submicron CMOS, IEEE
J. Solid-State Circuits, vol. 30, no. 2, pp. 101-108, Feb. 1995.
[Res98] P. Restle and A. Deutsch, Designing the Best Clock Distribution Net
work, Symp. VLSI Circuits Dig. Tech. Papers, pp. 2-5, June 1998.
[Sch71] I. Schmideg, Harmonic synchronization of nonlinear oscillators, Proc.
IEEE, pp. 1250-1251, Aug. 1971.
[Sed91] A. S. Sedra and K. C. Smith, Microelectronic Circuits, 3rd edition, Fort
Worth, TX: Saunders College Publishing, 1991, pp. 869-872.
[Sha97] D. Shaeffer and T. Lee, A 1.5-V, 1.5-GHz CMOS low noise amplifier,
IEEE J. Solid-State Circuits, vol. 32, pp. 745-759, May 1997.


206
The measured transducer gains for the antennas on the AIN substrate at 15 GHz
are -54.1 dB for antenna pair Z6c_6a (distance = 6.9 mm) and -60.1 dB for antenna pair
ZfcSc (distance 5.6 mm). Interestingly, the gain for the diagonal pair, z6c-6a> is larSer>
even though the distance is greater. This seems to indicate that the signal degradation due
to interference structures is greater for antenna pair Z6c_6c. With a 20.7-dBm available
input power to the transmitting antenna, and a 0.6-dB mismatch loss, the power delivered
to each receiver is as follows: -34 dBm for Rx2 and -40 dBm for Rx4. The 6-dB difference
in input power levels can result in as much as 2.1% of clock skew, according to section
5.4, equation (5.17) (i.e., skew = cos_1{ 10'6/20}). Thus, at least a total of between
17C
3.6-4.4% of systematic clock skew due to time-of-flight delay mismatch and amplitude
mismatch should be present in this measurement.
6.7.3 Measured Skew Between Two Clock Receivers
The skew between the two receivers (for the case with Z6b as the transmitting
antenna) was extrapolated from the response of the phase detector when the start-up states
of the dividers were varied across their 8 possible values. Since the skew induced by the
measurement system will be constant at a single frequency, the output of the phase detec
tor can be compared to the skew measured with the oscilloscope. This results in a volt-
age-versus-skew response for the phase detector which should be maximum at the point of
zero skew between the two input signals, and which should be similar to the response
shown in Figure 4-26(b).
As discussed in Chapter 4, the programmable divider has a state-dependent initial
ization failure which causes the divider to initialize to any of the 8 possible states for a
given input desired state. The programmed state depends on the state of the divider when


10
In the clock receiver, the frequency divider translates the ~20-GHz global clock
signal to a ~2.5-GHz local clock signal. The divider must be capable of high frequency of
operation and low input-sensitivity while consuming minimal power. Also the dividers
should be synchronized between receivers to reduce the clock skew. The design and
implementation of high-frequency CMOS frequency dividers are presented in Chapter 4.
This section will present a design methodology based on injection locking for dividers
implemented with source-coupled logic (SCL). These dividers oscillate in the absence of
an input signal; hence, this oscillation can be injection-locked to provide frequency divi
sion from a low input-swing voltage signal. Measured results of SCL dividers operating up
to 10 and 15.8 GHz, at 2.5 and 1.5 V, using 0.25- and 0.18-|im CMOS technologies are
presented. Also, a dual-phase dynamic pseudo-NMOS divider [Yan99a] implemented in a
partially-scaled 0.1 -gm CMOS technology is presented, which operates up to 18.75 and
15.4 GHz on SOI and bulk substrates at 1.5 V, respectively. Finally, an initialization and
start-up methodology to synchronize the dividers in each clock receiver is presented,
which allows for decreasing the systematic skew.
In Chapter 5, the system requirements for the wireless clock distribution system
are developed, translating the clock metrics of skew and jitter into standard RF metrics. To
maximize the power transfer from the clock source to the local clock system, matching,
gain, and antenna requirements are discussed. Clock skew and jitter are then used to set
requirements for the power-level at the input of the frequency divider, the signal-to-noise
ratio (SNR) at the input of the frequency divider, the noise figure for the LNA with
source-follower buffers, and an DP3 for the LNA and source-follower buffers. The overall
system requirements are summarized and tabulated.


272
[HunOOb]
C.-M. Hung, L. Shi, I. Lagnado and K. K. 0, 25.9-GHz voltage-controlled
oscillator fabricated in a CMOS process, Symp. VLSI Circuits Dig. Tech.
Papers, pp. 100-101, June 2000.
[HunOOc]
C.-M. Hung and Kenneth K. O, Packaged 1.1-GHz CMOS VCO with
phase noise of -126 dBc/Hz at a 600-kHz offset, IEEE J. Solid-State Cir
cuits, vol. 35, no. 1, pp. 100-103, Jan. 2000.
[HunOl]
C.-M. Hung, B. Floyd, and K. 0, Fully integrated CMOS VCOs and pres-
calers, IEEE Trans, on Microwave Theory and Techniques, vol. 49, no. 1,
pp. 17-22, Jan. 2001.
[Jin85]
R. P. Jindal, High frequency noise in fine line NMOS field effect transis
tors, IEDM Technical Digest, pp. 68-71, Dec. 1985.
[Kae98]
V. R. von Kaenel, A high-speed, low-power clock generator for a micro
processor application, IEEE J. Solid-State Circuits, vol. 33, no. 11, pp.
1634-1639, Nov. 1998
[KimOOa]
K. Kim, Design and characterization of RF components for inter- and
intra-chip wireless communications, Ph.D. Dissertation, University of
Florida, Gainesville, FL, 2000.
[KimOOb]
K. Kim, H. Yoon, and K. K. 0, On-chip wireless interconnection with
integrated antennas, IEDM Technical Digest, pp. 485-488, Dec. 2000.
[Kis99]
S. V. Kishore, G. Chang, G. Asmanis, C. Hull, and F. Stubbe, Substrate-
induced high-frequency noise in deep sub-micron MOSFETs for RF appli
cations, Proc. Custom Integrated Circuits Conference, pp. 365-368, May
1999.
[Kle99]
B. Kleveland, C. H. Diaz, D. Vook. L. Madden, T. H. Lee, and S. S. Wong,
Monolithic CMOS distributed amplifier and oscillator, ISSCC Digest
Technical Papers, pp. 70-71, Feb. 1999.
[KnaOO]
H. Knapp, T. F. Meister, M. Wurzer, D. Zoschg, K. Aufinger, and L. Treit-
inger, A 79GHz dynamic frequency divider in SiGe bipolar technology,
ISSCC Digest Technical Papers, pp. 208-209, Feb. 2000.
[KuoOl]
T. C. Kuo and B. B. Lusignan, A 1.5W class-F RF power amplifier in
0.2pm CMOS technology, ISSCC Digest Technical Papers, pp. 154-155,
Feb. 2001.
[Kur68]
K. Kurokawa, Noise in synchronized oscillators, IEEE Trans, on Micro-
wave Theory and Techniques, vol. 16, no. 4, pp. 234-240, Apr. 1968.


183
between. The transmission gain changes by 5 to 10 dB, depending on the antenna and fre
quency. Also, the antennas have slightly different impedances (not shown), since the
reflection coefficient is a function of the metal structures reflecting the transmitted signal
back into the antenna. This shows that the structures located around the antenna affect the
antenna performance. The implication on system design is that this degradation in signal
level has to be accounted for when estimating the worst-case signal-to-noise ratio. Also,
this shows the need for electromagnetic simulation tools which can be used to try to esti
mate the effects of these structures beforehand, allowing design rules to be developed.
Much greater analysis of this problem is contained in [YooOO, KimOOa, KimOOb].
Figure 6-15(a) shows the reflection coefficients of the loop and zigzag antennas on
an impedance Smith chart, while Figure 6-15(b) shows the measured antenna impedance
for a zigzag (Z6b) antenna and a loop (LPj) antenna. The zigzag impedance is -100 Q
with a capacitive reactance; the loop impedance is -50 Q with an inductive reactance. The
mismatch loss between the receiving zigzag antenna and the LNA input (which was
matched to -100 Q) is approximately 0.3 dB at 15 GHz. Therefore, virtually all of the
power from the receiving antenna is transferred to the LNA.
Figure 6-16(a) shows the measured antenna gain for zigzag dipole antennas imple
mented in either metal 1, 2, or 3. At 15 GHz, the gains are -43, -49, and -49 dB, for anten
nas Zj, Z2, and Z3, respectively. The gain for the antennas in metal 1 is about 5 dB greater
than that in metals 2 and 3 for frequencies above -12 GHz. The antennas in metals 2 and 3
have nearly the same characteristic, where the gain decreases and then plateaus with fre
quency. Finally, the gain for these antennas at 3.6-mm separation is close to the gain of the
metal-6 antenna, Z6f (-45 dB at 3.2 mm). More work is required to understand these
effects, both with and without metal interference structures present.


182
Figure 6-14 (a) Loop-zigzag transmission gains for d=5.2 (LPpZ6f) and 4.8 mm (LPj-Z6a).
Also shown is the pad transducer gain (IS2il2) at d=5.6 mm for reference, (b)
Zigzag transmission gains for metal-6 antennas (Zga6b6c) at d=6.7 mm,
demonstrating the effect of interference structures on gain.
conclusively that the signal is propagating from one antenna to the other, and that these
waves are launched much more efficiently from the antennas than from the pads alone.
Figure 6-14(a) shows Ga versus frequency for a loop-zigzag antenna pair at 5.2-
and 4.8-mm separations (corresponding to antenna pairs LPrZ6f and LPrZ6a, respec
tively). At 15 GHz, Ga is -58 and -54 dB, respectively, which is less than the zigzag-zigzag
gains. These results are promising, though, and show that at high enough frequencies, loop
antennas are useful for the transmitting antennas.
Figure 6-14(b) shows Ga versus frequency for three pairs of zigzag antennas. Each
pair is separated by 6.7 mm, corresponding to antennas Z6a bb bc. These results demon
strate the effect of interference structures on antenna gain. As can be seen from Figure
6-10, each pair has different types and densities of metal and active structures located in


31
resistance (Rsub), and all of the capacitances (Cgs, Cgd, Cgb, Cdb and Csb) is an arduous
(albeit doable) task, yielding cumbersome equations. These equations can be found in
Appendix D, both for the general case considering all second-order effects, as well as
case-studies examining individual second-order effects. Rather than show these equations
here, the resultant parameters are evaluated numerically using Matlab. Figures 2-7(a)-(d)
show the effects of Qgs, Rsub, each of the capacitancesgate-to-drain (Cgd), gate-to-body
(Cgb), drain-to-body (Cdb), and source-to-body (Csb)and gmb on derived noise parame
ters for a single MOS transistor with GIN turned on and off. Each noise parameter is plot
ted versus the following normalized parameters: Qgs ranging from 0 to 10 with GIN both
on and off (a first-order effect including width dependence of drain noise), Rsub ranging
from 0 to 1000, and gmb/gm, Cgb/Cgs Cgd/Cgs, and Cdb sb/Cgs each ranging from 0 to 1.
The following are assumed: y = 1.2, f1 = \, and 8 = 2y. For each individual effect
VCD TJ 4
examined, the other parameters are held constant according to the assumptions listed in
Figure 2-7.
The results for Fmin, plotted in Figure 2-7(a) show a strong dependence on each of
the effects considered. First, to reduce Fmin, Rsub should be either very small (short) or
very large (open). Substrate resistance can be minimized by surrounding the transistor
with large-area substrate contacts very close to the channel. Second, to reduce Fm¡n, the
mechanisms which couple substrate noise into the transistor (i.e., gmb, Cgb, and Cdb)
should in general be minimized, where Cgb and Cdb are layout-dependent. Third, Fmin
decreases with increasing Cgd. This is due to feedback through Cgd decreasing the uncor
related noise resistance, Ru, caused by GIN. Finally, since Qgs is inversely proportional to
device width, the channel thermal noise and the coupled substrate noise are reduced as Qgs


45
Vdd=2.7V
Figure 3-2 Schematic of 900-MHz source-degenerated CMOS LNA including on-chip and
package parasitics.
3.3 A 900-MHz. 0.8-pm CMOS LNA
3.3.1 Circuit Implementation
To verify the design methodology for CMOS source-degenerated LNAs, a
900-MHz version was implemented in a 0.8-p.m CMOS technology with 3 metal layers.
This technology was obtained through MOSIS. Figure 3-2 shows the schematic of the
LNA. The important on-chip and package parasitics are shown, including series resis
tances for inductors and capacitors (rs), pad capacitance (Cj^), gate resistance (rg), induc
tor parasitic capacitance (cox), and bondwire/ground inductance (L^/Lg^). Both bias
voltages, Vgl and Vg2, are generated off chip for flexibility in testing. Figure 3-3(a) shows
a die photograph of the 900-MHz LNA [Flo99a], The circuit area is 720x720 fim2. This


30
.2
= 4kTb
A/
(2C
2\
gs
5Sdo
(2.30)
where 5 is another bias-dependent parameter (1.33 for long-channel devices). Since the
origin of both the drain and gate noise currents is the channel, these noise sources are cor
related, having a correlation coefficient (c) of j0.395 for long-channel devices. Due to the
gate-to-channel capacitance providing a 90 phase shift, c is complex.
Aside from GIN, there are three other second-order noise sources within the LNA,
as follows: channel noise of M2, thermal noise associated with gate resistance, and ther
mal noise generated in the substrate [Kis99]. To understand the relative importance of
these second-order effects, the noise parameters for a single transistor will be derived in
the following section.
2.6.2 Noise Parameters of Single Transistor
Figure 2-6 shows the high-frequency, quasi-static model for the intrinsic MOSFET
[Tsi99], including noise sources. Deriving the noise parameters for this single transistor
including drain and gate-induced noise, the body transconductance (gmb), substrate


155
where the PSD is fixed at the center-frequency value across the noise bandwidth. The
noise bandwidth agrees very well with the estimated value of 2 GHz given in (5.41).
The SNR can now be calculated assuming that the input signal to the receiver is
-54 dBm, and that the total input noise is dominated by thermal noise (as was similarly
done in [BravOOb]). The -54-dBm input power is the desired sensitivity, which again is a
result of 10-dBm transmitted power, -56-dB antenna-to-antenna gain at 1.5 cm, 7 dB of
interference effects, and 1 dB of mismatch loss. The signal power level at the input of the
divider is -54 dBm plus the gain of the LNA (22.8 dB), or -31.2 dBm. Therefore, the sim
ulated SNR is -31.2 dBm (-53 dBm) = 21.8 dB. Applying the 21.8-dB SNR to (5.35), the
expected RMS cycle-to-cycle jitter is 1.6 ps at the output clock frequency of 1 GHz, corre
sponding to 0.16% RMS jitter. The peak-to-peak jitter is 9.7 ps, which is ~1% of the clock
period. This easily meets the specification of less than 3% of peak-to-peak jitter.
5.8 Linearity Specification
Nonlinearity in the LNA (with source-followers) will result in the following main
effects: generation of harmonics, gain compression specified in terms of a 1-dB compres
sion point (P¡dB)> and intermodulation specified in terms of a third-order intercept point
(IP3). A difficulty encountered in the clock receiver is the fact that the signals outside of
the band of interest are not heavily attenuated. In other words, the response of the LNA
is very broadband due to the limited inductor Q in the LNA. Therefore, interfering signals
themselves pose more of a problem than the harmonics or intermodulation products these
signals generate. In fact, if the interfering signals become too large, they can lock the
divider at a different frequency, even with the reduced gain of the LNA at the interferers
frequency. At the very least, these interfering signals will degrade the SNR, increasing the


221
Figure 7-4 Process variation of single-ended LNAs gain (S2j), noise figure (NF), and input
matching (Su), obtained with Monte Carlo analysis (30 iterations).
7.3.2 LNA and Frequency Divider Variation
Figure 7-4 shows the simulated gain, noise figure (NF), and input-matching (Su)
variation for a single-ended low noise amplifier (LNA) from a Monte Carlo simulation
with 30 iterations. At 8.5 GHz, the gain (S2j) varies by 0.7 dB, the NF varies by 0.2 dB,
and the Su varies by 1 dB. The variation in the phase (not shown) of S2i is 6, resulting in
a 0.2% skew referenced to the output local clock signal.
Figure 7-5 shows the variation obtained with Monte Carlo simulations of the 8:1
frequency divider for a 500-mV amplitude, 8.5 GHz input signal at Vdd = 2 V. At 1 V, the
skew is 26 ps at the rising edge and 42 ps at the falling edge. Since the output clock period
is -940 ps (8/8.5 GHz), the skew due to process variation in the frequency divider is 4.5%.
To obtain the variability in the self-resonant frequencies of the dividers, a fast-fourier
transform was taken on the output signals of the first 2:1 divider when it was


16
Figure 2-1 Potential CMOS LNA topologies: (a) common-gate or (b) common-source with
inductive degeneration.
network composed essentially of the source inductance (Ls), the gate-to-source capaci
tance (Cgs), and one over the transconductance (l/gm). The input impedance is designed to
be ~50 Q; however, to maximize the gain and minimize the noise figure, the input imped
ance can be set to ~35 Q while still achieving an Sjj of -15 dB. Thus, gm is set to approx
imately 1/35 Q"1, while Ls is chosen to parallel-resonate with Cgs at the operating
frequency. In addition to easily providing the input match, the common-gate topology
exhibits good linearity, due to the source degeneration of the transistor provided by Rs.
A drawback of this topology is its higher noise figure. As shown in Appendix A,
the noise factor for this topology is approximately the following:
F = 1 +
a \8mRs)
(2.4)
where a is the ratio between the device transconductance (gm) and the short-circuit drain
conductance (gd0). In the long-channel limit, y and a are 2/3 and 1, respectively, while y
can be significantly larger than 2/3 for short channel lengths, due to hot-electron effects
[Jin85]. Thus, in the long-channel limit and with gm=l/35 Of1, the noise factor is 1.47,
yielding a noise figure1 of 1.66 dB. However the noise figure can be significantly larger in


14
matching, and output matching. Implementation results for four different LNAs, operating
at 0.9, 8, 14, and 23.8 GHz, respectively, are then presented in the next chapter.
2.1.2 Performance Metrics of LNAs
As mentioned above, to minimize the receivers noise figure, the LNA is required
to have moderate to high gain and low noise figure. Examining (2.1), it would seem that
the LNAs gain can be increased arbitrarily, thereby minimizing the noise figure and
improving the sensitivity of the receiver. However, this is not the case due to the effect of
LNA gain on receiver linearity. Linearity is typically measured in terms of a third-order
intercept point (IP3) which is usually input-referred (IIP3). The IP3 is the output power at
which the third-order intermodulation products are equal to the desired linear component.
The expression for the total IIP3 for a cascaded system can be expressed as follows:
1 i Gi Gi Go
IIP3t IIP3j IIP32 IIP33 V
where HP3¡ and G¡ are the input-referred IP3 (in watts) and the power gain (in watts/watt)
of each individual stage of the receiver. For gains greater than one, the linearity of latter
stages will dominate the total receiver linearity. Therefore, to maximize the receivers
IIP3, the latter stages linearity should be maximized while reducing or limiting the total
preceding gain. Thus, the gain of the LNA (i.e., Gj) cannot be increased arbitrarily, since
that would degrade the receivers IIP3. Therefore, (2.1) and (2.3) imply an acceptable
range of LNA gains which will meet both the systems noise figure and linearity require
ments. Also from (2.3), it can be seen that the requirement for the LNAs IIP3 is not as
stringent as that for subsequent stages (e.g., mixer). Typically, the LNA is designed to have
a power gain of approximately 15 dB, a noise figure of less than 2 dB, and an IIP3 of -5
dBm, all the while consuming less than ~10 mW of power.


237
operation, where at low input power levels, the system does not operate across all process
variations. This necessitates the need to control the gain in the receiver, through adjusting
the resonant frequency of the amplifier and adjusting the self-resonant frequency of the
injection-locked divider. Fourth, the latency of the system is -340 ps for a 15-GHz,
0.18-p.m wireless interconnect, and future technology scaling will benefit wireless inter
connect latency over conventional interconnect latency. Fifth, the wireless clock distribu
tion system will consume approximately 1% of the projected microprocessor area;
however, this is offset by the top-level metal area that will be conserved. Sixth, design ver
ification of a wireless clock distribution system is currently difficult, and further work is
required. All of these feasibility results are promising, in that none are dissuasive. On the
contrary, these results are promising enough to warrant further investigation of larger-scale
wireless clock and interconnect systems.
7.7.2 Conclusions for Wireless Clock Distribution Systems
Examining these results, it can be seen that comparable power consumption, skew,
and jitter can be obtained with a wireless clock distribution system, as compared to con
ventional systems, with potential costs of added area and more difficult design verifica
tion. The benefits of a wireless clock distribution system, as discussed in Chapter 1, can be
summarized as being able to provide high frequency clock signals with little to no disper
sion over large distances, where the latency of the interconnect is improving with technol
ogy. This should increase the maximum clock frequency of microprocessors. Additionally,
these benefits can be obtained using conventional CMOS technology. Thus, while there is
a circuit overhead for a wireless clock distribution system, such a system could be used to
provide global interconnects for microprocessors operating well above 2500 MHz. In


198
Table 6-4 Summary of measured characteristics for 0.18-p.m CMOS LNA and divider
LNA and
SF Buffer
Expected
Measured
Frequency
Divider
Expected
Measured
f0
21 GHz
14.4 GHz
Division Ratio
8:1
Gain
27 dB
21 dB
fmax(Vdd=1.5V)
(Vdd=2.1V)
35 GHz
42 GHz
15.8 GHz
20.4 GHz
NF
2.94 dB
8 dB
fiSO (Vdd=1.5V)
(Vdd=2.1V)
28 GHz
34 GHz
9.15 GHz
14.2 GHz
Power
11.2 mW
28 mW
Power (Vdd=1.5V)
(Vdd=2.1V)
6 mW
18.5 mW
4.5 mW
12.2 mW
increased from 1.5 to 2.1 V, increasing fiso from 9.2 to 14.2 GHz. Operation of the clock
receiver is demonstrated by transmission of a global clock signal across the chip, detection
of this signal, and generation of a local clock signal by a single clock receiver.
The same measurement setup shown in Figure 6-9(a), and again in Figure 6-26(a),
was used to test the clock receiver. The input signal is generated off-chip and externally
amplified. This signal is then transformed into a balanced signal and injected into the
transmitting antenna on-chip. The output of the receiver located across the chip is then
probed and examined using an oscilloscope. Two clock receivers (Rxl and Rx3 from Fig
ure 6-10) were tested, having antenna separations of 3.2- and 5.6-mm, respectively. Figure
6-26(b) shows plots of the input voltage to the transmitting antenna and the output voltage
for the wireless clock receiver, for a 5.6-mm transmission distance (Rx3), demonstrating
operation of the clock receiver with integrated antenna. The input global clock frequency
is 15 GHz, and the output local clock frequency is 1.875 GHz (15 GHz divided by 8). The
minimum power level needed at the transmitter for the clock receiver to lock to the input
signal is 20.33 dBm. Note that the input signal plotted in Figure 6-26(b) is taken before the


20
operating frequency band, while withstanding component variations in Lg, Ls, and Cgsj.
Qualitatively, the input impedance is more sensitive to component variations for high-Q
networks. Thus, there is a maximum value for the quality factor which ensures matching
for a given frequency band and for a given set of component tolerances. An alternative
input quality factor independent of Ls can be defined as follows:
1
Assuming that inductors have a tolerance of TL, Cgsl
tolerance of Tt, equation (2.5) becomes the following:
has a tolerance of T
O
(2.11)
and coj has a
zin = C tLs( 1 Tl){ 1 Tt) + ML +Ls)(1Tl) + -
= co
j(0Cgsl(lTc)
-L,(1Tl)(1 Tt) + JQ,a(£( 1 TL)-
.(2.12)
Note that the variation in cox (which has a |vjai|/L dependence in the velocity-saturated
regime [Tsi99], where vsat is the saturation velocity) is partially correlated to the variation
in Cgs (which has a WLC0X dependence), through variations in the length of the device.
The impedance in (2.12) has to satisfy the input matching condition at both the upper and
lower ends of the frequency band (i.e., co = C0o-, where B is the total bandwidth). The
input impedances which satisfy Sj j < -10 dB, can be visualized on a Smith chart as those
impedances located within a circle centered at the origin with radius 0.32 (10'10/2), as
shown in Figure 2-3. If the input impedance is normalized and written as z¡n=(r)+j(i), the
required value for i to result in a specific S j j for a given value of r is as follows:
i =
|S,l|V+l)2-(r-l)2
i-M
(2.13)


216
networks need to be adjusted. In this work, the former was chosen to keep the analysis
straightforward. This results in shifting some power from the global distribution system to
the local distribution system, resulting in an underestimation of the total global capaci
tance. However, since each clock distribution system drives an identical clock load, mean
ingful comparisons are still obtained.
Two cases are used in comparing these clock distribution systems for 0.1 -|im gen
eration microprocessors. Case 1 is for aluminum (Al) interconnects and conventional
dielectrics; case 2 is for copper (Cu) interconnects and low-K dielectrics. To obtain case 2,
Cw and the components of CG arising from wires were scaled by additional factors of 0.63
and 0.5, representing the decreased width of Cu-lines as compared to Al-lines for constant
resistance lines, and the reduction in K from 4 to 2, respectively.
7.2.2 Clock Distribution Systems
Grid-Based System. The grid-based system, shown in Figure 7-3(a) and based on
DECs 21264 [Gie97], consists of a global tree supplying the clock to different spine loca
tions, and the capacitance of this network is CG. These spines drive the clock grid, which
has a capacitance Cw, from the four edges of the grid. The local clock generators then tap
off of the grid. Due to the amount of wiring used to form the grid, Cw is large. Conse
quently, large sector buffers are needed, thereby increasing CG.
H-Tree Based System. The H-tree system, shown in Figure 7-3(b) and based on
IBMs S/390 [Web97], consists of a global tree supplying the clock to different sector
buffer locations. Each buffer drives a balanced H-tree, which drives the local clock gener
ators. Cw is smaller for the H-tree; therefore, a smaller sector buffer is required.


170
the interference structures between the antennas. Unfortunately, a wireless interconnect
designer has no control over these layers, and, thus, cant predict the performance for the
antenna pair. Another path for signal propagation is through the substrate and back-layer
dielectric. These mediums are unused in normal system design, and therefore are
well-suited for wireless interconnects. These mediums can actually support two modes.
One is a reflected path, where the signal refracts through the substrate and dielectric layer,
bounces off of the back-side metal layer, and then travels back up to the receiving antenna.
This path is labeled path C. A third path is the substrate surface wave (also known as a lat
eral wave), labeled path B. This wave is generated from antennas which have finite beam
widths [KimOOa] (i.e., it does not exist for plane waves). Analyses of this path are con
tained in [Boa82] and [KimOOa]. Experiments show that for the case when no interference
structures are present, the gain through the substrate paths (B and C) can be comparable to
the gain through the direct path A. The relative strength of paths B and C depends on the
type and size of the underlaying dielectric medium [GuoOl]. Much greater detail on the
modeling of each of these paths is contained in [KimOOa, GuoOl].
6.3 Antenna Characteristics in 0.25-[im CMOS
Figure 6-3 shows a photo-illustration of a test chip used to characterize both anten
nas and antennas with LNA, and to demonstrate an on-chip wireless interconnect. This test
chip was fabricated in a standard 0.25-|im CMOS process and is approximately 7.8 mm .
The 0.25-p.m LNA and frequency divider presented in chapters 2 and 3 have also been
implemented in this test chip. Located on the far left and far right are a pair of linear dipole
antennas, which are used to characterize the antennas as well as to transmit to the receiver
circuitry. Also, on the right is an LNA with an integrated antenna, while on the left is a


94
Figure 4-10 Measured output waveform from 0.25-p.m CMOS 128:1 frequency divider. Input
frequency is 9.98 GHz for Vdd = 2.5 V and output frequency is 78 MHz.
maximum input frequency of 10 GHz. This input swing is almost equal to the supply volt
age; thus, the MDS of the receiver for this frequency would be very high. However, if the
input frequency is decreased to 7.4 GHz, then only a ~40-mVpk_pk input signal is required,
directly corresponding to an improvement in the receiver MDS. The measurements shown
in Figure 4-9(b) were not obtained at very low frequencies; thus, the characteristic input
sensitivity plot, as shown in Figure 4-4(a) was not completely obtained. Specifically, for
frequencies below fiSO the required input voltage swing should increase. This trend is
only beginning to be evident for the 2.0-V and 2.5-V plots in Figure 4-9(b). The fisos are
approximately 6.1 and 7.3 GHz for Vdd = 2.0 and 2.5 V, respectively. These numbers do
not agree well with the simulated values from SPICE, due to an inconsistency between the
simulated transistor model capacitance and the model capacitance fit to the actual wafer
data. Table 4-2 summarizes the measured results for the 0.25-|im frequency divider.


96
TSPC Divide-by-8s
Figure 4-11 Block diagram of 64:1 divider implemented in 0.18-pm CMOS technology.
Figure 4-12 Die photograph of 0.18-pm, 64:1, 15.8-GHz source-coupled divider.
Table 4-3 Transistor sizing for 0.18-pm 64:1 frequency divider
SCL 8:1 Divider
Transistors
First 2:1
Second 2:1
Third 2:1
Wj, w2, w11; W12
7 pm
1.94 pm
1.2 pm
W3, w4, w13, w14
5 pm
3 pm
2.4 pm
w5,w6,w15,w16
1.5 pm
1.7 pm
1.5 pm
W7,w8,w9,w10
2.6 pm
1.2 pm
0.94 pm
TSPC 8:1 Divider
Sizes Are Same for each 2:1 TSPC Divider
W2, W3, w4, w5, w6, w7, w8, w10
2 pm
Wj,Wn
1 pm
w9
4 pm
Outputs


185
Table 6-2 Comparison between measured and expected antenna characteristics, 0.18-p.m
Antenna Pair
Parameter at 15 GHz
Expected
Measured
Zigzag to Zigzag
Gain (d=6.7mm)
-48 dB
-53 dB
Impedance
(Zigzag)
125 j55 Q.
89 -j43 Q
Loop to Zigzag
Gain (d=5.2mm)
-51 dB
-58 dB
Impedance
(Loop)
30+j80Q
43 +jll4 £2
Figure 6-16(b) shows the measured gain versus frequency for the substrate
antennas. These substrate antennas are implemented in n+ in p-substrate and p+ in n-well,
overlaid with metal 1 and multiple contacts. At 15 GHz for a 3.65-mm separation, the
gains for a n-plus antenna implemented in p-substrate and for a p-plus antenna imple
mented in n-well are -51 and -49.5 dB, respectively. The gains are between 7 to 3 dB less
than that for the metal-6 zigzag (d=3.2 mm), for frequencies between 6 to 18 GHz, respec
tively. This result suggests that substrate antennas can be used for wireless interconnection
at frequencies above 18 GHz. Further feasibility studies are required
A comparison between the expected and measured antenna characteristics at 15
GHz is shown in Table 6-2. The expected characteristics were extracted from the results in
[KimOOa], These characteristics agree remarkably well, with the difference attributed to
the interference structures located between the antennas. While this result is encouraging,
the fact that the results on the 0.25-pm test chip for very similar antennas did not match is
equally disturbing. In particular, the impedance of the 0.25-pm antennas was about one
fifth of the 0.18-pm antenna impedance, resulting in a large mismatch. Both test chips
have metal-fill patterns which are required to maintain chemical-mechanical polishing
(CMP) uniformity. One difference between the chips is that the 0.18-|im chip allowed for


CHAPTER 3
CMOS LNA IMPLEMENTATION AND MEASUREMENTS
3.1 Overview
The previous chapter presented common-gate and source-degenerated LNAs
which can achieve high gain, low noise figure, and low power consumption. Also, a
detailed design methodology for source-degenerated LNAs was presented, considering
gain, NF, input matching, and output matching. Methodologies, however, are only as good
as their demonstrations; therefore, this chapter presents the implementation of CMOS
LNAs, including design of passive components. Low noise amplifiers operating at 0.9, 8,
14, and 24 GHz are presented. The LNAs are implemented in 0.8-, 0.25-, 0.18-, and
0.10-|im CMOS technologies, respectively.
3.2 Passive Components
3.2.1 Inductor Design Techniques
The quality factor of the on-chip spiral inductors directly affects both LNA gain
and noise figure. As already discussed, should be maximized to improve gain while
decreasing power consumption. Although the series resistance of Lg was not included in
(2.44), it can easily be shown that the series resistance of Lg (rLg) increases F by the quan
tity rLg/Rs. For these reasons, high quality-factor inductors are required for Lg and Ld.
Depending on the inductance and the operating frequency, it may not be practical to inte
grate Lg, which is the case for a 900-MHz LNA.
42


168
can have the same gain as a longer linear dipole antenna, while being more area efficient at
the cost of added capacitance [KimOOa]. In other words, for the same axial length, the zig
zag dipole antenna will have higher gain than a linear dipole antenna.
Loop antennas are also being considered for wireless interconnects. A loop
antenna whose total circumference (including multiple turns) is less than have a
small Rr Since the radiation resistance is much smaller than the loss resistance, loop
antennas have very low efficiencies. This is a drawback of using loop antennas for clock
transmission. To increase Rp additional loops can be included. A horizontally oriented
loop antenna is omnidirectional in the horizontal plane. Also, the loop antenna has an
aspect ratio of 1, compared with the zigzag dipole antenna which has an aspect ratio of 25,
indicating that loop antennas are more compatible to integrated circuit layout. Thus, a
multi-loop antenna is a potential candidate for the transmitting antenna, provided that Rr
can be sufficiently increased. As a first step towards evaluating this potential, a single-loop
antenna with a 200-|lm diameter and a 10-|im line width has been implemented [KimOOa].
6.2.3 Propagation Paths for On-Chip Antennas
Multiple propagation paths exist between two on-chip antennas. Figure 6-2 shows
an illustration of a cross-section of a chip with antennas. These antennas are implemented
in the top-level metal above an oxide layer. Located between the antennas are multiple
metal test structures which interfere with the transmitted clock signal. Also, the chip will
be flip-chip bonded onto a package substrate using solder balls. These ~100-|im diameter
metal balls will interfere with the global clock signals. Finally, within the PC board there
will be ground planes, which will again result in reduced antenna gain.


166
6.2.2 Types of Antennas
The ideal on-chip transmitting antenna would have an omnidirectional radiation
pattern for the plane parallel to the wafer surface. Also, the transmitting antenna would
have a large power handling capability. For clock distribution, in which there is only one
transmitter whose position is fixed, the ideal on-chip receiving antenna would be directive,
maximizing the antenna gain and minimizing interference and noise. Also, the receiving
antenna should be area efficient since multiple receivers are included on chip.
Currently, dipole and loop antenna structures are being considered for single-chip
wireless interconnects [KimOOa]. A horizontal dipole antenna is omnidirectional in the
vertical plane perpendicular to the antenna axis. For a dipole antenna whose electrical
length is less than approximately 0.85A., the radiation resistance increases with antenna
length [Bal97]. Also, the current distribution within the antenna increases with antenna
length; thus, longer dipole antennas radiate more power. Half-wavelength dipole antennas
(length = A,/2) are very common due to their radiation resistance being ~ 75 £2, which is a
traditional transmission-line characteristic impedance. The corresponding frequency at
which the antenna is a half-wavelength long is the resonant frequency of the antenna.
For on-chip antennas, the antenna size is limited by the size of the chip. Therefore,
to maximize the antennas radiation while limiting the physical size of the antenna
requires operating at higher frequencies (e.g., > 15 GHz), corresponding to smaller A,. For
reference purposes, Table 6-1 shows the wavelength versus frequency for the following
mediums: free-space, silicon dioxide, and silicon substrate. The dipole antenna length has
been limited to 2 mm, corresponding to A/10 and A/5 at 7.4 and 15 GHz, respectively, in


29
2.6 Noise Parameters
The noise factor of any two-port network can be classified in terms of four noise
parameters as follows:
F = Fmh + ^l\Zs-Zopl\2}, (2.28)
where Fmin is the minimum obtainable noise factor, Gn is the noise conductance, Zopt -
Ropt + jXopt is the optimum source impedance resulting in F = Fmin, and Zs = Rs + jXs is
the actual source impedance. Note that these noise parameters are more convenient than
the four traditional noise parameters (Fmin, Rn, Gopt, Bopt) [Gon97] for the source-degen
erated LNA, due to the source being in an impedance form rather than an admittance form.
A review of this set of noise parameters in impedance form is contained in Appendix C.
2.6.1 Noise Sources
The primary noise source within the CMOS LNA is thermal noise generated by the
distributed channel of the MOS transistor. A secondary noise source is thermal noise gen
erated by the series resistance of spiral inductors. Channel noise can be lumped into two
current-noise sourcesone at the drain and one at the gate. The drain noise-current has the
following power spectral density:
~2
Jf = MTygJc, (2.29)
where y is a bias-dependent parameter (0.67 for long-channel devices), k is Boltzmanns
constant (1.38 x 10'23 J/K), T is absolute temperature, and gdo is the short-circuit drain
conductance. The gate noise-current, a second-order effect known as gate-induced noise
(GIN) [Zie86, Sha97, Tsi99], has the following power spectral density:


172
Port 1
Network
Analyzer
Port 2
Baiun
\
Balanced
Probe
f ^ Antenna d
oLi v<
Semi-Rigid Cables
/
Balanced
!M!
-f Dielectric
/
Baiun
J I 1
\ /
Antenna
*/
Wafer
Glass mm
Figure 6-4 Measurement setup used to characterize on-chip antenna pair. On-wafer
measurements are performed with the die mounted on a glass slide above a
dielectric insulator.
Figure 6-5 (a) Transducer and transmission gain of linear dipole antenna pair,
(b) Phase versus frequency for linear dipole antenna pair.
antennas are implemented in metal 5 and separated from the substrate by -7 Jim of oxide.
These results are for a substrate resistivity of ~ 8 Q-cm. As can be seen, the antenna is
more efficient at higher frequencies, or when signal wavelengths are smaller. A transducer
gain of -64 dB is attained at 7.4 GHz, while the transmission gain is -49 dB. The large
difference between the transducer and transmission gain is due to the antenna impedance
being significantly different than the 100-Q differential impedance (i.e., differential volt
age over differential current). Since the 0.25-|im LNA, presented in section 3.4, is


262
be at rest with respect to the observer; thus, the plane itself is rotating at an angular fre
quency of coL. Alternately, the actual sinusoidal voltage signals will be the projection of
these phasors onto a line which is rotating at coL. Both VQ and E will rotate in this refer
ence plane at a velocity of ^, corresponding to an angular frequency of coL+^.
E.2.2 Differential Locking Equation
From Figure E-l(b), ()) can be related to 0, VL, and V0 by the following:
tan(j) =
V^sinG
V q + VlcosQ
(E.l)
The open-loop oscillator circuitthat which is represented by H(s)is assumed to be gen
erating a phase shift which decreases linearly versus frequency around the band of inter
est. The slope of this phase versus frequency response is
A =
d(j)
d(£>
(E.2)
For a tuned circuit with quality factor, Q, A = (2<2)/cdo Therefore, at an input frequency
dB
of to=coL+, H(s) generates a phase shift (with respect to the phase at co0) of
<)> = -A(to- co0) = -a(l + j-t ~ = Ai0o) (E3)
where A(0o is the difference between the natural and locking frequencies (g0-col). Plug
ging this into (E.l) and making a small-angle approximation of tan <)> = <)>, results in the
locking equation for an ILO, as follows [Adl46, Pac65]:
dQ _i\_ sin9
dt~ 0)0 U' V0J' l + (VL/yo)cos0
(E.4)


160
different global clock frequency is to be used, then these specifications can be modified
using the formulas presented in this chapter.
Some general requirements for the system which have not been discussed include
power consumption, balanced topology, and passive components. First, the power con
sumption is specified to be 40 mW per receiver at 1.5 Vdd. This number is based on the
analyses performed in Chapter 7, resulting in the power consumption of the wireless clock
distribution being comparable to that of conventional grid and H-tree distribution systems.
Second, the receiver should employ a balanced topology. This will obviate the need for a
balanced-to-unbalanced conversion between the antenna and LNA, reject common-mode
noise [Meh98, BravOOa], and provide dual-phase clock signals to the frequency divider.
Common-mode noise rejection will reduce the interference picked up by the receiver;
hence the SNR will improve and the jitter will decrease. Finally, the quality factors of the
passive components should be large. In silicon integrated circuits, typically the inductor Q
will limit the performance of the system. This is the case with CMOS LNAs, as shown in
Chapters 2 and 3. This will be further exemplified in Chapter 6 in the presentation of the
0.18-|im clock transmitter and receiver results. As the inductor Q improves, the equivalent
noise bandwidth will decrease, noise figure will decrease, receiver gain will increase, SNR
will increase, and clock jitter will decrease. Clearly, Q is integral to the performance of the
entire system. Based on the results presented in Chapter 6, inductor Qs of ~30 at 15 GHz
are recommended. However, as has been shown in Chapter 2 and will be shown in Chapter
7, high-Q circuits are more susceptible to process variations; thus, there are trade-offs
involved in choosing the inductor Q. More work is required to develop the optimal Q and/
or to develop high-Q circuits which are not as susceptible to process variations.


134
M =
1
1
-1
*
c
ant
1 -
r r
cl ant
(5.8)
As can be seen, Teq is a bilinear transform of rc. A circle with radius IFeql (corresponding
to Mant, Lmm, or VSWR) in the req plane will, therefore, map into a circle in the rc plane,
allowing the circuit impedance to be chosen for a given antenna impedance. Appendix A
of [Gon97] contains equations for circle mapping due to bilinear transforms; thus, the der
ivations are not repeated here. The equation for the circle in the Tc plane for a given Tant is
|rc-c| = r, (5.9)
where c and r are the center and radius of the circle in the Tc plane, given as follows:
c = r*
c 1 ant
f i ir I2 ^
r eg\
Hre/lrj7
r =
l£JHT
anti
i ir r i2
I eq| |x ant\
(5.10)
(5.11)
As can be seen, for perfect matching (i.e., M=1 or Teq = 0), the circle in the Tc
plane is centered at T*ant with a radius of 0, corresponding to conjugate matching. Figure
5-2 shows a family of circles in the rc plane corresponding to mismatch losses of {-1,
-0.5, -0.25, and -0.1} dB, VSWRs of {2.7,2.0,1.6, and 1.4}, and Teqs of {-6.9 -9.6, -12.5,
and -16.4} dB, when reading from the outside circle to the inside circle. These circles are
for an antenna impedance of 89-j45 £2, plotted on a Smith chart normalized by 100 £1 The
second circle from the outside corresponds to VSWR=2; therefore, all circuit impedances
located within this circle will meet the system specification, resulting in less than 0.5-dB
mismatch loss per antenna.


4.5 A 15.8-GHz, 0.18-p.m CMOS SCL Divider 95
4.5.1 Circuit Implementation 95
4.5.2 Measured Results 97
4.6 Programmable Frequency Divider Using SCL 98
4.6.1 Motivation and General Concept 98
4.6.2 System Start-Up Methodology 102
4.6.3 Circuitry for Programmable SCL Divider 104
4.6.4 Simulated Results 109
4.6.5 Testing Methodology 114
4.6.6 Measured Results 116
4.7 A 0.1-p.m CMOS [DP]2 Divider on SOI and Bulk Substrates 118
4.7.1 Circuit Implementation 118
4.7.2 Effect of SOI on Circuit Performance 120
4.7.3 Measured Results 123
4.8 Summary 125
5 SYSTEM REQUIREMENTS FOR WIRELESS CLOCK DISTRIBUTION 127
5.1 Overview 127
5.2 Power Transfer from Clock Source to Local Clock System 128
5.2.1 Transmitter Power 129
5.2.2 Antenna Specifications 130
5.2.3 Receiver Gain 131
5.2.4 Matching Between Antennas and Circuits 132
5.3 Definition of Clock Skew and Jitter 135
5.4 Clock Skew Versus Amplitude Mismatch 136
5.4.1 Latched Mode 137
5.4.2 Injection-Locked Mode 140
5.5 Clock Jitter Versus Signal-to-Noise Ratio 142
5.5.1 Phase Noise of Frequency Dividers 143
5.5.2 Conversion From Additive Noise to Phase Noise 144
5.5.3 Output Phase Noise of Clock Receiver 146
5.5.4 Jitter in Clock Receiver 147
5.6 Sensitivity and Noise Requirements 151
5.6.1 Receiver Sensitivity 151
5.6.2 Receiver Noise Requirements 152
5.7 Estimation of Total Noise, SNR, and Jitter for 0.25-|im Receiver 154
5.8 Linearity Specification 155
5.9 Summary 159
6 WIRELESS INTERCONNECTS FOR CLOCK DISTRIBUTION 162
6.1 Overview 162
6.2 On-Chip Antennas 163
6.2.1 Antenna Fundamentals 163
6.2.2 Types of Antennas 166
6.2.3 Propagation Paths for On-Chip Antennas 168
v


119
than the pull-up action of the PMOS transistor. This is done by choosing the widths of Mnl
W
and Mn2 such that their |iC0xy is approximately 4 times that of Mpl.
L/
Figure 4-29 shows a schematic of a 4:1 frequency divider, employing [DP]2 logic
in a divide-by-2N topology. The synchronous divide-by-2N circuit is formed by placing N
edge-triggered DFFs with an inverter in a loop. Each edge-triggered DFF is composed of
two level-sensitive [DP]2 latches with opposite clock polarities, and each [DP]2 latch is
simply a clocked pseudo-NMOS inverter [Biy96]. The circuit basically operates as a rip
ple chain, where a change at a single divider node will ripple through each stage, returning
to the same node N clock-periods later with opposite polarity, taking 2N clock cycles for a
full output period. The [DP]2 latch delay can be extracted from the speed of the divider,
since the maximum clock frequency is limited by the delays through the final [DP]2
level-sensitive latch and the pseudo-NMOS inverter (which are approximately equal).
Therefore, the delay through both of these will be approximately one half the minimum
input clock period, and the [DP]2 latch delay is then one fourth the minimum input clock
mX
In
' M
CLK|
CLK|
P2
CLK
>M
H
pi
^Out ,
clkH
Mn2
(a)
Mn1
1
m H
M
p2
Q
M
n4
1
M
n3
Figure 4-28 Schematics of (a) C2MOS inverter, (b) [DP]2 inverter or level-sensitive latch and
(c) [DP]2 edge-triggered D-Flip-Flop


253
where y and 8 are bias-dependent parameters, k is Boltzmanns constant, T is absolute
temperature, and gdo is the short-circuit drain conductance. The polarity of the noise
sources id, and iRb is arbitrary; however, the polarity of ig should be taken opposite to id to
yield a correlation coefficient with the correct polarity, equal to
(D.4)
c
For long-channel devices, c is purely imaginary, taking a value c=j0.395.
As discussed in Appendix C, any noisy two-port network can be represented as a
noiseless two-port network which has equivalent voltage (vn) and current (in) noise gener
ators at the input of the network. Once these noise generators are obtained, the noise
parameters can be derived as discussed in Appendix C. To obtain vn, the output short-cir
cuit current (isc) is equated for both a noisy (Figure D-l as shown) and noiseless network
(Figure D-l containing vn and in at the input and excluding ig, id, and iRb) when the input
is short-circuited. To obtain in, the same procedure is followed except the input is now
open-circuited. The results are as follows:
(D.5)
(D.6)
A transadmittance, y2j, referring output current to input voltage, is found to be
y2i = = gJi+M)-MCgd+ACdb). (D.7)
* V =0
vout w


163
circuitry to reduce the clock skew. Finally, a wireless clock transmitter (not in a PLL) has
been implemented, characterized, and demonstrated for a 5.6-mm interconnection dis
tance. These results conclusively demonstrate system operation and plausibility.
6.2 On-Chip Antennas
6.2.1 Antenna Fundamentals
To evaluate on-chip antennas, a review of basic antenna fundamentals is required,
such as radiation resistance, antenna impedance, directivity, antenna gain, and near/
far-field definitions. Radiation resistance (Rr) is equal to the total power radiated by the
antenna divided by the root-mean-squared (rms) current in the antenna. Hence, Rr repre
sents the radiation of an antenna. An antenna with a small radiation resistance indicates
that a large current would be required to radiate power. Keep in mind that antennas are
reciprocal, which means that the radiation properties apply both to an antennas transmit
ting and receiving capabilities.
The antenna impedance is the impedance seen looking into the antenna, defined
as the ratio of voltage to current at the antenna terminals. The antenna resistance is com
posed of the radiation resistance and loss resistance of the antenna, as well as substrate
components. The antenna reactance is composed of the capacitance and inductance of the
antenna lines as well as the substrate components. To maximize the power transfer
between the antenna and the circuitry, the circuit impedance should be the conjugate of the
antenna impedance.
Directivity (D) is a figure-of-merit used to quantify the directional properties of an
antenna. The directivity of an antenna is the radiation intensity (radiated power per unit
solid angle) for that antenna in a given direction divided by the radiation intensity of a


71
Figure 3-19 Input conductance to source-follower buffer driving a capacitive load versus
frequency and control voltage.
where the factors 8 and n have been defined in (2.24) and (2.25), and \ sf is defined in
(3.4). An effective quality factor has been introduced for Ldi, to account for the negative
input conductance of the source-follower buffer, as follows:
Qu =
Ql<1
(3.6)
i-(\Gsf\-za Impedance Z^ is the conductance of the drain inductor of the first stage at (tank) reso
nance. Such a substitution for the quality factor can be used whenever a tuned gain stage is
followed by a source-follower buffer. With the input impedance designed to be Rs=50 Q.
(i.e., gmj = 0.02 Q'1) and with Ldl=Ld2, the total forward transducer gain becomes
|S2I| = 2 \A
v|co = co.
ISmlj-.eff 2,2 i i
= 2 R QLdlQLd2CdoLd\ |AV,5/|
s
S(co0)
nW
(3.7)
It can be seen that the denominator of (3.6) can become zero, indicating that infi
nite gain is obtained. Thus, a finite output signal can be obtained when no input is present
to the amplifier, meaning that the circuit will oscillate. Obviously, this condition is to be


130
5.2.2 Antenna Specifications
Probably the most critical component for the wireless interconnect system is the
antenna. As the loss through the antennas increases, the transmission power and/or the
receiver gain have to be increased. Also, any impedance mismatch between the antenna
and the circuit (transmitter or receiver) causes reflection or mismatch loss. Finally, the sig
nals will be distributed across a digital chip, which has multiple metal interference and
package structures between the antennas. This will further degrade the antenna gain. Due
to its importance to the system, the antenna gain and impedance should be accurately pre
dicted for a given antenna structure, distance, and interference pattern. Currently, these
predictions have been based on experimental results, however simulation capability is
being developed and verified.
Much greater detail on antenna fundamentals and measured antenna characteristics
can be found in [KimOOa, KimOOb], as well as in Chapter 6. However, a target antenna
gain is needed for subsequent analyses. The 1999 ITRS [SIA99] projects the chip size for
a high-performance microprocessor at the 50-nm technology node to be 817 mm (refer to
Table 1-1). For a square die with the transmitter at the center and an evenly-spaced grid of
16 clock receivers, the antenna propagation distances will be either 0.5, 1.3, or 1.5 cm.
Based on the experimental work presented in [KimOOb], antenna-to-antenna transmission
gains of -50 dB and -60 dB are assumed for 1- and 2-cm separation distances at 15 GHz1,
without interference structures. This shows a 10-dB decrease in gain per octave of dis
tance. Using these results, the gains at 0.5, 1.3, and 1.5 cm are extrapolated to be -40, -53,
1. This gain is for a 2-mm long zigzag dipole antenna, 3 pm above a 20-fi-cm silicon
substrate, with a 30 bend-angle and a 30-gm trace width.


131
and -56 dB, respectively. Metal interference structures have been shown experimentally
(and in Chapter 6) to degrade the gain by 5 to 10 dB [YooOO, KimOOb], Applying simple
design rules can limit the gain degradation to 7 dB. Thus, antenna-to-antenna gains of-47
and -63 dB are specified for 0.5- and 1.5-cm distances with interference structures at 15
GHz. Of course, larger gain at a given frequency will improve the system performance by
increasing the signal-to-noise ratio or allowing less power to be transmitted by the power
amplifier. Finally, if the loss for a certain distance is too large, preventing system opera
tion, the operating frequency should be increased, since antenna gain increases with fre
quency. A more advanced technology could be required, though, for the circuits to operate
at the higher frequency.
5.2.3 Receiver Gain
A general rule for the receiver is that the amount of gain in the clock receiver has
to approximately equal the amount of loss through the antenna-pair. Referring to Figure
5-1, the signal at the receiver output is approximately -1.9 dBV, while the signal at the
transmitting antenna is ~ 0 dBV. Thus, the gain from the transmitting antenna to the output
of the clock receiver is on the order of 0 dB. This is reasonable, since the signals at the
input and output of the system should have a peak-to-peak amplitude of ~Vdd. Also, this is
similar to conventional interconnect systems, where the input and output signals of the
interconnect are approximately the same level. The difference is that the wireless intercon
nect system has considerable loss due to the antennas. Thus, for the nominal antenna gain
just specified, the receiver gain should be ~ 63 dB at 15 GHz.
The receiver gain is composed of the LNA and source-follower gain cascaded with
the divider conversion gain. Here the utility of using divider conversion gain rather than


140
will be small, however, since the offset voltage due to circuit mismatches should be less
than 100 mV. In the latched mode, Vmin is on the order of 0.5-1 V; thus, 8 should be less
than 0.3, meaning that amplitude mismatch will not cause more than 0.5% of clock skew.
As a result, automatic gain control (AGC) should not be required for SCL dividers operat
ing in the latched mode; however, as will be shown, AGC is required for dividers operat
ing in the injection-locked mode.
5.4.2 Injection-Locked Mode
When the divider is operating as a superharmonic injection-locked oscillator
(ILO), a phase difference exists between the locking signal and oscillator output signal.
This phase difference is a function of the input signal levels, as well as the difference
between the locking frequency and the natural frequency of the ILO. Appendix E reviews
the basic theory behind ILOs. In particular, the steady-state phase relationship between the
input and output signals of the ILO was shown to be [Pac65]:
f \
QSs = sin 1
Acd0A
W,
+ sm
-l
Am0A
a/1 + (Ac0
A)2/
(5.18)
o
in equation (E.8). Here, Aco0 is the difference between the natural and locking frequencies
(co0-coL), A is the derivative of the phase response of the ILO (i.e., 2:1 divider) with
respect to frequency (A = ^) from which an equivalent quality factor can be defined,
and VL and V0 are the locking and oscillator output voltage swings, respectively. A phase
difference exists because an oscillator generates 360 of phase-shift around the loop at its
natural oscillation frequency. When an input signal is injected into the loop, a phase differ
ence results at that injection point. Therefore, the loop adjusts the oscillation frequency to
counteract this phase difference. This is discussed in more detail in Appendix E.


118
output phase-difference was quantized does verify partial operation of the circuit. It is
expected (and simulations indicate) that this circuit will operate correctly when the new
DFFs given in Figure 4-24 are used.
4.7 A 0.1 -|lm CMOS TDP1- Divider on SOI and Bulk Substrates
4.7.1 Circuit Implementation
Dividers utilizing SCL do not require or generate full-swing input and output sig
nals. As a result, their operating speed can be higher than full-swing dividers, since it takes
longer to charge a given capacitive load to a larger voltage swing using a given current
CV
(i.e., x = ). A benefit of full-swing dividers is that they generally can be implemented
with fewer transistors, and, hence, the layouts are more compact. However, in addition to
their lower speed, full-swing dividers can have a reduced input sensitivity (or conversion
gain) since they require larger input signals.
Full-swing dividers are commonly implemented using true-single phase clocked
(TSPC) logic [Yua89] or clocked CMOS (C2MOS) logic [Wes92]. An alternative type of
high-speed logic is dual-phase dynamic pseudo-NMOS ([DP]2) logic [Biy96, Yan99a],
The C2MOS and [DP]2 inverters are shown in Figures 4-28 (a) and (b), while a [DP]2 DFF
is shown in Figure 4-28(c). The [DP]2 inverter is evolved from a C2MOS inverter by elim
inating the PMOS transistor used for logic evaluation. This reduces the input and load
capacitance of the inverter, increasing the speed for a given power dissipation. Also, the
removal of the PMOS transistor decreases the series resistance for charging the output
node, further increasing the speed. Since all three transistors of a [DP]2 inverter are on
during evaluation, the pull-down action of the NMOS transistors has to be made stronger


201
global clock signal. From this test, the clock skew and jitter can be evaluated. Also, this
will demonstrate a larger and more realistic portion of the system, in that time-of-flight
delays, amplitude mismatch, and process variation will all be present.
The receivers used for the double-receiver test are Rx2 and Rx4, located in the
upper and lower right-hand comers of the 0.18-pm UMC testchip, as shown in Figure
6-10. These receivers contain programmable dividers, allowing the startup phase of the
circuit to be adjusted in 45 increments, as presented in section 4.6. This programming is
used to tune out any time-invariant delay mismatch between the receivers, limiting the
clock skew to under 6.25% for an 8:1 divider. The transmitting antennas used for the dou
ble-receiver test are Z6b and Z6c, positioned on the left-hand side of the chip. Antenna Z6b
is located 5.95 mm from both Rx2 and Rx4, while Z6c is located 6.8 mm from Rx2 and 5.6
mm from Rx4.
6.7.1 Measurement Setup
The double-receiver measurement is illustrated in Figure 6-27. The input global
clock signal (GCLK) is generated off-chip and externally amplified. Before the amplifier,
GCLK is split, allowing it to be measured with an oscilloscope. The GCLK signal is then
transformed into a balanced signal and injected into the transmitting antenna, Z6b or Z6c.
The receivers, Rx2 and Rx4, are probed and their outputs are examined using an HP
54120B digitizing oscilloscope mainframe and HP 54121A four-channel test set. Since
each receiver produces out-of-phase local clock signals, the in-phase signals from Rx2 and
Rx4 were measured, while the out-of-phase signal from Rx4 was used as the trigger for
the oscilloscope. The out-of-phase signal from Rx2 was terminated with a 50-Q load.


44
One way to minimize the substrate resistance of the inductor is to use patterned
ground shields [Yue98, Che98]. Referring to Figure 3-1(b), ground shields provide a
low-resistance path to ground for capacitively coupled currents through termination of the
E-field. This reduces Rsub to approximately 5-10 f, and virtually eliminates Csub. The
shield is patterned to prevent eddy currents from flowing in the shield by not terminating
the B-field, which would decrease L and increase rs. Instead, the eddy currents are forced
into the substrate, with the magnitude of these currents being inversely proportional to the
substrate resistivity [FloOOb]. In the commonly used substrate resistivity range, the sub
strate loss increases with decreasing resistivity. This effect shows up as an increase in the
series resistance (rs) in the inductor model. This effect is reduced for smaller-area induc
tors, since the magnetic field does not penetrate as deeply into the substrate, thereby
reducing the eddy currents. The net result is an increase in inductor quality factor at the
cost of increased parasitic capacitance or decreased self-resonant frequency.
3.2.2 Capacitor implementation
Capacitors Cj and C2 in Figure 2-2, as well as the bypass capacitors for bias volt
ages are implemented using accumulation-mode MOS capacitors [Hun98]. These capaci
tors achieve high Q-factors, and are compatible with standard digital CMOS technologies.
The output dc voltage across Cj is 0 V, consistent with driving an image-rejection filter. A
concern with using this capacitor structure is the voltage-dependent capacitance and its
effect on PldB and IP3 performance. However, measurements show that PjdB and IP3 vary
by less than 0.5 dB for the output dc voltage ranging from ground to Vdd.


144
Figure 5-6
= VnCOSY|/
Phasor diagram for additive noise, showing how phase noise is generated.
the divider. Since a frequency divider basically samples the input phase during a zero
crossing of the signal, the phase noise spectrums will be replicated and aliased at multiples
of the sampling frequency [Ega90]. Since there are two zero-crossings per cycle, the sam
pling rate is twice the output frequency of the divider. The measured results in [Ega90]
show that the dominant sources are the output additive noise and the sampled input addi
tive noise. Therefore, these sources will be used to derive the output phase noise as a func
tion of the SNR at the divider input.
5.5.2 Conversion From Additive Noise to Phase Noise
Additive noise, as its name implies, adds to the signals present in the system. This
results in both an amplitude and phase fluctuation in the signal. Although this seems
straightforward enough, a quick derivation for this case will be presented. In addition, the
difference between single-sideband phase-noise density, L, and phase power-spectral-den
sity (PSD), S^, will be explained to clear up any misconceptions.
Figure 5-6 shows a phasor diagram representing an example where noise is added
to a sinusoidal signal with a root-mean-squared (RMS) voltage amplitude, A. The additive
noise is represented as vneJ't,) where vn is an RMS voltage. This noise can be divided into
orthogonal components, where the component parallel to A is defined as amplitude noise,
and the component perpendicular to A is defined as phase noise. The phase shift resulting
from the additive noise is as follows:


26
The voltage gain can then be written as
I*
vl CD = £0_
r. Qin 8ml (&Ld ^d)
8(J
(co0)
(2.23)
where
5 (co) =
8 m2
8 m2 + jCd\
(2.24)
n(co) =
1 +j(d(Cl + C2)Rl
(2.25)
ycoC2/?
Here, 8(co) is a low-pass filter response, where the low-pass filter is composed of Cdj and
l/gm2; (co) is a complex capacitive transformer ratio. At sufficiently high frequencies,
ci
n(co) reduces to 1+ ; however, this approximation must be justified and is rarely appro-
C2
priate for typical values of C2. By inspection, the ratio 15/nl is always less than 1. Further
simplification can be applied to (2.23) to obtain a more useful equation. Substituting (2.8)
and noting that cop = gmi/Cgsl, it can be shown that
M = =
03T Q-Ld Ld
Rs + CO TLS
S(0)
(2.26)
2.5.2 Methods to Maximize Gain
To maximize the gain (S2j), four steps can be taken. First, examining the effect of
the drain inductor on S21 requires examining the term QLdLd/n(co0), where n(co0) depends
on Qlj and Ld via Cj and C2. However, since n(C0o) is approximately a capacitance ratio,
its dependence on and Ld is weak. Therefore, S2j increases with increasing Qj^j and
increasing Ld. This can be seen in Figure 2-5, where S21 is plotted (a) versus Ld for
constant and (b) versus for constant Ld. For each data point, the output is
matched to 50 £2 Therefore, the LNA should be designed such that Ld and its quality


89
repeated for every desired input frequency. As can be seen from Figure 4-4(a), the
required input swing dips at the self-oscillation frequency (12.2 GHz), and then rapidly
increases for higher frequencies The locking frequency range for a given input swing is
equal to the distance between the left and right portions of the curve. This simulation veri
fies that larger input voltage swings result in larger locking ranges.
4.3.4 Implications of Injection Locking for Clock Distribution
The frequency divider will operate in the injection-locked mode for small-level
input signals, and in the latched mode for large input signals. Therefore, to minimize the
minimum detectable signal (MDS) of the receiver, the divider should be injection-locked,
with its input-referred self-oscillation frequency near the operating frequency of the
receiver. This provides an alternative interpretation for the wireless clock distribution sys
tem. Using this interpretation, each receiver is initially a free-running oscillator. The trans
mitted global clock signal is then used to injection-lock each oscillator and therefore
synchronize the system. Noise and interference can corrupt the global reference signal as
well as the phase of the oscillators themselves, resulting in phase noise or clock jitter.
4.4 A 10-GHz. 0.25-fim CMOS SCL Divider
4.4.1 Circuit implementation
A 10-GHz 128:1 source-coupled frequency divider has been implemented using a
standard 0.25-|im CMOS technology. A block diagram of the divider is shown in Figure
2. Note that for wireless interconnects, the region corresponding to input frequencies above
fiSO should be avoided since the receivers MDS would be severely degraded.


146
L(f) is 3 dB lower than S^(f). Units of dBr/Hz (decibels above 1 rad2/Hz) are typically used
for Sfy(f) and dBc/Hz (decibels above the power of the carrier) for Lif).
5.5.3 Output Phase Noise of Clock Receiver
Since the frequency divider in the clock receiver is preceded by an LNA (with
source-follower buffers), the additive noise spectrum at the input of the divider is
divadd = kT F(f) G{f) Af, (5.26)
where F(f) and G(f) are the noise factor and power gain of the LNA. Integrating this PSD
over all frequencies yields the total noise power, NdivT, as follows:
NdivT = £[- F(f) G(f)]df. (5.27)
An equivalent noise bandwidth (in Hertz) can be defined as
BWn =
kT-fQF(f)G(f)df
kT F(f 0)G(fD)
(5.28)
where f0 is the resonant frequency of the LNA. The additive PSD in (5.26) is converted to
a phase PSD using (5.25) to yield the phase noise at the input of the divider, as follows:
kT F(f) G{f)
2 P
sig
(5.29)
The phase noise at the output of a divider operating in the latched mode with divi
sion ratio N due to the input additive noise is [Ega90]
sout(f) = kT F(f) G(f)
* 2 N2 P
Slg
(5.30)
When the divider is an ILFD, the noise transfer function from the input to the output is
low-pass filtered, like a first-order phase-locked loop, with an equivalent loop bandwidth


BIOGRAPHICAL SKETCH
Brian A. Floyd was bom in West Palm Beach, Florida, on August 18, 1974. He
received the B.S. with highest honors and the M.Eng. degrees in electrical and computer
engineering from the University of Florida, Gainesville, in 1996 and 1998, respectively.
During the summers of 1994, 1995, and 1996, he worked for Motorola Inc., Boynton
Beach, Florida, in the areas of RF product development and IC design. In 1997, he joined
the Silicon Microwave Integrated Circuits and Systems (SiMICS) Research Group, at the
University of Florida Department of Electrical and Computer Engineering. From
1996-1997, he held the Robert C. Pittman Graduate Fellowship, while from 1998-2001 he
held the Intersil and Semiconductor Research Corporation Graduate Fellowship. His
research interests include RF integrated circuit design in both CMOS and BiCMOS tech
nologies, wireless communications, and high-speed clock and data distribution.
278


6
20-GHz signal using integrated antennas, and then amplify and synchronously divide it
down to a ~2.5-GHz local clock frequency. These local clock signals are then buffered and
distributed to adjacent circuitry. Figure l-2(b) shows an illustration of a multi-chip or
inter-chip wireless clock distribution system. Here the transmitter is located off-chip, uti
lizing an external antenna, potentially with a reflector. Integrated circuits located on either
a board or a multi-chip module each have integrated receivers which detect the transmitted
global clock signal and generate synchronized local clock signals.
1.2.3 Clock Receiver and Transmitter Architectures
Figure l-3(a) shows a simplified block diagram for an integrated clock transmitter.
The ~20-GHz signal is generated using a voltage-controlled oscillator (VCO). The signal
Transmitting
Antenna
(~20 GHz)
Local Clock
Output
(~2.5 GHz)
Figure 1-3
(b)
Block diagrams of (a) integrated clock transmitter and (b) integrated clock
receiver.


104
clock receiver detects GCLK with INI = 0 and CNT = 0, and generates a local clock
(LCLK). These LCLKs have random and systematic clock skews. Third, the skew is mea
sured against the receivers neighbors with phase detectors and converted to a digital word
with a 3-bit analog-to-digital converter (3 bits for 8 states). Note that a method is needed
to prevent mode-locking [Pra95], where systematic offsets can occur if the phase offsets
from multiple receivers are simply added (i.e., offset from one receiver is +90 and offset
from another receiver is -90 resulting in a net of zero phase offset, however all of the
receivers would be out of phase). To prevent mode-locking a non-monotonic phase detec
tor response is required [Pra95]. Fourth, GCLK is turned off, and each receiver is initial
ized to the appropriate state, with INI = 1 and CNT = 3-bit word from previous step. Fifth,
either all of the clock receivers report back to the transmitter when they have been ini
tialized, or the system waits a certain amount of time for initialization to occur. Sixth, fol
lowing training, GCLK is released and the receivers should begin detecting the signal and
generating the synchronized LCLKs. However a method is still required to release the ini
tialization signal once GCLK is present at the input to the frequency divider. This is com
plicated by the fact that the frequency divider self-oscillates without the input signals
present. Approaches to synchronously release the dividers from the initialized states will
be the subject of future work.
4.6.3 Circuitry for Programmable SCL Divider
The dividers used up to this point have been asynchronous, consisting of cascaded
2:1 dividers. Therefore, an initialization methodology for a 2:1 divider is required before a
2n:1 divider can be initialized. Figure 4-17(a) shows a block diagram of a 2:1 divider


199
Figure 6-26 (a) Measurement setup used to characterize wireless clock receiver.
(b) Input and output wave-forms for clock receiver, demonstrating operation of
receiver for 5.6-mm separation at 15 GHz.
external amplifier. The reduced voltage swing at the output results from driving a 50-2
load. The total power consumption for the receiver is 40 mW. Inferring from the measured
zigzag-zigzag antenna pair transmission gain (~ -53 dB) and accounting for mismatch loss
in the system, the power delivered to the input of the receiver is -34.3 dBm. As expected,
for the 3.2-mm separation, the input power level to the transmitting antenna can be 10 dB
lower, since the antenna transmission gain is 10 dB higher.
6.6.6 Simultaneous Transmitter and Receiver Operation
By comparing the results from both the transmitter and receiver circuits with inte
grated antennas, it can be seen that simultaneous operation of these two circuits (i.e., inte
grated transmitter communicating across the chip to integrated receiver) is currently not
possible. The input power level for the receiver is -34.3 dBm, while the power received
from the transmitter is -60 and -69 dBm for half-chip and full-chip transmission,


128
5.2 Power Transfer from Clock Source to Local Clock System
An illustration of the wireless interconnect is shown in Figure 5-1. The wireless
interconnect system has to generate a global clock signal (GCLK), broadcast this signal
across a distance containing metal interference structures, receive GCLK with the receiv
ing antenna, amplify and divide this down to the local clock signal (LCLK), and then dis
tribute this signal to adjacent circuits. The signal levels in Figure 5-1 are represented as
voltage amplitudes in dBV (decibels with respect to 1 V, which is 201og10{amplitude}).
Voltage should be used at points in the system where the load is capacitive. At points in
the system where the impedance is fixed, having a real component, the signal can be
expressed in power (i.e., dBm) rather than voltage. The conversion from dBm to dBV
depends on the resistance at that point (= R), and is given as follows:
(5.1)
For a 50-Q. (single-ended) impedance, the conversion is dBV = dBm 10 dB.
The total gain of the interconnect is the cascaded gains of each component, includ
ing any mismatch loss. The output LCLK voltage swing is then the input voltage swing
times the total system voltage gain, limited by the supply voltage. Thus, the main things
which effect the LCLK swing are as follows: output power of the transmitter, antenna-to-
antenna gain including interference effects, total receiver gain, and mismatch loss at the
antennas/circuit interface. Each component will be addressed in this section, and nominal
values will be recommended.


263
E.2.3 Locking Range and Locking Signal Level
The ILO is locked when 0 is constant, which is when ^ = 0. Therefore, the lock-
dt
ing range, Aco0, can be obtained from (E.4), as follows:
sin0
,A V0J\ + (Vl/V0)cosQ'
For small input locking signals (VLVo), this locking range reduces to
(E.5)
= (i'^)sin0 (E'6)
As can be seen, the locking range increases as VL increases. Also, as the slope of the phase
shift versus frequency decreases, the locking range increases. This corresponds to a lower
effective Q of the circuit. The required input locking signal level can be obtained from
(E.4), by knowing that Isin 01 is always less than 1, resulting in the following:
^ > AcoA
v o
(E.7)
E.2.4 Steady-State Phase Error
The steady-state phase error between the input locking signal and the ILO output
signal is given as follows [Pac65]:
( \
Qss = sin
-l
Aco0A
-Jl + (AaoA)2
VV,
+ sin
-l
Aco0A
a/i + (A oAY
(E.8)
o
At the natural frequency of the ILO, 0SS is zero. This steady-state phase error will result in
a skew for the ILO when it is used as a frequency divider in the clock receiver. This skew
will depend on the signal level of the injected signal, VL. For a superharmonic ILO used as


TABLE OF CONTENTS
page
ACKNOWLEDGMENTS ii
ABSTRACT viii
CHAPTERS
1 INTRODUCTION 1
1.1 Global Interconnect Challenges 1
1.2 Proposed Interconnect System 4
1.2.1 Potential Solutions 4
1.2.2 Description of Wireless Interconnect System 5
1.2.3 Clock Receiver and Transmitter Architectures 6
1.2.4 Potential Benefits 7
1.2.5 Areas of Research 8
1.3 Overview of Dissertation 9
2 CMOS LOW NOISE AMPLIFIERS 12
2.1 Overview 12
2.1.1 Scope of LNA Research 12
2.1.2 Performance Metrics of LNAs 14
2.2 Possible LNA Topologies 15
2.2.1 Common-Gate CMOS LNA 15
2.2.2 Source-Degenerated CMOS LNA 17
2.3 Input Matching for Source-Degenerated LNA 18
2.3.1 Input Impedance 18
2.3.2 Input Match to Withstand Component Variations 19
2.4 Output Matching for Source-Degenerated LNA 22
2.5 Gain of Source-Degenerated LNA 25
2.5.1 Gain Driving Resistive Load 25
2.5.2 Methods to Maximize Gain 26
2.5.3 Gain Driving Capacitive Load 28
2.6 Noise Parameters 29
2.6.1 Noise Sources 29
2.6.2 Noise Parameters of Single Transistor 30
2.6.3 Optimum Qgs for Minimum Noise 33


43
Figure 3-1 (a) General two-port lumped-element model for spiral inductor, (b) Illustration of
a spiral inductor above a patterned-ground shield. The shield terminates most of
E-field while allowing B-field to pass through, increasing Q.
A two-port lumped-element model for a general spiral inductor is shown in Figure
3-1(a). Using this model, the quality factor (Q) for an inductor with one port shorted to
ground can be written as follows [Lon97], where substrate capacitance (Csub) and
port-to-port capacitance (Cs) have been neglected:
Q
-(?
i+2c>L)
1 +
(3.1)
This accounts for loss in both the series resistance of the spiral and in the conductive sili
con substrate. Note that the interaction of the substrate with the magnetic field is not cap
tured with this equation (i.e., eddy-current effects). To maximize Q, the series resistance
(rs) should be minimized, oxide capacitance (Cox) should be minimized, and the substrate
resistance (Rsu/,) should either be very small or very large. This trend for Rsub is the same
as that seen in the noise-figure analysis. In fact, this is a general trend which can be
applied to any RF circuit or component [FloOOb].


152
receiver. Note that if the input interference is very large, then the equivalent antenna tem
perature can be much larger than 298 K [Smi98]. Equivalently, the noise figure can be
considered that of the antenna and receiver, where a noise figure for the antenna is defined
by the total noise (and interference) at the input of the receiver, divided by the available
thermal noise (kT). Note that for very large antenna noise figures, the noise figure of the
receiver becomes unimportant, meaning that the noise added by the receiver is a small
fraction of the noise already present at the input.
Converting (5.37) to dB and dBm yields the following for T = 353 K (80 C, which
is assumed to be the temperature of the microprocessor):
P*\dBm
dBm
(5.38)
-173 ^£ + NF(f0) + G(f0)\dB+moglQ(BWN) + SNRmin\dB.
Subtracting the LNA gain (G(f0)\B) from the signal power at the divider input and substi
tuting SNRmin = 12 dB, yields the input sensitivity for the clock receiver, as follows:
= 161
dBm
+ NF(fa)+10log. JBWn)
(5.39)
I dBm Hz
Referring back to section 5.2, the transmitter power level is specified to be 10 dBm, the
antenna-to-antenna gain is specified to be -63 dB at a 1.5-cm distance, and the mismatch
loss is specified to be 1 dB. Therefore, the desired sensitivity is -54 dBm (10 dBm 63 dB
- 1 dB = -54).
5.6.2 Receiver Noise Requirements
The noise performance of the receiver can now be specified by plugging in the
-54-dBm sensitivity into (5.39). This yields the noise figure of the receiver at the center
frequency and the noise bandwidth, as follows:


257
D.3.2 Case 2 Effect of GIN
The effect of GIN is now analyzed, while neglecting substrate resistance and all
other capacitances except Cgs. Under these assumptions, y2\ = gm> Cgt = Cgs, and the
input noise generators become
Vn2 =
&n
(D.26)
(D.27)
The noise parameters can then be derived, and are given in the following table, D-3.
Table D-3 Noise parameters considering GIN
Alternative Noise Parameters
Impedance-Based Noise Parameters
R2 = -f- (D.28)
(D.29)
2 = 2 (>C/b2 = YS,(^;)2f>2 (D.30)
Pn 2 = Rn2(-jCgs)bi (D.31)
The following variables have been introduced:
b\ = 1 + A|c|,
(D.33)
b2 = 1 +2A|c| +A2,
(D.34)
A = /§<
V 5y
(D.35)
Examining this set of noise parameters, it can first be seen that the inclusion of
GIN causes the optimal source reactance to slightly decrease. This means that series


209
-0.4 I * * *
16.515 16.52 16.525 16.53 16.535 16.54 16.545 16.55
Time (ns)
Figure 6-31 Illustration of the measured jitter of the wireless clock receiver. This jitter was
measured off of the 15-GHz input signal, yielding the jitter of the 1.875-GHz
output signal used as the trigger. The measured peak-to-peak jitter was 6.6 ps.
signal, which should have very small jitter or phase noise, yields the jitter of the triggered
1.875-GHz waveform. An illustration of the resultant plot with infinite persistence is
shown in Figure 6-31. This is an illustration, rather than the real data, since a screen cap
ture of the oscilloscope could not be obtained. The measured peak-to-peak jitter is 6.6 ps
at 1.875 GHz, corresponding to 1.24% of the local clock period. This is under the system
requirement of 3% peak-to-peak jitter. Note that this jitter is for a quiet system, meaning
that the supply voltages are not being modified by digital circuits. In the literature, this is
known as the quiet-supply jitter. In an operating microprocessor, the jitter will be
increased, due to larger switching noise and noisy supplies.
The signal-to-noise ratio at the input of the divider can be determined using (5.35)
and the measured jitter, and assuming that the jitter is dominated by input additive noise.
Assuming that the RMS jitter is one-sixth of the peak-to-peak jitter (or 1.1 ps = 0.2%),
then the SNR at the divider input is calculated to be 19.7 dB. This agrees remarkably well



PAGE 1

$ &026 :,5(/(66 ,17(5&211(&7 6<67(0 )25 08/7,*,*$+(57= &/2&. ',675,%87,21 %\ %5,$1 $ )/2<' $ ',66(57$7,21 35(6(17(' 72 7+( *5$'8$7( 6&+22/ 2) 7+( 81,9(56,7< 2) )/25,'$ ,1 3$57,$/ )8/),//0(17 2) 7+( 5(48,5(0(176 )25 7+( '(*5(( 2) '2&725 2) 3+,/2623+< 81,9(56,7< 2) )/25,'$

PAGE 2

$&.12:/('*0(176 ZRXOG OLNH WR EHJLQ E\ WKDQNLQJ P\ DGYLVRU 3URIHVVRU .HQQHWK 2 ZKRVH FRQVWDQW HQFRXUDJHPHQW JXLGDQFH DQG IULHQGVKLS KDYH EHHQ LQYDOXDEOH DOVR ZRXOG OLNH WR WKDQN 3URIHVVRUV )R[ )RVVXP DQG )UDQN IRU WKHLU LQWHUHVW LQ WKLV ZRUN DQG WKHLU WLPH FRPPLWPHQW LQ VHUYLQJ RQ P\ FRPPLWWHH 0XFK DSSUHFLDWLRQ JRHV WR WKH 6HPLFRQGXFWRU 5HVHDUFK &RUSRUDWLRQ 65&f IRU IXQGLQJ WKLV ZRUN DQG WR WKH 65& DQG ,QWHUVLO IRU VSRQVRULQJ P\ IHOORZVKLS 7KDQNV DOVR JR WR 65& PHQWRU 'U 6FRWW /LVW RI ,QWHO DQG WR IHOORZVKLS OLDLVRQ 'U .HQ 3RUWV RI ,QWHUVLO )XUWKHUPRUH ZRXOG OLNH WR WKDQN WKH VSRQVRUV RI WKH &RSSHU 'HVLJQ &KDOOHQJHf§WKH 65& 1RYHOOXV 6SHHG)DP,3(& DQG 80&f§ DV ZHOO DV ,%0 DQG WKH 1DY\ IRU SURYLGLQJ DFFHVV WR DGYDQFHG &026 WHFKQRORJLHV ZRXOG OLNH WR WKDQN P\ FROOHDJXHV &KLK0LQJ +XQJ DQG .LKRQJ .LP ZKRVH EHQHILFLDO GLVFXVVLRQV DGYLFH DQG IULHQGVKLS KDYH FRQWULEXWHG LPPHQVHO\ WR WKLV ZRUN $OVR ZRXOG OLNH WR UHFRJQL]H IRUPHU 65& JURXS PHPEHUV .LP 0HKWD +
PAGE 3

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

PAGE 4

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

PAGE 5

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

PAGE 6

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

PAGE 7

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

PAGE 8

$EVWUDFW RI 'LVVHUWDWLRQ 3UHVHQWHG WR WKH *UDGXDWH 6FKRRO RI WKH 8QLYHUVLW\ RI )ORULGD LQ 3DUWLDO )XOILOOPHQW RI WKH 5HTXLUHPHQWV IRU WKH 'HJUHH RI 'RFWRU RI 3KLORVRSK\ $ &026 :,5(/(66 ,17(5&211(&7 6<67(0 )25 08/7,*,*$+(57= &/2&. ',675,%87,21 %\ %ULDQ $ )OR\G 0D\ &KDLU .HQQHWK 2 0DMRU 'HSDUWPHQW (OHFWULFDO DQG &RPSXWHU (QJLQHHULQJ $V WKH FORFN IUHTXHQF\ DQG FKLS VL]H RI KLJKSHUIRUPDQFH PLFURSURFHVVRUV LQFUHDVH LW EHFRPHV LQFUHDVLQJO\ GLIILFXOW WR GLVWULEXWH VLJQDOV DFURVV WKH FKLS GXH WR LQFUHDVLQJ SURSDJDWLRQ GHOD\V DQG GHFUHDVLQJ DOORZDEOH FORFN VNHZ 7KLV GLVVHUWDWLRQ SUHn VHQWV WKH GHVLJQ LPSOHPHQWDWLRQ DQG IHDVLELOLW\ RI D ZLUHOHVV LQWHUFRQQHFW V\VWHP IRU FORFN GLVWULEXWLRQ 7KH V\VWHP FRQVLVWV RI WUDQVPLWWHUV DQG UHFHLYHUV ZLWK LQWHJUDWHG DQWHQn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

PAGE 9

XVLQJ DQG _LP WHFKQRORJLHV UHVSHFWLYHO\ $ GHVLJQ PHWKRGRORJ\ EDVHG RQ LQMHFWLRQ ORFNLQJ LV GHYHORSHG IRU &026 IUHTXHQF\ GLYLGHUV DQG D SURJUDPPDEOH GLYLGHU ZKLFK OLPLWV FORFN VNHZ LV SUHVHQWHG 'LYLGHUV RSHUDWLQJ XS WR DQG *+] DUH GHPRQVWUDWHG LPSOHPHQWHG LQ DQG SP WHFKQRORJLHV UHVSHFWLYHO\ 5HVXOWV IRU WKH RYHUDOO ZLUHOHVV LQWHUFRQQHFW V\VWHP DUH WKHQ SUHVHQWHG 6\VWHP UHTXLUHPHQWV JDLQ PDWFKLQJ QRLVH OLQHDULW\f IRU ZLUHOHVV FORFN GLVWULEXWLRQ DUH GHULYHG LQFOXGLQJ VSHFLILFDWLRQV IRU VLJQDOWRQRLVH UDWLR YHUVXV FORFN MLWWHU DQG DPSOLWXGH PLVn PDWFK YHUVXV FORFN VNHZ :LUHOHVV LQWHUFRQQHFW V\VWHPV DUH GHPRQVWUDWHG IRU WKH ILUVW WLPH XVLQJ RQFKLS DQWHQQD SDLUV FORFN UHFHLYHUV DQG FORFN WUDQVPLWWHUV 7KH LQWHUFRQQHFWV RSHUDWH DFURVV PP DW *+] XVLQJ D SP WHFKQRORJ\ DQG DFURVV PP DW *+] XVLQJ D _LP WHFKQRORJ\ 8VLQJ WKH PP *+] LQWHUFRQQHFW D SV FORFN VNHZ DQG SV SHDN MLWWHU KDYH EHHQ PHDVXUHG DW *+] IRU WZR UHFHLYHUV VHSDUDWHG E\ a PP )LQDOO\ WKH ZLUHOHVV LQWHUFRQQHFW V\VWHP LV DQDO\]HG LQ WHUPV RI SRZHU GLVVLSDn WLRQ V\QFKURQL]DWLRQ SURFHVV YDULDWLRQ ODWHQF\ DQG DUHD 7KHVH UHVXOWV LQGLFDWH WKH IHDVLELOLW\ RI DQ LQWUDFKLS ZLUHOHVV LQWHUFRQQHFW V\VWHP XVLQJ LQWHJUDWHG DQWHQQDV ,;

PAGE 10

&+$37(5 ,1752'8&7,21 *OREDO ,QWHUFRQQHFW &KDOOHQJHV $FFRUGLQJ WR WKH ,QWHUQDWLRQDO 7HFKQRORJ\ 5RDGPDS IRU 6HPLFRQGXFWRUV ,756f >6,$@ DW WKH DQG _LP WHFKQRORJ\ JHQHUDWLRQV FKLS DUHDV IRU KLJKSHUn IRUPDQFH PLFURSURFHVVRUV DUH SURMHFWHG WR EH DSSUR[LPDWHO\ DQG PP UHVSHFn WLYHO\ 2QFKLS JOREDO FORFN IUHTXHQFLHV DUH SURMHFWHG WR EH DQG *+] ZKLOH ORFDO FORFN IUHTXHQFLHV DUH SURMHFWHG WR EH DQG *+] UHVSHFWLYHO\ 7KHVH WUHQGV IRU KLJKSHUIRUn PDQFH PLFURSURFHVVRUV DUH VKRZQ LQ 7DEOH ,Q VXFK LQWHJUDWHG FLUFXLWV ,&Vf WKH GHOD\ DVVRFLDWHG ZLWK JOREDO LQWHUFRQQHFWVf§WKRVH ZKLFK FRQQHFW IXQFWLRQDO XQLWV DFURVV WKH ,&f§KDV EHFRPH PXFK ODUJHU WKDQ WKH GHOD\ IRU D VLQJOH ORJLF JDWH KHUHLQ WHUPHG JDWHGHOD\f 7KLV LV VKRZQ LQ )LJXUH ZKLFK SORWV WKH JOREDO LQWHUFRQQHFW GHOD\ DQG JDWHGHOD\ YHUVXV PLQLPXP IHDWXUH VL]H >6,$ %RK@ $V IHDWXUH VL]H GHFUHDVHV JDWHGHOD\ GHFUHDVHV LOOXVWUDWLQJ WKH ZHOONQRZQ IDFW WKDW WUDQVLVWRU VFDOLQJ LPSURYHV GHYLFH SHUIRUPDQFH DQG FKLS GHQVLW\ VLPXOWDQHRXVO\ +RZHYHU WKH SURSDJDWLRQ GHOD\ RI D YROWDJH ZDYH LV DSSUR[LPDWHO\ HTXDO WR 5& >:HV@ ZKHUH LV WKH OHQJWK RI D ZLUH DQG 5 DQG & DUH LWV UHVLVWDQFH DQG FDSDFLWDQFHSHUXQLWOHQJWK $V WKH &026 WHFKQRORJ\ LV VFDOHG WR VPDOOHU IHDWXUH VL]HV ERWK WKH 5& WLPHFRQVWDQW >%RK 7DX@ DQG WKH FKLS DUHD Of LQFUHDVH 7KHUHIRUH WKH JOREDO LQWHUFRQQHFW GHOD\ LQFUHDVHV DQG TXLFNO\ EHJLQV WR GRPLQDWH WKH RYHUDOO V\VWHP GHOD\

PAGE 11

7DEOH 7HFKQRORJ\ 7UHQGV RI 6HPLFRQGXFWRU ,QGXVWU\ IRU 0LFURSURFHVVRUVn 6,$@ )LJXUH *OREDO LQWHUFRQQHFW GHOD\ DQG JDWHGHOD\ YHUVXV WHFKQRORJ\ JHQHUDWLRQ IRU DOXPLQXP DQG FRSSHU PHWDOOL]DWLRQ DQG FRQYHQWLRQDO DQG ORZ. GLHOHFWULFV 7R RIIVHW WKLV SUREOHP FRSSHU &Xf LQWHUFRQQHFWV DQG ORZ. GLHOHFWULFV KDYH EHHQ LQWURGXFHG GHFUHDVLQJ WKH JOREDO SURSDJDWLRQ GHOD\ E\ D UDWLR RI DSSUR[LPDWHO\ 3FX.ORY S$M.FRQYf ZKHUH S LV WKH UHVLVWLYLW\ RI FRSSHU DOXPLQXPf DQG LV WKH UHODWLYH GLHOHFWULF

PAGE 12

FRQVWDQW IRU ORZ. FRQYHQWLRQDOf GLHOHFWULFV 7KLV FRUUHVSRQGV WR D GRZQZDUG VKLIW LQ WKH JOREDO GHOD\ SORWWHG LQ )LJXUH 8QIRUWXQDWHO\ VLQFH WKH GHOD\ VFDOHV ZLWK FKLS DUHD GHVSLWH WKH DGGLWLRQ RI FRSSHU DQG ORZ. GLHOHFWULFV JOREDO LQWHUFRQQHFW GHOD\ ZLOO FRQn WLQXH WR LQFUHDVH ZLWK VXFFHHGLQJ JHQHUDWLRQV RI PLFURSURFHVVRUV &RSSHU DQG ORZ. GLHOHFWULFV RQO\ H[WHQG WKH OLIHWLPH RI FRQYHQWLRQDO LH WUDGLWLRQDO FRQGXFWRUf LQWHUFRQn QHFW V\VWHPV D IHZ WHFKQRORJ\ JHQHUDWLRQV ,Q SDUWLFXODU WKH JOREDO GHOD\ LV VWLOO PXFK ODUJHU WKDQ WKH JDWHGHOD\ IRU WKH SP WHFKQRORJ\ QRGH DQG EH\RQG 7KH JOREDO LQWHUFRQQHFW GHOD\ LV HVSHFLDOO\ GHWULPHQWDO WR JOREDO FORFN VLJQDO GLVn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n DWHG ZLWK LQWHUFRQQHFW UHVLVWDQFH $ QRQ]HUR LQWHUFRQQHFW UHVLVWDQFH FDXVHV WKH KDUPRQLFV RI WKH FORFN VLJQDO WR WUDYHO DW GLIIHUHQW YHORFLWLHV WKURXJK WKH LQWHUFRQQHFW UHVXOWLQJ LQ DQ LQFUHDVH LH VORZLQJf RI WKH ULVH DQG IDOO WLPHV RI WKH VLJQDO $V WKH LQWHUFRQQHFW OHQJWK LV LQFUHDVHG WKH ULVH DQG IDOO WLPHV LQFUHDVH DQG WKH\ FDQ XOWLPDWHO\ OLPLW WKH PD[LPXP IUHn TXHQF\ RI WKH VLJQDO >'HX@ %RWK RI WKH SUHYLRXVO\ VWDWHG SUREOHPV KDYH EHHQ VRPHZKDW FLUFXPYHQWHG LQ WKH ,756 IRU SP JHQHUDWLRQV DQG EH\RQG E\ GLVWULEXWLQJ D ORZHU IUHTXHQF\ JOREDO FORFN DQG DOORZLQJ IXQFWLRQDO XQLWV WR RSHUDWH RII RI D KLJKHU IUHTXHQF\ ORFDO FORFN +RZHYHU

PAGE 13

UHIHUULQJ WR 7DEOH WKHUH LV DQ LQFUHDVLQJ JDS EHWZHHQ WKH SURMHFWHG ORFDO DQG JOREDO FORFN VLJQDO IUHTXHQFLHV XQGHUVFRULQJ WKH VKRUWFRPLQJV RI FXUUHQW JOREDO FORFN GLVWULEXn WLRQ V\VWHPV &OHDUO\ DGYDQFHG LQWHUFRQQHFW V\VWHPV FDSDEOH RI GLVWULEXWLQJ KLJK IUHn TXHQF\ VLJQDOV ZLWK VKRUW SURSDJDWLRQ GHOD\V DQG PLQLPDO SRZHU GLVVLSDWLRQ DUH QHHGHG WR DGGUHVV WKHVH FRQFHUQV 3URSRVHG ,QWHUFRQQHFW 6\VWHP 3RWHQWLDO 6ROXWLRQV 3RWHQWLDO VROXWLRQV WR DGGUHVV WKH OLPLWDWLRQV RI FRQYHQWLRQDO JOREDO LQWHUFRQQHFW V\VWHPV FDQ FXUUHQWO\ EH FDWHJRUL]HG DV IROORZV f IXUWKHU PRGLI\LQJ WKH SURSHUWLHV RI FRQGXFWRUV VXFK DV FRROHG PHWDO >$@ RU VXSHUFRQGXFWLYH PHWDO f VKRUWHQLQJ WKH GLVn WDQFH RI JOREDO LQWHUFRQQHFWV E\ XVLQJ WKUHHGLPHQVLRQDO VWUXFWXUHV RU f VKLIWLQJ DZD\ IURP V\QFKURQRXV FRPSXWHU DUFKLWHFWXUHV WRZDUGV DV\QFKURQRXV DUFKLWHFWXUHV $ ILQDO FDWn HJRU\ RI VROXWLRQV UHTXLUHV WKLQNLQJ HYHQ PRUH fRXWVLGH RI WKH ER[f DQG HQWDLOV UHVHDUFK RI D PRUH IXQGDPHQWDO QDWXUH DV IROORZV f XVLQJ DOWHUQDWLYH PHGLXPV WR GLVWULEXWH VLJQDOV VXFK DV RSWLFDO >0LO@ RU RUJDQLF PHGLXPV +RZHYHU DOO RI WKH H[DPSOHV MXVW OLVWHG UHTXLUH VLJQLILFDQW GHYHORSPHQW DQGRU FKDQJHV WR HLWKHU WKH VHPLFRQGXFWRU PDWHULDOV PDQXIDFWXULQJ SURFHVV RU FLUFXLW GHVLJQ SURFHVV $Q DOWHUQDWLYH JOREDO LQWHUFRQQHFW V\Vn WHP RI WKH IRXUWK FDWHJRU\ LV WR GLVWULEXWH VLJQDOV DW WKH VSHHG RI OLJKW XVLQJ PLFURZDYHV DQG DQWHQQDV HPSOR\LQJ D FRQYHQWLRQDO &026 WHFKQRORJ\ 7KLV V\VWHP LV WHUPHG D ZLUHn OHVV RU UDGLRIUHTXHQF\ 5)f LQWHUFRQQHFW V\VWHP > )OR22D@

PAGE 14

Df Ef )LJXUH &RQFHSWXDO V\VWHP LOOXVWUDWLRQV RI Df LQWUDFKLS DQG Ef LQWHUFKLS ZLUHOHVV LQWHUFRQQHFW V\VWHPV IRU FORFN VLJQDO GLVWULEXWLRQ 'HVFULSWLRQ RI :LUHOHVV ,QWHUFRQQHFW 6\VWHP 7KH ZLUHOHVV LQWHUFRQQHFW V\VWHP FRQVLVWV RI LQWHJUDWHG UHFHLYHUV DQG WUDQVPLWWHUV ZLWK RQFKLS DQWHQQDV ZKLFK FRPPXQLFDWH DFURVV D VLQJOH FKLS RU EHWZHHQ PXOWLSOH FKLSV YLD HOHFWURPDJQHWLF ZDYHV :LUHOHVV LQWHUFRQQHFWV FDQ EH XVHG IRU ERWK GDWD DQG FORFN VLJn QDOV +RZHYHU IRU ZLUHOHVV GDWD D PRGXODWLRQ VFKHPH LV UHTXLUHG ZKLOH IRU D ZLUHOHVV FORFN RQO\ D VLQJOH WRQH LV UHTXLUHG 7KHUHIRUH ZLUHOHVV FORFN GLVWULEXWLRQ LV D QDWXUDO ILUVW VWHS IRU HYDOXDWLQJ WKH SRWHQWLDO RI ZLUHOHVV LQWHUFRQQHFWV LQ JHQHUDO DV ZHOO DV IRU GHYHORSn LQJ WKH NH\ FRPSRQHQWV RI D ZLUHOHVV LQWHUFRQQHFW V\VWHP )RU WKHVH UHDVRQV ZLUHOHVV FORFN GLVWULEXWLRQ LV VWXGLHG LQ WKLV ZRUN $ FRQFHSWXDO LOOXVWUDWLRQ RI D VLQJOHFKLS RU LQWUDFKLS ZLUHOHVV LQWHUFRQQHFW V\Vn WHP IRU FORFN GLVWULEXWLRQ LV VKRZQ LQ )LJXUH ODf $Q DSSUR[LPDWHO\ *+] VLJQDO LV JHQHUDWHG RQFKLS DQG DSSOLHG WR DQ LQWHJUDWHG WUDQVPLWWLQJ DQWHQQD ZKLFK LV ORFDWHG DW RQH SDUW RI WKH ,& &ORFN UHFHLYHUV GLVWULEXWHG WKURXJKRXW WKH ,& GHWHFW WKH WUDQVPLWWHG

PAGE 15

*+] VLJQDO XVLQJ LQWHJUDWHG DQWHQQDV DQG WKHQ DPSOLI\ DQG V\QFKURQRXVO\ GLYLGH LW GRZQ WR D a*+] ORFDO FORFN IUHTXHQF\ 7KHVH ORFDO FORFN VLJQDOV DUH WKHQ EXIIHUHG DQG GLVWULEXWHG WR DGMDFHQW FLUFXLWU\ )LJXUH OEf VKRZV DQ LOOXVWUDWLRQ RI D PXOWLFKLS RU LQWHUFKLS ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP +HUH WKH WUDQVPLWWHU LV ORFDWHG RIIFKLS XWLn OL]LQJ DQ H[WHUQDO DQWHQQD SRWHQWLDOO\ ZLWK D UHIOHFWRU ,QWHJUDWHG FLUFXLWV ORFDWHG RQ HLWKHU D ERDUG RU D PXOWLFKLS PRGXOH HDFK KDYH LQWHJUDWHG UHFHLYHUV ZKLFK GHWHFW WKH WUDQVPLWWHG JOREDO FORFN VLJQDO DQG JHQHUDWH V\QFKURQL]HG ORFDO FORFN VLJQDOV &ORFN 5HFHLYHU DQG 7UDQVPLWWHU $UFKLWHFWXUHV )LJXUH ODf VKRZV D VLPSOLILHG EORFN GLDJUDP IRU DQ LQWHJUDWHG FORFN WUDQVPLWWHU 7KH a*+] VLJQDO LV JHQHUDWHG XVLQJ D YROWDJHFRQWUROOHG RVFLOODWRU 9&2f 7KH VLJQDO 7UDQVPLWWLQJ $QWHQQD a *+]f /RFDO &ORFN 2XWSXW a *+]f )LJXUH Ef %ORFN GLDJUDPV RI Df LQWHJUDWHG FORFN WUDQVPLWWHU DQG Ef LQWHJUDWHG FORFN UHFHLYHU

PAGE 16

IURP WKH 9&2 LV WKHQ DPSOLILHG XVLQJ D SRZHU DPSOLILHU 3$f DQG IHG WR WKH WUDQVPLWWLQJ DQWHQQD 7KH 9&2 LV SKDVHORFNHG WR DQ H[WHUQDO UHIHUHQFH XVLQJ D SKDVHORFNHG ORRS 3//f SURYLGLQJ IUHTXHQF\ VWDELOLW\ 7KH 3// FRQVLVWV RI D SKDVHIUHTXHQF\ GHWHFWRU 3)'f D ORRS ILOWHU WKH 9&2 DQG D IUHTXHQF\ GLYLGHU )LJXUH OEf VKRZV D EORFN GLDJUDP IRU DQ LQWHJUDWHG FORFN UHFHLYHU 7KH JOREDO FORFN VLJQDO LV GHWHFWHG ZLWK D UHFHLYLQJ DQWHQQD DPSOLILHG XVLQJ D ORZ QRLVH DPSOLILHU /1$f DQG GLYLGHG GRZQ WR WKH ORFDO FORFN IUHTXHQF\ 7KHVH ORFDO FORFN VLJQDOV DUH WKHQ EXIIHUHG DQG GLVWULEXWHG WR DGMDFHQW FLUFXLWV 7KH DPSOLILHU LV WXQHG WR WKH FORFN WUDQVPLVn VLRQ IUHTXHQF\ WR UHGXFH LQWHUIHUHQFH DQG QRLVH 6LQFH WKH PLFURSURFHVVRU LV H[WUHPHO\ QRLV\ DW WKH ORFDO FORFN IUHTXHQF\ DQG LWV KDUPRQLFV WUDQVPLWWLQJ WKH JOREDO FORFN DW D IUHn TXHQF\ KLJKHU WKDQ WKH ORFDO FORFN IUHTXHQF\ SURYLGHV D OHYHO RI QRLVH LPPXQLW\ IRU WKH V\VWHP >0HK@ $OVR RSHUDWLQJ DW D KLJKHU IUHTXHQF\ GHFUHDVHV WKH UHTXLUHG DQWHQQD VL]H 7KH UHFHLYHU LV LPSOHPHQWHG LQ D IXOO\ GLIIHUHQWLDO DUFKLWHFWXUH ZKLFK UHMHFWV FRPn PRQPRGH QRLVH HJ VXEVWUDWH QRLVHf >0HK %UDY22D@ REYLDWHV WKH QHHG IRU D EDO DQFHGWRXQEDODQFHG FRQYHUVLRQ EHWZHHQ WKH DQWHQQD DQG WKH /1$ DQG SURYLGHV GXDOSKDVH FORFN VLJQDOV WR WKH IUHTXHQF\ GLYLGHU 3RWHQWLDO %HQHILWV 7KH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP ZRXOG DGGUHVV WKH LQWHUFRQQHFW QHHGV RI WKH VHPLFRQGXFWRU LQGXVWU\ LQ SURYLGLQJ KLJKIUHTXHQF\ FORFN VLJQDOV ZLWK VKRUW SURSDJDWLRQ GHOD\V 7KHVH QHHGV ZRXOG EH PHW ZKLOH SURYLGLQJ PXOWLSOH EHQHILWV )LUVW VLJQDO SURSDJDn WLRQ RFFXUV DW WKH VSHHG RI OLJKW VKRUWHQLQJ WKH JOREDO LQWHUFRQQHFW GHOD\ ZLWKRXW UHTXLUn LQJ LQWHJUDWHG RSWLFDO FRPSRQHQWV 6HFRQG WKH JOREDO LQWHUFRQQHFW ZLUHV XVHG LQ FRQYHQWLRQDO FORFN GLVWULEXWLRQ V\VWHPV DUH HOLPLQDWHG IUHHLQJ XS WKHVH PHWDO OD\HUV IRU

PAGE 17

RWKHU DSSOLFDWLRQV 7KLUG UHIHUULQJ WR )LJXUH OEf WKH LQWHUFKLS FORFN GLVWULEXWLRQ V\Vn WHP FDQ SURYLGH JOREDO FORFN VLJQDOV ZLWK D VPDOO VNHZ WR DQ DUHD PXFK JUHDWHU WKDQ WKH SURMHFWHG ,& VL]H 7KLV LV DQ DGGLWLRQDO EHQHILW SRVVLEO\ DOORZLQJ V\QFKURQL]DWLRQ RI DQ HQWLUH 3& ERDUG RU D PXOWLFKLS PRGXOH 0&0f )RXUWK LQ WKH ZLUHOHVV V\VWHP GLVSHUVLYH HIIHFWV DUH PLQLPL]HG VLQFH D PRQRWRQH JOREDO FORFN VLJQDO LV WUDQVPLWWHG )LIWK DQRWKHU EHQHILW LV D PRUH XQLIRUPO\ GLVWULEXWHG SRZHU ORDG HTXDOL]LQJ WHPSHUDWXUH JUDGLHQWV DFURVV WKH FKLS 6L[WK E\ DGMXVWLQJ WKH GLYLVLRQ UDWLR LQ WKH UHFHLYHU KLJKHU IUHTXHQF\ ORFDO FORFN VLJQDOV >6,$@ FDQ EH REWDLQHG ZKLOH PDLQWDLQLQJ V\QFKURQL]DWLRQ ZLWK D ORZHU IUHn TXHQF\ V\VWHP FORFN 6HYHQWK DQ LQWDQJLEOH EHQHILW RI ZLUHOHVV LQWHUFRQQHFW V\VWHPV LV WKH HIIHFW WKH\ FRXOG KDYH RQ PLFURSURFHVVRU RU V\VWHP LPSOHPHQWDWLRQV SRWHQWLDOO\ DOORZLQJ SDUDGLJP VKLIWV VXFK DV GUDVWLFDOO\ LQFUHDVHG FKLS VL]H )LQDOO\ FRPSDUHG WR RWKHU SRWHQn WLDO EUHDNWKURXJK LQWHUFRQQHFW WHFKQLTXHV VXFK DV RSWLFDO VXSHUFRQGXFWLYH RU RUJDQLF D ZLUHOHVV DSSURDFK EDVHG RQ VLOLFRQ VHHPV WR EH D SRWHQWLDO VROXWLRQ ZKLFK LV FRPSDWLEOH ZLWK WKH WHFKQRORJ\ WUHQGV RI WKH VHPLFRQGXFWRU LQGXVWU\ $UHDV RI 5HVHDUFK 7KH PDLQ DUHDV RI UHVHDUFK IRU ZLUHOHVV FORFN GLVWULEXWLRQ DUH DV IROORZV LQWHJUDWn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

PAGE 18

GLVVHUWDWLRQV >0HK %UDY22D .LP22D@ DQG DUH VWLOO EHLQJ DFWLYHO\ SXUVXHG ZKLOH WKH ODVW WZR LWHPV DUH WKH VXEMHFW RI WKLV GLVVHUWDWLRQ 2YHUYLHZ RI 'LVVHUWDWLRQ 7KLV GLVVHUWDWLRQ IRFXVHV RQ WKH LPSOHPHQWDWLRQ RI DQ LQWUDFKLS ZLUHOHVV FORFN GLVn WULEXWLRQ V\VWHP DQG HYDOXDWLRQ RI LWV IHDVLELOLW\ VHUYLQJ DV D QDWXUDO ILUVW VWHS IRU HYDOXDWn LQJ WKH SRWHQWLDO RI ERWK LQWUD DQG LQWHUFKLS ZLUHOHVV LQWHUFRQQHFWV 7KLV ZRUN ZLOO HPSKDVL]H WKH GHVLJQ DQG LPSOHPHQWDWLRQ RI 5) &026 UHFHLYHU FLUFXLWU\ DQG ZLOO HYDOXn DWH WKH V\VWHP IHDVLELOLW\ E\ LPSOHPHQWLQJ ERWK VLQJOHOLQN RQH WUDQVPLWWHURQH UHFHLYHUf DQG PXOWLOLQN RQH WUDQVPLWWHUPXOWLSOH UHFHLYHUVf ZLUHOHVV LQWHUFRQQHFWV 7KH ORZ QRLVH DPSOLILHU /1$f LV WKH ILUVW FRPSRQHQW LQ WKH FORFN UHFHLYHU DQG WKURXJK LWV ORZ QRLVH ILJXUH DQG PRGHUDWH JDLQ LW DSSUR[LPDWHO\ VHWV WKH VLJQDOWRQRLVH UDWLR RI WKH HQWLUH UHFHLYHU *HQHUDOO\ &026 /1$V KDYH KDG LQIHULRU SHUIRUPDQFH FRPn SDUHG WR VLOLFRQ ELSRODU RU JDOOLXPDUVHQLGH *D$Vf /1$V $OVR WKHUH DUH YHU\ IHZ H[DPn SOHV RI &026 /1$V RSHUDWLQJ DERYH *+] &KDSWHU SUHVHQWV D QHZ GHVLJQ PHWKRGRORJ\ IRU VRXUFHGHJHQHUDWHG &026 /1$V H[DPLQLQJ LQSXW DQG RXWSXW PDWFKLQJ JDLQ DQG QRLVH SDUDPHWHUV 7KH HIIHFWV RI VXEVWUDWH UHVLVWDQFH JDWHLQGXFHG QRLVH DQG LQSXWPDWFKLQJ YDULDWLRQV DUH FRQVLGHUHG 7KHVH PHWKRGRORJLHV DUH GHPRQVWUDWHG LQ &KDSn WHU ZKHUH WKH LPSOHPHQWDWLRQ DQG PHDVXUHG UHVXOWV RI VRXUFHGHJHQHUDWHG &026 /1$V RSHUDWLQJ DW DQG *+] DQG LPSOHPHQWHG LQ DQG SP &026 WHFKQRORJLHV DUH SUHVHQWHG $OVR D *+] WXQHG DPSOLILHU ZLWK D FRPPRQJDWH LQSXW LPSOHPHQWHG LQ D SDUWLDOO\VFDOHG VLOLFRQRQLQVXODWRU 62,f _LP &026 WHFKQRORJ\ LV SUHVHQWHG 7KHVH /1$V DUH LQWHQGHG IRU XVH LQ FORFN UHFHLYHUV KRZHYHU WKH UHVXOWV DUH DOVR DSSOLFDEOH WR VWDQGDUG &026 UHFHLYHUV

PAGE 19

,Q WKH FORFN UHFHLYHU WKH IUHTXHQF\ GLYLGHU WUDQVODWHV WKH a*+] JOREDO FORFN VLJQDO WR D a*+] ORFDO FORFN VLJQDO 7KH GLYLGHU PXVW EH FDSDEOH RI KLJK IUHTXHQF\ RI RSHUDWLRQ DQG ORZ LQSXWVHQVLWLYLW\ ZKLOH FRQVXPLQJ PLQLPDO SRZHU $OVR WKH GLYLGHUV VKRXOG EH V\QFKURQL]HG EHWZHHQ UHFHLYHUV WR UHGXFH WKH FORFN VNHZ 7KH GHVLJQ DQG LPSOHPHQWDWLRQ RI KLJKIUHTXHQF\ &026 IUHTXHQF\ GLYLGHUV DUH SUHVHQWHG LQ &KDSWHU 7KLV VHFWLRQ ZLOO SUHVHQW D GHVLJQ PHWKRGRORJ\ EDVHG RQ LQMHFWLRQ ORFNLQJ IRU GLYLGHUV LPSOHPHQWHG ZLWK VRXUFHFRXSOHG ORJLF 6&/f 7KHVH GLYLGHUV RVFLOODWH LQ WKH DEVHQFH RI DQ LQSXW VLJQDO KHQFH WKLV RVFLOODWLRQ FDQ EH LQMHFWLRQORFNHG WR SURYLGH IUHTXHQF\ GLYLn VLRQ IURP D ORZ LQSXWVZLQJ YROWDJH VLJQDO 0HDVXUHG UHVXOWV RI 6&/ GLYLGHUV RSHUDWLQJ XS WR DQG *+] DW DQG 9 XVLQJ DQG _LP &026 WHFKQRORJLHV DUH SUHVHQWHG $OVR D GXDOSKDVH G\QDPLF SVHXGR1026 GLYLGHU >
PAGE 20

7KH PDMRU JRDO RI WKLV UHVHDUFK LV WR GHPRQVWUDWH WKH RSHUDWLRQ RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP DQG HYDOXDWH LWV IHDVLELOLW\ &KDSWHU WDFNOHV WKH ILUVW RI WKHVH WZR JRDOV E\ GHPRQVWUDWLQJ WKH RSHUDWLRQ DQG SODXVLELOLW\ RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\Vn WHP )LUVW DQ RYHUYLHZ RI DQWHQQD IXQGDPHQWDOV LV SUHVHQWHG DORQJ ZLWK PHDVXUHG RQFKLS DQWHQQD FKDUDFWHULVWLFV IRU WHVWFKLSV LPSOHPHQWHG LQ DQG SP &026 WHFKQRORn JLHV $QWHQQD WUDQVPLVVLRQ JDLQ SKDVH DQG LPSHGDQFH UHVXOWV IRU OLQHDU GLSROH ]LJ]DJ GLSROH ORRS DQWHQQDV DQG DQWHQQDV LQ GLUHFW FRQWDFW ZLWK WKH VXEVWUDWH DUH SUHVHQWHG 7ZR VLQJOHUHFHLYHU LQWHUFRQQHFWV DQG RQH VLQJOHWUDQVPLWWHU LQWHUFRQQHFW DUH WKHQ SUHVHQWHG GHPRQVWUDWLQJ ZLUHOHVV LQWHUFRQQHFWV IRU WKH ILUVW WLPH 7KHVH LQWHUFRQQHFWV RSHUDWH DW *+] DFURVV PP XVLQJ WKH IMP WHFKQRORJ\ DQG DW *+] DFURVV DQG PP XVLQJ WKH _LP WHFKQRORJ\ $OVR D GRXEOHUHFHLYHU ZLUHOHVV LQWHUFRQQHFW ZLWK D SURn JUDPPDEOH GLYLGHU LV GHPRQVWUDWHG DQG WKH FORFN VNHZ DQG MLWWHU DUH REWDLQHG &KDSWHU HYDOXDWHV WKH IHDVLELOLW\ RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP LQ WHUPV RI SRZHU FRQVXPSWLRQ SURFHVV YDULDWLRQ V\QFKURQL]DWLRQ ODWHQF\ DUHD DQG GHVLJQ YHULILn FDWLRQ 7KH ZRUVWFDVH FORFN VNHZ DQG MLWWHU DUH HVWLPDWHG DQG FRPSDUHG WR WKH PHDVXUHG VNHZ DQG MLWWHU SUHVHQWHG LQ &KDSWHU 7KLV FKDSWHU ZLOO VKRZ WKDW FRPSDUDEOH SRZHU FRQn VXPSWLRQ VNHZ DQG MLWWHU FDQ EH REWDLQHG ZLWK D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP DV FRPSDUHG WR FRQYHQWLRQDO V\VWHPV ZLWK SRWHQWLDO FRVWV RI DGGHG DUHD DQG PRUH GLIILFXOW GHVLJQ YHULILFDWLRQ )LQDOO\ WKLV FKDSWHU FRQWDLQV FRQFOXVLRQV RQ WKH RYHUDOO IHDVLELOLW\ RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP DQG RQ WKLV ZRUN LQ JHQHUDO DQG VXJJHVWV IXWXUH ZRUN

PAGE 21

&+$37(5 &026 /2: 12,6( $03/,),(56 2YHUYLHZ 6FRSH RI /1$ 5HVHDUFK $ NH\ EXLOGLQJ EORFN IRU WKH FORFN UHFHLYHU DV ZHOO DV IRU WKH IURQWHQG RI VXSHUn KHWHURG\QH DQG GLUHFW FRQYHUVLRQ UHFHLYHUV LV WKH ORZ QRLVH DPSOLILHU /1$f 'XH WR LWV PRGHUDWH WR KLJK JDLQ DV ZHOO DV LWV ORZ QRLVH ILJXUH WKH /1$ DSSUR[LPDWHO\ VHWV WKH RYHUn DOO VLJQDOWRQRLVH UDWLR RI WKH UHFHLYHU E\ UHGXFLQJ WKH LPSDFW RI QRLVH IURP VXEVHTXHQW VWDJHV 7KLV LV HTXLYDOHQW WR VD\LQJ WKDW ZKHQ WKH /1$ LV SURSHUO\ GHVLJQHG WKH WRWDO UHFHLYHU QRLVH ILJXUH LV URXJKO\ WKDW RI WKH /1$ )RU D FDVFDGHG V\VWHP WKH WRWDO QRLVH IDFn WRU )f DQG QRLVH ILJXUH 1)f DUH JLYHQ E\ WKH IROORZLQJ IRUPXODV >*RQ@ f f 1) RJ)f ZKHUH 61f LV D VLJQDOWRQRLVH UDWLR DQG )c DQG *M DUH WKH QRLVH IDFWRUV DQG DYDLODEOH SRZHU JDLQV UHVSHFWLYHO\ RI LQGLYLGXDO VWDJHV LQ WKH UHFHLYHU $V FDQ EH VHHQ WKH QRLVH FRQWULEXWLRQV IURP ODWWHU VWDJHV DUH GLYLGHG E\ WKH WRWDO JDLQ SUHFHGLQJ WKHPf§D SURFHVV NQRZQ DV fLQSXWUHIHUULQJf 7KXV WR PLQLPL]H WKH WRWDO QRLVH ILJXUH WKH ILUVW VWDJH LQ WKH UHFHLYHU VKRXOG DPSOLI\ WKH LQSXW VLJQDO ZKLOH DGGLQJ PLQLPDO QRLVH )RU WKH ZLUHOHVV LQWHUFRQQHFW DSSOLFDWLRQ D &026 WHFKQRORJ\ LV UHTXLUHG WR EH FRQVLVWHQW ZLWK WKH ,756 +RZHYHU 5) &026 FLUFXLWU\ KDV RQO\ UHFHQWO\ EHHQ XQGHU

PAGE 22

LQYHVWLJDWLRQ E\ UHVHDUFKHUV ZLWK PRVW RI WKH UHVHDUFK RFFXUULQJ DW IUHTXHQFLHV EHORZ *+] 7KH OLPLWHG ZRUN LQ &026 LV GXH WR WKH VWULQJHQW SHUIRUPDQFH UHTXLUHPHQWV RI FRQn YHQWLRQDO ZLUHOHVV VWDQGDUGV HJ *OREDO 6\VWHP IRU 0RELOH *60f RU *OREDO 3RVLWLRQLQJ 6\VWHPV *36ff ZKLFK KDV QHFHVVLWDWHG WKH XVH RI *D$V RU VLOLFRQ ELSRODU WHFKQRORJLHV +RZHYHU &026 WHFKQRORJLHV FDQ SRWHQWLDOO\ UHGXFH WKH RYHUDOO FRVW RI WKH WUDQVFHLYHU E\ DOORZLQJ LQFUHDVHG OHYHOV RI LQWHJUDWLRQ ZLWK WKH VLQJOHFKLS UDGLR EHLQJ DQ XOWLPDWH JRDO )RU &026 WHFKQRORJ\ WR FRPSHWH ZLWK VLOLFRQ ELSRODU RU *D$V WHFKQRORJLHV IRU ZLUHOHVV DSSOLFDWLRQV LW PXVW DW OHDVW GHOLYHU WKH PLQLPXP QHFHVVDU\ SHUIRUPDQFH DW D UHGXFHG V\Vn WHPOHYHO FRVW )URP D SHUIRUPDQFH VWDQGSRLQW WR EH FRPSHWLWLYH ZLWK ELSRODU RU *D$V /1$V &026 /1$V PXVW HTXDO RU VXUSDVV WKHLU ORZ SRZHU FRQVXPSWLRQV RI DSSUR[Ln PDWHO\ P: DQG WKHLU ORZ QRLVH ILJXUHV RI DSSUR[LPDWHO\ G% >6KD@ )URP D FRVW VWDQGSRLQW WKH /1$ VKRXOG EH LPSOHPHQWHG LQ D VWDQGDUG GLJLWDO &026 WHFKQRORJ\ ZLWKn RXW DQ\ VSHFLDOW\ SDVVLYH FRPSRQHQWV DQG WKH /1$ VKRXOG UHTXLUH D PLQLPDO QXPEHU RI H[WHUQDO FRPSRQHQWV :KLOH UHFHQW ZRUNV KDYH GHPRQVWUDWHG WKH SRWHQWLDO RI &026 /1$V IRU aO*+] DSSOLFDWLRQV WKH\ JHQHUDOO\ KDYH GLIILFXOW\ LQ DWWDLQLQJ ERWK ORZ QRLVH ILJXUH DQG ORZ SRZHU FRQVXPSWLRQ VLPXOWDQHRXVO\ >6KD 6WX +XD@ $OWHUQDWHO\ LI WKH\ GR PHHW WKH ORZ QRLVH DQG ORZ SRZHU LW LV PDGH SRVVLEOH E\ XVLQJ HLWKHU PRGLILHG &026 WHFKQROn RJLHV RU H[WHUQDO PDWFKLQJ QHWZRUNV >*UD22 +D\@ $OVR DW WKLV WLPH RWKHU WKDQ ZRUN WKDW WKH DXWKRU KDV SDUWLFLSDWHG LQ RU RULJLQDWHG &026 /1$V RSHUDWLQJ DERYH a *+] KDYH QRW EHHQ UHSRUWHG 7KH IROORZLQJ WZR FKDSWHUV SUHVHQW WKH GHVLJQ DQG LPSOHPHQWDn WLRQ RI &026 /1$V RSHUDWLQJ EHWZHHQ DQG *+] $Q H[SOLFLWO\ GHILQHG GHVLJQ PHWKRGRORJ\ LV SUHVHQWHG LQ WKLV FKDSWHU ZKLFK FRQVLGHUV JDLQ QRLVH ILJXUH LQSXW

PAGE 23

PDWFKLQJ DQG RXWSXW PDWFKLQJ ,PSOHPHQWDWLRQ UHVXOWV IRU IRXU GLIIHUHQW /1$V RSHUDWLQJ DW DQG *+] UHVSHFWLYHO\ DUH WKHQ SUHVHQWHG LQ WKH QH[W FKDSWHU 3HUIRUPDQFH 0HWULFV RI /1$V $V PHQWLRQHG DERYH WR PLQLPL]H WKH UHFHLYHUfV QRLVH ILJXUH WKH /1$ LV UHTXLUHG WR KDYH PRGHUDWH WR KLJK JDLQ DQG ORZ QRLVH ILJXUH ([DPLQLQJ f LW ZRXOG VHHP WKDW WKH /1$fV JDLQ FDQ EH LQFUHDVHG DUELWUDULO\ WKHUHE\ PLQLPL]LQJ WKH QRLVH ILJXUH DQG LPSURYLQJ WKH VHQVLWLYLW\ RI WKH UHFHLYHU +RZHYHU WKLV LV QRW WKH FDVH GXH WR WKH HIIHFW RI /1$ JDLQ RQ UHFHLYHU OLQHDULW\ /LQHDULW\ LV W\SLFDOO\ PHDVXUHG LQ WHUPV RI D WKLUGRUGHU LQWHUFHSW SRLQW ,3f ZKLFK LV XVXDOO\ LQSXWUHIHUUHG ,,3f 7KH ,3 LV WKH RXWSXW SRZHU DW ZKLFK WKH WKLUGRUGHU LQWHUPRGXODWLRQ SURGXFWV DUH HTXDO WR WKH GHVLUHG OLQHDU FRPSRQHQW 7KH H[SUHVVLRQ IRU WKH WRWDO ,,3 IRU D FDVFDGHG V\VWHP FDQ EH H[SUHVVHG DV IROORZV L *L *L *R ,,3W ,,3M ,,3 ,,3 9 f ZKHUH +3c DQG *c DUH WKH LQSXWUHIHUUHG ,3 LQ ZDWWVf DQG WKH SRZHU JDLQ LQ ZDWWVZDWWf RI HDFK LQGLYLGXDO VWDJH RI WKH UHFHLYHU )RU JDLQV JUHDWHU WKDQ RQH WKH OLQHDULW\ RI ODWWHU VWDJHV ZLOO GRPLQDWH WKH WRWDO UHFHLYHU OLQHDULW\ 7KHUHIRUH WR PD[LPL]H WKH UHFHLYHUfV ,,3 WKH ODWWHU VWDJHVf OLQHDULW\ VKRXOG EH PD[LPL]HG ZKLOH UHGXFLQJ RU OLPLWLQJ WKH WRWDO SUHFHGLQJ JDLQ 7KXV WKH JDLQ RI WKH /1$ LH *Mf FDQQRW EH LQFUHDVHG DUELWUDULO\ VLQFH WKDW ZRXOG GHJUDGH WKH UHFHLYHUfV ,,3 7KHUHIRUH f DQG f LPSO\ DQ DFFHSWDEOH UDQJH RI /1$ JDLQV ZKLFK ZLOO PHHW ERWK WKH V\VWHPfV QRLVH ILJXUH DQG OLQHDULW\ UHTXLUHn PHQWV $OVR IURP f LW FDQ EH VHHQ WKDW WKH UHTXLUHPHQW IRU WKH /1$fV ,,3 LV QRW DV VWULQJHQW DV WKDW IRU VXEVHTXHQW VWDJHV HJ PL[HUf 7\SLFDOO\ WKH /1$ LV GHVLJQHG WR KDYH D SRZHU JDLQ RI DSSUR[LPDWHO\ G% D QRLVH ILJXUH RI OHVV WKDQ G% DQG DQ ,,3 RI G%P DOO WKH ZKLOH FRQVXPLQJ OHVV WKDQ a P: RI SRZHU

PAGE 24

,Q VXSHUKHWHURG\QH DQG GLUHFWFRQYHUVLRQ UHFHLYHUV WKH /1$ LV SUHFHGHG E\ WKH DQWHQQD D GXSOH[HU ILOWHU RU WUDQVPLWUHFHLYH VZLWFK DQG DQ RSWLRQDO SUHVHOHFW ILOWHU 6LQFH WKHVH FRPSRQHQWV DUH QRW W\SLFDOO\ LQWHJUDWHG WKH LQSXW WR WKH /1$ LV GULYHQ WKURXJK D 4 WUDQVPLVVLRQ OLQH 7KHUHIRUH WKH /1$f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fV RXWSXW VKRXOG EH PDWFKHG WR e/ 7KH LQSXW DQG RXWSXW PDWFKLQJ FULWHULD IRU D + PDWFK DUH VSHFLILHG LQ WHUPV 6MM DQG 6 ZKHUH ERWK VKRXOG EH OHVV WKDQ G% 3RVVLEOH /1$ 7RSRORJLHV 'HVLJQLQJ DQ /1$ FRQVLVWV RI PHHWLQJ WKH JDLQ QRLVH PDWFKLQJ DQG OLQHDULW\ SHUn IRUPDQFH PHWULFV ZKLOH PLQLPL]LQJ SRZHU FRQVXPSWLRQ DQG FRVW ZKHUH DOO RI WKH DIRUHn PHQWLRQHG WHQG WR WUDGHRII WR D FHUWDLQ H[WHQW ZLWK RQH DQRWKHUf 7RZDUGV WKLV HQG WKHUH DUH WZR PDLQ FLUFXLW WRSRORJLHV IRU &026 WKDW ZLOO EH GLVFXVVHGf§FRPPRQJDWH DQG FRPn PRQVRXUFH ZLWK LQGXFWLYH GHJHQHUDWLRQ &RPPRQ*DWH &026 /1$ 7KH ILUVW SRVVLEOH WRSRORJ\ HPSOR\V D FRPPRQJDWH DPSOLILHU ZLWK VRXUFH LQGXFn WDQFH VKRZQ LQ )LJXUH Df $SSHQGL[ $ FRQWDLQV GHULYDWLRQV IRU WKH LQSXW LPSHGDQFH JDLQ DQG QRLVH ILJXUH IRU WKLV FRPPRQJDWH WRSRORJ\ +HUH WKH LQSXW LV D SDUDOOHO UHVRQDQW

PAGE 25

)LJXUH 3RWHQWLDO &026 /1$ WRSRORJLHV Df FRPPRQJDWH RU Ef FRPPRQVRXUFH ZLWK LQGXFWLYH GHJHQHUDWLRQ QHWZRUN FRPSRVHG HVVHQWLDOO\ RI WKH VRXUFH LQGXFWDQFH /Vf WKH JDWHWRVRXUFH FDSDFLn WDQFH &JVf DQG RQH RYHU WKH WUDQVFRQGXFWDQFH OJPf 7KH LQSXW LPSHGDQFH LV GHVLJQHG WR EH a 4 KRZHYHU WR PD[LPL]H WKH JDLQ DQG PLQLPL]H WKH QRLVH ILJXUH WKH LQSXW LPSHGn DQFH FDQ EH VHW WR a 4 ZKLOH VWLOO DFKLHYLQJ DQ 6MM RI G% 7KXV JP LV VHW WR DSSUR[n LPDWHO\ 4 ZKLOH /V LV FKRVHQ WR SDUDOOHOUHVRQDWH ZLWK &JV DW WKH RSHUDWLQJ IUHTXHQF\ ,Q DGGLWLRQ WR HDVLO\ SURYLGLQJ WKH LQSXW PDWFK WKH FRPPRQJDWH WRSRORJ\ H[KLELWV JRRG OLQHDULW\ GXH WR WKH VRXUFH GHJHQHUDWLRQ RI WKH WUDQVLVWRU SURYLGHG E\ 5V $ GUDZEDFN RI WKLV WRSRORJ\ LV LWV KLJKHU QRLVH ILJXUH $V VKRZQ LQ $SSHQGL[ $ WKH QRLVH IDFWRU IRU WKLV WRSRORJ\ LV DSSUR[LPDWHO\ WKH IROORZLQJ ) D ?P5Vff f ZKHUH D LV WKH UDWLR EHWZHHQ WKH GHYLFH WUDQVFRQGXFWDQFH JPf DQG WKH VKRUWFLUFXLW GUDLQ FRQGXFWDQFH JGf ,Q WKH ORQJFKDQQHO OLPLW \ DQG D DUH DQG UHVSHFWLYHO\ ZKLOH \ FDQ EH VLJQLILFDQWO\ ODUJHU WKDQ IRU VKRUW FKDQQHO OHQJWKV GXH WR KRWHOHFWURQ HIIHFWV >-LQ@ 7KXV LQ WKH ORQJFKDQQHO OLPLW DQG ZLWK JP O 2I WKH QRLVH IDFWRU LV \LHOGLQJ D QRLVH ILJXUH RI G% +RZHYHU WKH QRLVH ILJXUH FDQ EH VLJQLILFDQWO\ ODUJHU LQ

PAGE 26

WKH VKRUWFKDQQHO UHJLPH DQG ZKHQ WDNLQJ LQWR DFFRXQW RWKHU VRXUFHV RI QRLVH HJ VXEn VWUDWH UHVLVWDQFH LQGXFWRU SDUDVLWLF UHVLVWDQFH DQG JDWHLQGXFHG QRLVH >=LH 6KD 7VL@f ZLWK W\SLFDO H[SHULPHQWDO YDOXHV H[FHHGLQJ G% 7KH KLJK QRLVH ILJXUH IRU WKH FRPPRQJDWH WRSRORJ\ LV D PDMRU GLVDGYDQWDJH )XQn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f ZKLFK FRQVLVWV RI D FRPPRQVRXUFH DPSOLILHU ZLWK LQGXFWLYH GHJHQHUDWLRQ 7R REWDLQ WKH H[WUD GHJUHH RI IUHHGRP LQ WKH GHVLJQ DQ LQGXFWRU /Jf LV DGGHG LQ VHULHV ZLWK WKH JDWH $V ZLOO EH VKRZQ LQ WKH QH[W VHFWLRQ WKLV WRSRORJ\ DOORZV WKH SRZHU DQG QRLVH PDWFKLQJ FRQGLWLRQV WR EH PHW VLPXOWDQHRXVO\ ZKLOH H[KLELWLQJ VXIILFLHQW OLQHDULW\ DQG DOORZLQJ IRU ORZ SRZHU FRQVXPSWLRQ 6RXUFH'HJHQHUDWHG &026 /1$ )LJXUH VKRZV D VLPSOLILHG VFKHPDWLF RI D &026 /1$ ZLWK LQGXFWLYH VRXUFH GHJHQHUDWLRQ 7KLV /1$ LV PDWFKHG WR 4 DW ERWK WKH LQSXW DQG WKH RXWSXW $ VLQn JOHVWDJH WRSRORJ\ LV XVHG WR PLQLPL]H WKH SRZHU GLVVLSDWLRQ DQG WR LPSURYH G% FRPn SUHVVLRQ SRLQW 3OG%f DQG ,3 SHUIRUPDQFH 7KH FLUFXLW JDLQ LV SURYLGHG E\ D FDVFRGH ,I JP n WKHQ WKH QRLVH ILJXUH ZRXOG EH DSSUR[LPDWHO\ G% IRU WKH ORQJFKDQQHO OLPLW

PAGE 27

9GG )LJXUH 6FKHPDWLF RI WKH VRXUFHGHJHQHUDWHG &026 /1$ DPSOLILHU ZKLFK KDV D UHGXFHG 0LOOHU HIIHFW $OVR D FDVFRGH H[KLELWV KLJK UHYHUVH LVRODn WLRQ ZKLFK VLPSOLILHV WKH GHVLJQ SURFHGXUH E\ GHFRXSOLQJ WKH LQSXW DQG RXWSXW PDWFKLQJ FRQGLWLRQV :KLOH WKLV WRSRORJ\ LV VLPLODU WR RWKHU UHSRUWHG &026 /1$V IXQGDPHQWDO GLIIHUHQFHV LQFOXGH WKH RPLVVLRQ RI D RXWSXW EXIIHU WKH XVH RI VKLHOGLQJ VWUXFWXUHV DQG WKH XVH RI D FDSDFLWLYH WUDQVIRUPHU DW WKH RXWSXW ,QSXW 0DWFKLQJ IRU 6RXUFH'HJHQHUDWHG /1$ ,QSXW ,PSHGDQFH ,W FDQ UHDGLO\ EH VKRZQ WKDW WKH LQSXW LPSHGDQFH RI WKH VRXUFHGHJHQHUDWHG /1$ QHJOHFWLQJ JDWHWRGUDLQ FDSDFLWDQFH &JGLf LV DV IROORZV ] Lf&$Wf V UOW -p/JV? /JV? f

PAGE 28

$ VDOLHQW IHDWXUH RI WKLV WRSRORJ\ LV WKDW VRXUFH GHJHQHUDWLRQ SURYLGHV D UHDO WHUP WR WKH LQSXW LPSHGDQFH ZKLFK FDQ WKHQ EH XVHG WR PDWFK WR  7KLV UHDO WHUP LV QRW D UHVLVn WDQFH SHU VH KHQFH LW ZLOO QRW JHQHUDWH WKHUPDO QRLVHf EXW UDWKHU ZKHQ D FXUUHQW LV DSSOLHG WR WKH LQSXW QRGH WKH YROWDJH WKDW GHYHORSV DW WKDW QRGH KDV FRPSRQHQWV ERWK LQ SKDVH UHDO WHUPf DQG s GHJUHHV RXW RI SKDVH LPDJLQDU\ WHUPV GXH WR /J /V DQG &JVOf 7KH LQSKDVH FRPSRQHQW UHVXOWV IURP WKH VHULHV IHHGEDFN LQ WKH VRXUFH RI 0M $V FDQ EH VHHQ IURP f WKLV QHWZRUN WDNHV WKH IRUP RI D VHULHV UHVRQDQW FLUFXLW ZLWK UHVRQDQW IUHTXHQF\ BL >6 /f&6-@ f 7KXV DW VHULHV UHVRQDQFH WKH LQSXW LPSHGDQFH EHFRPHV U\ B 6P M =cQ a U /Va W/V f ZKLFK LV D IXQFWLRQ RI WKH ELDV FRQGLWLRQ WKH FKDQQHO OHQJWK RI 0M DQG /V 7KH TXDOLW\ IDFWRU RI WKLV QHWZRUN LQFOXGLQJ WKH VRXUFH UHVLVWDQFH 5V LV LQ p&JLO f5 FRU/-ff 7KXV WR GHVLJQ WKH LQSXW PDWFKLQJ QHWZRUN IRU D JLYHQ &JVL WR DFKLHYH DQ DQG VHULHV UHVRQDQFH WKH IROORZLQJ UHODWLRQV DUH XVHG f G% f§ / f§ U M p7 /V f7 /V rR&} ,QSXW 0DWFK WR :LWKVWDQG &RPSRQHQW 9DULDWLRQV f f $ PHWKRGRORJ\ LV WKHQ QHHGHG WR FKRRVH &JVO LH WKH ZLGWK RI 0Mf DQG \LHOG /V DQG /J 2QH ZD\ LV WR FKRRVH &JVL WR PHHW WKH LQSXW PDWFKLQJ FRQGLWLRQ RYHU DQ HQWLUH

PAGE 29

RSHUDWLQJ IUHTXHQF\ EDQG ZKLOH ZLWKVWDQGLQJ FRPSRQHQW YDULDWLRQV LQ /J /V DQG &JVM 4XDOLWDWLYHO\ WKH LQSXW LPSHGDQFH LV PRUH VHQVLWLYH WR FRPSRQHQW YDULDWLRQV IRU KLJK4 QHWZRUNV 7KXV WKHUH LV D PD[LPXP YDOXH IRU WKH TXDOLW\ IDFWRU ZKLFK HQVXUHV PDWFKLQJ IRU D JLYHQ IUHTXHQF\ EDQG DQG IRU D JLYHQ VHW RI FRPSRQHQW WROHUDQFHV $Q DOWHUQDWLYH LQSXW TXDOLW\ IDFWRU LQGHSHQGHQW RI /V FDQ EH GHILQHG DV IROORZV $VVXPLQJ WKDW LQGXFWRUV KDYH D WROHUDQFH RI 7/ &JVO WROHUDQFH RI 7W HTXDWLRQ f EHFRPHV WKH IROORZLQJ KDV D WROHUDQFH RI 7 2 f DQG FRM KDV D ]LQ &œ W/V s 7Of^ s 7Wf 0/ /Vfs7Of FR M&JVOOs7Ff /s7Ofs 7Wf -4De s 7/f f 1RWH WKDW WKH YDULDWLRQ LQ FR[ ZKLFK KDV D _YMDL_/ GHSHQGHQFH LQ WKH YHORFLW\VDWXUDWHG UHJLPH >7VL@ ZKHUH YVDW LV WKH VDWXUDWLRQ YHORFLW\f LV SDUWLDOO\ FRUUHODWHG WR WKH YDULDWLRQ LQ &JV ZKLFK KDV D :/&; GHSHQGHQFHf WKURXJK YDULDWLRQV LQ WKH OHQJWK RI WKH GHYLFH 7KH LPSHGDQFH LQ f KDV WR VDWLVI\ WKH LQSXW PDWFKLQJ FRQGLWLRQ DW ERWK WKH XSSHU DQG ORZHU HQGV RI WKH IUHTXHQF\ EDQG LH FR &Rs ZKHUH % LV WKH WRWDO EDQGZLGWKf 7KH LQSXW LPSHGDQFHV ZKLFK VDWLVI\ 6M M G% FDQ EH YLVXDOL]HG RQ D 6PLWK FKDUW DV WKRVH LPSHGDQFHV ORFDWHG ZLWKLQ D FLUFOH FHQWHUHG DW WKH RULJLQ ZLWK UDGLXV nrf DV VKRZQ LQ )LJXUH ,I WKH LQSXW LPSHGDQFH LV QRUPDOL]HG DQG ZULWWHQ DV ]cQ UfMLf WKH UHTXLUHG YDOXH IRU L WR UHVXOW LQ D VSHFLILF 6 M M IRU D JLYHQ YDOXH RI U LV DV IROORZV L s _6O_9OfUOf L0 f

PAGE 30

7KH ZRUVW FDVH PDWFKLQJ FRQGLWLRQ RFFXUV HLWKHU DW WKH ORZHU HQG RI WKH IUHTXHQF\ EDQG ZKHQ WKH FRPSRQHQW WROHUDQFHV FDXVH WKH UHVRQDQW IUHTXHQF\ WR LQFUHDVH RU DW WKH XSSHU HQG RI WKH IUHTXHQF\ EDQG ZKHQ WKH FRPSRQHQW WROHUDQFHV FDXVH WKH UHVRQDQW IUHTXHQF\ WR GHFUHDVH %RWK FRQGLWLRQV \LHOG DSSUR[LPDWHO\ WKH VDPH YDOXH IRU 4JV /RRNLQJ DW WKH ILUVW FRQGLWLRQ LH / /7Of &JVO &JVOO7Ff WR FR4 DQG FRW k[7Wff WKH UHTXLUHG YDOXH RI 4JV FDQ EH VROYHG E\ QRUPDOL]LQJ f DQG WKHQ DSSO\LQJ f DV IROORZV 4JV 7Ff R 7/f 7FffL? _Q_9Of f§ Uf _6 f OL ZKHUH U [fW/V 7Of7Wf f 4Ea IL f 4E f f

PAGE 31

7DEOH 5HTXLUHG 9DOXH RI 4JV IRU &RPSRQHQW 9DULDWLRQV ZLWK 6 M W G% %DQG *60 0+] *60 0+] *60 0+] ,60 *+] ,60 *+] ,60 *+] 4E ,7O, ,7&, ,7W, b b b b b b pW/V 4JV 7DEOH VKRZV WKH UHTXLUHG YDOXH RI 4JV IRU WKH *60 DQG ,60 ,QGXVWULDO 6FLHQn WLILF DQG 0HGLFDOf EDQGV IRU YDU\LQJ FRPSRQHQW WROHUDQFHV DQG YDOXHV IRU !S/V ,W FDQ EH VHHQ WKDW DV 4% WKH TXDOLW\ IDFWRU LPSOLHG E\ WKH RSHUDWLQJ IUHTXHQF\ EDQGf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b RI FRPSRQHQW YDULDWLRQV 2QFH 4JV LV VSHFLILHG &JVO DQG KHQFH FDQ EH GHWHUPLQHG DOORZLQJ WKH LQGXFWRU YDOXHV WR EH FKRVHQ DV JLYHQ E\ f DQG f 2XWSXW 0DWFKLQJ IRU 6RXUFH'HJHQHUDWHG /1$ 'ULYLQJ D  ORDG ZKLOH SURYLGLQJ VXIILFLHQW SRZHU JDLQ UHTXLUHV HLWKHU DQ RXWn SXW PDWFKLQJ QHWZRUN RU D  EXIIHU 6LQFH D EXIIHU W\SLFDOO\ ZRXOG GLVVLSDWH DGGLWLRQDO SRZHU ZKLOH DOVR GHJUDGLQJ WKH FLUFXLW OLQHDULW\ D VLQJOHVWDJH GHVLJQ ZLWK DQ RXWSXW

PAGE 32

9GG )LJXUH 2XWSXW PDWFKLQJ HTXLYDOHQW FLUFXLWV IRU &026 /1$ PDWFKLQJ QHWZRUN LV SUHIHUDEOH $V VKRZQ LQ )LJXUH DQG )LJXUH Df WKH RXWSXW LPSHGDQFH RI 0 LV WUDQVIRUPHG WR 4 XVLQJ D WKUHHHOHPHQW PDWFKLQJ QHWZRUN FRQVLVWLQJ RI D FDSDFLWLYH WUDQVIRUPHU &M DQG &f DQG D VKXQW LQGXFWRU /Gf 7KLV WKUHHHOHPHQW QHWZRUN LV DOVR NQRZQ DV D Q PDWFKLQJ QHWZRUN &RPSDUHG WR D WZRHOHn PHQW PDWFKLQJ QHWZRUN RU f/f QHWZRUNf D ,, QHWZRUN DOORZV PXOWLSOH VHWV RI FRPSRQHQW YDOXHV IRU /G &M DQG & (DFK VHW FRUUHVSRQGV WR D GLIIHUHQW ORDGHG TXDOLW\ IDFWRU IRU WKH ,, QHWZRUN 7KH WZRHOHPHQW QHWZRUN RQ WKH RWKHU KDQG DOORZV RQO\ RQH VHW RI FRPSRn QHQW YDOXHV IRU /G DQG & &M LV QR ORQJHU SUHVHQWf ZLWK WKH TXDOLW\ IDFWRU RI WKH ORDGHG QHWZRUN IL[HG E\ WKH LPSHGDQFH WUDQVIRUPDWLRQ UDWLR >%RZ@ 7KHUHIRUH WKH ,, QHWZRUN SURYLGHV DGGHG IOH[LELOLW\ WR WKH GHVLJQHU DOORZLQJ PXOWLSOH YDOXHV IRU WKH GUDLQ LQGXFWRU

PAGE 33

)LJXUH Ef VKRZV DQ HTXLYDOHQW FLUFXLW IRU WKH RXWSXW PDWFKLQJ QHWZRUN LQFOXGn LQJ WKH SDUDVLWLF VHULHV UHVLVWDQFH U/Gf DQG VKXQW FDSDFLWDQFH &/Gf DVVRFLDWHG ZLWK /G 7KH RXWSXW LPSHGDQFH RI WKH FDVFRGH LV UHSUHVHQWHG DV D UHVLVWDQFH 53 RQ WKH RUGHU RI NLORRKPVf LQ SDUDOOHO ZLWK D FDSDFLWDQFH &3 ZKHUH &3 FRQVLVWV RI &JG DQG &GE )RU RSWLPDO SRZHU WUDQVIHU WKH PDWFKLQJ QHWZRUN ZRXOG WUDQVIRUP 5SA WR  +RZHYHU WKLV LPSOLHV D TXDOLW\ IDFWRU RI 4 M5S IRU WKH PDWFKLQJ QHWZRUN ZKLFK LV RQ WKH RUGHU RI IRU 5A NLO 6LQFH WKH PDWFKLQJ QHWZRUN LQFOXGHV DQ RQFKLS VSLUDO LQGXFn WRU WKH TXDOLW\ IDFWRU RI WKH QHWZRUN LV OLPLWHG E\ WKDW RI WKH LQGXFWRU 4/G 7R LQFOXGH WKH ILQLWH 4 RI WKH VSLUDO LQGXFWRU UA FDQ EH WUDQVIRUPHG WR DQ HTXLYDOHQW UHVLVWDQFH VKXQWLQJ WKH LQGXFWRU DV IROORZV 5S/G 4/G AfU/G 4/G U/G f f 6LQFH 53/G LV W\SLFDOO\ PXFK OHVV WKDQ 53 WKH PDWFKLQJ QHWZRUN WUDQVIRUPV D FRPSOH[ VRXUFH FRQVLVWLQJ RI 53/G LQ SDUDOOHO ZLWK WKH FDSDFLWDQFH &SM&MAM LQWR WKH  ORDG $ GLOHPPD QRZ DSSHDUV LQ WKDW WKH LPSHGDQFH WR EH WUDQVIRUPHG 53/Gf LV D IXQFn WLRQ RI WKH QHWZRUN SURYLGLQJ WKH LPSHGDQFH WUDQVIRUPDWLRQ /Gf 7KHUHIRUH DQ LWHUDWLYH DSSURDFK FDQ EH WDNHQ XVLQJ JDLQ DQG RXWSXW PDWFKLQJ DV FULWHULD DV IROORZV f DQ LQLWLDO YDOXH LV FKRVHQ IRU /G f 53/G DQG DUH WKHQ HVWLPDWHG DORQJ ZLWK =0 LQ )LJXUH Df f YDOXHV IRU &M DQG & DUH GHVLJQHG WR DFKLHYH D 4 PDWFK DQG f 6 DQG 6 DUH FKHFNHG DJDLQVW WKHLU VSHFLILFDWLRQV DQG /G LV XSGDWHG LI QHFHVVDU\ 7KH LQLWLDO YDOXH IRU /G LV FKRVHQ EDVHG RQ WKH GHVLUHG JDLQ XVLQJ IRUPXODV GHULYHG LQ WKH QH[W VHFWLRQ 7KH HTXDWLRQV IRU 4 DQG & FDQ EH UHDGLO\ GHULYHG >+R22@ DQG DUH JLYHQ LQ $SSHQGL[ % 7KXV DQ\ VWDQGDUG WZRHOHPHQW PDWFKLQJ WHFKQLTXH FDQ EH XVHG WR FKRRVH &M DQG & IRU FRPSOHWLQJ WKH PDWFK RI =0 WR 4 HJ >%RZ@f

PAGE 34

*DLQ RI 6RXUFH'HJHQHUDWHG /1$ )RU DQ DPSOLILHU GULYHQ E\ D VRXUFH 9V ZLWK VRXUFH LPSHGDQFH 5V DQG ORDGHG DW WKH RXWSXW E\ 5/ YLD WUDQVPLVVLRQ OLQHV ZLWK FKDUDFWHULVWLF LPSHGDQFHV RI 5V DQG 5/ UHVSHFWLYHO\ WKH IRUZDUG WUDQVGXFHU SRZHU JDLQ LV JLYHQ E\ >*RQ@ *W a S /RDG $ 96 _VLI 9 2XW A A6 L 56 f ZKHUH LV WKH SRZHU GHOLYHUHG WR WKH ORDG DQG 3$96 LV WKH SRZHU DYDLODEOH IURP WKH VRXUFH )RU 56 5/ e WKH WUDQVGXFHU SRZHU JDLQ LV WKHQ *W Of6n B f OY_ f *DLQ 'ULYLQJ 5HVLVWLYH /RDG ,W FDQ EH VKRZQ WKDW WKH YROWDJH JDLQ IURP WKH VRXUFH WR WKH ORDG $A DVVXPLQJ WKDW WKH LQSXW LV DW VHULHV UHVRQDQFH DQG DJDLQ QHJOHFWLQJ &JGO LV DV IROORZV ,D 9_ f *nf_r} Mr!&-O f MpR*A/ &R& &f5O ZKHUH &GO LV WKH WRWDO FDSDFLWDQFH DW WKH QRGH FRQQHFWLQJ WKH GUDLQ RI 0M WR WKH VRXUFH RI 0 DQG 4LQ LV WKH TXDOLW\ IDFWRU RI WKH LQSXW VHULHV UHVRQDQW FLUFXLW DV JLYHQ E\ HTXDWLRQ f +HUH =AT LV WKH WRWDO HTXLYDOHQW LPSHGDQFH DW WKH GUDLQ QRGH RI 0 DV VKRZQ LQ )LJXUH 6LQFH WKH RXWSXW PDWFKLQJ QHWZRUN HVVHQWLDOO\ PDWFKHV WKH HTXLYDOHQW SDUDOOHO UHVLVWDQFH RI LQGXFWRU /G WR WKH ORDG UHVLVWDQFH 5/ WKHQ LV DSSUR[LPDWHO\ HTXDO WR WKH IROORZLQJ A/HTARf 4/GU/G A/GARAGf f

PAGE 35

7KH YROWDJH JDLQ FDQ WKHQ EH ZULWWHQ DV ,r YO &' eB U f 4LQ f PO n t/G f fAGf pmFRf f ZKHUH FRf P P Mp&G? f QFRf MG&O &f5O f \FR&" +HUH FRf LV D ORZSDVV ILOWHU UHVSRQVH ZKHUH WKH ORZSDVV ILOWHU LV FRPSRVHG RI &GM DQG OJP mFRf LV D FRPSOH[ FDSDFLWLYH WUDQVIRUPHU UDWLR $W VXIILFLHQWO\ KLJK IUHTXHQFLHV FL QFRf UHGXFHV WR f§ KRZHYHU WKLV DSSUR[LPDWLRQ PXVW EH MXVWLILHG DQG LV UDUHO\ DSSUR & SULDWH IRU W\SLFDO YDOXHV RI & %\ LQVSHFWLRQ WKH UDWLR QO LV DOZD\V OHVV WKDQ )XUWKHU VLPSOLILFDWLRQ FDQ EH DSSOLHG WR f WR REWDLQ D PRUH XVHIXO HTXDWLRQ 6XEVWLWXWLQJ f DQG QRWLQJ WKDW FRS JPL&JVO LW FDQ EH VKRZQ WKDW 0 7 n 4/G n /G 5V &2 7/6 6mf f 0HWKRGV WR 0D[LPL]H *DLQ 7R PD[LPL]H WKH JDLQ 6Mf IRXU VWHSV FDQ EH WDNHQ )LUVW H[DPLQLQJ WKH HIIHFW RI WKH GUDLQ LQGXFWRU RQ 6 UHTXLUHV H[DPLQLQJ WKH WHUP 4/G/GQFRf ZKHUH QFRf GHSHQGV RQ 4OM DQG /G YLD &M DQG & +RZHYHU VLQFH Q&Rf LV DSSUR[LPDWHO\ D FDSDFLWDQFH UDWLR LWV GHSHQGHQFH RQ DQG /G LV ZHDN 7KHUHIRUH 6M LQFUHDVHV ZLWK LQFUHDVLQJ 4MAM DQG LQFUHDVLQJ /G 7KLV FDQ EH VHHQ LQ )LJXUH ZKHUH 6 LV SORWWHG Df YHUVXV /G IRU FRQVWDQW DQG Ef YHUVXV IRU FRQVWDQW /G )RU HDFK GDWD SRLQW WKH RXWSXW LV PDWFKHG WR e 7KHUHIRUH WKH /1$ VKRXOG EH GHVLJQHG VXFK WKDW /G DQG LWV TXDOLW\

PAGE 36

IDFWRU LV PD[LPL]HG 6HFRQG WR PD[LPL]H WKH JDLQ FRM VKRXOG EH PD[LPL]HG 7KLV FDQ EH DFKLHYHG LQ WZR ZD\V DV IROORZV f ELDV WKH WUDQVLVWRU IRU PD[LPXP &'M ZKLFK FDQ LQFUHDVH WKH SRZHU FRQVXPSWLRQ RU f XVH D PRUH DGYDQFHG LH FRVWO\f WHFKQRORJ\ VKRUWHU FKDQQHO OHQJWKf IRU D JLYHQ SRZHU FRQVXPSWLRQ 7KXV FRVW DQG SRZHU FRQVXPSn WLRQ VKRXOG EH EDODQFHG WR DFKLHYH WKH FRUUHFW WUDGHRII IRU R!S 7KLUG ORRNLQJ LQ WKH GHQRPLQDWRU RI f LW FDQ EH VHHQ WKDW WKH JDLQ DQG WKH LQSXW PDWFKLQJ FULWHULD WUDGHRII YLD WKH WHUP !S/V 6LQFH VHULHV IHHGEDFN LV XVHG LQ WKH VRXUFH RI 0" WKH JDLQ LV UHGXFHG ZKLOH WKH LQSXW VHULHV UHVLVWDQFH 5Vf LV LQFUHDVHG E\ D IDFWRU RI RQH SOXV WKH ORRS JDLQ O2M/M95Vf 5HWXUQLQJ WR f WKH LQSXW PDWFKLQJ FRQGLWLRQ FDQ VWLOO EH PHW E\ FKRRVn LQJ F2M/V 4 DV RSSRVHG WR  7KXV E\ DOORZLQJ D VPDOO DPRXQW RI PLVPDWFK DW WKH LQSXW LH 6M@ G%f WKH JDLQ LV LQFUHDVHG E\ DSSUR[LPDWHO\ b RU G% )LQDOO\ WKH WHUP QO VKRXOG EH PD[LPL]HG ZKHUH FDQ EH PD[LPL]HG E\ UHGXFLQJ &GO WKURXJK MXGLFLRXV OD\RXW DQG Q FDQ EH PLQLPL]HG E\ GHFUHDVLQJ RU HYHQ HOLPLQDWLQJ &M Df Ef )LJXUH 6L DW UHVRQDQFH YHUVXV Df /G IRU FRQVWDQW DQG Ef IRU FRQVWDQW /G 7KH RXWSXW LV DOZD\V PDWFKHG WR e 1RWH WKDW LI EHFRPHV WRR ODUJH WKH /1$ FDQ EHFRPH LQWROHUDQW WR SURFHVV YDULDWLRQV

PAGE 37

/RRNLQJ DW WKHVH IRXU RSWLRQV 4/G DQG b KDYH WKH PRVW VLJQLILFDQW LPSDFW RQ JDLQ ,QWXLWLYHO\ KLJKTXDOLW\ IDFWRU LQGXFWRUV UHVXOW LQ /1$V ZKLFK FRQVXPH ORZHU SRZHU 7KLV LV WKH FDVH IRU >+D\@ ZKHUH H[WHUQDO PDWFKLQJ QHWZRUNV DUH XVHG WR REWDLQ KLJK JDLQ DQG ORZ QRLVH DW D YHU\ VPDOO SRZHU +RZHYHU IRU D IXOO\ LQWHJUDWHG /1$ H[WHUn QDO PDWFKLQJ QHWZRUNV DUH WDERR VLQFH WKH H[WUD FRPSRQHQWV LQFUHDVH WKH FRVW DQG VLQFH WKHLU KLJK 4fV DUH QRW YHU\ UHSHDWDEOH 7KXV WR REWDLQ WKH GHVLUHG JDLQ XVLQJ RQFKLS LQGXFWRUV ZLWK OLPLWHG TXDOLW\ IDFWRUV WKH SRZHU FRQVXPSWLRQ KDV WR EH LQFUHDVHG 7KH TXDOLW\ IDFWRU RI /G LV OLPLWHG E\ WKH QXPEHU RI PHWDO OD\HUV SUHVHQW WKH R[LGH WKLFNQHVV EHWZHHQ WKH WRSOHYHO PHWDO DQG WKH VXEVWUDWH WKH VXEVWUDWH UHVLVWDQFH DQG WKH W\SH RI PDWHULDO XVHG IRU WKH PHWDO LQWHUFRQQHFW DOXPLQXP RU FRSSHUf *DLQ 'ULYLQJ &DSDFLWLYH /RDG :KHQ WKH /1$ GRHV QRW GULYH  DQG LQVWHDG GULYHV D FDSDFLWLYH ORDG LH PL[HU RU IUHTXHQF\ GLYLGHUf WKH JDLQ LV VLJQLILFDQWO\ LQFUHDVHG 7KXV ZKHQ WKH FDSDFLWLYH WUDQVIRUPHU &c DQG &f LV UHPRYHG FRQVLVWV RQO\ RI WKH GUDLQ LQGXFWRU ZLWK LWV SDUD VLWLFV LQ SDUDOOHO ZLWK DQ HTXLYDOHQW ORDG FDSDFLWDQFH FRPSRVHG RI &A &0 DQG WKH FDSDFLWDQFH RI WKH VXEVHTXHQW VWDJH 7KXV =/HTFRf EHFRPHV WZLFH WKDW JLYHQ E\ HTXDWLRQ f $OVR ZLWK WKH RXWSXW PDWFKLQJ QHWZRUN UHPRYHG QRQf LV QR ORQJHU UHTXLUHG 7KH YROWDJH JDLQ WKHQ EHFRPHV WKH IROORZLQJ f :LWK WKH HOLPLQDWLRQ RI UFFRf DQG WKH DGGLWLRQDO IDFWRU RI WZR WKH JDLQ FDQ EH VLJQLILFDQWO\ ODUJHU 7KLV PHDQV WKDW WKH GHVLUHG JDLQ RI a G% FDQ EH DFKLHYHG DW D UHGXFHG SRZHU FRQn VXPSWLRQ LH 2Sf $OVR VLQFH 4 DQG & DUH UHPRYHG WKH WRWDO FDSDFLWDQFH UHVRQDWLQJ ZLWK /G LV VPDOO 7KXV /G FDQ EH LQFUHDVHG IXUWKHU LQFUHDVLQJ WKH JDLQ

PAGE 38

1RLVH 3DUDPHWHUV 7KH QRLVH IDFWRU RI DQ\ WZRSRUW QHWZRUN FDQ EH FODVVLILHG LQ WHUPV RI IRXU QRLVH SDUDPHWHUV DV IROORZV ) )PK AO?=V=RSO?` f ZKHUH )PLQ LV WKH PLQLPXP REWDLQDEOH QRLVH IDFWRU *Q LV WKH QRLVH FRQGXFWDQFH =RSW 5RSW M;RSW LV WKH RSWLPXP VRXUFH LPSHGDQFH UHVXOWLQJ LQ ) )PLQ DQG =V 5V M;V LV WKH DFWXDO VRXUFH LPSHGDQFH 1RWH WKDW WKHVH QRLVH SDUDPHWHUV DUH PRUH FRQYHQLHQW WKDQ WKH IRXU WUDGLWLRQDO QRLVH SDUDPHWHUV )PLQ 5Q *RSW %RSWf >*RQ@ IRU WKH VRXUFHGHJHQn HUDWHG /1$ GXH WR WKH VRXUFH EHLQJ LQ DQ LPSHGDQFH IRUP UDWKHU WKDQ DQ DGPLWWDQFH IRUP $ UHYLHZ RI WKLV VHW RI QRLVH SDUDPHWHUV LQ LPSHGDQFH IRUP LV FRQWDLQHG LQ $SSHQGL[ & 1RLVH 6RXUFHV 7KH SULPDU\ QRLVH VRXUFH ZLWKLQ WKH &026 /1$ LV WKHUPDO QRLVH JHQHUDWHG E\ WKH GLVWULEXWHG FKDQQHO RI WKH 026 WUDQVLVWRU $ VHFRQGDU\ QRLVH VRXUFH LV WKHUPDO QRLVH JHQn HUDWHG E\ WKH VHULHV UHVLVWDQFH RI VSLUDO LQGXFWRUV &KDQQHO QRLVH FDQ EH OXPSHG LQWR WZR FXUUHQWQRLVH VRXUFHVf§RQH DW WKH GUDLQ DQG RQH DW WKH JDWH 7KH GUDLQ QRLVHFXUUHQW KDV WKH IROORZLQJ SRZHU VSHFWUDO GHQVLW\ a -I 07\J-F f ZKHUH \ LV D ELDVGHSHQGHQW SDUDPHWHU IRU ORQJFKDQQHO GHYLFHVf N LV %ROW]PDQQfV FRQVWDQW [ n -.f 7 LV DEVROXWH WHPSHUDWXUH DQG JGR LV WKH VKRUWFLUFXLW GUDLQ FRQGXFWDQFH 7KH JDWH QRLVHFXUUHQW D VHFRQGRUGHU HIIHFW NQRZQ DV JDWHLQGXFHG QRLVH *,1f >=LH 6KD 7VL@ KDV WKH IROORZLQJ SRZHU VSHFWUDO GHQVLW\

PAGE 39

N7E $ & ? JV 6GR f ZKHUH LV DQRWKHU ELDVGHSHQGHQW SDUDPHWHU IRU ORQJFKDQQHO GHYLFHVf 6LQFH WKH RULJLQ RI ERWK WKH GUDLQ DQG JDWH QRLVH FXUUHQWV LV WKH FKDQQHO WKHVH QRLVH VRXUFHV DUH FRUn UHODWHG KDYLQJ D FRUUHODWLRQ FRHIILFLHQW Ff RI M IRU ORQJFKDQQHO GHYLFHV 'XH WR WKH JDWHWRFKDQQHO FDSDFLWDQFH SURYLGLQJ D r SKDVH VKLIW F LV FRPSOH[ $VLGH IURP *,1 WKHUH DUH WKUHH RWKHU VHFRQGRUGHU QRLVH VRXUFHV ZLWKLQ WKH /1$ DV IROORZV FKDQQHO QRLVH RI 0 WKHUPDO QRLVH DVVRFLDWHG ZLWK JDWH UHVLVWDQFH DQG WKHUn PDO QRLVH JHQHUDWHG LQ WKH VXEVWUDWH >.LV@ 7R XQGHUVWDQG WKH UHODWLYH LPSRUWDQFH RI WKHVH VHFRQGRUGHU HIIHFWV WKH QRLVH SDUDPHWHUV IRU D VLQJOH WUDQVLVWRU ZLOO EH GHULYHG LQ WKH IROORZLQJ VHFWLRQ 1RLVH 3DUDPHWHUV RI 6LQJOH 7UDQVLVWRU )LJXUH VKRZV WKH KLJKIUHTXHQF\ TXDVLVWDWLF PRGHO IRU WKH LQWULQVLF 026)(7 >7VL@ LQFOXGLQJ QRLVH VRXUFHV 'HULYLQJ WKH QRLVH SDUDPHWHUV IRU WKLV VLQJOH WUDQVLVWRU LQFOXGLQJ GUDLQ DQG JDWHLQGXFHG QRLVH WKH ERG\ WUDQVFRQGXFWDQFH JPEf VXEVWUDWH

PAGE 40

UHVLVWDQFH 5VXEf DQG DOO RI WKH FDSDFLWDQFHV &JV &JG &JE &GE DQG &VEf LV DQ DUGXRXV DOEHLW GRDEOHf WDVN \LHOGLQJ FXPEHUVRPH HTXDWLRQV 7KHVH HTXDWLRQV FDQ EH IRXQG LQ $SSHQGL[ ERWK IRU WKH JHQHUDO FDVH FRQVLGHULQJ DOO VHFRQGRUGHU HIIHFWV DV ZHOO DV FDVHVWXGLHV H[DPLQLQJ LQGLYLGXDO VHFRQGRUGHU HIIHFWV 5DWKHU WKDQ VKRZ WKHVH HTXDWLRQV KHUH WKH UHVXOWDQW SDUDPHWHUV DUH HYDOXDWHG QXPHULFDOO\ XVLQJ 0DWODE )LJXUHV DfGf VKRZ WKH HIIHFWV RI 4JV 5VXE HDFK RI WKH FDSDFLWDQFHVf§JDWHWRGUDLQ &JGf JDWHWRERG\ &JEf GUDLQWRERG\ &GEf DQG VRXUFHWRERG\ &VEff§DQG JPE RQ GHULYHG QRLVH SDUDPHn WHUV IRU D VLQJOH 026 WUDQVLVWRU ZLWK *,1 WXUQHG RQ DQG RII (DFK QRLVH SDUDPHWHU LV SORWn WHG YHUVXV WKH IROORZLQJ QRUPDOL]HG SDUDPHWHUV 4JV UDQJLQJ IURP WR ZLWK *,1 ERWK RQ DQG RII D ILUVWRUGHU HIIHFW LQFOXGLQJ ZLGWK GHSHQGHQFH RI GUDLQ QRLVHf 5VXE UDQJLQJ IURP WR DQG JPEJP &JE&JV &JG&JV DQG &GE VE&JV HDFK UDQJLQJ IURP WR 7KH IROORZLQJ DUH DVVXPHG \ If§ ? DQG \ )RU HDFK LQGLYLGXDO HIIHFW 9&' 7H[DPLQHG WKH RWKHU SDUDPHWHUV DUH KHOG FRQVWDQW DFFRUGLQJ WR WKH DVVXPSWLRQV OLVWHG LQ )LJXUH 7KH UHVXOWV IRU )PLQ SORWWHG LQ )LJXUH Df VKRZ D VWURQJ GHSHQGHQFH RQ HDFK RI WKH HIIHFWV FRQVLGHUHG )LUVW WR UHGXFH )PLQ 5VXE VKRXOG EH HLWKHU YHU\ VPDOO VKRUWf RU YHU\ ODUJH RSHQf 6XEVWUDWH UHVLVWDQFH FDQ EH PLQLPL]HG E\ VXUURXQGLQJ WKH WUDQVLVWRU ZLWK ODUJHDUHD VXEVWUDWH FRQWDFWV YHU\ FORVH WR WKH FKDQQHO 6HFRQG WR UHGXFH )PcQ WKH PHFKDQLVPV ZKLFK FRXSOH VXEVWUDWH QRLVH LQWR WKH WUDQVLVWRU LH JPE &JE DQG &GEf VKRXOG LQ JHQHUDO EH PLQLPL]HG ZKHUH &JE DQG &GE DUH OD\RXWGHSHQGHQW 7KLUG )PLQ GHFUHDVHV ZLWK LQFUHDVLQJ &JG 7KLV LV GXH WR IHHGEDFN WKURXJK &JG GHFUHDVLQJ WKH XQFRUn UHODWHG QRLVH UHVLVWDQFH 5X FDXVHG E\ *,1 )LQDOO\ VLQFH 4JV LV LQYHUVHO\ SURSRUWLRQDO WR GHYLFH ZLGWK WKH FKDQQHO WKHUPDO QRLVH DQG WKH FRXSOHG VXEVWUDWH QRLVH DUH UHGXFHG DV 4JV

PAGE 41

Df Ef Ff Gf $668037,216 \ OFO r!7 FR 56XE \ 4JV PE JP &JE &JV &JG &J6 &GE &JV )LJXUH 1RLVH SDUDPHWHUV Df )A Ef *Q Ff 5RSW DQG Gf ;A IRU WKH IROORZLQJ QRUPDOL]HG SDUDPHWHUV 4JV ZLWK *,1 RQ DQG RII 5VXE JPE &GE &JE DQG &JG )RU HDFK SDUDPHWHU VWXGLHG DOO RWKHU SDUDPHWHUV DUH KHOG FRQVWDQW XVLQJ WKH YDOXHV OLVWHG LQ WKH DVVXPSWLRQ WDEOH DQG *,1 LV RQ H[FHSW REYLRXVO\ IRU 4JV *,1 RII FDVHf LQFUHDVHV 7KHUHIRUH )PcQ GHFUHDVHV DFFRUGLQJO\ )RU YHU\ ORZ YDOXHV RI 4JV )PMQ GHFUHDVHV GXH WR WKH WUDQVLVWRU DQG LWV SDUDVLWLF FDSDFLWDQFHV EHLQJ VR ODUJH 6SHFLILFDOO\ &VE EHJLQV WR VKXQW WKH VXEVWUDWH QRLVH WR JURXQG GHFUHDVLQJ LWV FRQWULEXWLRQ WR )PLQ 7KH YDOXH RI 4JV IRU ZKLFK )PLQ LV PD[LPXP GHSHQGV RQ WKH YDOXH RI 5A

PAGE 42

2SWLPXP IRU 0LQLPXP 1RLVH :KLOH )PLQ LV VWURQJO\ GHSHQGHQW RQ PXOWLSOH HIIHFWV *,1 5VXE JPE &JG DQG &JEf *Q DQG ;RSW DUH RQO\ VWURQJO\ GHSHQGHQW RQ 4JV ZLWK *(1 RQ RU RII DQG 5RSW LV RQO\ VWURQJO\ GHSHQGHQW RQ 4JV ZLWK *,1 RQ 7KHUHIRUH RQO\ *,1 DQG 4JV QHHG WR EH FRQVLGn HUHG IRU *Q DQG =RSW :LWK WKLV VLPSOLILFDWLRQ D GHVLJQ SURFHGXUH IRU PLQLPXP QRLVH FDQ EH GHYHORSHG )LUVW )PLQ LV PLQLPL]HG ZLWK UHVSHFW WR 5VXE DQG HDFK RI WKH FDSDFLWDQFHV 6HFRQG WKH DFWXDO QRLVH IDFWRU )f LV PLQLPL]HG ZLWK UHVSHFW WR 4JV LQFOXGLQJ WKH HIIHFWV RI *,1 WKURXJK FKRRVLQJ WKH QRLVH PDWFKLQJ FRQGLWLRQ WR DSSUR[LPDWHO\ FRLQFLGH ZLWK WKH SRZHU PDWFKLQJ FRQGLWLRQ )LJXUH VKRZV )PLQ DQG ) YHUVXV 4JV ZLWK *,1 RQ +HUH 6MM LV FRQVWUDLQHG WR EH RU G% ZLWK !S/V VHW WR e &RQVWUDLQLQJ 6M M SUHn YHQWV =V IURP EHLQJ HTXDO WR ARSW UHVXOWLQJ LQ D KLJKHU QRLVH ILJXUH $V FDQ EH VHHQ IURP )LJXUH RSWLPL]LQJ ) ZLWK UHVSHFW WR 4JV LV PXFK PRUH YDOXDEOH WKDQ RSWLPL]LQJ )PLQ $OVR VLQFH )PLQ LV UHODWLYHO\ IODW FRPSDUHG WR ) RQO\ WKH TXDQWLW\ ))PLQ KDV WR EH )LJXUH 0LQLPXP QRLVH ILJXUH DQG DFWXDO QRLVH ILJXUH ZKHQ WKH LQSXW PDWFKLQJ LV FRQVWUDLQHG E\ 6Qr 7KH GLIIHUHQFH EHWZHHQ ) DQG )AQ LQGLFDWHV WKDW FRQVWUDLQLQJ 6X FDXVHV D QRLVH PLVPDWFK

PAGE 43

RSWLPL]HG )URP WKH UHVXOWV VKRZQ LQ )LJXUH LW FDQ EH VHHQ WKDW ))PLQ D IXQFWLRQ RI *Q =RSW DQG =Vf FDQ EH RSWLPL]HG FRQVLGHULQJ *,1 DQG 4JV RQO\ 7KH QRLVH SDUDPHWHUV RI D FDVFRGH 0M DQG 0f ZLWK DQG ZLWKRXW VRXUFH GHJHQHUn DWLRQ DUH VKRZQ LQ 7DEOH LQFOXGLQJ *(1 IRU 0M +HUH &JGO! JDWH DQG VXEVWUDWH 7DEOH 1RLVH 3DUDPHWHUV LQ ,PSHGDQFH )RUPDW &DVFRGH 0M DQG 0f &DVFRGH ZLWK /V D$E BI7L 9f}< UQ *Q f 7OAO8UEO 8L -84JV5V an ` 5RS .SW f 5 B$-L0B$-?F .rSW P F K a K AVV.6 pR/JV?E \ E f f ;RSc ;R3W GR/V f nL8f9FfZO )PLQ 5 PLQ f UHVLVWDQFH &JEO DQG JDWHLQGXFHG QRLVH RI 0 KDYH DOO EHHQ QHJOHFWHG 7KH QRLVH SDUDPn HWHUV IRU D VLQJOH WUDQVLVWRU XQGHU WKH VDPH DVVXPSWLRQV DUH GHULYHG LQ $SSHQGL[ VHFWLRQ 7KH IROORZLQJ YDULDEOHV KDYH EHHQ LQWURGXFHG f $ 9 \ f f E[ e $_F_ f E & $_F_$ f

PAGE 44

D f§ f tGR )URP ff DQG f LW FDQ EH VHHQ WKDW *Q LV LQYHUVHO\ SURSRUWLRQDO WR 4JV ZKLOH 5RSW DQG ;RSW DUH GLUHFWO\ SURSRUWLRQDO WR 4JV 7KLV DJUHHV YHU\ ZHOO ZLWK WKH QRLVH SDUDPHWHUV SORWWHG LQ )LJXUH 6RXUFH GHJHQHUDWLRQ PRGLILHV WKH QRLVH SDUDPHWHUV >+DU@ )RU WKH VSHFLILF FDVH FRUUHVSRQGLQJ WR QHJOHFWLQJ &JG DQG WUDQVLVWRU RXWSXW LPSHGDQFH WKH PRGHO XVHG WR FDOFXn ODWH WKH RULJLQDO QRLVH SDUDPHWHUVf DOO EXW ;RSW UHPDLQ XQFKDQJHG D EHQHILW RI XVLQJ LPSHGDQFH QRLVH SDUDPHWHUVf 7KHVH PRGLILHG QRLVH SDUDPHWHUV DUH DOVR VKRZQ LQ 7DEOH 7KURXJK VHWWLQJ e DQG WKH QRLVH SDUDPHWHUV IRU D VLQJOH WUDQVLVWRU H[FOXGn LQJ *,1 DQG 5VXE FDQ EH REWDLQHG ([DPLQLQJ ff DQG f LW FDQ EH VHHQ WKDW DV ZLWK WKH JDLQ DQDO\VLV R!S VKRXOG EH PD[LPL]HG WR UHGXFH QRLVH ILJXUH 7KHVH QRLVH SDUDPHWHUV IRU WKH VRXUFHGHJHQHUDWHG FDVFRGH DUH VXEVWLWXWHG LQWR f DORQJ ZLWK =V 5V MFR/J \LHOGLQJ WKH IROORZLQJ )) PLQ f ZKHUH 4O pR/J /6f 5F f %\ LQVSHFWLRQ RI f VHWWLQJ 4/ 4JV FRUUHVSRQGLQJ WR VHULHV UHVRQDQFH DW WKH LQSXWf GRHV QRW UHVXOW LQ PLQLPXP QRLVH WKXV WKHUH LV D WUDGHRII EHWZHHQ WKH QRLVH PDWFKLQJ DQG WKH SRZHU PDWFKLQJ FRQGLWLRQV )RU D JLYHQ YDOXH RI 6M` DQG DVVXPLQJ R![/V 5 WKH IROORZLQJ UHTXLUHG YDOXH RI 4O IRU PLQLPXP QRLVH FDQ EH REWDLQHG WKURXJK WKH DLG RI f

PAGE 45

f 7KXV WKH LQSXW VKRXOG EH GHVLJQHG VXFK WKDW WKH VHULHV UHVRQDQFH RFFXUV DW D IUHTXHQF\ KLJKHU WKDQ WKH RSHUDWLQJ IUHTXHQF\ 1RWH WKDW IRU 6Q 4/ 4JV DV H[SHFWHG 6XEVWLWXWLQJ f LQWR f WDNLQJ WKH SDUWLDO GHULYDWLYH ZLWK UHVSHFW WR 4JV DQG VHWWLQJ WKH UHVXOW WR ]HUR \LHOGV WKH RSWLPXP 4JV IRU PLQLPXP ))PLQ DV IROORZV f )RU WKH DVVXPSWLRQV OLVWHG LQ )LJXUH ZLWK 6X G% DQG VHWWLQJ W O WKHQ 4rJSVf HTXDOV DQG 4/ LV )RU G% 4rSW LV DQG 4/ LV ZKLOH IRU 6ML rR G% FRUUHVSRQGLQJ WR D SHUIHFW LQSXW PDWFK 4rJ3Vf 4O $W WKHVH RSWLPDO 4JV YDOXHV ))PLQ LV DQG IRU 6Q DQG m G% UHVSHFWLYHO\ ZKHQ FRQVLGHULQJ *,1 RQO\ LH ff :KHQ FRQVLGHULQJ DOO RI WKH VHFRQGRUGHU HIIHFWV LH DV LQ )LJXUH f ))PLQ LV DQG UHVSHFWLYHO\ 7KHUHIRUH PDWFKLQJ WKH LQSXW IRU SHUIHFW VHULHV UHVRQDQFH UHVXOWV LQ D KLJKHU 1) DV LV WKH FDVH IRU >6KD@ 2SWLPXP :LGWK RI 0R 7KH RSWLPXP VL]LQJ RI 0 FDQ EH REWDLQHG E\ PLQLPL]LQJ e:f 6LQFH JP DQG JGR DUH GLUHFWO\ SURSRUWLRQDO WR : ZKLOH FGO KDV FRPSRQHQWV GLUHFWO\ SURSRUWLRQDO WR :M DQG WR : WKHQ W:f KDV FRPSRQHQWV SURSRUWLRQDO WR DQG LQYHUVHO\ SURSRUWLRQDO WR 7KH LPSDFW LQ GHFLEHOV GHSHQGV RQ WKH SDUWLFXODU YDOXH RI )PLQ )RU )PcQ G% WKH FRUn UHVSRQGLQJ QRLVH ILJXUHV DUH DQG G% UHVSHFWLYHO\ )RU )MMMM G% WKH FRUUHVSRQGLQJ QRLVH ILJXUHV DUH DQG G% UHVSHFWLYHO\

PAGE 46

: 7KXV DQ RSWLPXP YDOXH RI : H[LVWV ZKLFK PLQLPL]HV & 7KH FDSDFLWDQFH DW WKH GUDLQ RI 0@ &GOf FDQ EH ZULWWHQ DV IROORZV QHJOHFWLQJ LQWHUFRQQHFW FDSDFLWDQFH &G? &G??: A &G?An f ZKHUH &GOO:@ FRQVLVWV RI &GEO SOXV D IUDFWLRQ RI &JGO DQG &GO: FRQVLVWV RI &VE DQG &JV 'LIIHUHQWLDWLQJ f ZLWK UHVSHFW WR : DVVXPLQJ FRQVWDQW GUDLQ FXUUHQW DQG FRQn VWDQW 9GVO \LHOGV WKH IROORZLQJ : A9L f 7KXV WKH FDSDFLWDQFH DW WKH GUDLQQRGH RI 0@ VKRXOG EH HTXDOO\ GLYLGHG EHWZHHQ 0M DQG 0 )LJXUH VKRZV D SORW RI VLPXODWHG QRLVH IDFWRU YHUVXV : IRU D _LP WHFKQROn RJ\ )RU WKLV VLPXODWLRQ WKH GHIDXOW QRLVH PRGHO IRU %6,0Y LV EHLQJ XVHG 7KH RSWLPXP ZLGWK RI 0 LV _LP ZKHQ : M LV VHW WR _OP &DOFXODWLRQV VKRZ WKDW IRU WKLV VDPH WHFKn QRORJ\ WKH UDWLR &GOO&GL LV HTXDO WR \LHOGLQJ D FDOFXODWHG RSWLPDO : RI LP )LJXUH 1RLVH IDFWRU YHUVXV ZLGWK RI FDVFRGHG WUDQVLVWRU 0f 7KH RSWLPXP ZLGWK RFFXUV ZKHQ WKH FDSDFLWDQFH DW WKH LQWHUPHGLDWH QRGH EHWZHHQ 0c DQG 0 LV HTXDOO\ EDODQFHG EHWZHHQ 0c DQG 0

PAGE 47

7KXV WKH VLPXODWHG DQG FDOFXODWHG RSWLPDO YDOXHV IRU : DJUHH UHDVRQDEO\ ZHOO $OVR DV FDQ EH VHHQ IURP )LJXUH FKRRVLQJ ::M UDQJLQJ IURP WR ZLOO UHVXOW LQ PLQLPDO QRLVH FRQWULEXWLRQ IURP 0 7KHUHIRUH WKH QRLVH SDUDPHWHUV \LHOG WKH RSWLPXP YDOXHV IRU :M DQG : WKURXJK HTXDWLRQV f DQG f UHVSHFWLYHO\ DV ZHOO DV WKH RSWLPXP JDWH LQGXFWDQFH IRU D JLYHQ LQSXW UHIOHFWLRQ FRHIILFLHQW WKURXJK HTXDWLRQ f 'HVLJQ 0HWKRGRORJLHV IRU 6RXUFH'HJHQHUDWHG /1$ 'HULYDWLRQ%DVHG 0HWKRGRORJ\ 1RZ D FRPSOHWH GHVLJQ PHWKRGRORJ\ EDVHG RQ WKH SUHYLRXV GHULYDWLRQV KDV EHHQ GHYHORSHG IRU WKH VRXUFHGHJHQHUDWHG &026 /1$ DVVXPLQJ D JLYHQ WHFKQRORJ\ )LUVW FKRRVH 4JV WR EH QRPLQDOO\ 7KLV YDOXH VKRXOG EH WKH PLQLPXP RI WKRVH REWDLQHG IURP f DQG f \LHOGLQJ PLQLPXP QRLVH DQG UREXVWQHVV DJDLQVW FRPSRQHQW YDULDWLRQV 2QFH 4JV LV GHILQHG :M FDQ EH GHWHUPLQHG EDVHG RQ &R[ 6HFRQG XVLQJ WKH JDLQ DQG QRLVH HTXDWLRQV DORQJ ZLWK WKH GHVLUHG SRZHU FRQVXPSWLRQ HVWLPDWH WKH UHTXLUHG FR\ \LHOGLQJ DQ DSSUR[LPDWH RSHUDWLQJ SRLQW IRU 0M 7KLUG GHWHUPLQH : XVLQJ f )RXUWK VHW /V QRPLQDOO\ WR W2[ ZKHUH ORZHU YDOXHV FDQ EH XVHG WR VOLJKWO\ ERRVW WKH JDLQ ZKLOH VDFULILFLQJ LQSXW PDWFKLQJ UREXVWQHVV )LIWK GHWHUPLQH /J XVLQJ f DQG f EDVHG RQ WKH GHVLUHG YDOXH IRU 6MM 6L[WK XVLQJ JDLQ HTXDWLRQ f HVWLPDWH WKH YDOXH RI /G QHHGHG WR DFKLHYH D GHVLUHG YDOXH IRU 6 QRWH QO FDQ EH HVWLPDWHG WR EH af 6HYHQWK H[WUDFW SDUDVLWLFV RI /G DQG FDVFRGH DQG PDWFK WKH RXWSXW WR e XVLQJ DQ\ VWDQGDUG LGHDO PDWFKLQJ WHFKQLTXH GHWHUPLQLQJ YDOXHV IRU &M DQG &

PAGE 48

Df 5V Lf§:9 ,QSXW 0DWFKLQJ 1HWZRUN U 2XWSXW 0DWFKLQJ 1HWZRUN 5 )LJXUH ,OOXVWUDWLRQ RI FRQVWDQW DYDLODEOH JDLQ PHWKRGV ZKHUH IRU WKH JDLQ VWDJH HLWKHU Df D VLQJOH WUDQVLVWRU LV XVHG RU Ef D VRXUFHGHJHQHUDWHG FDVFRGH ZLWK LQGXFWLYH ORDG LV XVHG $OWHUQDWLYH &RQVWDQW $YDLODEOH *DLQ 0HWKRGRORJ\ 7KH DERYH GHVLJQ PHWKRGRORJ\ LV VHQVLWLYH WR WKH GHULYHG QRLVH SDUDPHWHUV $V 63,&( LQWULQVLF QRLVH PRGHOV FRQWLQXH WR LPSURYH LW LV HDVLHU DQG fPRUHf DFFXUDWH WR XVH &$' WRROV WR REWDLQ WKH QRLVH SDUDPHWHUV RI DQ LQWULQVLF PRGHO ZLWK DGGLWLRQDO H[WULQVLF SDUDVLWLFV DOVR PRGHOHG $ PHWKRGRORJ\ IRU FKRRVLQJ /J 4 DQG & ZHOOVXLWHG WR &$' LV WKH FRQVWDQW DYDLODEOH JDLQ PHWKRG 7UDGLWLRQDOO\ WKH FRQVWDQW DYDLODEOH JDLQ PHWKRG LV XVHG WR JHQHUDWH LGHDO PDWFKLQJ QHWZRUNV ZKHQ JLYHQ WKH 6SDUDPHWHUV DQG QRLVH SDUDPn HWHUV RI D VLQJOH WUDQVLVWRU >*RQ@ DV LOOXVWUDWHG LQ )LJXUH Df +RZHYHU VLQFH ORVV\ LQGXFWRUV DUH XVHG DW ERWK WKH GUDLQ DQG WKH VRXUFH D FRPSRVLWH FLUFXLW LQFOXGLQJ /V /G DQG WKH FDVFRGH LV XVHG LQ SODFH RI WKH DFWLYH GHYLFH DV LOOXVWUDWHG LQ )LJXUH Ef 9DOn XHV IRU :M :! /V DQG /G DUH FKRVHQ XVLQJ WKH SUHYLRXV PHWKRGRORJ\ 7KH 6SDUDPHWHUV DQG QRLVH SDUDPHWHUV RI WKH FRPSRVLWH FLUFXLW LQFOXGLQJ WKH LQGXFWRUV DUH H[WUDFWHG XVLQJ

PAGE 49

UV 3 1) &LUFOHV )LJXUH &RQVWDQW DYDLODEOH JDLQ DQG 1) FLUFOHV IRU FDVFRGH ZLWK /V DQG /G 63,&( 8VLQJ WKHVH SDUDPHWHUV FRQVWDQW DYDLODEOH JDLQ *$f FLUFOHV DQG FRQVWDQW QRLVH FLUFOHV DUH SORWWHG RQ WKH 6PLWK FKDUW DORQJ ZLWK LQSXW DQG RXWSXW VWDELOLW\ FLUFOHV 8VLQJ WKHVH FLUFOHV WKH UHIOHFWLRQ FRHIILFLHQW RI WKH VRXUFH DQG WKH ORDG 7V DQG 7/ FDQ EH GHWHUPLQHG 2QFH 7V DQG 7/ DUH GHWHUPLQHG WKH LQSXW DQG RXWSXW PDWFKLQJ QHWZRUNV FDQ UHDGLO\ EH GHVLJQHG )LJXUH VKRZV D SORW RI VLPXODWHG FRQVWDQW JDLQ DQG QRLVH FLUFOHV RQ WKH 6PLWK FKDUW $VVXPLQJ 4JV LV FKRVHQ XVLQJ f 5RSW VKRXOG EH FORVH WR WKH 4 FLUFOH 7KH YDOXH RI /J FORVHVW WR URSW ZKLOH VWLOO PHHWLQJ WKH GHVLUHG 6Q VSHFLILFDWLRQ FDQ WKHQ EH UHDG RII WKH 6PLWK FKDUW 7KH RXWSXW LV WKHQ FRQMXJDWHO\ PDWFKHG LH 7O f§ 7RXW f XVLQJ WKH IROORZLQJ IRUPXOD 6LOWnVn f 7KLV GHWHUPLQHV WKH RXWSXW PDWFKLQJ QHWZRUN WKH VLPSOHVW RI ZKLFK LV D FDSDFLWLYH WUDQVn IRUPHU &M DQG & 7R REWDLQ &M D FRQVWDQW FRQGXFWDQFH FLUFOH LV WUDYHUVHG IURP WKH FHQWHU RI WKH 6PLWK FKDUW ZKLOH WR REWDLQ & D FRQVWDQW UHVLVWDQFH FLUFOH LV WUDYHUVHG IURP WKH HQGSRLQW RI WKH SUHYLRXV SDWK WR A/f§ARXW DV VKRZQ LQ )LJXUH

PAGE 50

)LJXUH 2XWSXW PDWFKLQJ QHWZRUN GHVLJQ WR PDWFK 4 WR URXW f 6XPPDU\ &RPPRQJDWH DQG VRXUFHGHJHQHUDWHG /1$V KDYH EHHQ SUHVHQWHG $Q H[SOLFLW GHVLJQ PHWKRGRORJ\ KDV EHHQ GHYHORSHG IRU D VRXUFHGHJHQHUDWHG &026 /1$ FRQVLGHUn LQJ JDLQ 1) LQSXW PDWFKLQJ DQG RXWSXW PDWFKLQJ ,Q SDUWLFXODU WKH RSWLPDO VL]LQJ RI WKH WUDQVLVWRUV KDV EHHQ SUHVHQWHG IRU UREXVWQHVV DJDLQVW LQSXW PDWFKLQJ YDULDWLRQV DQG IRU PLQLPDO QRLVH ILJXUH 7KH QRLVH SDUDPHWHUV LQ LPSHGDQFH IRUPf IRU D 026)(7 KDYH EHHQ GHULYHG WR XQGHUVWDQG WKH HIIHFW RI YDULRXV VHFRQGRUGHU HIIHFWV RQ QRLVH SHUIRUPDQFH *DWHLQGXFHG QRLVH DQG VXEVWUDWH UHVLVWDQFH ERWK DIIHFW WKH QRLVH SDUDPHWHUV VLJQLILFDQWO\ 8VLQJ WKHVH UHVXOWV D GHVLJQ PHWKRGRORJ\ IRU PLQLPXP QRLVH ZKLOH FRQn VWUDLQLQJ WKH LQSXW PDWFKLQJ FRQGLWLRQ KDV EHHQ SUHVHQWHG ZKLFK \LHOGV DQ RSWLPDO YDOXH IRU WKH VL]H RI WKH SULPDU\ WUDQVLVWRU LQ WKH FDVFRGH $Q RSWLPXP VL]LQJ IRU WKH FDVFRGHG GHYLFH ZDV REWDLQHG WKURXJK PLQLPL]DWLRQ RI QRLVH ILJXUH )LQDOO\ DQ DOWHUQDWLYH GHVLJQ PHWKRGRORJ\ EDVHG RQ WKH FRQVWDQW DYDLODEOH JDLQ PHWKRG KDV EHHQ SUHVHQWHG

PAGE 51

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f LW FDQ HDVLO\ EH VKRZQ WKDW WKH VHULHV UHVLVWDQFH RI /J U/Jf LQFUHDVHV ) E\ WKH TXDQn WLW\ U/J5V )RU WKHVH UHDVRQV KLJK TXDOLW\IDFWRU LQGXFWRUV DUH UHTXLUHG IRU /J DQG /G 'HSHQGLQJ RQ WKH LQGXFWDQFH DQG WKH RSHUDWLQJ IUHTXHQF\ LW PD\ QRW EH SUDFWLFDO WR LQWHn JUDWH /J ZKLFK LV WKH FDVH IRU D 0+] /1$

PAGE 52

)LJXUH Df *HQHUDO WZRSRUW OXPSHGHOHPHQW PRGHO IRU VSLUDO LQGXFWRU Ef ,OOXVWUDWLRQ RI D VSLUDO LQGXFWRU DERYH D SDWWHUQHGJURXQG VKLHOG 7KH VKLHOG WHUPLQDWHV PRVW RI (ILHOG ZKLOH DOORZLQJ %ILHOG WR SDVV WKURXJK LQFUHDVLQJ 4 $ WZRSRUW OXPSHGHOHPHQW PRGHO IRU D JHQHUDO VSLUDO LQGXFWRU LV VKRZQ LQ )LJXUH Df 8VLQJ WKLV PRGHO WKH TXDOLW\ IDFWRU 4f IRU DQ LQGXFWRU ZLWK RQH SRUW VKRUWHG WR JURXQG FDQ EH ZULWWHQ DV IROORZV >/RQ@ ZKHUH VXEVWUDWH FDSDFLWDQFH &VXEf DQG SRUWWRSRUW FDSDFLWDQFH &Vf KDYH EHHQ QHJOHFWHG 4 LfF!/f f 7KLV DFFRXQWV IRU ORVV LQ ERWK WKH VHULHV UHVLVWDQFH RI WKH VSLUDO DQG LQ WKH FRQGXFWLYH VLOLn FRQ VXEVWUDWH 1RWH WKDW WKH LQWHUDFWLRQ RI WKH VXEVWUDWH ZLWK WKH PDJQHWLF ILHOG LV QRW FDSn WXUHG ZLWK WKLV HTXDWLRQ LH HGG\FXUUHQW HIIHFWVf 7R PD[LPL]H 4 WKH VHULHV UHVLVWDQFH UVf VKRXOG EH PLQLPL]HG R[LGH FDSDFLWDQFH &R[f VKRXOG EH PLQLPL]HG DQG WKH VXEVWUDWH UHVLVWDQFH 5VXf VKRXOG HLWKHU EH YHU\ VPDOO RU YHU\ ODUJH 7KLV WUHQG IRU 5VXE LV WKH VDPH DV WKDW VHHQ LQ WKH QRLVHILJXUH DQDO\VLV ,Q IDFW WKLV LV D JHQHUDO WUHQG ZKLFK FDQ EH DSSOLHG WR DQ\ 5) FLUFXLW RU FRPSRQHQW >)OR22E@

PAGE 53

2QH ZD\ WR PLQLPL]H WKH VXEVWUDWH UHVLVWDQFH RI WKH LQGXFWRU LV WR XVH SDWWHUQHG JURXQG VKLHOGV >)OR22E@ ,Q WKH FRPPRQO\ XVHG VXEVWUDWH UHVLVWLYLW\ UDQJH WKH VXEn VWUDWH ORVV LQFUHDVHV ZLWK GHFUHDVLQJ UHVLVWLYLW\ 7KLV HIIHFW VKRZV XS DV DQ LQFUHDVH LQ WKH VHULHV UHVLVWDQFH UVf LQ WKH LQGXFWRU PRGHO 7KLV HIIHFW LV UHGXFHG IRU VPDOOHUDUHD LQGXFn WRUV VLQFH WKH PDJQHWLF ILHOG GRHV QRW SHQHWUDWH DV GHHSO\ LQWR WKH VXEVWUDWH WKHUHE\ UHGXFLQJ WKH HGG\ FXUUHQWV 7KH QHW UHVXOW LV DQ LQFUHDVH LQ LQGXFWRU TXDOLW\ IDFWRU DW WKH FRVW RI LQFUHDVHG SDUDVLWLF FDSDFLWDQFH RU GHFUHDVHG VHOIUHVRQDQW IUHTXHQF\ &DSDFLWRU LPSOHPHQWDWLRQ &DSDFLWRUV &M DQG & LQ )LJXUH DV ZHOO DV WKH E\SDVV FDSDFLWRUV IRU ELDV YROWn DJHV DUH LPSOHPHQWHG XVLQJ DFFXPXODWLRQPRGH 026 FDSDFLWRUV >+XQ@ 7KHVH FDSDFLn WRUV DFKLHYH KLJK 4IDFWRUV DQG DUH FRPSDWLEOH ZLWK VWDQGDUG GLJLWDO &026 WHFKQRORJLHV 7KH RXWSXW GF YROWDJH DFURVV &M LV 9 FRQVLVWHQW ZLWK GULYLQJ DQ LPDJHUHMHFWLRQ ILOWHU $ FRQFHUQ ZLWK XVLQJ WKLV FDSDFLWRU VWUXFWXUH LV WKH YROWDJHGHSHQGHQW FDSDFLWDQFH DQG LWV HIIHFW RQ 3OG% DQG ,3 SHUIRUPDQFH +RZHYHU PHDVXUHPHQWV VKRZ WKDW 3MG% DQG ,3 YDU\ E\ OHVV WKDQ G% IRU WKH RXWSXW GF YROWDJH UDQJLQJ IURP JURXQG WR 9GG

PAGE 54

9GG 9 )LJXUH 6FKHPDWLF RI 0+] VRXUFHGHJHQHUDWHG &026 /1$ LQFOXGLQJ RQFKLS DQG SDFNDJH SDUDVLWLFV $ 0+] SP &026 /1$ &LUFXLW ,PSOHPHQWDWLRQ 7R YHULI\ WKH GHVLJQ PHWKRGRORJ\ IRU &026 VRXUFHGHJHQHUDWHG /1$V D 0+] YHUVLRQ ZDV LPSOHPHQWHG LQ D SP &026 WHFKQRORJ\ ZLWK PHWDO OD\HUV 7KLV WHFKQRORJ\ ZDV REWDLQHG WKURXJK 026,6 )LJXUH VKRZV WKH VFKHPDWLF RI WKH /1$ 7KH LPSRUWDQW RQFKLS DQG SDFNDJH SDUDVLWLFV DUH VKRZQ LQFOXGLQJ VHULHV UHVLVn WDQFHV IRU LQGXFWRUV DQG FDSDFLWRUV UVf SDG FDSDFLWDQFH &MAf JDWH UHVLVWDQFH UJf LQGXFn WRU SDUDVLWLF FDSDFLWDQFH FR[f DQG ERQGZLUHJURXQG LQGXFWDQFH /A/JAf %RWK ELDV YROWDJHV 9JO DQG 9J DUH JHQHUDWHG RII FKLS IRU IOH[LELOLW\ LQ WHVWLQJ )LJXUH Df VKRZV D GLH SKRWRJUDSK RI WKH 0+] /1$ >)ORD@ 7KH FLUFXLW DUHD LV [ ILP 7KLV

PAGE 55

Df )LJXUH 'LH SKRWRJUDSKV RI Df SP 0+] VLQJOHHQGHG &026 /1$ DQG Ef SP *+] GLIIHUHQWLDO &026 /1$ /1$ ZDV SDFNDJHG WKHUHIRUH JROG ERQGZLUHV DUH XVHG WR LPSOHPHQW /V DV ZHOO DV WR LPSURYH WKH TXDOLW\ IDFWRU RI /G 7KH RQFKLS SRUWLRQ RI /G FRQVLVWHG RI D JURXQGVKLHOGHG WXP Q+ VSLUDO LQGXFWRU ZLWK D TXDOLW\ IDFWRU RI DW 0+] ,W ZDV LPSOHn PHQWHG XVLQJ PHWDOV DQG VKXQWHG WRJHWKHU 7KH ERQGZLUH DQG SLQ DGGHG DQ DGGLWLRQDO Q+ WR /G )RU WKLV RSHUDWLQJ IUHTXHQF\ DQG WHFKQRORJ\ /V LV a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

PAGE 56

0HDVXUHG 5HVXOWV 7KH /1$ ZDV FKDUDFWHUL]HG DW IRXU GLIIHUHQW RSHUDWLQJ SRLQWV WR VWXG\ WKH WUDGHRII EHWZHHQ SRZHU FRQVXPSWLRQ DQG 1) )LJXUH Df VKRZV D SORW RI PHDVXUHG JDLQ DQG 1) YHUVXV ELDV FXUUHQW 2SHUDWLQJ SRLQWV WR FRUUHVSRQG WR PLQLPXP WR PD[LPXP 1) DQG PD[LPXP WR PLQLPXP SRZHU UHVSHFWLYHO\ $W P$ RI FXUUHQW RSHUDWLQJ SRLQW f WKH /1$ DFKLHYHV D 1) RI G% ZKLFK LV FRPSHWLWLYH ZLWK PRVW *D$V DQG ELSRODU /1$V +RZHYHU WKH SRZHU FRQVXPSWLRQ LV P: ZKLFK LV TXLWH ODUJH $V WKH SRZHU FRQVXPSn WLRQ LV GHFUHDVHG WR P: RSHUDWLQJ SRLQW f WKH 1) RQO\ LQFUHDVHV WR G% 7KLV UHVXOW PHHWV ERWK RI WKH GHVLUHG SHUIRUPDQFH PHWULFV IRU &026 /1$V 1) G% DQG SRZHU P:f 0HDVXUHPHQWV VKRZ WKDW WR DFKLHYH D 1) RI G% RQO\ P: LV UHTXLUHG 2SHUDWLQJ SRLQWV DQG KDYH FRPSDUDEOH 1)fV WR >6WX@ DQG >+XD@ EXW DW RQO\ D IUDFWLRQ RI WKHLU SRZHU 0XFK RI WKLV SRZHU VDYLQJV FRPHV IURP WKH DELOLW\ WR GULYH 2SHUDWLQJ 3RLQW Df Ef )LJXUH Df 0HDVXUHG JDLQ 6Lf DQG QRLVH ILJXUH 1)f YHUVXV ELDV FXUUHQW )RXU GLIIHUHQW RSHUDWLQJ SRLQWV DUH VKRZQ QXPEHUHG WR FRUUHVSRQGLQJ WR PLQLPXP WR PD[LPXP 1) DQG PD[LPXP WR PLQLPXP FXUUHQW UHVSHFWLYHO\ Ef *DLQ DQG 1) YHUVXV IUHTXHQF\ IRU RSHUDWLQJ SRLQWV t 3f /6

PAGE 57

4 ZLWKRXW DQ DGGLWLRQDO RXWSXW EXIIHU )LJXUH Ef VKRZV D SORW RI WKH WUDQVGXFHU JDLQ 6f DQG 1) YHUVXV IUHTXHQF\ IRU WKH /1$ IRU RSHUDWLQJ SRLQWV DQG $W 0+] WKH PHDVXUHG WUDQVGXFHU JDLQV 6f DUH DQG G% DW DQG P: UHVSHFWLYHO\ )LJXUHV Df DQG Ef VKRZ WKH LQSXW DQG RXWSXW UHIOHFWLRQ FRHIILFLHQWV 6Q DQG 6f IRU RSHUDWLQJ SRLQWV DQG 7KH UHODWLYHO\ KLJK 6AfV DUH GXH WR WKH ERQGZLUH LQGXFn WDQFH EHLQJ WRR VPDOO JHQHUDWLQJ RQO\ a 4 RI UHDO LPSHGDQFH 7R LPSURYH WKH LQSXW PDWFK WR XQGHU G% /V VKRXOG EH VOLJKWO\ LQFUHDVHG ZKLOH /J FDQ EH GHFUHDVHG WR PDLQn WDLQ WKH QRLVH PDWFK 6 LV EHORZ G% RYHU D 0+] VSDQ 7KH YHU\ ORZ 6 FRUUHn VSRQGV WR D FRQMXJDWH PDWFK DW WKH RXWSXW 7R PHDVXUH )PLQ WKH PHDVXUHPHQW VHWXS VKRZQ LQ )LJXUH Df ZDV XVHG ([WHUn QDO WXQHUV DUH LQVHUWHG DW WKH LQSXW DQG RXWSXW RI WKH SDFNDJHG /1$ DQG WKH LPSHGDQFHV RI WKHVH WXQHUV DUH DGMXVWHG XQWLO WKH RSWLPXP QRLVH PDWFKLQJ FRQGLWLRQ LH URSW RU =RSWf RFFXUV FRUUHVSRQGLQJ WR )PLQ )LJXUH Ef VKRZV WKH RSWLPXP VRXUFHUHIOHFWLRQ FRHIILn FLHQW URSWf YHUVXV IUHTXHQF\ IRU RSHUDWLQJ SRLQWV DQG SORWWHG RQ 6PLWK FKDUWV )LJXUH )LJXUH Df ,QSXW 6Xf DQG Ef 2XWSXW 6f UHIOHFWLRQ FRHIILFLHQWV YHUVXV IUHTXHQF\ RSHUDWLQJ SRLQWV DQG

PAGE 58

1RLVH 3DUDPHWHU 0HDVXUHPHQW 6HWXS Df Ef &4 ‘' ( /Ln WHf 1RLVH 6RXUFH %LDV 7HH 7XQHU '87 :/Jf 7XQHU %LDV 7HH 1RLVH )LJXUH r 0HDV 6\VWHP aU '& '& 7XQHU &RQWURO URSW 2SHUDWLQJ 3RLQW URSW 2SHUDWLQJ 3RLQW 2S 3RLQW 2S 3RLQW ‘ ‘f§ f§‘ LL LU rf§ ‘ f§‘f§ ‘ r f f§f Q ff§ f§f ff§ f f ‘ )UHTXHQF\ *+]f )LJXUH Df 0HDVXUHPHQW V\VWHP XVHG WR REWDLQ )PLQ DQG 3RSW Ef 2SWLPXP VRXUFH UHIOHFWLRQ FRHIILFLHQW IRU PLQLPXP QRLVH URSWf Ff 0LQLPXP QRLVH ILJXUH YHUVXV IUHTXHQF\ 7DEOH 0HDVXUHG QRLVH SDUDPHWHUV DW 0+] IRU SP &026 /1$ 2SHUDWLQJ &XUUHQW ) f $ PP U )6Q RRG%f )6Q G%f 3RLQW P$f G%f RSW G%f G%f = r = r O= r = r Ff VKRZV WKH PHDVXUHG )PLQ YHUVXV IUHTXHQF\ IRU RSHUDWLQJ SRLQWV DQG DORQJ ZLWK D FXUYH ILW WR WKH PHDVXUHG GDWD 7DEOH VKRZV WKH PHDVXUHG QRLVH SDUDPHWHUV H[FOXGn LQJ 5Qf IRU HDFK RSHUDWLQJ SRLQW $W 0+] )PLQ LV DQG G% IRU RSHUDWLQJ

PAGE 59

SRLQWV DQG UHVSHFWLYHO\ 7R PDWFK WKH LQSXW WR REWDLQ )PLQ /J VKRXOG EH LQFUHDVHG DQG DQ RIIFKLS FDSDFLWRU VKRXOG EH DGGHG 7KLV WUDQVIRUPV WKH  VRXUFH LPSHGDQFH WR =RSW HTXLYDOHQW WR UHIOHFWLRQ FRHIILFLHQW URSWf +RZHYHU 7RSW IRU WKH /1$ LV QRW HTXDO WR 6M Mr WKHUHIRUH WKH QRLVH PDWFKLQJ FRQGLWLRQ DQG WKH SRZHU PDWFKLQJ FRQGLWLRQ GR QRW FRLQFLGH 7DEOH DOVR VKRZV WKH 1) FRUUHVSRQGLQJ WR SHUIHFW LQSXW SRZHU PDWFKLQJ LH 6M M rrG%f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n DJH ZLWK ORZ JURXQGLQGXFWDQFH JURXQGVKLHOGHG SDG VWUXFWXUHV D JURXQGVKLHOGHG LQGXFn WRU DQG D FDVFRGH DPSOLILHU 7KH PHDVXUHG 3OG% DQG ,3 GDWD IRU HDFK RSHUDWLQJ SRLQW DUH VKRZQ LQ )LJXUH $V FDQ EH VHHQ WKH SRZHU RI WKH LQWHUPRGXODWLRQ SURGXFW 3ILRf XQFKDQJHG E\ WKH FXUUHQW OHYHO LQ WKH /1$ +RZHYHU WKH RXWSXW SRZHU RI WKH IXQGDPHQWDO 3IOf GHSHQGV RQ WKH RSHUDWLQJ SRZHU JDLQ DW WKH VSHFLILF FXUUHQW OHYHO 7KHUHIRUH WKH ,,3 UHODWLYHO\ VFDOHV LQ DFFRUGDQFH ZLWK WKH JDLQ 7KH RXWSXW SRZHU OHYHO DW ZKLFK WKH /1$ FRPSUHVVHV LV D IXQFWLRQ RI WKH VXSSO\ YROWDJH ZKLFK LV 9 IRU RSHUDWLQJ SRLQWV DQG 9 IRU RSHUDWLQJ SRLQW 7KHUHIRUH WKH 3OG% RXWSXWf LV DSSUR[LPDWHO\ WKH VDPH IRU HDFK RSHUDWLQJ SRLQW ZKLOH 3OG% LQSXWf VFDOHV ZLWK WKH JDLQ 7KH PHDVXUHG +3fV DW P: DQG P: DUH G%P DQG G%P 'XH WR WKH XVH RI D VLQJOHVWDJH WRSRORJ\ WKH DPSOLILHU LV YHU\ OLQHDU DQG LV VXLWDEOH IRU *60 DSSOLFDWLRQV

PAGE 60

7DEOH 6XPPDU\ RI PHDVXUHG FKDUDFWHULVWLFV IRU 0+] &026 /1$ I 0+] 2S 3RLQW 2S 3RLQW 2S 3RLQW 2S 3RLQW 3RZHU P:f *DLQ G%f 1RLVH )LJXUH G%f )PLQ G%f )6Q G%f G%f 9''9f 6Q G%f 6 G%f G%f ,,3 G%Pf SOG% LQSXWf G%Pf 6XPPDU\ IRU 0+] /1$ $ VXPPDU\ RI WKH PHDVXUHG FKDUDFWHULVWLFV IRU HDFK RSHUDWLQJ SRLQW LV VKRZQ LQ 7DEOH 2SHUDWLQJ SRLQW KDV DQ )PLQ ZKLFK LV G% KLJKHU WKDQ D 0+] &026

PAGE 61

/1$ LPSOHPHQWHG LQ D OHYHOPHWDO _LP %L&026 SURFHVV >*UD22@ ZKLFK LV FXUn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a *+] 7KLV /1$ ZDV DOVR LPSOHPHQWHG ZLWK DQ HQWLUH FORFN UHFHLYHU DV ZLOO EH SUHVHQWHG LQ &KDSWHU 7KH WHFKQRORJ\ SURYLGHV PHWDO OD\HUV DQG ERWK Sn a *FPf DQG S ZLWK HSLn WD[LDO OD\HU *FPf VXEVWUDWHV DOORZLQJ WKH HIIHFWV RI VXEVWUDWH UHVLVWLYLW\ RQ /1$ SHUIRUPDQFH WR EH VWXGLHG 7KLV WHFKQRORJ\ ZDV REWDLQHG RQFH DJDLQ WKURXJK 026,6 )LJXUH Ef VKRZV D GLH SKRWRJUDSK RI WKH *+] GLIIHUHQWLDO /1$ ZKLOH )LJXUH VKRZV D VFKHPDWLF RI WKH GLIIHUHQWLDO /1$ LQFOXGLQJ LPSRUWDQW SDUDVLWLFV SDG FDSDFLn WDQFHV DUH LPSOLHG IRU HDFK ODEHOOHG LQSXWf 7KH GLH VL]H LV [ ILP $OO WKUHH LQGXFWRUV /G /J DQG /V DUH LQWHJUDWHG DV ZHOO DV S) E\SDVV FDSDFLWRUV 6\PPHWU\ ZDV PDLQWDLQHG WRSWRERWWRP IRU WKH GLIIHUHQWLDO OD\RXW WR DYRLG DQ\ V\VWHPDWLF HUURUV IURP OD\RXW PLVPDWFK $JDLQ JURXQGVKLHOGHG SDGV DUH XVHG WKURXJKRXW WKH OD\RXW

PAGE 62

/ Q+ )LJXUH 6FKHPDWLF RI *+] IXOO\ GLIIHUHQWLDO /1$ VKRZLQJ SDUDVLWLFV 7KLV /1$ LV D IXOO\ GLIIHUHQWLDO YHUVLRQ RI WKDW JLYHQ LQ )LJXUH $V PHQWLRQHG LQ WKH ILUVW FKDSWHU D GLIIHUHQWLDO WRSRORJ\ LV UHTXLUHG IRU WKH FORFN UHFHLYHU DSSOLFDWLRQ 7UDQVLVWRU 0 SURYLGHV WDLO FXUUHQW WR WKH GLIIHUHQWLDO DPSOLILHU ZKHUH 9ELDV rV H[WHUQDOO\ JHQHUDWHG $Q LPSRUWDQW GLIIHUHQFH EHWZHHQ WKLV /1$ DQG WKH 0+] /1$ LV WKDW LQGXFWRUV /JO DQG /VO DUH LQWHJUDWHG 7KLV /1$ ZDV RULJLQDOO\ GHVLJQHG WR SURYLGH G% RI JDLQ ZKLOH GULYLQJ e DQG WKH DQWLFLSDWHG 1) ZDV DSSUR[LPDWHO\ G% 7KH KLJKHU DQWLFLSDWHG 1) FRPSDUHG WR WKH 0+] UHVXOW LV GXH WR VKRUWFKDQQHO HIIHFWV LQFUHDVLQJ \ >-LQ@ DV ZHOO DV WKH LQWHJUDWLRQ RI /J $FFRUGLQJ WR VLPXODWLRQV WKH LQWHJUDn WLRQ RI /J LQFUHDVHV WKH 1) E\ DSSUR[LPDWHO\ G% )LUVW WKH LQFUHDVH LV GXH WR WKHUPDO QRLVH JHQHUDWHG E\ WKH VHULHV UHVLVWDQFH RI /J U/JLf ZKRVH 4 IDFWRU LV f IRU Sn Sf VXEVWUDWHV 6HFRQG WKH SDUDVLWLF FDSDFLWDQFHV DW WKH LQSXW DQG RXWSXW SRUWV RI /J &/Jcf PRYH WKH LQSXW PDWFK DZD\ IURP WKH RSWLPDO QRLVH PDWFK 7KHUHIRUH WKH LQWHJUDWLRQ RI /J

PAGE 63

LPSRVHV VHYHUH UHVWULFWLRQV RQ WKH DFKLHYDEOH QRLVH ILJXUH )RU VWDQGDUG ZLUHOHVV DSSOLFDn WLRQV DW WKLV IUHTXHQF\ UDQJH /J LV DSSUR[LPDWHO\ Q+ ZKLFK LV VXLWDEOH IRU ERQGZLUH LPSOHPHQWDWLRQ 7KLV ZRXOG LQFUHDVH WKH 4 GUDPDWLFDOO\ DQG KHOS WKH QRLVH ILJXUH +RZn HYHU IRU WKH ZLUHOHVV LQWHUFRQQHFW DSSOLFDWLRQ VLQFH WKH DQWHQQD LV LQWHJUDWHG /J KDV WR EH LQWHJUDWHG DV ZHOO ,QGXFWRU &KDUDFWHULVWLFV 7R GHPRQVWUDWH WKH HIIHFW RI VXEVWUDWH UHVLVWDQFH RQ LQGXFWRU TXDOLW\ IDFWRU Q+ VSLUDO LQGXFWRUV KDYH EHHQ LPSOHPHQWHG LQ WKH OHYHOPHWDO SP &026 SURFHVV RQ ERWK Sn DQG S ZLWK HSLWD[LDO OD\HUf VXEVWUDWHV 7KHVH LQGXFWRUV ZHUH XVHG IRU /J LQ WKH /1$ 7KH LQGXFWRU LV LPSOHPHQWHG XVLQJ VKXQWHG PHWDOV DQG OD\HUV ZLWK WXUQV DQG DQ DUHD RI [ ILP 7KH TXDOLW\ IDFWRU LV HVWLPDWHG XVLQJ ERWK WKH KDOIEDQGZLGWK PHWKRG >@ DQG WKH FRQYHQWLRQDO PHWKRG 4 LPDJ\XfUHDO\Lff )LJXUH VKRZV )LJXUH 4XDOLW\ IDFWRUV DQG LQGXFWDQFH RI /J YHUVXV IUHTXHQF\ IRU GLIIHUHQW VXEVWUDWH UHVLVWLYLWLHV S a LFP Sn ZLWK HSL a LOFPf 7KH LQGXFWRU LV LPSOHPHQWHG LQ D PHWDOOD\HU _LP &026 SURFHVV

PAGE 64

WKH PHDVXUHG 4 DQG LQGXFWDQFH RI WKHVH LQGXFWRUV IRU ERWK VXEVWUDWH W\SHV 7KH TXDOLW\ IDFWRU RI WKH LQGXFWRU RQ S VXEVWUDWHV 4 f LV b RI WKDW RQ Sn VXEVWUDWHV 4 f )RU S VXEVWUDWHV WKH LQGXFWDQFH VOLJKWO\ GHFUHDVHV ZKLOH WKH VHULHV UHVLVWDQFH QHDUO\ GRXEOHV GXH WR WKH LQGXFHG HGG\ FXUUHQWV 0HDVXUHPHQWV DOVR VKRZ WKDW IRU D Q+ LQGXFWRU ZLWK DQ DUHD RI [ _LP WKH 4 UHGXFWLRQ LV RQO\ b >)OR22E@ 7KLV VLJQLILHV WKDW VPDOOHU DUHD LQGXFWRUV DUH EHQHILFLDO IRU ORZUHVLVWLYLW\ VXEVWUDWHV 0HDVXUHG 5HVXOWV )LJXUH Df VKRZV WKH PHDVXUHG 6SDUDPHWHUV IRU WKH GLIIHUHQWLDO /1$ 7R REWDLQ WKLV EDODQFHG PHDVXUHPHQW XVLQJ DQ XQEDODQFHG QHWZRUN DQDO\]HU r FRXSOHUV ZHUH SODFHG DW WKH LQSXW DQG RXWSXW RI WKH /1$ DFWLQJ DV EDOXQV 'ULYLQJ 4 GLIIHUHQn WLDOO\ WKH /1$ SURYLGHV G% RI JDLQ ZLWK LQSXW DQG RXWSXW UHIOHFWLRQ FRHIILFLHQWV RI OHVV WKDQ G% FRQVXPLQJ P: IURP D 9 VXSSO\ 7KLV JDLQ LV G% ORZHU WKDQ WKDW H[SHFWHG GXH WR D GLVFUHSDQF\ EHWZHHQ WKH VLPXODWHG WUDQVLVWRU FKDUDFWHULVWLFV DQG WKH PHDVXUHG WUDQVLVWRU FKDUDFWHULVWLFV ,Q SDUWLFXODU &JV ZDV b ODUJHU WKDQ WKDW VLPXODWHG WKXV 2M ZDV ORZHU DQG WKH JDLQ ZDV GHFUHDVHG DV JLYHQ E\ f )LJXUH Ef VKRZV WKH PHDVXUHG JDLQ DQG 1) RI WKH VLQJOHHQGHG /1$ IRU ERWK Sn DQG S VXEVWUDWHV (DFK /1$ FRQVXPHV P: IURP D 9 VXSSO\ 7KH SRZHU JDLQ RI WKH /1$ RQ D S VXEVWUDWH LV G% ZKLOH WKDW IRU WKH /1$ RQ D S VXEVWUDWH LV G% )RU WKH Sn VXEVWUDWH WKHUH LV D G% GLIIHUHQFH EHWZHHQ WKH JDLQ IRU WKH VLQJOHHQGHG DQG GLIIHUHQWLDO /1$V ZKRVH RULJLQ LV XQNQRZQ 7KHRUHWLFDOO\ WKH JDLQV VKRXOG EH WKH VDPH ZKHQ HDFK KDOIFLUFXLW RI WKH GLIIHUHQWLDO /1$ LV RSHUDWHG DW WKH VDPH SRZHU FRQVXPSWLRQ DV WKH VLQJOHHQGHG /1$ 5HIHUULQJ WR )LJXUH WKH VSLUDO LQGXFWRU 4 IRU S VXEVWUDWH LV

PAGE 65

)LJXUH Df 0HDVXUHG 6SDUDPHWHUV RI GLIIHUHQWLDO /1$ Ef 0HDVXUHG JDLQ DQG 1) IRU VLQJOHHQGHG 1) IRU ERWK S DQG S VXEVWUDWHV ab RI WKDW IRU Sn VXEVWUDWH DQG WKH LQGXFWDQFH LV VPDOOHU 7KHUHIRUH WKH UHVRQDQW IUHn TXHQF\ RI WKH S /1$ LV DERXW b KLJKHU GXH WR WKH GHFUHDVHG LQGXFWDQFH 6LPXODWLRQ UHVXOWV VKRZ WKDW a G% RI WKH JDLQ UHGXFWLRQ FDQ EH DWWULEXWHG WR WKH b UHGXFWLRQ LQ WKH 4 IRU /G $QRWKHU a G% FRPHV IURP WKH DGGLWLRQDO ORVV GXH WR WKH LQFUHDVH LQ U/J 7KH 1) RI WKH VLQJOHHQGHG S /1$ LV a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f

PAGE 66

7DEOH 6XPPDU\ RI PHDVXUHG UHVXOWV IRU *+] _OP /1$V 'LIIHUHQWLDO /1$ Sn VXEVWUDWHf 6LQJOH(QGHG /1$V YGG 9 YGG 9 3RZHU &RQVXPSWLRQ P: 3RZHU &RQVXPSWLRQ P: 5HVRQDQW )UHTXHQF\ *+] 5HV )UHTXHQF\ Sn VXEVWUDWHf *+] *DLQ G% *DLQ Sf VXEVWUDWHf G% VOO G% 1) Sn VXEVWUDWHf G% 6 G% 5HV )UHTXHQF\ S VXEVWUDWHf *+] 5HYHUVH ,VRODWLRQ G% *DLQ S VXEVWUDWHf G% 1) S VXEVWUDWHf G% $ *+] PQ &026 /1$ &LUFXLW ,PSOHPHQWDWLRQ $ ILQDO LPSOHPHQWDWLRQ RI D VRXUFHGHJHQHUDWHG /1$ ZDV GRQH LQ D SP &026 WHFKQRORJ\ ZLWK OD\HUV RI FRSSHU LQWHUFRQQHFWV >)OR2OD@ 7KLV WHFKQRORJ\ ZDV REWDLQHG IURP 80& DV SDUW RI WKH 65& &RSSHU 'HVLJQ &KDOOHQJH VSRQVRUHG E\ WKH 6HPLn FRQGXFWRU 5HVHDUFK &RUSRUDWLRQ 65&f 1RYHOOXV 6SHHG)DP,3(& DQG 80& 7KH VXEn VWUDWH UHVLVWLYLW\ LV 2FP 7KH FLUFXLWV ZHUH GHVLJQHG WR RSHUDWH DW a *+] ZKLOH PHDVXUHPHQWV UHYHDO a *+] RSHUDWLRQ )RU WKLV WHFKQRORJ\ ERWK D VLQJOHVWDJH GLIIHUHQWLDO /1$ DQG DQ /1$ ZLWK VRXUFHIROORZHU EXIIHUV ZHUH LPSOHPHQWHG 7KH RSHUDWLRQ RI VRXUFHIROORZHUV ZLOO EH SUHn VHQWHG LQ VHFWLRQ 7KH VFKHPDWLF RI WKH VLQJOHVWDJH /1$ LV LGHQWLFDO WR WKH *+] YHUVLRQ VKRZQ LQ )LJXUH ZKHUH FRPSRQHQW YDOXHV DQG VL]HV ZHUH PRGLILHG DV UHIOHFWHG LQ )LJXUH )LJXUH VKRZV D VFKHPDWLF RI WKH IXOO\ GLIIHUHQWLDO /1$ ZLWK

PAGE 67

/ Q+ )LJXUH 6FKHPDWLF RI *+] IXOO\ GLIIHUHQWLDO /1$ VKRZLQJ SDUDVLWLFV VRXUFHIROORZHU EXIIHUV 7KLV FLUFXLW LV XVHG LQ WKH FORFN UHFHLYHU GULYLQJ WKH IUHTXHQF\ GLYLGHU 7ZR GLIIHUHQW LQSXW PDWFKLQJ FRQGLWLRQV ZHUH LPSOHPHQWHG 7KH LQSXW PDWFKLQJ FRQGLWLRQ ZDV  GLIIHUHQWLDOO\f ZKHQ WKH /1$ ZDV PHDVXUHG XVLQJ  WHVW HTXLSn PHQW DQG M  ZKHQ WKH /1$ ZDV GULYHQ E\ WKH RQFKLS DQWHQQD 7KHUHIRUH WKH UHDFWDQFH RI WKH JDWH LQGXFWRUV IRU WKH ODWWHU FDVH ZDV LQFUHDVHG E\ 4 HDFK VLGH RU 4 GLIIHUHQWLDOO\f 7KH VLQJOHVWDJH GLIIHUHQWLDO /1$ ZDV GLIIHUHQWLDOO\ PDWFKHG DW WKH RXWSXW WR 4 XVLQJ FDSDFLWLYH WUDQVIRUPHUV )LJXUH Df VKRZV D GLH SKRWRJUDSK RI WKH LP GLIIHUHQWLDO /1$ 7KH GLH VL]H LV [ _LP ,QGXFWRU &KDUDFWHULVWLFV 7KH JDWH DQG GUDLQ LQGXFWRUV IRU WKH *+] /1$ KDYH WKH IROORZLQJ GHVLJQ SDUDPHWHUV LQGXFWDQFH Q+ WXUQV DUHD [ _-P ZLGWK SP DQG WXUQ

PAGE 68

9GG )LJXUH 6FKHPDWLF RI GLIIHUHQWLDO *+] /1$ ZLWK VRXUFHIROORZHUV Df Ef )LJXUH 'LH SKRWRJUDSKV RI Df _[P *+] GLIIHUHQWLDO &026 /1$ DQG Ef _LP *+] 62, &026 WXQHG DPSOLILHU

PAGE 69

VSDFLQJ _LP 3DWWHUQHGJURXQG VKLHOGV ZHUH DJDLQ XVHG WR GHFUHDVH WKH VXEVWUDWH UHVLVWDQFH 6LQFH WKLV WHFKQRORJ\ KDV D KLJKUHVLVWLYLW\ VXEVWUDWH WKH LQFUHDVH LQ VHULHV UHVLVWDQFH DQG KHQFH GHJUDGDWLRQ LQ 4 FDXVHG E\ HGG\ FXUUHQWV LQ WKH VXEVWUDWH LV VPDOO DV ZDV VKRZQ LQ VHFWLRQ /G ZDV LPSOHPHQWHG XVLQJ PHWDOV DQG VKXQWHG WRJHWKHU GHFUHDVLQJ WKH VHULHV UHVLVWDQFH DQG PD[LPL]LQJ 4 6LPXODWLRQV VKRZ WKDW IRU /J PLQLn PL]LQJ WKH SDUDVLWLF FDSDFLWDQFH KDV D JUHDWHU HIIHFW RQ 1) WKDQ PLQLPL]LQJ WKH VHULHV UHVLVWDQFH SURYLGHG WKH 4 LV f 7KLV LV GXH WR WKH FDSDFLWDQFH PRYLQJ WKH LQSXW PDWFKn LQJ FRQGLWLRQ DZD\ IURP WKH RSWLPXP QRLVH PDWFK 7KHUHIRUH GXH WR WKH DYDLODELOLW\ RI FRSSHU RQO\ PHWDO ZDV XVHG IRU /J $ NH\ EHQHILW RI FRSSHU PHWDOOL]DWLRQ LV LWV UHGXFWLRQ RI LQWHUFRQQHFW VHULHV UHVLVn WDQFH )RU LQGXFWRUV WKH 4 ZLOO EH DOPRVW GRXEOHG DV FRPSDUHG ZLWK WKH VDPHWKLFNQHVV DOXPLQXP HTXDO WR WKH UDWLR RI WKH $ DQG &X UHVLVWLYLWLHVf %RWK PHWDOV DQG ZHUH WR KDYH D QRPLQDO WKLFNQHVV RI SP +RZHYHU GXULQJ IDEULFDWLRQ WKHVH WZR PHWDO OD\HUV KDG WR EH WKLQQHG IURP SP WR DSSUR[LPDWHO\ SP 7KHUHIRUH UV ZDV LQFUHDVHG DQG 4 ZDV GHFUHDVHG )LJXUH VKRZV WKH PHDVXUHG LQGXFWDQFH DQG TXDOLW\ IDFWRU 4EZ >@f

PAGE 70

7DEOH &RPSDULVRQ %HWZHHQ ([SHFWHG DQG 0HDVXUHG ,QGXFWRU &KDUDFWHULVWLFV ,QGXFWRU *+]f / Q+f UV4f FR[ P 5VXE Af 4 &JDWH ([SHFWHG 0HDVXUHG a /GUDLQ ([SHFWHG 0HDVXUHG a YHUVXV IUHTXHQF\ IRU /J DQG /G 7KH EHVW 4 LV DSSUR[LPDWHO\ DW *+] DQG LQFUHDVHV WR DSSUR[LPDWHO\ DW *+] 7DEOH VKRZV D FRPSDULVRQ EHWZHHQ WKH H[SHFWHG DQG PHDVXUHG LQGXFWRU FKDUDFWHULVWLFV )LUVW WKH PHDVXUHG LQGXFWDQFH LV a Q+ ODUJHU WKDQ WKDW H[SHFWHG UHVXOWLQJ LQ GHFUHDVHG FLUFXLW UHVRQDQW IUHTXHQF\ 6HFRQG WKH PHDVXUHG VHULHV UHVLVWDQFH LV PRUH WKDQ WZLFH WKDW H[SHFWHG GXH WR LQFUHDVHG PHWDO VKHHW UHVLVWDQFH ZKLOH WKH PHDVXUHG 4 LV DERXW KDOI WKDW H[SHFWHG $OWKRXJK WKH 4fV DUH QRW DV KLJK DV H[SHFWHG WKH\ DUH FHUWDLQO\ DGHTXDWH IRU LPSOHPHQWLQJ 5) FLUFXLWU\ SURYLGHG WKDW WKH LPSDFW RI ORZHU 4 LV DFFRXQWHG IRU GXULQJ WKH GHVLJQ SKDVH 7KHVH UHVXOWV LOOXVWUDWH WKH GUDPDWLF HIIHFW WKDW GHFUHDVLQJ WKH WRSOHYHO PHWDOVf VKHHW UHVLVWDQFH KDV RQ LQGXFWRU FKDUDFWHULVWLFV 1HYHUWKHOHVV IRU FRQVWDQW PHWDO WKLFNQHVV FRSSHU PHWDOOL]DWLRQ JUHDWO\ LPSURYHV LQGXFWRU FKDUDFWHULVWLFV RYHU DOXPLQXP PHWDOOL]DWLRQ 7KXV WKHVH UHVXOWV DUH DSSUR[LPDWHO\ WZLFH DV JRRG DV DQ DOODOXPLQXP SURn FHVV ZLWK WKH VDPH PHWDO WKLFNQHVVHV RU DERXW FRPSDUDEOH WR DQ DOXPLQXP SURFHVV ZLWK WZLFH WKH PHWDO WKLFNQHVV 7KHVH PHDVXUHPHQW UHVXOWV VXJJHVW WKDW JUHDWO\ LPSURYHG LQGXFWRU FKDUDFWHULVWLFV FDQ EH DWWDLQHG E\ WKLFNHQLQJ PHWDO DQG OD\HUV

PAGE 71

7UDQVLVWRU &KDUDFWHUL]DWLRQ 7KH _LP /1$ DQG WKH HQWLUH ZLUHOHVV LQWHUFRQQHFW V\VWHP ZHUH GHVLJQHG WR RSHUDWH DW a *+] +RZHYHU DV ZLOO EH VKRZQ LQ WKH VXEVHTXHQW VHFWLRQV WKH RSHUDWLQJ IUHTXHQF\ IRU WKHVH FLUFXLWV KDV VKLIWHG GRZQ WR a *+] 5HIHUULQJ WR 7DEOH LW FDQ EH VHHQ WKDW WKH PHDVXUHG LQGXFWDQFH DQG LWV SDUDVLWLF FDSDFLWDQFH ZHUH VOLJKWO\ KLJKHU WKDQ H[SHFWHG ZLWKLQ Q+f $W a *+] WKHVH GLVFUHSDQFLHV FDQ DFFRXQW IRU DW PRVW D *+] IUHTXHQF\ VKLIW 7R XQGHUVWDQG WKH UHPDLQLQJ IUHTXHQF\ VKLIW WKH GHVLJQV KDYH EHHQ UHYLHZHG WR PDNH VXUH WKDW DOO RI WKH PHWDO SDUDVLWLFV KDYH EHHQ SURSHUO\ DFFRXQWHG DQG WKDW WKH WUDQVLVWRUV KDYH EHHQ VSHFLILHG FRUUHFWO\ LQFOXGLQJ WKH DUHD DQG SHULPHWHU RI GUDLQV DQG VRXUFHV 7KH UHPDLQLQJ IUHTXHQF\ VKLIW LV WKHUHIRUH EHOLHYHG WR EH SULPDULO\ GXH WR WKH GLIIHUHQFHV EHWZHHQ WKH &$' LQWULQVLF WUDQVLVWRU PRGHO XVHG GXULQJ WKH GHVLJQ SURn FHVV DQG WKH DFWXDO IDEULFDWHG WUDQVLVWRU FKDUDFWHULVWLFV 1RWH WKDW WKH PRGHOV XVHG WR GHVLJQ WKHVH 5) FLUFXLWV ZHUH WKH VWDQGDUG %6,0 PRGHOV WR ZKLFK H[WULQVLF SDUDVLWLF FRPn SRQHQWV LPSRUWDQW DW UDGLR IUHTXHQFLHV DUH DGGHG HJ 5JDWH 5VXEf DQG FDSDFLWDQFHV DVVRn FLDWHG ZLWK LQWHUFRQQHFWLQJ WKH PXOWLSOH ILQJHUV RI WKH WUDQVLVWRUf 7KLV WHFKQLTXH KDV EHHQ XVHG YHU\ HIIHFWLYHO\ LQ WKH SDVW WR \LHOG VLPXODWHG UHVXOWV ZKLFK DJUHH YHU\ FORVHO\ ZLWK PHDVXUHG UHVXOWV 7KH 6SDUDPHWHUV RI LQGLYLGXDO 1026 DQG 3026 WUDQVLVWRUV KDYH EHHQ PHDVXUHG WR TXDQWLI\ WKH GLIIHUHQFH LQ FDSDFLWDQFH EHWZHHQ WKH PRGHO DQG WKH DFWXDO WUDQVLVWRUV %RWK WKH 1026 DQG 3026 WUDQVLVWRUV KDYH : _LP DQG / _LP DQG DUH LPSOHn PHQWHG ZLWK M[P ORQJ ILQJHUV WRWDO ILQJHUV f 7KH 6SDUDPHWHUV RI WKH WUDQVLVWRUV DUH WKHQ WUDQVIRUPHG LQWR \SDUDPHWHUV 3DUDVLWLFV DVVRFLDWHG ZLWK WKH SDGV DUH GHHPEHGGHG WKURXJK \SDUDPHWHU VXEWUDFWLRQ RI DQ RSHQ WHVW VWUXFWXUH :LWK WKH WUDQVLV

PAGE 72

k0HDVXUHG \ f§ 6LPXODWHG \ 3026 U9JV 9 9GV 9 I) )UHTXHQF\ *+]f )LJXUH 0HDVXUHG YHUVXV VLPXODWHG \X ZLWK WUDQVLVWRU LQ OLQHDU UHJLRQf DQG \ ZLWK WUDQVLVWRU RIIf IRU 1026 DQG 3026 WUDQVLVWRUV 7KH VORSH RI WKHVH OLQHV RYHU N LV HTXDO WR WKH FDSDFLWDQFHV &JJ DQG &GG IRU \X DQG \ UHVSHFWLYHO\ WRU ELDVHG LQ WKH OLQHDU UHJLRQ \Q \LHOGV WKH WRWDO JDWH FDSDFLWDQFH &JJ &JV &JG &JE :/&R[ :&9f :LWK WKH WUDQVLVWRU WXUQHG RII WKH FKDQQHO FRQGXFWDQFH LV ]HUR DQG \ \LHOGV WKH WRWDO GUDLQ FDSDFLWDQFH &GG &GE &JGf ZKLOH \ \LHOGV &JG $OWKRXJK WKHUH DUH PRUH DFFXUDWH ZD\V WR REWDLQ WKHVH FDSDFLWDQFHV WKLV PHDVXUHPHQW SURYLGHV D TXLFN DQG HDV\ ZD\ WR REVHUYH DQ\ FDSDFLWDQFH WUHQGV )LJXUH VKRZV WKH PHDVXUHG YHUVXV VLPXODWHG \X DQG \ IRU WKH 1026 DQG 3026 WUDQVLVWRUV 7KH VORSH RI HDFK RI WKHVH OLQHV LV SURSRUWLRQDO WR FDSDFLWDQFH WKHUHIRUH WKH VWHHSHU WKH OLQH WKH KLJKHU WKH FDSDFLWDQFH $V FDQ EH VHHQ WKH PHDVXUHG FDSDFLWDQFHV DUH DOO KLJKHU WKDQ WKH VLPXODWHG FDSDFLWDQFHV QRWHf§EHFDXVH WKH RSHQ SDG VWUXFWXUH XVHG IRU GHHPEHGGLQJ GRHV QRW LQFOXGH WKH LQWHUFRQQHFWV WR WKH WUDQVLVWRU WKH PHDVXUHPHQWV RYHUHVWLPDWH WKH FDSDFLWDQFHV

PAGE 73

E\ D IHZ I)f )RU 1026 WKH JDWH FDSDFLWDQFH LV a WLPHV ODUJHU ZKLOH IRU 3026 LW LV WLPHV ODUJHU )RU ERWK 1026 DQG 3026 WKH WRWDO GUDLQ FDSDFLWDQFH LV DSSUR[LPDWHO\ WZLFH WKDW ZKLFK LV VLPXODWHG ZLWK WKH 3026 GLIIHUHQFH EHLQJ PRUH VHYHUH 7KLV FDQ UHVXOW LQ DV PXFK DV D WLPHV -Of UHGXFWLRQ LQ UHVRQDQW IUHTXHQF\ IRU D WXQHG FLUFXLW 0HDVXUHG 5HVXOWV )LJXUH Df VKRZV WKH PHDVXUHG JDLQ 6f DQG QRLVH ILJXUH 1)f IRU WKH VLQn JOHVWDJH /1$ PDWFKHG WR 4 DW WKH LQSXW DQG RXWSXW ZKLOH )LJXUH Ef VKRZV WKH PHDVXUHG LQSXW DQG RXWSXW UHIOHFWLRQ FRHIILFLHQWV 6Q DQG 6f 'ULYLQJ 4 GLIIHUHQWLDOO\ WKH /1$ SURYLGHV G% RI JDLQ DW *+] ZLWK 6Q DQG 6 HTXDO WR DQG G% UHVSHFWLYHO\ 7KH /1$ FRQVXPHV P: IURP D 9 VXSSO\ 7KH UHYHUVH LVRODWLRQ LV EHWWHU WKDQ G% )LQDOO\ WKH QRLVH ILJXUH LV G% DW UHVRQDQFH 7KH UHVXOWV DUH VXPPDUL]HG LQ 7DEOH )UHTXHQF\ *+]f )UHTXHQF\ *+]f Df Ef )LJXUH Df 0HDVXUHG JDLQ DQG QRLVH ILJXUH IRU JP /1$ Ef 0HDVXUHG LQSXW DQG RXWSXW UHIOHFWLRQ FRHIILFLHQWV IRU SP /1$

PAGE 74

7DEOH 6XPPDU\ RI PHDVXUHG UHVXOWV IRU *+] _LP /1$V ZLWK FRSSHU 6LQJOH6WDJH 'LIIHUHQWLDO /1$ 'LIIHUHQWLDO /1$ ZLWK 6RXUFH)ROORZHUV &HQWHU )UHTXHQF\ *+] &HQWHU )UHTXHQF\ *+] *DLQ G% *DLQ DW 9JF 9 G% 6Q G% *DLQ DW 9JF 9 G% 6 G% 6Q G% 5HYHUVH ,VRODWLRQ G% 5HYHUVH ,VRODWLRQ G% 1RLVH )LJXUH G% 1RLVH )LJXUH G% YGG 9 YGG 9 3RZHU &RQVXPSn WLRQ P: 3RZHU &RQVXPSn WLRQ 9JF 9f P: 6LPXODWLRQV SUHGLFWHG D G% JDLQ D G% 1) DQG D *+] UHVRQDQW IUHTXHQF\ 7KH UHGXFHG UHVRQDQW IUHTXHQF\ KDV DOUHDG\ EHHQ DFFRXQWHG IRU GXH WR WKH LQFUHDVHG LQGXFWDQFH DQG WKH ODUJHU WUDQVLVWRU FDSDFLWDQFHV 5HIHUULQJ WR HTXDWLRQ f D b UHGXFWLRQ LQ 4 UHVXOWV LQ D G% UHGXFWLRQ LQ O6LO 7KLV DFFRXQWV IRU WKH GURS LQ 6L IURP WKH H[SHFWHG YDOXH RI G% WR WKH PHDVXUHG YDOXH RI G% 7KH XQGHUVWDQGLQJ RI WKH KLJKHU QRLVH ILJXUH LV QRW DV VWUDLJKWIRUZDUG WKRXJK )LUVW WKHVH QRLVH PHDVXUHPHQWV ZHUH REWDLQHG WKURXJK RQZDIHU SURELQJ XVLQJ H[WHUQDO EDOXQV ZKLFK \LHOGV KLJKHU QRLVH ILJXUHV RQ WKH RUGHU RI G%f DV FRPSDUHG WR SDFNDJHG FKLSV GXH WR YDULDWLRQV LQ WKH FRQWDFW UHVLVWDQFH RI WKH SUREH 6HFRQG JHQHUDOO\ VSHDNLQJ DV WKH JDLQ LV GHFUHDVHG WKH QRLVH ILJXUH LQFUHDVHV GXH WR KLJKHU LQSXWUHIHUUHG QRLVH 7KLUG GHFUHDVHG LQGXFWRU 4 UHVXOWV LQ PRUH WKHUPDO QRLVH JHQHUDWHG E\ WKH UHVLVWLYH ORVV RI WKH LQGXFWRUV 7KLV LV SDUn WLFXODUO\ VHYHUH IRU /J DQG /V ZKLFK DUH ORFDWHG DW WKH LQSXW RI WKH /1$ )RXUWK WKH PRGn HOLQJ RI WKH WUDQVLVWRUfV WKHUPDO QRLVH FRHIILFLHQW \ ZKLFK ZDV DVVXPHG WR EH LQ VLPXODWLRQV EDVHG RQ I[P PHDVXUHPHQW UHVXOWV LV H[SHFWHG WR LQFUHDVH ZLWK

PAGE 75

GHFUHDVLQJ FKDQQHO OHQJWK GXH WR LQFUHDVHG KRWFDUULHULQGXFHG QRLVHf DQG WKH ERG\ UHVLVn WDQFH FDQ JUHDWO\ LQIOXHQFH WKH QRLVH ILJXUH $FFRUGLQJ WR 63,&( VLPXODWLRQV LQFUHDVLQJ \ WR DQG 5VXE IURP WR 4 LQFUHDVHV WKH QRLVH ILJXUH E\ G% )LQDOO\ %6,0YfV PRGHOLQJ RI WKH WUDQVLVWRUfV QRLVH SDUDPHWHUV LV QRW YHU\ DFFXUDWH WKHUHIRUH %6,0YfV RSWLPDO QRLVH PDWFKLQJ FRQGLWLRQ YHUVXV WKH DFWXDO RSWLPXP FDQ EH YHU\ GLIIHUHQW 7KLV FDQ UHVXOW LQ RSHUDWLQJ DW D VXERSWLPDO LQSXW QRLVHPDWFKLQJ FRQGLWLRQ $OO RI WKHVH HIIHFWV WDNHQ WRJHWKHU FDQ \LHOG WKH G% LQFUHDVH LQ QRLVH ILJXUH )LJXUH VKRZV WKH PHDVXUHG 6SDUDPHWHUV IRU WKH GLIIHUHQWLDO /1$ ZLWK VRXUFHIROORZHU EXIIHUV GULYLQJ D  ORDGf )RU 9JF 9 WKH JDLQ RI WKH /1$EXIIHU FLUFXLW LQFUHDVHV WR G% ZKLOH FRQVXPLQJ D WRWDO RI P: IURP D 9 VXSSO\ 7DEOH VXPPDUL]HV WKH PHDVXUHG UHVXOWV 7KH UHVRQDQW IUHTXHQF\ KDV LQFUHDVHG WR *+] 7KH JDLQ LV LQFUHDVHG FRPSDUHG WR WKH VLQJOHVWDJH /1$ UHVXOW GXH WR UHPRYDO RI WKH FDSDFLWLYH WUDQVIRUPHU DQG WKH SUHVHQFH RI QHJDWLYH FRQGXFWDQFH DW WKH RXWSXW QRGHV RI WKH /1$ 7RR PXFK QHJDWLYH FRQGXFWDQFH KRZHYHU FDQ UHVXOW LQ XQVWDEOH RSHUDWLRQ E\

PAGE 76

FDXVLQJ WKH WRWDO UHVLVWDQFH DW WKH RXWSXW QRGH RI WKH /1$ WR EH QHJDWLYH %\ LQFUHDVLQJ 9JF WR 9 WKH JDLQ RI WKH /1$EXIIHU FLUFXLW FRXOG EH LQFUHDVHG WR G% KRZHYHU 6 EHFRPHV JUHDWHU WKDQ LQGLFDWLQJ XQVWDEOH RSHUDWLRQ 7KHUHIRUH 9JF LV FKRVHQ VXFK WKDW VWDELOLW\ LV HQVXUHG LH 6 G%f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f &026 WHFKQRORJ\ IURP ,%0 7KH SURFHVV XVHV D SP GHVLJQ UXOH VHW IRU DOO GLPHQVLRQV H[FHSW IRU WKH SP HIIHFWLYH JDWH OHQJWK M[P GUDZQ JDWH OHQJWKf DQG QP JDWH

PAGE 77

R[LGH WKLFNQHVV 7ZR PHWDO OD\HUV _OP HDFKf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n WDLQHG LQ VHFWLRQ DQG $SSHQGL[ $ $W UHVRQDQFH WKH LQSXW LPSHGDQFH WR WKH DPSOLILHU LV OJPO 7KHUHIRUH JPO LV GHVLJQHG WR EH 4n WR SURYLGH D 4 PDWFK )ROORZLQJ WKH LQSXW VWDJH LV D VRXUFHIROORZHU EXIIHU ZKLFK WKHQ GULYHV D FRPPRQVRXUFH FDVFRGH VWDJH ZLWK D WXQHG ORDG 7KH RXWSXW RI WKLV VWDJH LV PDWFKHG WR  XVLQJ D FDSDFLWLYH WUDQVn IRUPHU 7KLV VWDJH LV D QRQGHJHQHUDWHG YHUVLRQ RI WKH /1$V GLVFXVVHG HDUOLHU )LJXUH

PAGE 78

Ef VKRZV D GLH SKRWRJUDSK RI WKH _LP /1$ 7KH GLH VL]H LV [ PP *URXQGVKLHOGHG GLDPRQGVKDSHG SDGV DUH XVHG DW WKH LQSXW DQG RXWSXW WR UHGXFH WKH SDUDn VLWLF FDSDFLWDQFH WR WKH VXEVWUDWH 7KUHH RQFKLS VSLUDO LQGXFWRUV RI Q+ DUH XVHG LQ WKH FLUFXLW )LQDOO\ RQFKLS DFFXPXODWLRQPRGH 026 FDSDFLWRUV RI S) DUH XVHG WR E\SDVV 9GG! 9JL} DQG 9JF 6RXUFH)ROORZHU 7KH VRXUFHIROORZHU LV XVHG WR VKLIW WKH GF ELDV OHYHO IURP WKH RXWSXW RI WKH ILUVW VWDJH DW 9GGf WR WKH LQSXW RI WKH WKLUG VWDJH 7KH ELDV SRLQW RI WKH VRXUFHIROORZHU LV FRQn WUROOHG WKURXJK LQSXW 9JF ZKLFK FRQWUROV WKH FXUUHQW WKURXJK 0 DQG VHWV WKH 9JV RI 0 $ VRXUFHIROORZHU ZLWK D FDSDFLWLYH ORDG FDQ KDYH D QHJDWLYH LQSXW FRQGXFWDQFH 6LQFH LQGXFWLYH VRXUFHGHJHQHUDWLRQ SURYLGHV D SRVLWLYH UHDO LQSXW LPSHGDQFH IRU WKH WUDQVLVWRU E\ UHDVRQ LW IROORZV WKDW FDSDFLWLYH VRXUFHGHJHQHUDWLRQ SURYLGHV D QHJDWLYH UHDO LQSXW LPSHGDQFH IRU WKH WUDQVLVWRU 5HIHUULQJ WR )LJXUH LW FDQ EH VKRZQ WKDW WKH LQSXW DGPLWWDQFH ORRNLQJ LQWR 0 LV DV IROORZV 5H^7&JVa 6P&JV&/7@ t 6Uf p&VV &nLUff f ,P
PAGE 79

ZKLFK KDV D PDJQLWXGH OHVV WKDQ RQH D SROH DW FR JPJ7f&J6&/7f !S DQG D ]HUR DW f b 1HJDWLYH FRQGXFWDQFH LQGLFDWHV WKDW HQHUJ\ LV EHLQJ JHQHUDWHG :KHQ D VRXUFHIROn ORZHU LV XVHG DIWHU DQ DPSOLILHU FRQWDLQLQJ DQ LQGXFWLYH ORDG WKLV HQHUJ\ DFWXDOO\ UHSOHQn LVKHV VRPH RI WKH HQHUJ\ ORVW LQ WKH LQGXFWRU FDXVLQJ WKH TXDOLW\ IDFWRU WR LQFUHDVH +HQFH WKH QHJDWLYH FRQGXFWDQFH RI WKH VRXUFHIROORZHU LQFUHDVHV WKH JDLQ RI WKH /1$ WKURXJK LQFUHDVLQJ WKH ORDG TXDOLW\ IDFWRU )LJXUH VKRZV WKH PHDVXUHG LQSXW FRQGXFWDQFH YHUVXV FRQWURO YROWDJH 9JFf YHUVXV IUHTXHQF\ IRU D VRXUFHIROORZHU LPSOHPHQWHG LQ WKH SP &026 WHVW FKLS 7KLV EXIIHU LV GULYLQJ DQ RSHQ SDG VWUXFWXUH KDYLQJ D FDSDFLWDQFH RI a I) 7KH FRQGXFWDQFH EHFRPHV PRUH QHJDWLYH ZLWK LQFUHDVLQJ IUHTXHQF\ LQ DJUHHPHQW ZLWK f $OVR WKH FRQn GXFWDQFH EHFRPHV PRUH QHJDWLYH DV 9JF LQFUHDVHV WR 9 DQG WKHQ EHJLQV WR EHFRPH OHVV QHJDWLYH IRU IXUWKHU LQFUHDVHV LQ 9JF 7KLV LV GXH WR JP LQFUHDVLQJ IRU 9JF EHWZHHQ WR 9 FDXVLQJ *VI WR GHFUHDVH +RZHYHU WKH WUDQVLVWRUV LQ WKH VRXUFHIROORZHU HYHQWXn DOO\ HQWHU WKH OLQHDU UHJLPH FDXVLQJ J7 WR LQFUHDVH DQG JP WR GHFUHDVH WKXV *VI EHFRPHV OHVV QHJDWLYH *DLQ RI *+] /1$ 7KH YROWDJH JDLQ RI WKH WKUHHVWDJH /1$ FDQ EH UHDGLO\ GHULYHG XVLQJ HTXDWLRQV $f f DQG f $VVXPLQJ *VI WKH WRWDO YROWDJH JDLQ LV DV IROORZV Y_FR D OAY HJ n AY VI n AY FM_ f QRf PMr&JV P W M£&JV &OWf VPIL/Ge!r/G FRf }*2

PAGE 80

)LJXUH ,QSXW FRQGXFWDQFH WR VRXUFHIROORZHU EXIIHU GULYLQJ D FDSDFLWLYH ORDG YHUVXV IUHTXHQF\ DQG FRQWURO YROWDJH ZKHUH WKH IDFWRUV DQG Q KDYH EHHQ GHILQHG LQ f DQG f DQG ? VI LV GHILQHG LQ f $Q HIIHFWLYH TXDOLW\ IDFWRU KDV EHHQ LQWURGXFHG IRU /GL WR DFFRXQW IRU WKH QHJDWLYH LQSXW FRQGXFWDQFH RI WKH VRXUFHIROORZHU EXIIHU DV IROORZV 4X 4O f L?*VI?]D}Rf\ ,PSHGDQFH =A LV WKH FRQGXFWDQFH RI WKH GUDLQ LQGXFWRU RI WKH ILUVW VWDJH DW WDQNf UHVRn QDQFH 6XFK D VXEVWLWXWLRQ IRU WKH TXDOLW\ IDFWRU FDQ EH XVHG ZKHQHYHU D WXQHG JDLQ VWDJH LV IROORZHG E\ D VRXUFHIROORZHU EXIIHU :LWK WKH LQSXW LPSHGDQFH GHVLJQHG WR EH 5V 4 LH JPM 4nf DQG ZLWK /GO /G WKH WRWDO IRUZDUG WUDQVGXFHU JDLQ EHFRPHV _6,_ ‘ ?$ Y_FR FR ,6POMHII f L L 5 4/GO4/G&GR/G? n _$9_ V 6FRf Q: f ,W FDQ EH VHHQ WKDW WKH GHQRPLQDWRU RI f FDQ EHFRPH ]HUR LQGLFDWLQJ WKDW fLQILn QLWHf JDLQ LV REWDLQHG 7KXV D ILQLWH RXWSXW VLJQDO FDQ EH REWDLQHG ZKHQ QR LQSXW LV SUHVHQW WR WKH DPSOLILHU PHDQLQJ WKDW WKH FLUFXLW ZLOO RVFLOODWH 2EYLRXVO\ WKLV FRQGLWLRQ LV WR EH

PAGE 81

DYRLGHG 7KXV WKH FRQGXFWDQFH RI WKH VRXUFHIROORZHU LV DGMXVWHG VXFK WKDW WKH JDLQ LV LPSURYHG ZKLOH VWDELOLW\ LV PDLQWDLQHG 0HDVXUHG 5HVXOWV 7KH PHDVXUHG 6SDUDPHWHUV RI WKH 62, DPSOLILHU DUH VKRZQ LQ )LJXUH Df $W *+] WKH DPSOLILHU LV SHUIHFWO\ PDWFKHG WR 4 DW WKH LQSXW ZLWK DQ RI G% DQG ZHOOPDWFKHG DW WKH RXWSXW ZLWK DQ 6 RI G% 7KH WUDQVGXFHU JDLQ 6Lf LV G% ZKLOH WKH UHYHUVH LVRODWLRQ LV G% ZKLFK DUH ERWK H[FHOOHQW IRU D &026 FLUFXLW DW WKLV IUHTXHQF\ UDQJH $OVR WKH JDLQ LV JUHDWHU WKDQ G% EH\RQG *+] 7KH WRWDO VXSSO\ FXUn UHQW LV P$ IRU D 9 VXSSO\ ZKLFK LV KLJK GXH WR WKH XVH RI PXOWLSOH DPSOLILHU VWDJHV 7KH PHDVXUHG TXDOLW\ IDFWRU 4f RI WKH RQFKLS VSLUDO LQGXFWRUV LV a DW *+] 7KLV YDOXH LV ORZHU WKDQ WKH H[SHFWHG YDOXH RI ZKLFK UHVXOWHG LQ WKH JDLQ EHLQJ a G% RJ^`f ORZHU WKDQ LWV VLPXODWHG YDOXH RI G% 7KH UHDVRQ IRU WKLV 4 GHJUDGDn WLRQ KRZHYHU LV QRW IXOO\ XQGHUVWRRG Df )UHTXHQF\ *+]f Ef )UHTXHQF\ *+]f )LJXUH Df 0HDVXUHG 6SDUDPHWHUV DQG Ef PHDVXUHG RSHUDWLQJ SRZHU JDLQ *Sf DQG QRLVH ILJXUH 1)f YV IUHTXHQF\

PAGE 82

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ff $FFRUGLQJ WR HTXDWLRQ f WKH LQGXFWRU 4 LV JUHDWO\ UHGXFHG DW *+] GXH WR WKH VXEVWUDWH UHVLVn WDQFH ,Q IDFW KDG SDWWHUQHG JURXQG VKLHOGV EHHQ XVHG WKHQ WKH VXEVWUDWH UHVLVWDQFH ZRXOG KDYH EHHQ UHGXFHG WR 4 ZKLOH WKH SDUDVLWLF FDSDFLWDQFH ZRXOG KDYH LQFUHDVHG 7KLV ZRXOG KDYH DSSUR[LPDWHO\ GRXEOHG WKH LQGXFWRU 4 SRWHQWLDOO\ LQFUHDVLQJ WKH JDLQ E\ RJ^` G% IRU WKH FDVH ZKHQ WKH VRXUFHIROORZHU LV DGMXVWHG IRU FORVH WR ]HUR LQSXW FRQGXFWDQFHf )XUWKHU LPSURYHPHQWV FRXOG EH REWDLQHG E\ LPSURYLQJ WKH EDFNHQG SURn FHVV E\ LQFUHDVLQJ WKH WRSOHYHO PHWDO WKLFNQHVV LQFUHDVLQJ WKH GLVWDQFH EHWZHHQ WKH WRSOHYHO PHWDO DQG WKH VXEVWUDWH DQG XVLQJ FRSSHU UDWKHU WKDQ DOXPLQXP IRU WKH PHWDO DV ZLOO EH WKH FDVH IRU IXOO\VFDOHG _LP WHFKQRORJ\f 6XFK LPSURYHPHQWV VKRXOG HDVLO\ OHDG WR LQGXFWRU 4 >%XU@ DW *+] DV KDV EHHQ VKRZQ LQ WKLV FKDSWHU 7KLV ZRXOG GUDPDWLFDOO\ LQFUHDVH WKH DPSOLILHU JDLQ DQG GHFUHDVH WKH SRZHU FRQVXPSWLRQ 8VLQJ VLPXODWHG YDOXHV IRU DOO TXDQWLWLHV H[FHSW LQ f DQG WKHQ LQVHUWLQJ WKH H[WUDFWHG YDOXH RI UHVXOWV LQ 6L G% ZKLFK LV FORVH WR WKH PHDVXUHG 6L 7KH GLVFUHSDQF\ LV DWWULEXWHG WR GLIIHUHQFHV EHWZHHQ WKH WUDQVLVWRU PRGHO DQG WKH DFWXDO GHYLFH 7KH HVWLPDWHG LQSXW FRQGXFWDQFH WR WKH VRXUFHIROORZHU LV P6 UHVXOWLQJ LQ D 4HII RI 7KHUHIRUH WKH QHJDWLYH FRQGXFWDQFH RI WKH VRXUFHIROORZHU LPSURYHV 6 E\ D IDFWRU RI

PAGE 83

a RU G% +RZHYHU ,$YVIO LV WKHUHIRUH WKH QHW JDLQ LQ 6L GXH WR WKH VRXUFHIROn ORZHU LV r V G%f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f ZKLOH WKH '3 LV G%P 7KHVH OLQHDULW\ GDWD DUH FRQVLGHUDEO\ EHWWHU WKDQ WKRVH REWDLQHG LQ >+R@ DQ HQFRXUDJLQJ UHVXOW LQGLFDWLQJ WKDW DFFHSWDEOH OLQn HDULW\ FDQ EH DFKLHYHG ZLWK DPSOLILHUV XWLOL]LQJ VRXUFHIROORZHUV )LJXUH Ef VKRZV WKH PHDVXUHG QRLVH ILJXUH 1)f DQG DVVRFLDWHG JDLQ *DVf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fV SHUIRUPDQFH LV WKDW DOWKRXJK WKH 1) LV ODUJHU WKDQ WKH JDLQ LW GRHV QRW PHDQ WKDW WKH DPSOLILHU LV XQXVDEOH LH WKDW WKH RXWSXW VLJQDOWRQRLVH UDWLR 615f ZLOO DOZD\V EH f 1RLVH ILJXUH LV D UHODWLYH TXDQWLW\ UHSUHVHQWLQJ GHJUDGDWLRQ RI 615 7KHUHIRUH D KLJK 1) GHJUDGHV WKH G\QDPLF UDQJH RI D FLUFXLW WKURXJK UDLVLQJ WKH PLQLPXP GHWHFWDEOH VLJQDO

PAGE 84

7DEOH 6XPPDU\ RI PHDVXUHG UHVXOWV IRU *+] 62, &026 WXQHG DPSOLILHU 3DUDPHWHU 5HVXOW 3DUDPHWHU 5HVXOW 5HVRQDQW )UHTXHQF\ *+] 1) G% 6L G% 3LG%LQf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

PAGE 85

LPSOHPHQWHG RQ ERWK S DQG Sn VXEVWUDWHV 7KH GLIIHUHQWLDO /1$ RQ S VXEVWUDWH SURYLGHV G% RI JDLQ ZLWK D G% QRLVH ILJXUH FRQVXPLQJ P: IURP D 9 VXSSO\ 8VLQJ D S VXEVWUDWH GHJUDGHV WKH JDLQ E\ a G% DQG WKH QRLVH ILJXUH E\ a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n FLHV $OWKRXJK FRSSHU LQWHUFRQQHFWV KDYH LQFUHDVHG WKH LQGXFWRU 4 LQ WKH SP WHFKQROn RJ\ DQ XQSODQQHG WKLQQLQJ RI WKH PHWDO OD\HU GXULQJ IDEULFDWLRQ KDV PLWLJDWHG WKH RYHUDOO LPSURYHPHQW (YHQ VWLOO LQGXFWRU 4fV RI a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
PAGE 86

&+$37(5 &026 )5(48(1&< ',9,'(56 2YHUYLHZ ,Q WKH FORFN UHFHLYHU WKH IUHTXHQF\ GLYLGHU WUDQVODWHV WKH JOREDO FORFN VLJQDO WR WKH ORFDO FORFN VLJQDO 7KH GLYLGHU VKRXOG KDYH D PD[LPXP RSHUDWLQJ IUHTXHQF\ DERYH *+] DQG VKRXOG GLVVLSDWH D PLQLPDO DPRXQW RI SRZHU ,Q DGGLWLRQ WKH GLYLGHU VKRXOG EH FDSDEOH RI ORFNLQJ WR YHU\ ORZOHYHO LQSXW VLJQDOVf§VSHFLILHG LQ WHUPV RI LQSXW VHQVLWLYn LW\f§ZKLFK ZLOO LPSURYH WKH PLQLPXP GHWHFWDEOH VLJQDO 0'6f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n JLHV GLYLGHUV RSHUDWLQJ XS WR *+] KDYH EHHQ UHSRUWHG >:DQ22 :HW22 5D] .XU@ &OHDUO\ VLOLFRQ LV FDSDEOH RI VXSSRUWLQJ GLYLGHU FLUFXLWV RSHUDWLQJ DERYH *+] ,Q WKLV FKDSWHU WKH GHVLJQ DQG LPSOHPHQWDWLRQ RI KLJKIUHTXHQF\ &026 IUHTXHQF\ GLYLGn HUV RSHUDWLQJ XS WR *+] DQG HPSOR\LQJ HLWKHU VRXUFHFRXSOHG ORJLF 6&/f RU GXDOSKDVH G\QDPLF SVHXGR1026 >'3@ ORJLF >%L\
PAGE 87

XS WR DQG *+] UHVSHFWLYHO\ 7KH >'3@ GLYLGHU LV LPSOHPHQWHG LQ D SDUn WLDOO\VFDOHG _LP &026 WHFKQRORJ\ ZLWK EXON DQG 62, VXEVWUDWHV $ GHVLJQ PHWKRGn RORJ\ IRU 6&/ GLYLGHUV EDVHG RQ LQMHFWLRQ ORFNLQJ ZKLFK DOORZV IRU PD[LPL]LQJ WKH RSHUDWLQJ IUHTXHQF\ DQG PLQLPL]LQJ WKH UHTXLUHG YROWDJH VZLQJ RI WKH GLYLGHUfV LQSXW VLJn QDO LV GLVFXVVHG $OVR XVLQJ 6&/ D QHZ SURJUDPPDEOH GLYLGHU LV GHYHORSHG ZKLFK DOORZV WKH VWDUWXS VWDWH RI WKH GLYLGHU WR EH FRQWUROOHG 6XFK SURJUDPPLQJ ZLOO GHFUHDVH WKH V\Vn WHPDWLF FORFN VNHZ SURYLGHG WKDW WKH GLYLGHUV DUH SURSHUO\ LQLWLDOL]HG DQG V\QFKURQRXVO\ UHOHDVHG IURP WKH VWDUWXS VWDWH 7KH LQLWLDOL]DWLRQ DQG VWDUWXS PHWKRGRORJ\ IRU WKH GLYLGn HUV DUH DOVR SUHVHQWHG LQ WKLV FKDSWHU )UHTXHQF\ 'LYLGHU 8VLQJ 6RXUFH&RXSOHG /RJLF &LUFXLW 'HVFULSWLRQ )LJXUH Df VKRZV D EORFN GLDJUDP RI D GLYLGHE\HLJKW f IUHTXHQF\ GLYLGHU HPSOR\LQJ 6&/ 7KH GLYLGHU FRQVLVWV RI WKUHH FDVFDGHG GLYLGHUV ZLWK GXDOSKDVH LQSXWV DQG RXWSXWV (DFK GLYLGHU VKRZQ LQ )LJXUHV Ef DQG Ff FRQVLVWV RI WZR 6&/ 'ODWFKHV LQ D PDVWHUVODYH FRQILJXUDWLRQ >:DU )OR22D +XQ2O@ 7KH RXWSXWV DUH WLHG WR WKH LQSXWV ZLWK LQYHUWHG SKDVH WR SHUIRUP D WRJJOH RSHUDWLRQ 6RXUFHFRXSOHG ORJLF DOVR NQRZQ DV 026 FXUUHQWPRGH ORJLF 0&0/ff KDV VRPH GLVWLQFW DGYDQWDJHV IRU WKH ZLUHOHVV LQWHUFRQQHFW DSSOLFDWLRQ )LUVW D VPDOOOHYHO LQSXW YROWDJH FDQ EH GHWHFWHG RQ WKH FORFN LQSXW 7KLV LV GXH WR WKH GLIIHUHQWLDO VWUXFWXUH RI WKH ODWFK DV ZHOO DV WKH PHFKDQLVP RI GLYLVLRQ ZKLFK ZLOO EH GLVFXVVHG VKRUWO\ 7KHUHIRUH WKH LQSXW VHQVLWLYLW\ IRU WKH GLYLGHU LV KLJK LPSURYLQJ WKH 0'6 RI WKH UHFHLYHU 6HFRQG WKH ODWFK FDQ EH GHVLJQHG VXFK WKDW WKH RXWSXWV GR QRW VZLQJ UDLOWRUDLO GHFUHDVLQJ WKH

PAGE 88

Ef 'LYLGHU 7RJJOH ))f 4E 4 0DVWHU 6ODYH )LJXUH %ORFN GLDJUDPV RI Df GLYLGHU FRQVLVWLQJ RI WKUHH FDVFDGHG GLYLGHUV DQG Ef GLYLGHU FRQVLVWLQJ RI WZR 'ODWFKHV LQ PDVWHUVODYH Ff 6FKHPDWLF RI GLYLGHU LPSOHPHQWHG XVLQJ VRXUFHFRXSOHG ORJLF

PAGE 89

SRZHU FRQVXPSWLRQ IRU D JLYHQ RSHUDWLQJ IUHTXHQF\ 7KLUG WKH XVH RI D GLIIHUHQWLDO VWUXFn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f 7KH YROWDJH 9ELDV LV JURXQGHG WR LQFUHDVH WKH PD[LPXP RSHUDWLQJ IUHTXHQF\ E\ RSHUDWLQJ 0 LQ WKH OLQHDU UHJLRQ ZKLFK ORZHUV WKH 5& WLPH FRQVWDQW IRU QRGHV 4 DQG 4E )XUWKHUPRUH ZKHQ WKH YROWDJH RI 4 LV LQFUHDVLQJ GXULQJ D ORZ WR KLJK WUDQVLWLRQf WKH GUDLQWRVRXUFH YROWDJH DQG WKH RXWSXW UHVLVWDQFH RI 0 LQ OLQHDU UHJLRQf DUH GHFUHDVLQJ 7KLV QRQOLQHDULW\ RI 0 KHOSV WR SXOO XS 4 WR LWV ORJLF KLJK ZKLFK LQ WXUQ KHOSV WR LQFUHDVH WKH PD[LPXP RSHUDWLQJ IUHTXHQF\ DW WKH VDPH SRZHU FRQVXPSWLRQ 6LPLODUO\ ZKHQ 4 LV GHFUHDVLQJ WKH QRQOLQHDULW\ WHQGV WR SXVK GRZQ 4 WR LWV ORJLF ORZ ,Q DGGLWLRQ WR WKH DGYDQWDJH RI KLJK VSHHG RSHUDWLRQ WKLV WRSRORJ\ KDV UHGXFHG SRZHU FRQVXPSWLRQ FRPn SDUHG WR WKH WRSRORJ\ LQ >&UD@ GXH WR WKH HOLPLQDWLRQ RI WKH IROGHG GLRGHFRQQHFWHG WUDQVLVWRUV IURP 4 DQG 4E WR JURXQG /DWFKHG 2SHUDWLQJ 0RGH 7KH GLYLGHU LPSOHPHQWHG ZLWK 6&/ FDQ RSHUDWH LQ RQH RI WZR PRGHV 7KH ILUVW PRGH LV WKH VWDQGDUG GLJLWDO RSHUDWLRQ RI ODWFKLQJ ZKLOH WKH VHFRQG LV DQ LQMHFWLRQORFNHG PRGH ZKLFK ZLOO EH SUHVHQWHG LQ WKH QH[W VHFWLRQ 7KH ODWFKLQJ PRGH RI RSHUDWLRQ RFFXUV

PAGE 90

ZKHQ WKH LQSXW FORFN VLJQDOV KDYH D ODUJH YROWDJH VZLQJ +HUH 0 DFW DV DQ HYDOXDWLRQ VWDJH ZKHQ WKH ODWFK LV WUDQVSDUHQW DQG UHJHQHUDWLYH SDLU 0 DFW DV D KROG VWDJH ZKHQ WKH ODWFK LV RSDTXH 7KXV IRU &/. KLJK WKH PDVWHU LV WUDQVSDUHQW DQG WKH VODYH LV RSDTXH ZLWK 4c DQG 4EL WDNLQJ WKH YDOXH RI DQG 'E )RU &/. ORZ WKH PDVWHU LV RSDTXH DQG WKH VODYH LV WUDQVSDUHQW DQG QRGHV 4c DQG 4EL DUH SDVVHG WR WKH RXWSXW $V LV HYLGHQW IURP WKLV GHVFULSWLRQ FDVFDGLQJ WZR OHYHOVHQVLWLYH 'ODWFKHV LQ D PDVWHUVODYH FRQILJXUDWLRQ UHVXOWV LQ DQ HGJHWULJJHUHG 'IOLSIORS '))f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n UHQWV WKURXJK 0M DQG 0 DUH QRW QHFHVVDULO\ HTXDO 7KH FXUUHQW YDOXHV GHSHQG RQ WKH VL]n LQJ RI 0M DQG 0 DQG WKH LQSXW FORFN VLJQDOV 7KH FXUUHQW SXOOLQJ 4Ec ORZ GXULQJ HYDOXDWLRQ IRU WKH PDVWHU ')) LV WKHQ PRGHOHG DV IM,Lf ZKHUH IM GHSHQGV RQ WKH VL]LQJ RI 0 DQG 0 WKH YROWDJHV 4 DQG 4E DQG WKH HYDOXDWLRQ FXUUHQW ,M 7KXV 4EL KDV D YROWDJH GXULQJ HYDOXDWLRQ HTXDO WR /2:(YDO 9P:25SP f

PAGE 91

/LNHZLVH GXULQJ KROG RSHUDWLRQ RI WKH PDVWHU 'ODWFK WKH FXUUHQW SXOOLQJ 4EL ORZ LV PRGn HOHG DV IOf ZKHUH I GHSHQGV RQ WKH VL]LQJ RI 0 DQG 0 WKH YROWDJHV 4c DQG 4EL DQG WKH KROG FXUUHQW 7KXV 4EL KDV D YROWDJH GXULQJ KROG HTXDO WR /2:+ROG 9GGI,f5S0 f ,I IOf IL,Mf WKH RXWSXW YROWDJH VZLQJ RI WKH ODWFK LV GHFUHDVHG DQG WKH YROWDJHV 4c DQG 4EL PLJKW QRW EH ODUJH HQRXJK WR GULYH WKH QH[W PDVWHUVODYH IOLSIORS )RU D JLYHQ IM,Mf WR IOf UDWLR D ORZHU IUHTXHQF\ OLPLW IRU WKH GLYLGHU H[LVWV VHW E\ KRZ IDVW WKH QRGHV FDQ FKDQJH IURP /2:(YDL WR /2:+RLG $V ORQJ DV WKH LQSXW IUHn TXHQF\ LV KLJK HQRXJK WKHQ /2:(YDL FDQQRW IXOO\ FKDUJH WR /2:+ROG DQG WKH GLYLGHU IXQFWLRQV SURSHUO\ 7KLV ORZHUIUHTXHQF\ OLPLW ZDV REVHUYHG LQ VLPXODWLRQV ZKHUH WKH PLQLPXP IUHTXHQF\ ZDV DSSUR[LPDWHO\ RQHHLJKWK RI WKH PD[LPXP IUHTXHQF\ IRU WKH FKRVHQ WUDQVLVWRU ZLGWKV 7R PDNH WKH TXDVLG\QDPLF ODWFK IXOO\ VWDWLF ILUVW WDLO FXUUHQW VRXUFHV VKRXOG EH DGGHG DQG :M VKRXOG EH VHW HTXDO WR : 7KLV WKHQ PDNHV ,M HTXDO WR 6HFRQG WKH ZLGWKV RI 0 VKRXOG DOO EH HTXDO >0XU@ IRUFLQJ IM,Lf WR EH HTXDO WR IOf ,QMHFWLRQ /RFNLQJ RI 6&/ )UHTXHQF\ 'LYLGHU 7KH VHFRQG PRGH RI RSHUDWLRQ IRU WKH 6&/ GLYLGHU LV LQMHFWLRQ ORFNLQJ ,QMHFn WLRQ ORFNLQJ LV WKH SURFHVV RI V\QFKURQL]LQJ DQ RVFLOODWRU ZLWK DQ LQFLGHQW VLJQDO 2VFLOODn WRUV FDQ EH LQMHFWLRQORFNHG WR WKH IXQGDPHQWDO >$GO 3DF 8]X@ VXEKDUPRQLFV >=KD@ DQG VXSHUKDUPRQLFV >6FK 5DW@ RI WKH LQSXW VLJQDO 7KLV VHFWLRQ ZLOO H[DPn LQH WKH RVFLOODWLRQ RI WKH 6&/ GLYLGHU DQG GLVFXVV KRZ WKH GLYLGHU FDQ EH LQMHFWLRQORFNHG 7KH DFWXDO WKHRU\ EHKLQG LQMHFWLRQ ORFNLQJ LV SUHVHQWHG LQ $SSHQGL[ ( 7KLV DSSHQGL[ UHYLHZV WKH EDVLF RSHUDWLRQ RI LQMHFWLRQORFNHG RVFLOODWRUV ,/2fVf DQG SUHVHQWV WKH

PAGE 92

ORFNLQJ UDQJH LWV GHSHQGHQFH RQ VLJQDO OHYHO WKH VWHDG\VWDWH SKDVH HUURU EHWZHHQ LQSXW DQG RXWSXW VLJQDOV DQG WKH SKDVH QRLVH RI ,/2fV 2VFLOODWLRQ RI 6&/ 'LYLGHU 7R XQGHUVWDQG WKH SKHQRPHQRQ RI LQMHFWLRQ ORFNLQJ WKH RVFLOODWLRQ RI WKH GLYLGHU ZLWK QR LQSXW SUHVHQW PXVW ILUVW EH XQGHUVWRRG :KHQ WKH LQSXW FORFN VLJQDO VZLQJV DUH YHU\ VPDOO WKHQ &/. DQG &/.E DUH DSSUR[LPDWHO\ HTXDO DW WKHLU FRPPRQPRGH YDOXH 7KXV ERWK WKH PDVWHU DQG VODYH ODWFKHV DUH VHPLWUDQVSDUHQW DOORZLQJ VLJQDOV WR SURSDJDWH WKURXJK ERWK ODWFKHV 5HIHUULQJ WR )LJXUH Ef LI WKH GHOD\ IURP WKH JDWH WR WKH GUDLQ RI 0 LV HTXDO WR [SG WKHQ WKH WRWDO GHOD\ DURXQG WKH ORRS LV HTXDO WR [SG 7KXV WKH GLYLGHU RVFLOODWHV DW D IUHTXHQF\ HTXDO WR f§ DQG WKH VLJQDO DW WKH GUDLQ RI 0 ZLOO ODJ WKH VLJQDO DW WKH JDWH RI 0 E\ r 7KH GLYLGHU FLUFXLW FDQ WKXV EH OLNHQHG WR D WZRVWDJH ULQJ RVFLOODWRU ZLWK DQ LQYHUVLRQ LQ WKH ORRS $GGLWLRQDO XQGHUVWDQGLQJ RI WKH RVFLOODWRU FDQ EH JDLQHG E\ VROYLQJ IRU WKH VPDOOVLJQDO ORRS JDLQ RI WKH GLYLGHU )LUVW DVVXPH WKH WZR ODWFKHV DUH LGHQWLFDO DQG WKDW HDFK ODWFK LV V\PPHWULF LH : : : : DQG : :f &RQVLGHU WKH FDVH ZKHQ &/. DQG &/.M DUH KHOG DW D FRQVWDQW DQG HTXDO YDOXH )RU WKLV FDVH 0@ DQG 0 EHFRPH FXUUHQW VRXUFHV &URVVFRXSOHG WUDQVLVWRUV 0 FDQ EH UHSUHVHQWHG DV QHJDWLYH UHVLVWDQFHV ZLWK YDOXHV RI OJFF ZKHUH JFF LV WKH WUDQVFRQGXFWDQFH RI 0 7KH 3026 ORDG FDQ EH PRGHOHG DV D FRQGXFWDQFH JS )LQDOO\ WKH WRWDO FDSDFLWDQFH DW WKH RXWSXW QRGHV RI WKH ODWFK FDQ EH UHSUHVHQWHG E\ &/ )LJXUH Ff FDQ WKHQ EH VLPSOLILHG WR )LJXUH E\ $Q DFWXDO H[SUHVVLRQ IRU WKH SURSDJDWLRQ GHOD\ WKURXJK HDFK ODWFK ZDV QRW GHULYHG VLQFH WKLV GHOD\ FDQ UHDGLO\ DQG PRUH DFFXUDWHO\ EH REWDLQHG ZLWK 63,&(

PAGE 93

)LJXUH 5HSUHVHQWDWLRQ RI VRXUFHFRXSOHG ODWFK DV D IXOO\GLIIHUHQWLDO DPSOLILHU DQG WKH GLYLGHU DV DQ RVFLOODWRU UHSUHVHQWLQJ HDFK VRXUFHFRXSOHG ODWFK DV D IXOO\GLIIHUHQWLDO DPSOLILHU 7KH VPDOOVLJQDO ORRS JDLQ 7f RI WKH GLYLGHU FDQ EH VKRZQ WR EH DV IROORZV 7 6P 6SJFF -A/ f )RU RVFLOODWLRQV WR EHJLQ WKH SKDVH RI 7 VKRXOG EH r DQG ,7, &KRRVLQJ 7, HQVXUHV VWDUWXS DQG FDXVHV WKH DPSOLWXGH RI RVFLOODWLRQ WR LQFUHDVH XQWLO QRQOLQHDULWLHV ZLWKLQ WKH DPSOLILHU OLPLW WKH DPSOLWXGH 7KLV DPSOLWXGH FRQGLWLRQ UHTXLUHV JP fR&/ )RU WKH SKDVH RI 7 WR EH r WKH SKDVHVKLIW WKURXJK HDFK GLIIHUHQWLDO DPSOLILHU VKRXOG EH r 7KLV FRUUHVSRQGV WR J3 JFF 7KLV %DUNKDXVHQ FULWHULRQ IRU RVFLOODWRUV DSSOLHV WR VPDOOVLJQDO VLQXVRLGDO RVFLOODWLRQV +RZHYHU ZLWK D ORRS JDLQ JUHDWHU WKDQ RQH WKH RVFLOn ODWRU ZLOO EHFRPH QRQOLQHDU DQG LWV DPSOLWXGH DQG IUHTXHQF\ ZLOO EH GHWHUPLQHG E\ WKH ODUJHVLJQDO FKDUDFWHULVWLFV RI WKH FLUFXLW 7KH GLYLGHU WKHQ EHKDYHV DV D UHOD[DWLRQ RVFLOODn WRU DOVR NQRZQ DV DQ DVWDEOH PXOWLYLEUDWRUf >6HG@ 5HOD[DWLRQ RVFLOODWRUV DUH K\VWHUHWLF FRPSDUDWRUV ZKLFK FKDUJH DQG GLVFKDUJH FDSDFLWDQFHV ORFDWHG LQ D IHHGEDFN ORRS 7KH RVFLOODWLRQ IUHTXHQF\ RI D UHOD[DWLRQ RVFLOODn WRU GHSHQGV RQ WKH FDSDFLWDQFH EHLQJ FKDUJHG DV ZHOO DV WKH FXUUHQW GULYH RI WKH

PAGE 94

FRPSDUDWRU ,W LV LQWHUHVWLQJ WR QRWH WKDW DQ\ ULQJ RVFLOODWRU ZKRVH SURSDJDWLRQ GHOD\ LV VHW E\ WKH FKDUJLQJ RI 5& FLUFXLWV FDQ EH FRQVLGHUHG WR EH D UHOD[DWLRQ RVFLOODWRU >(JD@ WKXV WKH WZR PRGHOV XVHG WR GHVFULEH WKH GLYLGHUfV RVFLOODWLRQ DUH HTXLYDOHQW 'HVFULSWLRQ RI ,QMHFWLRQ /RFNLQJ 6LQFH WKH GLYLGHU FDQ VHOIRVFLOODWH LW LV SRVVLEOH WR LQMHFWLRQORFN WKH GLYLGHU WR DQ LQSXW VLJQDO ZKRVH IUHTXHQF\ LV FORVH WR HLWKHU WKH IXQGDPHQWDO VXEKDUPRQLF RU VXSHUn KDUPRQLF RI WKH RVFLOODWLRQ IUHTXHQF\ 7KXV IRU D LQMHFWLRQORFNHG IUHTXHQF\ GLYLGHU (/)'f WKH LQSXW VLJQDO LV FORVH WR WKH VHFRQG KDUPRQLF RI WKH RVFLOODWLRQ IUHTXHQF\ KHUHLQ UHIHUUHG WR DV IL62 7KH ORFNLQJ PHFKDQLVP RI WKH RVFLOODWRU IRU DQ LQSXW VLJQDO DW WZLFH WKH QDWXUDO IUHTXHQF\ D VXSHUKDUPRQLFf LV FUHDWHG E\ QRQOLQHDU SURSHUWLHV LQ WKH RVFLOODWRU ZKLFK JHQHUDWH LQWHUPRGXODWLRQ SURGXFWV FORVH WR WKH QDWXUDO IUHTXHQF\ >5DW@ 5HIHUULQJ WR )LJXUH OFf WKH GUDLQ QRGH RI 0M RVFLOODWHV DW WZLFH WKH QDWXUDO IUHTXHQF\ KHQFH WKLV QRGH LV QDWXUDOO\ VXLWHG WR LQMHFWLQJ D VLJQDO DW WKLV GRXEOH IUHTXHQF\ $Q DOWHUn QDWLYH DQG HTXLYDOHQW XQGHUVWDQGLQJ RI WKH FLUFXLW LV WKDW LQMHFWLRQ ORFNLQJ LV EHLQJ XVHG WR V\QFKURQL]H WKLV fGRXEOHIUHTXHQF\f RVFLOODWRU GUDLQ RI 0Mf DW D IUHTXHQF\ QHDU IL6R 7KLV XQGHUVWDQGLQJ LV PRUH VWUDLJKWIRUZDUG DQG DYRLGV WKH FRPSOLFDWLRQV VXUURXQGLQJ VXSHUn KDUPRQLF LQMHFWLRQ ORFNLQJ 7KH UDQJH RI LQSXW IUHTXHQFLHV RYHU ZKLFK WKH ,/)' FDQ EH ORFNHG RU fSXOOHGf LV D IXQFWLRQ RI WKH LQSXW YROWDJH VZLQJ DQG WKH IUHTXHQF\GHSHQGHQW HQHUJ\ORVV LQ WKH RVFLOn ODWRU $SSHQGL[ ( VKRZV WKDW WKH YROWDJH VZLQJ RI WKH ORFNLQJ VLJQDO 9/f LV UHODWHG WR WKH ORFNLQJ IUHTXHQF\ UDQJH $FRf DQG WKH RVFLOODWRU RXWSXW YROWDJH VZLQJ 9f DV IROORZV A ‘ Y R f

PAGE 95

+HUH 94 LV WKH YROWDJH VZLQJ RI WKH GRXEOHIUHTXHQF\ RVFLOODWRU GUDLQ RI 0Mf DQG $ LV HTXDO WR WKH GHULYDWLYH RI SKDVHVKLIW DURXQG WKH RVFLOODWRU ZLWK UHVSHFW WR IUHTXHQF\ )RU D WXQHG RVFLOODWRU $ LV HTXDO WR 4R $SSHQGL[ ) VKRZV WKDW WKH HTXLYDOHQW TXDOLW\ IDFWRU RI D ULQJRVFLOODWRU LV DSSUR[LPDWHO\ HTXDO WR W 7KXV DQ LQMHFWLRQORFNHG ULQJ RVFLOODWRU KDV WKH IROORZLQJ LQSXWVZLQJ WR ORFNLQJUDQJH UHODWLRQVKLS DOVR NQRZQ DV WKH LQSXW VHQn VLWLYLW\ RI WKH GLYLGHUf f $ YROWDJH FRQYHUVLRQ JDLQ FDQ EH GHILQHG IRU WKH IUHTXHQF\ GLYLGHU HTXDO WR WKH UDWLR RI WKH RXWSXW VZLQJ WR WKH LQSXW VZLQJ 6LQFH WKH JDLQ RFFXUV DW WZR GLIIHUHQW IUHTXHQn FLHV LW LV UHIHUUHG WR DV FRQYHUVLRQ JDLQ 7KLV JDLQ FDQ EH UHODWHG WR f E\ UHODWLQJ WKH RXWSXW VZLQJ RI WKH GLYLGHU WR WKH VZLQJ RI WKH GRXEOHIUHTXHQF\ RVFLOODWRU 7KXV WKH FLUn FXLW KDV fLQILQLWHf JDLQ DW IL62 $OVR f VKRZV WKDW WKH IXUWKHU DZD\ WKH LQSXW IUHTXHQF\ LV IURP IL62 WKH ODUJHU WKH LQSXW YROWDJHVZLQJ KDV WR EH 7R FRQWURO WKH RVFLOODWLRQ IUHTXHQF\ DQG FRQYHUVLRQ JDLQ RI WKH GLYLGHU WKH WUDQVLVn WRU VL]HV LQSXW FRPPRQPRGH YROWDJH DQG VXSSO\ YROWDJH FDQ DOO EH DGMXVWHG 5HIHUULQJ WR )LJXUH Ff WKH VSHHG RI VZLWFKLQJ QRGH 4c EL +,*+ GHSHQGV RQ WKH VL]H DQG ,9JVO RI 0 QRWH 9GG GHSHQGHQFHf ZKLOH WKH VSHHG RI VZLWFKLQJ QRGH 4c EL /2: GHSHQGV RQ WKH VL]HV RI 0 DQG 0L DV ZHOO DV WKH FRPPRQPRGH YROWDJH RQ WKH FORFN LQSXWV )RU ERWK FDVHV WKH VZLWFKLQJ VSHHG LV GHFUHDVHG E\ WKH SUHVHQFH RI UHJHQHUDWLYH SDLU 0 VLQFH WKH SRVLWLYH IHHGEDFN KDV WR EH RYHUFRPH E\ WKH FKDUJLQJ RU GLVFKDUJLQJ PHFKDQLVP 7KHUHn IRUH DOO RI WKH WUDQVLVWRU VL]HV WKH VXSSO\ YROWDJH DQG WKH FORFN FRPPRQPRGH YROWDJH GHWHUPLQH WKH VHOIRVFLOODWLRQ IUHTXHQF\ IRU WKH (/)' 1RWH WKDW WKH WUDQVLVWRU VL]HV DUH

PAGE 96

XVHG WR DGMXVW WKH RVFLOODWLRQ IUHTXHQF\ GXULQJ GHVLJQ ZKLOH WKH YROWDJHV FDQ EH XVHG WR DGMXVW WKH RVFLOODWLRQ IUHTXHQF\ GXULQJ FLUFXLW WHVWLQJ 6LPXODWLRQ RI ,QMHFWLRQ/RFNHG 'LYLGHU 7KLV VHFWLRQ ZLOO SUHVHQW VLPXODWLRQ UHVXOWV IRU DQ ,/)' VKRZLQJ WKH IL62 DQG GHPRQVWUDWLQJ WKH GHSHQGHQFH RI WKH LQSXW VZLQJ RQ WKH ORFNLQJ UDQJH 7R H[WUDFW IL62f WKH GLYLGHU LV VLPXODWHG XVLQJ D WUDQVLHQW DQDO\VLV ZLWK ERWK &/. DQG &/.E KHOG DW DSSUR[LPDWHO\ PLGVXSSO\ $V ZLWK RWKHU RVFLOODWRU VLPXODWLRQV LQLWLDO FRQGLWLRQV DUH UHTXLUHG ZKHUH QRGHV 4c DQG 4 DUH LQLWLDOL]HG WR WKH VXSSO\ DQG 4EL DQG 4E DUH LQLWLDOL]HG WR JURXQG )LQDOO\ VLQFH WKLV LV DQ DQDORJ VLPXODWLRQ WKH 63,&( WROHUDQFHV VKRXOG EH GHFUHDVHG DV FRPSDUHG WR D VWDQGDUG GLJLWDO VLPXODWLRQ )LJXUH VKRZV WKH VLPXODWHG RVFLOODWLRQ RI WKH (/)' IRU D _LP &026 SURFHVV 6LQFH 9GG 9 &/. DQG &/.E DUH VHW WR a 9 7KH RXWSXW RI WKH ILUVW GLYLGHU LV RVFLOODWLQJ DW *+] ZKLOH WKH IROORZLQJ GLYLGHUV DUH ORFNHG DW DQG )LJXUH 6LPXODWLRQ RI VHOIRVFLOODWLRQ RI VRXUFHFRXSOHG GLYLGHU 7KH FORFN LQSXWV DUH KHOG DW 9 FDXVLQJ WKH ILUVW GLYLGHU WR RVFLOODWH DW *+] IRU 9GG 9IL62 *+]f

PAGE 97

'LYLGHU VWRSV ORFNLQJ 2XWSXW 7LPH QVf Ef )LJXUH Df 6LPXODWHG LQSXW VHQVLWLYLW\ RI _OP &026 GLYLGHU Ef ([SRQHQWLDOO\ GHFUHDVLQJ LQSXW VLJQDO DW *+] DQG RXWSXW VLJQDO RI GLYLGHU LOOXVWUDWLQJ GLYLGHUfV GHSHQGHQFH RQ LQSXW YROWDJH VZLQJ *+] UHVSHFWLYHO\ 7KXV IL62 LV HTXDO WR *+] 7KLV VLPXODWLRQ FDQ EH LWHUDWHG WKURXJK YDULRXV GHYLFH VL]HV WR VHW IL6 WR WKH GHVLUHG RSHUDWLQJ IUHTXHQF\ $GGLWLRQDOO\ WKLV VLPXODWLRQ FDQ EH UXQ IRU WKH VHFRQG DQG WKLUG GLYLGHUV WR VHW WKHLU UHVSHFWLYH VHOIRVFLOODWLRQ IUHTXHQFLHV WR RQHIRXUWK DQG RQHHLJKWK RI WKH GHVLUHG RSHUDWLQJ IUHn TXHQF\ 1RWH WKDW IRU DQ DV\QFKURQRXV GLYLGHU DQ\ RI WKH WKUHH GLYLGHUV FDQ RVFLOn ODWH SURYLGHG WKDW WKH LQSXW VLJQDO OHYHO WR HDFK GLYLGHU LV VPDOO +RZHYHU IRU WKH W\SLFDO FDVH ZKHUH WKH ILUVW GLYLGHU RVFLOODWHV ZLWK D ODUJH RXWSXW VLJQDO WKH IROORZLQJ GLYLGHUV RSHUDWH LQ D ODWFKHG PRGH 7KH LQSXWVZLQJ YHUVXV ORFNLQJIUHTXHQF\UDQJH UHODWLRQVKLS LV VKRZQ LQ )LJXUH Df ZKLFK SORWV WKH VLPXODWHG LQSXW SHDNWRSHDN YROWDJH VZLQJ YHUVXV LQSXW IUHn TXHQF\ 7R REWDLQ WKHVH LQSXW VZLQJV XVLQJ D VLQJOH VLPXODWLRQ D GDPSHG VLQXVRLG ZDV LQMHFWHG LQWR WKH GLYLGHU DV VKRZQ LQ )LJXUH Ef (YHQWXDOO\ WKH GLYLGHU FDQQRW ORFN WR WKH LQSXW VLJQDO DW ZKLFK SRLQW WKH LQSXW YROWDJH VZLQJ LV GHWHUPLQHG 7KLV VLPXODWLRQ LV

PAGE 98

UHSHDWHG IRU HYHU\ GHVLUHG LQSXW IUHTXHQF\ $V FDQ EH VHHQ IURP )LJXUH Df WKH UHTXLUHG LQSXW VZLQJ GLSV DW WKH VHOIRVFLOODWLRQ IUHTXHQF\ *+]f DQG WKHQ UDSLGO\ LQFUHDVHV IRU KLJKHU IUHTXHQFLHV 7KH ORFNLQJ IUHTXHQF\ UDQJH IRU D JLYHQ LQSXW VZLQJ LV HTXDO WR WKH GLVWDQFH EHWZHHQ WKH OHIW DQG ULJKW SRUWLRQV RI WKH FXUYH 7KLV VLPXODWLRQ YHULn ILHV WKDW ODUJHU LQSXW YROWDJH VZLQJV UHVXOW LQ ODUJHU ORFNLQJ UDQJHV ,PSOLFDWLRQV RI ,QMHFWLRQ /RFNLQJ IRU &ORFN 'LVWULEXWLRQ 7KH IUHTXHQF\ GLYLGHU ZLOO RSHUDWH LQ WKH LQMHFWLRQORFNHG PRGH IRU VPDOOOHYHO LQSXW VLJQDOV DQG LQ WKH ODWFKHG PRGH IRU ODUJH LQSXW VLJQDOV 7KHUHIRUH WR PLQLPL]H WKH PLQLPXP GHWHFWDEOH VLJQDO 0'6f RI WKH UHFHLYHU WKH GLYLGHU VKRXOG EH LQMHFWLRQORFNHG ZLWK LWV LQSXWUHIHUUHG VHOIRVFLOODWLRQ IUHTXHQF\ QHDU WKH RSHUDWLQJ IUHTXHQF\ RI WKH UHFHLYHU 7KLV SURYLGHV DQ DOWHUQDWLYH LQWHUSUHWDWLRQ IRU WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\Vn WHP 8VLQJ WKLV LQWHUSUHWDWLRQ HDFK UHFHLYHU LV LQLWLDOO\ D IUHHUXQQLQJ RVFLOODWRU 7KH WUDQVn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fV 0'6 ZRXOG EH VHYHUHO\ GHJUDGHG

PAGE 99

)LJXUH &/. 4 &/. 4 ,QSXWV f f &/.E 4E !f§ &/.E 4E &/. 4 f &/.E 4E 6&/ 'LYLGHE\ F?N 2 UON 2 UON 2 +N Q 4E LU 4E LU 4E aLU 4E +! 2XWSXW 763& 'LYLGHE\ %ORFN GLDJUDP RI GLYLGHQ )LJXUH 6FKHPDWLF RI WUXH VLQJOHSKDVH FORFNHG 763&f ODWFK 7KH ILUVW GLYLGHU LV LPSOHPHQWHG ZLWK 6&/ ZKLOH WKH ORZHU IUHTXHQF\ DV\Qn FKURQRXV GLYLGHU LV LPSOHPHQWHG XVLQJ WUXHVLQJOHSKDVHFORFNHG 763&f ORJLF >
PAGE 100

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

PAGE 101

6LJQDO *HQHUDWRU )LJXUH %LDV7 $ %LDV7 $ *QG6LJ r &RXSOHU 6LJ*QG W 3UREH 6LJQDO *URXQG 3UREH 2fVFRSH 4 6HPL5LJLG &DEOHV 0HDVXUHPHQW VHWXS XVHG WR FKDUDFWHUL]H GXDOSKDVH IUHTXHQF\ GLYLGHU WKH RXWSXW EXIIHU VZLWFKHV LWV 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r FRXSOHU ZKLFK DFWV DV D EDOXQ 7KXV WKH XQEDODQFHG VLJQDO IURP WKH VZHHSHU LV FRQn YHUWHG WR GLIIHUHQWLDO LQSXW VLJQDOV IRU WKH GLYLGHU ZKRVH SRZHU LV LGHDOO\f G% OHVV WKDQ WKH SRZHU DW WKH LQSXW RI WKH EDOXQ 7R VHW WKH FRPPRQPRGH YROWDJH IRU WKH LQSXW VLJQDOV ELDV7V DUH SODFHG RQ WKH GLIIHUHQWLDO DQG FRPPRQPRGH LQSXWV IRU WKH r FRXSOHU 7KLV LV SUHIHUDEOH WR SODFLQJ WKH ELDV7V RQ WKH RXWSXW RI WKH FRXSOHU VLQFH SKDVH PLVPDWFK LQ WKH ELDV7V ZLOO EH HOLPLQDWHG 6HPLULJLG FDEOHV DUH WKHQ XVHG WR FRQQHFW WKH EDOXQ WR WKH JURXQGVLJQDOVLJQDOJURXQG SUREH )LQDOO\ WKH GLYLGHU LV SUREHG DQG WKH RXWSXW LV PHDn VXUHG XVLQJ DQ RVFLOORVFRSH $ NH\ UHTXLUHPHQW IRU WKLV PHDVXUHPHQW VHWXS LV WKDW WKH

PAGE 102

Df Ef )LJXUH 0HDVXUHG UHVXOWV IRU _LP GLYLGHU Df 0D[LPXP LQSXW IUHTXHQF\ DQG SRZHU FRQVXPSWLRQ YHUVXV VXSSO\ YROWDJH Ef ,QSXW VHQVLWLYLW\ YHUVXV IUHTXHQF\ YHUVXV VXSSO\ YROWDJH SKDVH GLIIHUHQFH EHWZHHQ WKH WZR LQSXW VLJQDOV KDV WR EH DV FORVH WR r DV SRVVLEOH HOVH WKH GLYLGHU ZLOO QRW RSHUDWH XS WR LWV PD[LPXP RSHUDWLQJ IUHTXHQF\ 7KH PHDVXUHG PD[LPXP LQSXW IUHTXHQF\ YHUVXV VXSSO\ YROWDJH IRU WKH GLYLGHU LV VKRZQ LQ )LJXUH Df $OVR VKRZQ LV WKH PHDVXUHG SRZHU FRQVXPSWLRQ YHUVXV VXSSO\ YROWDJH IRU WKH GLYLGHU FRUH $W D 9 VXSSO\ WKH GLYLGHU RSHUDWHV XS WR *+] ZKLOH FRQVXPLQJ P: 7KH PHDVXUHG RXWSXW ZDYHIRUP IRU D *+] LQSXW VLJQDO LV VKRZQ LQ )LJXUH 7KH UHGXFHG RXWSXW VZLQJ LV GXH WR GULYLQJ D 4 RXWSXW ORDG $W 9 WKH GLYLGHU FDQ RSHUDWH XS WR a *+] ZKLOH FRQVXPLQJ P: 7KHVH DUH H[FHOOHQW VSHHG DQG SRZHU WUDGHRIIV IRU D _LP WHFKQRORJ\ 7R PHDVXUH WKH LQSXW VHQVLWLYLW\ RI WKH GLYLGHU WKH PLQLPXP LQSXW SRZHU OHYHO IRU HDFK LQSXW IUHTXHQF\ ZDV UHFRUGHG IRU PXOWLSOH VXSSO\ YROWDJHV )LJXUH Ef VKRZV WKH UHVXOWDQW SORWV )RU D 9 VXSSO\ D 9SOFBSN LQSXW VLJQDO LV UHTXLUHG WR RSHUDWH DW WKH

PAGE 103

)LJXUH 0HDVXUHG RXWSXW ZDYHIRUP IURP SP &026 IUHTXHQF\ GLYLGHU ,QSXW IUHTXHQF\ LV *+] IRU 9GG 9 DQG RXWSXW IUHTXHQF\ LV 0+] PD[LPXP LQSXW IUHTXHQF\ RI *+] 7KLV LQSXW VZLQJ LV DOPRVW HTXDO WR WKH VXSSO\ YROWn DJH WKXV WKH 0'6 RI WKH UHFHLYHU IRU WKLV IUHTXHQF\ ZRXOG EH YHU\ KLJK +RZHYHU LI WKH LQSXW IUHTXHQF\ LV GHFUHDVHG WR *+] WKHQ RQO\ D aP9SNBSN LQSXW VLJQDO LV UHTXLUHG GLUHFWO\ FRUUHVSRQGLQJ WR DQ LPSURYHPHQW LQ WKH UHFHLYHU 0'6 7KH PHDVXUHPHQWV VKRZQ LQ )LJXUH Ef ZHUH QRW REWDLQHG DW YHU\ ORZ IUHTXHQFLHV WKXV WKH FKDUDFWHULVWLF LQSXW VHQVLWLYLW\ SORW DV VKRZQ LQ )LJXUH Df ZDV QRW FRPSOHWHO\ REWDLQHG 6SHFLILFDOO\ IRU IUHTXHQFLHV EHORZ IL62 WKH UHTXLUHG LQSXW YROWDJH VZLQJ VKRXOG LQFUHDVH 7KLV WUHQG LV RQO\ EHJLQQLQJ WR EH HYLGHQW IRU WKH 9 DQG 9 SORWV LQ )LJXUH Ef 7KH ILVRfV DUH DSSUR[LPDWHO\ DQG *+] IRU 9GG DQG 9 UHVSHFWLYHO\ 7KHVH QXPEHUV GR QRW DJUHH ZHOO ZLWK WKH VLPXODWHG YDOXHV IURP 63,&( GXH WR DQ LQFRQVLVWHQF\ EHWZHHQ WKH VLPXODWHG WUDQVLVWRU PRGHO FDSDFLWDQFH DQG WKH PRGHO FDSDFLWDQFH ILW WR WKH DFWXDO ZDIHU GDWD 7DEOH VXPPDUL]HV WKH PHDVXUHG UHVXOWV IRU WKH _LP IUHTXHQF\ GLYLGHU

PAGE 104

7DEOH 6XPPDU\ RI PHDVXUHG FKDUDFWHULVWLFV IRU 6&/ IUHTXHQF\ GLYLGHUV )UHTXHQF\ 'LYLGHU _OP 'LYLGHU _LP 'LYLGHU 'LYLVLRQ 5DWLR I $PD[ *+] # 9 *+] # 9 *+] # 9 *+] # 9 *+] # 9 IL62 *+] # 9 *+] # 9 *+] # 9 *+] # 9 3RZHU P: # 9 P: # 9 P: # 9 P: # 9 P: # 9 $ *+] ALP &026 6&/ 'LYLGHU &LUFXLW ,PSOHPHQWDWLRQ %XLOGLQJ RQ WKH UHVXOWV DFKLHYHG LQ WKH LP &026 WHFKQRORJ\ D *+] VRXUFHFRXSOHG IUHTXHQF\ GLYLGHU ZDV LPSOHPHQWHG XVLQJ D _OP &026 WHFKQROn RJ\ ZLWK VL[ OD\HUV RI FRSSHU LQWHUFRQQHFWV >)OR2OD@ 7KLV WHFKQRORJ\ ZDV REWDLQHG IURP 80& DV SDUW RI WKH 65& &RSSHU 'HVLJQ &KDOOHQJH VSRQVRUHG E\ WKH 6HPLFRQGXFWRU 5HVHDUFK &RUSRUDWLRQ 65&f 1RYHOOXV 6SHHG)DP,3(& DQG 80& 7KH EORFN GLDJUDP VKRZQ LQ )LJXUH LV VLPLODU WR WKDW LQ )LJXUH H[FHSW WKH GLYLVLRQ UDWLR LV QRZ DQG WKHUH DUH WZR 763& ORZIUHTXHQF\ GLYLGHUV LQ RSSRVLWH SKDVH 7KLV LV GRQH WR SURn YLGH WZR RXWSXW VLJQDOV IRU WHVWLQJ DQG WULJJHULQJ DV ZHOO DV WR PLQLPL]H WKH VXSSO\OLQH ERXQFH $OO RI WKH GHYLFHV IRU WKH 6&/ DQG 763& GLYLGHUV VKRZQ LQ )LJXUHV DQG DUH LPSOHPHQWHG ZLWK FKDQQHO OHQJWKV RI LP DQG FKDQQHO ZLGWKV JLYHQ LQ 7DEOH )LJXUH VKRZV D GLH SKRWRJUDSK RI WKH GLYLGHU ZLWK D GLH VL]H RI [ LP 2QFKLS S) E\SDVV FDSDFLWRUV DUH LQFOXGHG EHWZHHQ 9GG DQG JURXQG IRU WKH

PAGE 105

763& 'LYLGHE\fV )LJXUH %ORFN GLDJUDP RI GLYLGHU LPSOHPHQWHG LQ SP &026 WHFKQRORJ\ )LJXUH 'LH SKRWRJUDSK RI SP *+] VRXUFHFRXSOHG GLYLGHU 7DEOH 7UDQVLVWRU VL]LQJ IRU SP IUHTXHQF\ GLYLGHU 6&/ 'LYLGHU 7UDQVLVWRUV )LUVW 6HFRQG 7KLUG :M Z Z : SP SP SP : Z Z Z SP SP SP ZZZZ SP SP SP :ZZZ SP SP SP 763& 'LYLGHU 6L]HV $UH 6DPH IRU HDFK 763& 'LYLGHU : : Z Z Z Z Z Z SP :M:Q SP Z SP 2XWSXWV

PAGE 106

)LJXUH 0HDVXUHG UHVXOWV RI ILP IUHTXHQF\ GLYLGHU Df 0D[LPXP LQSXW IUHTXHQF\ DQG VHOIRVFLOODWLRQ IUHTXHQF\ YHUVXV 9GG Ef ,QSXW VHQVLWLYLW\ YHUVXV IUHTXHQF\ IRU 9GG DQG 9 6&/ GLYLGHU 763& GLYLGHUV DQG RXWSXW EXIIHU $JDLQ VHSDUDWH VXSSOLHV DUH XVHG IRU WKH 6&/ DQG 763& GLYLGHUV WR SUHYHQW DQ\ ORZIUHTXHQF\ VXSSO\ QRLVH IURP WKH ODWWHU VWDJHV RI WKH GLYLGHU IURP FRUUXSWLQJ WKH KLJKIUHTXHQF\ RSHUDWLRQ RI WKH ILUVW GLYLGHU $OVR  UHVLVWRUV DUH LQFOXGHG EHWZHHQ WKH WZR LQSXWV RI WKH GLYLGHU WR JURXQG 0HDVXUHG 5HVXOWV 8VLQJ WKH PHDVXUHPHQW VHWXS LQ )LJXUH WKH PD[LPXP LQSXW IUHTXHQF\ DQG SRZHU FRQVXPSWLRQ IRU GLYLGHU FRUHf YHUVXV VXSSO\ YROWDJH SORWV IRU WKH GLYLGHU ZHUH REWDLQHG DQG DUH VKRZQ LQ )LJXUH Df )RU D 9 9GG WKH PD[LPXP LQSXW IUHn TXHQF\ LV *+] DQG WKH SRZHU FRQVXPSWLRQ LV P: 7KH PD[LPXP LQSXW IUHn TXHQF\ LQFUHDVHV WR *+] ZKHQ 9GG LV LQFUHDVHG WR 9 ZKLOH WKH SRZHU FRQVXPSWLRQ LQFUHDVHV WR P: 7KHVH DUH H[FHOOHQW SRZHUVSHHG WUDGHRIIV )LJXUH VKRZV WKH RXWSXW ZDYHIRUPV IRU ILQ DQG *+] DW 9GG DQG 9

PAGE 107

9GG 9 IcQ *+] 9GG 9 ILQ *+] UHVSHFWLYHO\ 7KH RXWSXW IUHTXHQFLHV DUH DQG 0+] UHVSHFWLYHO\ 2QFH DJDLQ WKH UHGXFHG RXWSXW VZLQJ LV IURP GULYLQJ D RXWSXW ORDG 7KH IcVR LV SORWWHG LQ )LJXUH Df YHUVXV VXSSO\ YROWDJH $W 9 ILV LV *+] ZKLOH DW 9 IL6 LV *+] ,W FDQ EH VHHQ WKDW WKH GLIIHUHQFH EHWZHHQ WKH PD[Ln PXP RSHUDWLQJ IUHTXHQF\ DQG ILVR LV a *+] )LJXUH Ef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

PAGE 108

SURSDJDWLRQ GHOD\V RI WKH FORFN VLJQDOV KDV WR EH XQGHU D FHUWDLQ WROHUDQFH W\SLFDOO\ OHVV WKDQ b RI D FORFN F\FOHf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
PAGE 109

PP Df ( ( LQ &?PP Ef )LJXUH ,OOXVWUDWLRQV WR VFDOHf RI JULGV XSRQ ZKLFK FORFN UHFHLYHUV VKRXOG EH SODFHG Df *ULG VSDFLQJ RI RQH ZDYHOHQJWK IURP WUDQVPLWWHU ZLWK VROLG IRU ZDYHOHQJWK LQ VLOLFRQ GLR[LGH DQG GDVKGRW IRU ZDYHOHQJWK LQ VLOLFRQ Ef *ULG VSDFLQJ RI RQH HLJKWK RI D ZDYHOHQJWK LQ R[LGH XVHG ZKHQ WKH GLYLGHUV DUH SURJUDPPHG WR LQLWLDO VWDWHV RI WR IRU GLYLGHE\f DOO RI WKH UHFHLYHUV WR EH D PXOWLSOH RI D ZDYHOHQJWK DZD\ IURP WKH WUDQVPLWWHU 7KLV FRQn FHSW LV LOOXVWUDWHG WR VFDOHf LQ )LJXUH Df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n HUV FDQ EH SODFHG RQ VXFK D JULG )URP H[DPLQLQJ )LJXUH Df LW FDQ EH VHHQ WKDW WKH JULG LV QRW GHQVH HQRXJK WR DFFRPPRGDWH FORFN UHFHLYHUV LQ D IHDVLEOH DUUDQJHPHQW 7KLV LV ZKHUH WKH FRQFHSW RI SURn JUDPPLQJ RU LQLWLDOL]LQJ WKH IUHTXHQF\ GLYLGHU LV XVHIXO )UHTXHQF\ GLYLGHUV DUH EDVLFDOO\

PAGE 110

FRXQWHUV ZKHUH WKH GLYLVLRQ UDWLR LV HTXDO WR WKH FRXQW HJ DQ GLYLGHU FRXQWV IURP WR ZKLOH RXWSXWWLQJ D RQH DQG IURP WR ZKLOH RXWSXWWLQJ D ]HURf ,I UHFHLYHUV FORVH WR WKH WUDQVPLWWHU DUH LQLWLDOL]HG WR D VPDOOHU FRXQW DQG UHFHLYHUV IDU IURP WKH WUDQVPLWWHU DUH LQLWLDOL]HG WR D ODUJHU FRXQW WKH PLVPDWFK LQ WLPHRIIOLJKW GHOD\V DPRQJ WKH UHFHLYHUV ZLOO WHQG WR EH FDQFHOOHG *UDSKLFDOO\ WKLV UHGXFHV WKH JULG VSDFLQJ E\ WKH GLYLVLRQ UDWLR +HQFH IRU DQ GLYLGHU WKH UHFHLYHUV FDQ EH SODFHG RQ D JULG ZLWK VSDFLQJ $ ZKLFK WKHQ HTXDOL]HV WKH WLPHRIIOLJKW GHOD\V DQG GHFUHDVHV WKH V\VWHPDWLF VNHZ WR 7KLV LV LOOXVWUDWHG LQ )LJXUH Ef ZKLFK VKRZV WKH JULG IRU ZDYHOHQJWK LQ VLOLFRQGLR[LGH 6LQFH UDQGRP FORFN VNHZ DOVR H[LVWV LW LV LPSRVVLEOH WR GHVLJQ D JULG WR HTXDOL]H WKH WRWDO VNHZ +RZHYHU DVVXPLQJ WKDW WKH WRWDO VNHZ LQ WKH V\VWHP FDQ EH HVWLPDWHG RU GHWHFWHG WKHQ WKH VWDUWXS VWDWH RI WKH GLYLGHU FDQ EH PRGLILHG DFFRUGLQJO\ 7KHUHIRUH WKH ZRUVWFDVH WRWDO VNHZ SHUFHQWDJH WKHQ EHFRPHV sf§A ZKHUH 0 LV WKH GLYLVLRQ UDWLR RI WKH GLYLGHU )RU DQ SURJUDPPDEOH GLYLGHU WKH PD[LPXP RXWSXW VNHZ LV b 1RWH WKDW LI WKH VNHZ FDQ EH HVWLPDWHG WKHQ WKH UHFHLYHUV GR QRW KDYH WR EH SODFHG RQ D JULG 7KH LQLWLDOL]DWLRQ PHWKRGRORJ\ LV FRQFHSWXDOO\ LOOXVWUDWHG LQ )LJXUH ZKLFK VKRZV WZR FORFN UHFHLYHUV ZLWK DQ LQLWLDO VNHZ RI b $OVR VKRZQ DUH WKH SRVVLEOH VWDUWXS VWDWHVSKDVHV IRU WKH GLYLGHUV LQ WKH FORFN UHFHLYHU $V FDQ EH VHHQ E\ VHOHFWn LQJ WKH FRUUHFW VWDWHV IRU HDFK GLYLGHU UHTXLULQJ LQLWLDO VNHZ HVWLPDWLRQ RU GHWHFWLRQf WKH VNHZ FDQ EH UHGXFHG IURP b WR b IRU WKLV SDUWLFXODU H[DPSOHf 7KHUHIRUH FLUFXLWV DUH UHTXLUHG WR LQLWLDOL]H WKH 6&/ GLYLGHU DQG D PHWKRGRORJ\ LV QHHGHG WR VWDUWXS WKH FORFN GLVWULEXWLRQ V\VWHP LQ SKDVH 7KLV LV FRPSOLFDWHG E\ WKH IDFW WKDW WKH 6&/ GLYLGHUV ZLOO VHOIRVFLOODWH LQ WKH DEVHQFH RI DQ LQSXW VLJQDO 7KHUHIRUH D ZD\ LV QHHGHG WR SUHYHQW WKHVH FLUFXLWV IURP VHOIRVFLOODWLQJ GXULQJ LQLWLDOL]DWLRQ DV ZHOO

PAGE 111

)LJXUH ,OOXVWUDWLRQV RI FRQFHSW RI GLYLGHU SURJUDPPLQJ WR UHGXFH V\VWHPDWLF VNHZ 7ZR UHFHLYHU VWDUW ZLWK D VNHZ RI b DQG E\ VHOHFWLQJ WKH FRUUHFW VWDWH RI HDFK WKH VNHZ FDQ EH UHGXFHG WR b IRU WKLV H[DPSOHf )LQDOO\ WKH LQLWLDOL]DWLRQ FLUFXLWU\ VKRXOG QRW ORDG GRZQ WKH GLYLGHU ZKLFK ZRXOG ORZHU WKH PD[LPXP IUHTXHQF\ DQG FKDQJH WKH FRQYHUVLRQ JDLQ YHUVXV IUHTXHQF\ 6\VWHP 6WDUW8S 0HWKRGRORJ\ ,W KDV EHHQ VKRZQ FRQFHSWXDOO\ KRZ SURJUDPPDEOH IUHTXHQF\ GLYLGHUV FDQ EH XVHG WR UHGXFH DQG SRWHQWLDOO\ HOLPLQDWH V\VWHPDWLF FORFN VNHZ 7KHVH SURJUDPPDEOH GLYLGHUV ZLOO UHO\ RQ VRPH LQLWLDOL]DWLRQ VLJQDO ,1,f ZKLFK ZLOO GHQRWH ZKHWKHU WKH GLYLGHU LV WR EH LQLWLDOL]HG RU WR RSHUDWH IUHHO\ 7KHUHIRUH D VWDUWXS PHWKRGRORJ\ LV QHHGHG IRU WKH V\VWHP WR LQLWLDOL]H DOO RI WKH UHFHLYHUV WR WKH FRUUHFW VWDWH RU FRXQW &17f DQG WKHQ WR UHOHDVH WKHVH UHFHLYHUV IURP WKH LQLWLDOL]HG VWDWH WR WKH IUHHUXQQLQJ VWDWH )RU SURJUDPPLQJ WKH GLYLGHUV WKHUH DUH WZR EDVLF PHWKRGVf§G\QDPLF DQG VWDWLF 2QH LV WR FRUUHFW WKH V\VWHP G\QDPLFDOO\ ZLWK IHHGEDFN +HUH WKH SKDVHRIIVHW EHWZHHQ HDFK FORFN UHFHLYHU DQG LWV QHDUHVW QHLJKERUV FDQ EH PHDVXUHG ZLWK D SKDVH GHWHFWRU DQG WKHQ WKH &17 RI WKH GLYLGHUV FDQ EH DGMXVWHG DFFRUGLQJO\ 7KH V\VWHP LV EDVLFDOO\

PAGE 112

RSHUDWLQJ DV D GHOD\ORFNHG ORRS HJ >6XW@f ZKHUH WKH SKDVH RI DOO RI WKH UHFHLYHUV ZLOO EH G\QDPLFDOO\ DGMXVWHG WR D VWHDG\VWDWH SKDVH ORFN 7KLV V\VWHP LV YHU\ VLPLODU WR D GLVn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n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n SULDWH &17 YDOXHV DUH KHOG LQ PHPRU\ DQG XVHG GXULQJ QRUPDO V\VWHP RSHUDWLRQ 7KLV SKDVHFRUUHFWLRQ LV VWDWLF UDWKHU WKDQ G\QDPLF KDSSHQLQJ RQO\ RQFH GXULQJ V\VWHP RSHUDn WLRQ 7KLV VWDWLF SKDVH FRUUHFWLRQ KDV WKH GUDZEDFN RI DFFXPXODWLQJ SKDVH HUURUV ,Q RWKHU ZRUGV RQFH WKH V\VWHP LV SURJUDPPHG WKHUH LV QR RWKHU ZD\ WR FRUUHFW IRU SKDVH HUURUV FDXVHG E\ QRLVH RU V\VWHP IOXFWXDWLRQV HJ SRZHU OLQH IOXFWXDWLRQVf 7KHUHIRUH WKH VLJ QDOWRQRLVH UDWLR RI WKH V\VWHP ZRXOG KDYH WR EH KLJK 7KH WUDLQLQJ PRGH ZRXOG KDYH WKH IROORZLQJ FRQFHSWXDO VWHSV )LUVW WKH FORFN WUDQVPLWWHU EURDGFDVWV WKH JOREDO FORFN *&/.f XVHG DV D WUDLQLQJ VLJQDO 6HFRQG HDFK

PAGE 113

FORFN UHFHLYHU GHWHFWV *&/. ZLWK ,1, DQG &17 DQG JHQHUDWHV D ORFDO FORFN /&/.f 7KHVH /&/.fV KDYH UDQGRP DQG V\VWHPDWLF FORFN VNHZV 7KLUG WKH VNHZ LV PHDn VXUHG DJDLQVW WKH UHFHLYHUfV QHLJKERUV ZLWK SKDVH GHWHFWRUV DQG FRQYHUWHG WR D GLJLWDO ZRUG ZLWK D ELW DQDORJWRGLJLWDO FRQYHUWHU ELWV IRU VWDWHVf 1RWH WKDW D PHWKRG LV QHHGHG WR SUHYHQW PRGHORFNLQJ >3UD@ ZKHUH V\VWHPDWLF RIIVHWV FDQ RFFXU LI WKH SKDVH RIIVHWV IURP PXOWLSOH UHFHLYHUV DUH VLPSO\ DGGHG LH RIIVHW IURP RQH UHFHLYHU LV r DQG RIIVHW IURP DQRWKHU UHFHLYHU LV r UHVXOWLQJ LQ D QHW RI ]HUR SKDVH RIIVHW KRZHYHU DOO RI WKH UHFHLYHUV ZRXOG EH RXW RI SKDVHf 7R SUHYHQW PRGHORFNLQJ D QRQPRQRWRQLF SKDVH GHWHFn WRU UHVSRQVH LV UHTXLUHG >3UD@ )RXUWK *&/. LV WXUQHG RII DQG HDFK UHFHLYHU LV LQLWLDOn L]HG WR WKH DSSURSULDWH VWDWH ZLWK ,1, DQG &17 ELW ZRUG IURP SUHYLRXV VWHS )LIWK HLWKHU DOO RI WKH FORFN UHFHLYHUV fUHSRUWf EDFN WR WKH WUDQVPLWWHU ZKHQ WKH\ KDYH EHHQ LQLn WLDOL]HG RU WKH V\VWHP ZDLWV D FHUWDLQ DPRXQW RI WLPH IRU LQLWLDOL]DWLRQ WR RFFXU 6L[WK IROn ORZLQJ WUDLQLQJ *&/. LV UHOHDVHG DQG WKH UHFHLYHUV VKRXOG EHJLQ GHWHFWLQJ WKH VLJQDO DQG JHQHUDWLQJ WKH V\QFKURQL]HG /&/.fV +RZHYHU D PHWKRG LV VWLOO UHTXLUHG WR UHOHDVH WKH LQLn WLDOL]DWLRQ VLJQDO RQFH *&/. LV SUHVHQW DW WKH LQSXW WR WKH IUHTXHQF\ GLYLGHU 7KLV LV FRPn SOLFDWHG E\ WKH IDFW WKDW WKH IUHTXHQF\ GLYLGHU VHOIRVFLOODWHV ZLWKRXW WKH LQSXW VLJQDOV SUHVHQW $SSURDFKHV WR V\QFKURQRXVO\ UHOHDVH WKH GLYLGHUV IURP WKH LQLWLDOL]HG VWDWHV ZLOO EH WKH VXEMHFW RI IXWXUH ZRUN &LUFXLWU\ IRU 3URJUDPPDEOH 6&/ 'LYLGHU 7KH GLYLGHUV XVHG XS WR WKLV SRLQW KDYH EHHQ DV\QFKURQRXV FRQVLVWLQJ RI FDVFDGHG GLYLGHUV 7KHUHIRUH DQ LQLWLDOL]DWLRQ PHWKRGRORJ\ IRU D GLYLGHU LV UHTXLUHG EHIRUH D Q GLYLGHU FDQ EH LQLWLDOL]HG )LJXUH Df VKRZV D EORFN GLDJUDP RI D GLYLGHU

PAGE 114

0DVWHU 6ODYH 4 R R 7f§ f§ R R 7 7 R 7 R 7f§ R f§ R 7f§ f &' e WR &2 WR &2 &2 &2 &2 &2 &2 &2 &2 &_ )LJXUH Df %ORFN GLDJUDP RI IUHTXHQF\ GLYLGHU XVLQJ 6&/ DQG Ef LGHDO LQSXW DQG RXWSXW ZDYHIRUPV IRU GLYLGHU DQG DVVRFLDWHG VWDWHV HPSOR\LQJ 6&/ $ VFKHPDWLF RI WKH 6&/ '))V LV VKRZQ LQ )LJXUH OFf )LJXUH Ef VKRZV LGHDO ZDYHIRUPV IRU WKH LQSXW FORFN VLJQDO DV ZHOO DV IRU DOO IRXU RI WKH fGUDLQf QRGHV LQ WKH GLYLGHU ODEHOOHG 4O 4O 4 DQG 4 ([DPLQLQJ WKHVH ZDYHIRUPV LW FDQ EH VHHQ WKDW WKH GLYLGHU KDV IRXU SRVVLEOH VWDWHV 6WDWH YDULDEOHV 6 DQG 62 DUH LQWURGXFHG WR LGHQWLI\ WKHVH VWDWHV ZKHUH 6, LV WKH GHVLUHG YDOXH RI WKH GLYLGHU LQSXW LH FORFN VLJQDOf DQG 62 LV GHVLUHG YDOXH RI WKH GLYLGHU RXWSXW GHILQHG WR EH 4 7KH LQLWLDOL]DWLRQ FLUFXLWU\ VKRXOG VHW DOO RI WKH GUDLQ QRGHV IRU HDFK RI WKH IRXU SRVVLEOH VWDWHV 7KLV UHVXOWV LQ WKH WUXWK WDEOH VKRZQ LQ 7DEOH ZKLFK LV REWDLQHG ZLWK WKH KHOS RI )LJXUH Ef 7KH GUDLQ QRGHV FDQ WKHQ EH GHILQHG ORJLFDOO\ E\ WKH VWDWH YDULn DEOHV DV VKRZQ LQ WKH WDEOH 7KHUHIRUH GXULQJ LQLWLDOL]DWLRQ GUDLQ QRGHV 4 DQG 4 DUH VHW WR 62 DQG 62 UHVSHFWLYHO\ ZKLOH GUDLQ QRGHV 4O DQG 4O DUH VHW WR WKH H[FOXVLYHQRU

PAGE 115

7DEOH 7UXWK WDEOH IRU LQLWLDOL]DWLRQ FLUFXLWU\ 6WDWH 2XWSXW 'UDLQ 1RGHV 6, &ORFNf 62 2XWSXWf 4 4 4O 4O O /RJLF )XQFWLRQ 4 62 4 62 4O 6, k62 T7 VL k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k62,1, f 9E 6, k62,1, f 9F ,1, f 9G ,1, f 7KH GUDLQ QRGHV DUH WKH LQYHUVH RI WKHVH IXQFWLRQV 6LQFH 3026 WUDQVLVWRUV ZLOO RQO\ EH DEOH WR SXOO WKH GUDLQ QRGHV KLJK ZKHQ WKHLU LQSXWV DUH ORZ 7KHUHIRUH YHU\ ZHDN 1026 WUDQVLVWRUV 0f DUH LQFOXGHG WR DFWLYDWH WKH UHJHQHUDWLYH SDLUV 0 f ZKHQ ,1, :KHQ

PAGE 116

6 62 RQH RI WKH 3026 WUDQVLVWRUV SXOOV LWV GUDLQ XS WR 9GG WKH SRVLWLYH IHHGEDFN LQ WKH UHJHQHUn DWLYH SDLU FDXVHV WKH RWKHU GUDLQ WR EH SXOOHG GRZQ WR JURXQG 6LQFH WKHVH 3026 WUDQVLVWRUV RULJLQDOO\ KDG WKHLU JDWHV JURXQGHG WR DFW DV SVHXGR1026 ORDGV WKH LQLWLDOn L]DWLRQ GDWD DUH PXOWLSOH[HG ZLWK D JURXQG VLJQDO ZKHUH WKH ,1, VLJQDO FRQWUROV WKH PXOWLn SOH[HU 7KXV IRU ,1, 9[ LV GHILQHG DV DERYH DQG IRU ,1, 9[ 7KH ORJLF EORFNV DUH LPSOHPHQWHG XVLQJ FRPSOLPHQWDU\ SDVVWUDQVLVWRU ORJLF &3/f >%HO@ LQ D ORJLF DUUD\ ,QSXWV WR WKH DUUD\ VKRZQ LQ )LJXUH DUH 6, 62 ,1, DQG WKHLU FRPSOHPHQWV ,QYHUWHUV DUH UHTXLUHG DW WKH RXWSXW RI HDFK ORJLF IXQFWLRQ WR

PAGE 117

6 6 6 6 62 62 62 VR )LJXUH 6FKHPDWLF RI LQLWLDOL]DWLRQ FLUFXLWU\ LPSOHPHQWHG ZLWK FRPSOHPHQWDU\ SDVVWUDQVLVWRU ORJLF UHVWRUH WKH YROWDJH OHYHOV $ EHQHILW RI &3/ LV WKDW RQO\ 1026 WUDQVLVWRUV DUH XVHG IRU WKH ORJLF DUUD\ DOORZLQJ IRU YHU\ FRPSDFW OD\RXW )LQDOO\ WR UHDOL]H WKH SURJUDPPDEOH GLYLGHU WKUHH SURJUDPPDEOH GLYLGHUV DUH FDVFDGHG 6LQFH WKH VWDWHFRQWURO LQSXWV 6 DQG 62 DUH WKH GHVLUHG LQSXW DQG RXWSXW YDOn XHV RI HDFK GLYLGHU WKHQ 62 IRU WKH ILUVW GLYLGHU LV HTXDO WR 6, IRU WKH VHFRQG GLYLGHU DQG 62 IRU WKH VHFRQG GLYLGHU LV HTXDO WR 6, IRU WKH WKLUG GLYLGHU 7KH UHVXOWDQW GLYLGHU EORFN GLDJUDP LV VKRZQ LQ )LJXUH ,QSXW 6, IRU WKH ILUVW GLYLGHU LV VHW WR JURXQG VLQFH VLPXODWLRQV VKRZ WKDW D WRWDO RI RQO\ HLJKW VWDWHV UDWKHU WKDQ VL[WHHQ DUH SRVVLEOH IRU WKLV GLYLGHU LH 6, \LHOGV VWDWHV DQG 6, \LHOGV WKH VDPH VWDWHVf 6WDWH FRQWURO LQSXWV ;;M;R IRUP D ELQDU\ QXPEHU HTXDO WR WKH GHVLUHG FRXQW

PAGE 118

;I 6WDWH ,QLWLDOL]DWLRQ ;L ; ,QSXWV 6 62 HON 4 3URJ 'LYLGHU HONE 4E ,1, f§2 6 62 HON 4 3URJ 'LYLGHU HONE 4E ,1, 6 62 HON 4 3URJ 'LYLGHU 2OFONE 4E ,1, 2XWSXWV )LJXUH %ORFN GLDJUDP RI SURJUDPPDEOH IUHTXHQF\ GLYLGHU &17f RI WKH GLYLGHU LH GHFLPDO f 7KLV &17 ZLOO FRPH IURP WKH ELW $' FRQYHUWHU DIWHU WKH SKDVH GHWHFWRU 6LPXODWHG 5HVXOWV %DVLF IXQFWLRQDOLW\ RI WKH LQLWLDOL]DWLRQ FLUFXLW LV GHPRQVWUDWHG LQ )LJXUH ZKLFK VKRZV WKH VLPXODWLRQ UHVXOWV IRU DQ GLYLGHU IRU ;;M; UDQJLQJ IURP ELQDU\ WR $V FDQ EH VHHQ WKH RXWSXW SKDVH FDQ EH YDULHG LQ LQFUHPHQWV RI W UF0f RU r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n WLDOL]H WR DQ\ RI WKH HLJKW SRVVLEOH VWDWHV 7KLV PHDQV WKDW WKH DFWXDO LQLWLDOL]HG YDOXH IRU D JLYHQ &17 GHSHQGV RQ WKH FXUUHQW VWDWH RI WKH GLYLGHU ZKHQ ,1, WUDQVLWLRQV KLJK

PAGE 119

)LJXUH 2XWSXW ZDYHIRUPV RI GLYLGHU IRU HDFK VWDUWXS FRQGLWLRQ 7KLV HIIHFW FDQ EH GHPRQVWUDWHG DQG DQDO\]HG XVLQJ 63,&( )LJXUH VKRZV WKH VLPXODWHG VLJQDOV IRU WZR SURJUDPPDEOH GLYLGHUV RSHUDWLQJ DW DQ LQSXW IUHTXHQF\ RI *+] ,Q WKLV VLPXODWLRQ WKH FORFN VLJQDOV DUH DOZD\V SUHVHQW DQG WKH ,1, VLJQDO LV EHLQJ VZLWFKHG UHSHDWHGO\ IURP KLJK WR ORZ DV LQGLFDWHG LQ WKH SORWV 2QH GLYLGHU LV DOZD\V LQLn WLDOL]HG WR &17 ZKLOH WKH RWKHU GLYLGHU LV DOZD\V LQLWLDOL]HG WR &17 ,GHn DOO\ WKLV VKRXOG FRUUHVSRQG WR D SKDVH RIIVHW RI r EHWZHHQ WKH WZR GLYLGHUV HDFK WLPH WKH GLYLGHU LV LQLWLDOL]HG $V FDQ EH VHHQ IURP )LJXUH DIWHU WKH ILUVW WZR LQLWLDOL]DWLRQV WKH RXWSXW ZDYHIRUPV WDNH WKH GHVLUHG r GLIIHUHQFH +RZHYHU DIWHU WKH WKLUG LQLWLDOL]DWLRQ WKH RXWSXW SKDVH GLIIHUHQFH LV r 7KH IDLOXUH RI WKH FLUFXLW FDQ EH VHHQ E\ H[DPLQLQJ WKH RXWSXW ZDYHIRUPV IURP WKH ILUVW VHFRQG DQG WKLUG GLYLGHUV UHIHUUHG WR DV WKH DQG RXWSXWVf 6LQFH RQH GLYLGHU LV LQLWLDOL]HG WR VWDWH DQG WKH RWKHU LV LQLWLDOL]HG WR VWDWH HDFK WLPH ,1, WKH DQG RXWSXWV VKRXOG EH DOO ]HURV RU DOO RQHV IRU WKH GLYLGHU DQG GLYLGHU UHVSHFWLYHO\ )URP WKH ZDYHIRUPV LQ )LJXUH

PAGE 120

,OO L ‘r 2XWSXWV L/M 7Q 8 Y/O UW 9 n £Q n n Q n ,1, Q m RVA LQ Q £Q Q Q 2XWSXWV Q f Q 2XWSXWV Q f Q f Q 2XWSXWV Q 7LPH Q 6LPXODWHG ZDYHIRUPV RI WZR SURJUDPPDEOH GLYLGHUVf§SURJUDPPHG WR VROLG OLQHVf DQG GDVKHG OLQHVfaIRU UHSHDWHG LQLWLDOL]DWLRQV 2Q WKH WKLUG LQLWLDOL]DWLRQ WKH FLUFXLW IDLOV DV LQGLFDWHG GXULQJ WKH WKLUG LQLWLDOL]DWLRQ WKH RXWSXW RI WKH GLYLGHU VWD\V ORZ FKDQJLQJ WKH SURJUDPPHG VWDWH IURP WR 7KLV DFFRXQWV IRU WKH H[WUD r SKDVH VKLIW 'HSHQGLQJ RQ WKH VWDWH WKDW HDFK GLYLGHU LV LQ ZKHQ ,1, WUDQVLWLRQV KLJK WKH FLUFXLW PD\ QRW LQLWLDOL]H WR WKH FRUUHFW VWDWH )LJXUH Df VKRZV SORWV RI WKH GUDLQ YROWn DJHV IRU WKH WKLUG GLYLGHU LH RXWSXWVf DURXQG WKH IDLOXUH WLPH RI WKH WKLUG

PAGE 121

LQ ‘ Q Df )LJXUH L L L ‘ L ‘ +& YD +,* AD 4 +,*+ f§>A AD 0 SXOO XS ,,0D 9E /2: 4 &/., L6 SXOOn GRZQ OE' nD W0D0__,1, A Af21fA f &/.E AD Q Q Q Q Q 7LPH Ef Df 6LPXODWHG ZDYHIRUPV RI SURJUDPPDEOH GLYLGHU GXULQJ VWDWHGHSHQ LQLWLDOL]DWLRQ IDLOXUH Ef 6FKHPDWLF RI ')) LQGLFDWLQJ FRQIOLFW RI VWDWHV LQLWLDOL]DWLRQ 7KHVH ZDYHIRUPV 4 4 DQG DUH LGHQWLILHG LQ WKH VFKHPDWLF VKRZQ LQ )LJXUH Ef 7KH GHVLUHG VWDWHV RI 4 DQG 4 GXULQJ LQLWLDOL]DWLRQ DUH KLJK DQG ORZ UHVSHFWLYHO\ WKXV 9E DQG 9D DUH DQG UHVSHFWLYHO\ )RU 4 WR EH KLJK GXULQJ LQLWLDOL]Dn WLRQ LW KDV WR EH SXOOHG XS WKURXJK 0D ZLWK 9E ORZ +RZHYHU GHSHQGLQJ RQ WKH YDOXHV RI 4 DQG WKHUH ZLOO DOVR EH D GLVFKDUJH SDWK WR JURXQG IRU 4 7KH SXOOGRZQ SDWKV DUH SRVn VLEOH VLQFH 0LD DD DOO RQ $V VHHQ LQ )LJXUH Df 4 LV a 9 WKXV WKH SXOOGRZQ SDWK WKURXJK 0D LV TXLWH VWURQJ 7KH DFWXDO YDOXH RI 4 LV VHW WKURXJK D YROWDJH GLYLGHU ZLWK WKH YDOXH GHSHQGLQJ RQ WKH FXUUHQW VWDWH RI WKH GLYLGHU ZKHQ ,1, VZLWFKHV KLJK 7R UHFWLI\ WKLV SUREOHP ERWK WKH HYDOXDWH DQG KROG SRUWLRQV RI WKH ')) VKRXOG EH WXUQHG RII ZKHQ ,1, $OVR ERWK KLJK DQG ORZ YDOXHV KDYH WR EH ZULWWHQ WR WKH GUDLQ QRGHV RI WKH GLYLGHU DV RSSRVHG WR WKH 3026 RQO\ EHLQJ DEOH WR SXOO WKH QRGH KLJKf )LJn XUH Df VKRZV WKH QHZ SURJUDPPDEOH ')) WR EH XVHG LQ WKH GLYLGHU ZLWK WKHVH

PAGE 122

YE Df 9 GG ')) IRU KLJKVSHHG Ef )LJXUH Df 6FKHPDWLF RI IL[HG SURJUDPPDEOH ')) IRU ORZHU IUHTXHQF\ GLYLGHE\ DQG GLYLGHE\f Ef 6FKHPDWLF RI LPSURYHG KLJKIUHTXHQF\ GLYLGHE\f SURJUDPPDEOH ')) PRGLILFDWLRQV 7KH HYDOXDWH DQG KROG SRUWLRQV DUH WXUQHG RII GXULQJ ,1, E\ LQFOXGLQJ WZR WUDQVLVWRUV GULYHQ E\ (1, LQ VHULHV ZLWK WKH &/. DQG &/.E WUDQVLVWRUV SHUIRUPLQJ D ORJLFDO $1' IXQFWLRQf 7R ZULWH ERWK KLJK DQG ORZ YDOXHV WR WKH GUDLQ QRGHV WKH 3026 WUDQVLVWRUV DUH UHSODFHG ZLWK LQYHUWHUV 7KH VL]H RI WKH 1026 WUDQVLVWRU LQ WKLV LQYHUWHU LV NHSW VPDOO 6LPXODWLRQV VKRZ WKDW XVLQJ WKLV QHZ ')) LQ WKH ORZHU IUHTXHQF\ DQG VWDJHV HOLPLQDWHV WKH VWDWHGHSHQGHQW LQLWLDOL]DWLRQ IDLOXUH +RZHYHU XVLQJ WKLV ')) LQ WKH KLJKIUHTXHQF\ VWDJH GHFUHDVHV WKH RSHUDWLQJ IUHTXHQF\ GXH WR WKH LQFUHDVHG FDSDFLn WLYH ORDG IURP WKH LQYHUWHU 7KXV WKH PRGLILHG ')) VKRZQ LQ )LJXUH Ef LV XVHG IRU WKH ILUVW VWDJH 6LPXODWLRQV VKRZ WKDW WKLV VWDJH ZLOO LQLWLDOL]H SURSHUO\ GXH WR WKH VL]LQJ UDWLR EHWZHHQ WKH 3026 DQG 1026 +RZHYHU LQ IXWXUH LPSOHPHQWDWLRQV RI WKLV GLYLGHU

PAGE 123

WKH LQLWLDOL]DWLRQ KDV WR EH FDUHIXOO\ FKHFNHG IRU WKH ILUVW GLYLGHU DQG WKH DSSURSULDWH ')) VKRZQ LQ )LJXUH Df RU Ef VKRXOG EH VHOHFWHG 7KH FRUUHFW RSHUDWLRQ RI WKH SURJUDPPDEOH GLYLGHU ZLWK WKH QHZ '))fV ZDV YHULILHG LQ VLPXODWLRQV 7KH VFHQDULR LV WKH VDPH DV LQ )LJXUH ZKHUH WZR GLYLGHUV DUH SURJUDPPHG WR &17 DQG ZLWK ,1, VZLWFKLQJ IURP KLJK WR ORZ UHSHDWHGO\ )LJXUH VKRZV WKH VLPXODWHG UHVXOWV $V FDQ EH VHHQ HDFK WLPH ,1, JRHV KLJK WKH RXWn SXWV RI WKH GLYLGHU JR KLJK DQG WKRVH RI WKH GLYLGHU JR ORZ 7KHUH LV QR ORQJHU DQ\ VLWXDWLRQ ZKHUH WKH RXWSXW QRGHV RI WKH '))fV DUH PLVWDNHQO\ SXOOHG ORZ ZKHQ WKH\ VKRXOG EH KLJK $V D UHVXOW WKH RXWSXWV EHWZHHQ WKH GLYLGHUV DOZD\V PDLQWDLQ D r SKDVH VKLIW 7KXV WKHVH QHZ '))fV UHPRYH WKH GHSHQGHQFH RI LQLWLDOL]DWLRQ RQ WKH FXUUHQW VWDWH RI WKH GLYLGHU DQG VKRXOG DOORZ WKH VWDUWXS VWDWH RI WKH GLYLGHU WR EH SURJUDPPHG LQ r LQFUHPHQWV DV RULJLQDOO\ SODQQHG 7HVWLQJ 0HWKRGRORJ\ $ PHWKRGRORJ\ LV QHHGHG WR FKDUDFWHUL]H WKH FORFN VNHZ EHWZHHQ LQGHSHQGHQW GLYLGHUV DQG WR GHPRQVWUDWH WKH LQLWLDOL]DWLRQ FLUFXLWU\ 0HDVXULQJ WKH FORFN VNHZ EHWZHHQ WZR FORFN FLUFXLWV XVLQJ ZDIHU SURELQJ ZRXOG UHTXLUH PDWFKLQJ WKH GHOD\ WKURXJK WKH H[WHUQDO PHDVXUHPHQW V\VWHP SUREHV PP FRD[LDO FDEOHV 5) FRQQHFWRUVf VXFK WKDW WKH PHDVXUHG FORFN VNHZ LV GXH RQO\ WR WKH VNHZ LQ WKH FLUFXLWV WKHPVHOYHV +RZHYHU PDWFKLQJ WKHVH H[WHUQDO FRPSRQHQWV WR ZLWKLQ b RI WKH UHTXLUHG VNHZ ZKLFK LV b RI WKH ORFDO FORFN SHULRGf LV D GLIILFXOW WDVN RU LQ RWKHU ZRUGV WKH H[WHUQDO FRPSRQHQWV QHHG WR EH PDWFKHG WR ZLWKLQ SV IRU D *+] ORFDO FORFN 7KHUHIRUH WKH FORFN VNHZ LV GHWHFWHG RQ FKLS XVLQJ D SKDVH GHWHFWRU ZKLFK DFWV DV D VNHZWRYROWDJH FRQYHUWHU

PAGE 124

% A R 2XWSXWV Z R Z R UP II> IU\ VU f§,f§ ,1, Q 2XWSXWV 2XWSXWV Q IW Q 2n ,Q f Q £Q Q Q L L L L L L LL fP^ ? V?$$ P Q Q n Q Q Q )LJXUH 6LPXODWHG ZDYHIRUPV RI WZR SURJUDPPDEOH GLYLGHUV ZLWK WKH QHZ '))VSURJUDPPHG WR VROLG OLQHVf DQG GDVKHG OLQHVfaIRU UHSHDWHG LQLWLDOL]DWLRQV 7KH FLUFXLW IXQFWLRQV FRUUHFWO\ DQG WKH RXWSXW SKDVH GLIIHUHQFH EHWZHHQ WKH WZR GLYLGHUV UHPDLQV FRQVWDQW )LJXUH Df VKRZV D VFKHPDWLF RI WKH SKDVH GHWHFWRU 3'f DQG ORZSDVV ILOWHU /3)f >5D]@ 7KH 3' IXQFWLRQV DV DQ H[FOXVLYH 125 FLUFXLW WUDQVLVWRUV 0AJf $ SVHXGR1026 ORDG LV XVHG $ NH\ IHDWXUH LV WKDW WKLV 3' LV V\PPHWULF IRU LQSXWV &/. DQG &/. 7KLV HOLPLQDWHV DQ\ V\VWHPDWLF SKDVH HUURU WKDW ZRXOG EH LQWURGXFHG XVLQJ

PAGE 125

)LJXUH Df 6FKHPDWLF RI SKDVH GHWHFWRU DQG ORZSDVV ILOWHU Ef 2XWSXW RI SKDVH GHWHFWRU YHUVXV LQSXW VNHZ FRQYHQWLRQDO ;25 RU JLOEHUW FHOO DUFKLWHFWXUHV $QRWKHU EHQHILW LV WKH XVH RI VLQJOHHQGHG LQSXWV WKXV RQO\ D VLQJOHHQGHG FORFN VLJQDO KDV WR EH URXWHG WR WKH 3' )LJXUH Ef VKRZV WKH VLPXODWHG RXWSXW YROWDJH YHUVXV LQSXW FORFN VNHZ IRU WKH 3' DQG /3) 7KH YROW DJHVNHZ SORW LV V\PPHWULF ZLWK UHVSHFW WR VNHZ KHQFH WKH DEVROXWH YDOXH RI VNHZ LV PHDVXUHG 6LQFH WKH RXWSXW RI WKH 3' LV FORVH WR OLQHDU LW FDQ EH FKDUDFWHUL]HG E\ PHDVXUn LQJ WKH RXWSXW YROWDJH IRU r DQG r SKDVH RIIVHW 7KH PHDVXUHG VNHZ FDQ WKHQ EH H[WUDSRODWHG EDVHG RQ WKHVH WZR GDWD SRLQWV 1RWH WKDW WKH YROWDJHVNHZ SORW UHSHDWV HYHU\ SHULRG WKHUHIRUH WKH PHDVXUHG FORFN VNHZ ZLOO EH PRGXOR 7F8F WVNHZ H[WUDFWHG WVNHZ PrG 7FLNf ZKHUH 7FLN LV WKH SHULRG RI WKH ORFDO FORFN VLJQDO 0HDVXUHG 5HVXOWV 7R HYDOXDWH WKH RSHUDWLRQ RI WKH LQLWLDOL]DWLRQ FLUFXLWU\ D SKDVHGHWHFWRU DQG D ORZSDVV ILOWHU ZHUH LPSOHPHQWHG RQFKLS ZLWK WZR LGHQWLFDO IUHTXHQF\ GLYLGHUV 7KHVH IUHTXHQF\ GLYLGHUV ZHUH LPSOHPHQWHG ZLWK WKH RULJLQDO '))fV )LJXUH f VLQFH WKH

PAGE 126

)LJXUH 0HDVXUHG RXWSXWV RI SKDVH GHWHFWRU XVHG WR FKDUDFWHUL]H LQLWLDOL]DWLRQ FLUFXLWU\ VWDWHGHSHQGHQW LQLWLDOL]DWLRQ IDLOXUH KDG QRW \HW EHHQ LGHQWLILHG 2QH GLYLGHU ZDV DOZD\V LQLWLDOL]HG WR VWDWH ZKLOH WKH RWKHU GLYLGHUfV VWDWH ZDV SURJUDPPDEOH 0HDVXUHPHQWV RI WKH WZRGLYLGHU WHVW VWUXFWXUH UHYHDO WKDW WKH RXWSXW SKDVH GLIIHUHQFH EHWZHHQ WKH WZR GLYLGHUV FDQ WDNH RQ RQH RI HLJKW GLVFUHWH YDOXHV LQGLFDWLQJ WKDW WKH VWDUWXS SKDVH RI WKH GLYLGHUV ZDV EHLQJ DGMXVWHG 7KHVH HLJKW YDOXHV DUH VKRZQ LQ )LJXUH YHUVXV WKH LGHDO SURJUDPPHG VWDWH +RZHYHU IRU HDFK SURJUDPPHG VWDWH WKH RXWSXW ZRXOG QRW WDNH RQ WKH VDPH SKDVH YDOXH IRU PXOWLSOH UHSHWLWLRQV RI WKH H[SHULPHQW LH ,1, VZLWFKLQJ RQ DQG RIIf )RU H[DPSOH ZKHQ WKH GLYLGHU ZDV SURJUDPPHG WR VWDWH WKLV VKRXOG FRUUHVSRQG WR D r SKDVH GLIIHUHQFH DQG WKH RXWSXW RI WKH SKDVH GHWHFWRU VKRXOG EH DW LWV PLQLPXP SRLQW +RZHYHU WKH SKDVH GHWHFWRU RXWSXW ZRXOG WDNH RQ DQ\ RI WKH HLJKW SRVVLEOH YDOXHV VLQFH WKH LQLWLDOL]HG YDOXH GHSHQGV RQ WKH FXUUHQW VWDWH RI WKH GLYLGHU 7KXV HYHU\ SURJUDPn PDEOH VWDWH \LHOGV DQ\ RI WKH HLJKW SRVVLEOH SKDVH YDOXHV LQGLFDWHG E\ GRWWHG OLQHV LQ )LJn XUH 7KLV VKRZV WKDW WKH LQLWLDOL]DWLRQ SURFHVV LV QRW ZRUNLQJ FRUUHFWO\ ZKLFK OHG WR WKH GHYHORSPHQW RI WKH QHZ '))fV WR UHPRYH WKLV VWDWHGHSHQGHQFH 7KH IDFW WKDW WKH

PAGE 127

RXWSXW SKDVHGLIIHUHQFH ZDV TXDQWL]HG GRHV YHULI\ SDUWLDO RSHUDWLRQ RI WKH FLUFXLW ,W LV H[SHFWHG DQG VLPXODWLRQV LQGLFDWHf WKDW WKLV FLUFXLW ZLOO RSHUDWH FRUUHFWO\ ZKHQ WKH QHZ '))V JLYHQ LQ )LJXUH DUH XVHG $ _OP &026 7'3 'LYLGHU RQ 62, DQG %XON 6XEVWUDWHV &LUFXLW ,PSOHPHQWDWLRQ 'LYLGHUV XWLOL]LQJ 6&/ GR QRW UHTXLUH RU JHQHUDWH IXOOVZLQJ LQSXW DQG RXWSXW VLJn QDOV $V D UHVXOW WKHLU RSHUDWLQJ VSHHG FDQ EH KLJKHU WKDQ IXOOVZLQJ GLYLGHUV VLQFH LW WDNHV ORQJHU WR FKDUJH D JLYHQ FDSDFLWLYH ORDG WR D ODUJHU YROWDJH VZLQJ XVLQJ D JLYHQ FXUUHQW &9 LH [ f§f§f§ f $ EHQHILW RI IXOOVZLQJ GLYLGHUV LV WKDW WKH\ JHQHUDOO\ FDQ EH LPSOHPHQWHG ZLWK IHZHU WUDQVLVWRUV DQG KHQFH WKH OD\RXWV DUH PRUH FRPSDFW +RZHYHU LQ DGGLWLRQ WR WKHLU ORZHU VSHHG IXOOVZLQJ GLYLGHUV FDQ KDYH D UHGXFHG LQSXW VHQVLWLYLW\ RU FRQYHUVLRQ JDLQf VLQFH WKH\ UHTXLUH ODUJHU LQSXW VLJQDOV )XOOVZLQJ GLYLGHUV DUH FRPPRQO\ LPSOHPHQWHG XVLQJ WUXHVLQJOH SKDVH FORFNHG 763&f ORJLF >:HV@ $Q DOWHUQDWLYH W\SH RI KLJKVSHHG ORJLF LV GXDOSKDVH G\QDPLF SVHXGR1026 >'3@f ORJLF >%L\ '3@ LQYHUWHUV DUH VKRZQ LQ )LJXUHV Df DQG Ef ZKLOH D >'3@ ')) LV VKRZQ LQ )LJXUH Ff 7KH >'3@ LQYHUWHU LV HYROYHG IURP D &026 LQYHUWHU E\ HOLPn LQDWLQJ WKH 3026 WUDQVLVWRU XVHG IRU ORJLF HYDOXDWLRQ 7KLV UHGXFHV WKH LQSXW DQG ORDG FDSDFLWDQFH RI WKH LQYHUWHU LQFUHDVLQJ WKH VSHHG IRU D JLYHQ SRZHU GLVVLSDWLRQ $OVR WKH UHPRYDO RI WKH 3026 WUDQVLVWRU GHFUHDVHV WKH VHULHV UHVLVWDQFH IRU FKDUJLQJ WKH RXWSXW QRGH IXUWKHU LQFUHDVLQJ WKH VSHHG 6LQFH DOO WKUHH WUDQVLVWRUV RI D >'3@ LQYHUWHU DUH RQ GXULQJ HYDOXDWLRQ WKH SXOOGRZQ DFWLRQ RI WKH 1026 WUDQVLVWRUV KDV WR EH PDGH VWURQJHU

PAGE 128

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n SOH FKDLQ ZKHUH D FKDQJH DW D VLQJOH GLYLGHU QRGH ZLOO ULSSOH WKURXJK HDFK VWDJH UHWXUQLQJ WR WKH VDPH QRGH 1 FORFNSHULRGV ODWHU ZLWK RSSRVLWH SRODULW\ WDNLQJ 1 FORFN F\FOHV IRU D IXOO RXWSXW SHULRG 7KH >'3@ ODWFK GHOD\ FDQ EH H[WUDFWHG IURP WKH VSHHG RI WKH GLYLGHU VLQFH WKH PD[LPXP FORFN IUHTXHQF\ LV OLPLWHG E\ WKH GHOD\V WKURXJK WKH ILQDO >'3@ OHYHOVHQVLWLYH ODWFK DQG WKH SVHXGR1026 LQYHUWHU ZKLFK DUH DSSUR[LPDWHO\ HTXDOf 7KHUHIRUH WKH GHOD\ WKURXJK ERWK RI WKHVH ZLOO EH DSSUR[LPDWHO\ RQH KDOI WKH PLQLPXP LQSXW FORFN SHULRG DQG WKH >'3@ ODWFK GHOD\ LV WKHQ RQH IRXUWK WKH PLQLPXP LQSXW FORFN P; ,Q n 0 &/.f§_ &/.f§_ 3 &/. !0 + SL f§A2XW FON+ 0Q Df m0Q P + 0 S 4 0 Q 0 Q )LJXUH 6FKHPDWLFV RI Df &026 LQYHUWHU Ef >'3@ LQYHUWHU RU OHYHOVHQVLWLYH ODWFK DQG Ff >'3@ HGJHWULJJHUHG ')OLS)ORS

PAGE 129

)LJXUH 6FKHPDWLF RI >'3@ V\QFKURQRXV GLYLGHU SHULRG 7KLV GHOD\ JLYHV DQ LQGLFDWLRQ RI WKH VSHHG RI VWDQGDUG GLJLWDO ODWFKHV XVHG LQ SLSHn OLQHG FLUFXLWU\ IRU WKLV SURFHVV 8VLQJ >'3@ ORJLF DQG IUHTXHQF\ GLYLGHUV KDYH EHHQ LPSOHPHQWHG >)OR2OF@ LQ D SDUWLDOO\VFDOHG SP &026 WHFKQRORJ\ RQ VLOLFRQRQLQVXODWRU 62,f DQG EXON VXEVWUDWHV IURP ,%0 7KH SURFHVV XVHV D SP GHVLJQ UXOH VHW IRU DOO GLPHQVLRQV H[FHSW IRU WKH _LP JDWH OHQJWK DQG QP JDWH R[LGH WKLFNQHVV ZKLOH SURYLGLQJ WZR PHWDO OD\HUV )RU 62, SDUWLDOO\GHSOHWHG IORDWLQJERG\ WUDQVLVWRUV DUH XVHG IRU DOO FLUn FXLWU\ $ GLH SKRWRJUDSK RI WKH GLYLGHU LV VKRZQ LQ )LJXUH 7KH GLH VL]H LV [ PP (IIHFW RI 62, RQ &LUFXLW 3HUIRUPDQFH $ EHQHILW RI XVLQJ IORDWLQJ ERG\ 62, WUDQVLVWRUV LV WKDW WKH WKUHVKROG YROWDJH LV UHGXFHG GXH WR 9EV 7KLV 9EV LV FDXVHG E\ FKDUJLQJ RI WKH ERG\ QRGH WKURXJK UHJHQHUn DWLRQ FXUUHQW >6XK@ :LWK D ORZHU WKUHVKROG WKH JDWHRYHUGULYH LQFUHDVHV UHVXOWLQJ LQ PRUH FKDQQHO FXUUHQW +HQFH WKH VZLWFKLQJ VSHHG LV LQFUHDVHG UHVXOWLQJ LQ KLJKHU PD[Ln PXP RSHUDWLQJ IUHTXHQFLHV IRU WKH 62, GLYLGHU DV FRPSDUHG WR EXON

PAGE 130

)LJXUH 'LH SKRWRJUDSK RI IUHTXHQF\ GLYLGHU 7KH DFWXDO PD[LPXP RSHUDWLQJ IUHTXHQF\ IRU WKH 62, GLYLGHU LV GLIILFXOW WR VLPXn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n LQJ D 5& WLPHFRQVWDQW 7KH FRPELQDWLRQ RI D VKRUW WLPHFRQVWDQW DW WKH LQSXW VZLWFKLQJ IUHTXHQF\ ZLWK D ORQJ 5& WLPHFRQVWDQW IRU WKH ERG\ UHVXOWV LQ YHU\ ORQJ FRPSXWDWLRQ WLPHV WR UHDFK D VWHDG\ VWDWH 9EV 7KHUHIRUH WKH PD[LPXP RSHUDWLQJ VSHHG RI DQ 62, GLYLGHU LV GLIILFXOW WR VLPXODWH GXH WR WKH DPRXQW RI WLPH WKDW LW WDNHV WR UHDFK D VWHDG\VWDWH FRQGLWLRQ IRU 9EV DQG WKH QXPEHU RI VLPXODWLRQV LW WDNHV WR VFDQ IRU WKH PD[LPXP RSHUDWn LQJ IUHTXHQF\ $ ILQDO HIIHFW RI 62, RQ WKH >'3@ GLYLGHU SHUIRUPDQFH LV DQ LQFUHDVHG ORZHU IUHn TXHQF\ OLPLW GXH WR WKH SUHVHQFH RI D SDUDVLWLF ELSRODU GHYLFH LQ WKH WUDQVLVWRU )LJXUH

PAGE 131

!\ Df VKRZV WKH VFKHPDWLF RI D >'3@ LQYHUWHU ZKLOH )LJXUH Ef VKRZV WKH UHOHYDQW QRGH YROWDJHV DQG GHYLFH FXUUHQWV VLPXODWHG XVLQJ 62,63,&( YHU >)RV@ :KHQ LQ D KROG VWDWH &/. ORZf D VLQJOH >'3@ LQYHUWHU VKRXOG UHWDLQ LWV SUHYLRXV YDOXH DW WKH RXWSXW QRGH UHJDUGOHVV RI WKH LQSXW YDOXH WR WKH LQYHUWHU +RZHYHU ZKHQ KROGLQJ D fKLJKf YDOXH DW WKH RXWSXW ZLWK D QRQ]HUR LQSXW 0Q LV WXUQHG RQ DQG LWV GUDLQ LV SXOOHG GRZQ WR JURXQG 7KLV LV WKH VLWXDWLRQ VKRZQ LQ )LJXUH Ef ZKHUH WKH LQSXW LV KHOG DW a 9 D ORZ OHYHO IURP SVHXGR1026f IRU KROG DQG HYDOXDWH 6LQFH WKH DYHUDJH G\QDPLF 9EV RI 0Q LV 9 DQG 9G Q LV QRZ DW JURXQG D WUDQVLHQW ELSRODU OHDNDJH FXUUHQW ,EMWQf LV LQGXFHG LQ 0Q GHVSLWH WKH LQSXW RI 0Q EHLQJ KHOG ORZ >3HO@ 7KLV GUDLQV WKH FKDUJH VWRUHG DW WKH RXWSXW FDXVLQJ WKH LQYHUWHU WR EHJLQ WR FKDQJH VWDWHV ZKHQ LW VKRXOG EH KROGLQJ 7KHVH HIIHFWV FDQ EH VHHQ LQ )LJXUH Ef ZKHUH D %-7 FXUUHQW LV QRWLFHDEOH DQG WKH RXWSXW YROWDJH LV GHFUHDVLQJ 7KH WLPH FRQVWDQW DVVRFLDWHG ZLWK WKLV ELSRODU OHDNDJH LV RQ WKH RUGHU RI QDQRVHFRQGV IRU WKLV WHFKQRORJ\ ,I WKH GLYLGHU LV EHLQJ VZLWFKHG DW D UDWH PXFK Ef WLPH QVf )LJXUH 7UDQVLHQW ELSRODU HIIHFWV RI >'3@ ORJLF Df 6FKHPDWLF RI >'3@ LQYHUWHU DQG Ef FRUUHVSRQGLQJ QRGH YROWDJHV DQG FXUUHQWV

PAGE 132

2 ( &4 ‘R ( &4 ‘R )LJXUH KLJKHU WKDQ WKLV WKHQ WKH ELSRODU OHDNDJH VKRXOG QRW GHJUDGH SHUIRUPDQFH +RZHYHU WKH ORZHU IUHTXHQF\ FXWRII IRU DQ 62, GLYLGHU LV LQFUHDVHG ZKLFK FDQ SRWHQWLDOO\ UHGXFH WKH RYHUDOO RSHUDWLQJ IUHTXHQF\ UDQJH $QRWKHU LPSOLFDWLRQ RI WKLV LV WKDW IRU DQ 62, >'3@ GLYLGHU ZLWK D KLJK GLYLVLRQUDWLR D GLIIHUHQW ')) WRSRORJ\ LV QHHGHG IRU WKH ODWWHU VWDJHV 0HDVXUHG 5HVXOWV $ 62, GLYLGHU ZDV PHDVXUHG RQZDIHU XVLQJ KLJKIUHTXHQF\ SUREHV $ VSHFn WUXP DQDO\]HU ZDV UHTXLUHG WR H[DPLQH WKH RXWSXW VLJQDO GXH WR IUHTXHQF\ OLPLWDWLRQV RI WKH RVFLOORVFRSHfV WULJJHU WULJJHU IUHTXHQF\ OLPLWHG WR *+]f $ r FRXSOHU ZDV XVHG DV D EDOXQ WR JHQHUDWH WKH FORFN DQG HON VLJQDOV 0HDVXUHPHQWV VKRZ WKDW WKH 62, GLYLGHU FDQ RSHUDWH EHWZHHQ IUHTXHQFLHV RI DQG *+] IRU D 9 VXSSO\ )LJXUH )UHTXHQF\ 2IIVHW IURP *+] N+]f 2XWSXW VSHFWUXP RI 62, IUHTXHQF\ GLYLGHU IRU ILQ *+] DQG 9GG 9

PAGE 133

)LJXUH 2XWSXW ZDYHIRUP IRU EXON IUHTXHQF\ GLYLGHU IRU *+] LQSXW VLJQDO DQG 9GG 9 2XWSXW IUHTXHQF\ LV 0+] VKRZV WKH RXWSXW VSHFWUXP RI WKH GLYLGHU IRU DQ *+] LQSXW VLJQDO DQG D *+] RXWSXW VLJQDO 7KH VSHFWUXP DOVR FRQWDLQV KLJKHU RUGHU KDUPRQLFV RI WKH RXWSXW VLJQDO 7KH SRZHU FRQVXPSWLRQ RI WKH 62, GLYLGHU LV P: )URP WKH PD[LPXP RSHUDWLQJ IUHn TXHQF\ WKH H[WUDFWHG 62, >'3@ OHYHOVHQVLWLYH ODWFK GHOD\ LV SV DW 9 $ EXON GLYLGHU ZDV DOVR PHDVXUHG RQZDIHU 7KLV GLYLGHU FRQVLVWV RI WKH >'3@ GLYLGHU IROORZHG E\ DQ DV\QFKURQRXV GLYLGHU LPSOHPHQWHG ZLWK 763& ORJLF $W 9 WKH EXON GLYLGHU FDQ RSHUDWH EHWZHHQ DQG *+] $V H[SHFWHG WKH ORZHU IUHTXHQF\ FXWRII RI WKH EXON GLYLGHU LV OHVV WKDQ WKDW IRU WKH 62, GLYLGHU GXH WR 62,fV WUDQn VLHQW ELSRODU OHDNDJH $W *+] D 0+] RXWSXW VLJQDO LV JHQHUDWHG VKRZQ LQ )LJ XUH 7KH EXON GLYLGHU SRZHU FRQVXPSWLRQ LV P: 7KH H[WUDFWHG EXON >'3@ OHYHOVHQVLWLYH ODWFK GHOD\ LV SV DW 9 $V H[SHFWHG WKH EXON ODWFKLQJ GHOD\ LV ORQJHU WKDQ WKH 62, ODWFKLQJ GHOD\

PAGE 134

7DEOH 3HUIRUPDQFH VXPPDU\ IRU SP GLYLGHUV 62, 'LYLGHU %XON 'LYLGHU 'LYLVLRQ 5DWLR YGG 9 9 0D[ ,QSXW )UHT *+] *+] 2XWSXW )UHT *+] 0+] 0LQ ,QSXW )UHT *+] *+] 3RZHU RI FRUHf P: P: >'3@A ODWFK GHOD\ SV SV 7KH UHVXOWV DUH VXPPDUL]HG LQ 7DEOH 7KLV UHVXOW LV WKH KLJKHVW RSHUDWLQJ IUHn TXHQF\ IRU D IXOOVZLQJ IUHTXHQF\ GLYLGHU WR GDWH $OVR WKHVH UHVXOWV VKRZ WKDW D OP &026 WHFKQRORJ\ FDQ VXSSRUW GLJLWDO FLUFXLWV RSHUDWLQJ DERYH *+] DQG VXJJHVW KLJKHU RSHUDWLQJ IUHTXHQFLHV IRU IXOO\VFDOHG _LP WHFKQRORJ\ DQG EH\RQG 6XPPDU\ $Q GLYLGHU WRSRORJ\ EDVHG RQ D QHZ VRXUFHFRXSOHG ORJLF ODWFK KDV EHHQ GHYHOn RSHG 7ZR GLIIHUHQW PRGHV RI RSHUDWLRQ KDYH EHHQ LGHQWLILHG LQFOXGLQJ D VWDQGDUG GLJLWDO ODWFKLQJ PRGH DQG DQ LQMHFWLRQORFNHG PRGH ,QMHFWLRQ ORFNLQJ LV XVHG WR V\QFKURQL]H DQ RVFLOODWRU WR D ORZOHYHO LQSXW VLJQDO ZKRVH IUHTXHQF\ LV D VXSHUKDUPRQLF RI WKH RVFLOODWLRQ IUHTXHQF\ 8VLQJ LQMHFWLRQ ORFNLQJ DOORZV WKH GLYLGHU WR ORFN WR YHU\ ORZOHYHO LQSXW VLJn QDOV UHVXOWLQJ LQ FRQYHUVLRQ JDLQ IRU WKH GLYLGHU 8VLQJ D VWDQGDUG _LP &026 SURFHVV D GLYLGHU RSHUDWLQJ XS WR a *+] IRU 9GG 9 KDV EHHQ LPSOHPHQWHG )RU WKLV VXSSO\ YROWDJH WKH LQSXWUHIHUUHG VHOIRVFLOODWLRQ IUHTXHQF\ LV a *+] 7KXV WKH GLYLGHU FDQ GHWHFW YHU\ ORZOHYHO LQSXW VLJQDOV a P9SNBSNf FORVH WR WKLV IUHTXHQF\ 8VLQJ D _LP &026 WHFKQRORJ\ ZLWK FRSSHU LQWHUFRQQHFWV D GLYLGHU RSHUDWLQJ XS WR *+] DW 9 DQG *+] DW

PAGE 135

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n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n HUV 7KHVH UHVXOWV VKRZ WKDW >LP &026 FDQ VXSSRUW GLJLWDO FLUFXLWV RSHUDWLQJ DERYH *+] DQG VXJJHVW HYHQ KLJKHU IUHTXHQFLHV IRU IXOO\VFDOHG _LP WHFKQRORJ\ DQG EH\RQG

PAGE 136

&+$37(5 6<67(0 5(48,5(0(176 )25 :,5(/(66 &/2&. ',675,%87,21 2YHUYLHZ )XQGDPHQWDOO\ WKH JRDOV RI DQ\ LQWHUFRQQHFW V\VWHP DUH WR UHOLDEO\ FRQYH\ D VLJQDO IURP RQH ORFDWLRQ WR DQRWKHU WR ZLWKVWDQG QRLVH DQG LQWHUIHUHQFH WR HIILFLHQWO\ XVH ERWK DUHD DQG EDQGZLGWK LH ORZ FRVWf DQG WR GLVVLSDWH PLQLPDO SRZHU 1RW VXUSULVLQJO\ WKHVH DUH WKH VDPH JRDOV IRU ZLUHOHVV FRPPXQLFDWLRQV V\VWHPV +RZHYHU ZKLOH WKH JRDOV DUH WKH VDPH WKH WHUPLQRORJ\ DQG SHUIRUPDQFH PHWULFV XVHG WR FKDUDFWHUL]H HDFK W\SH RI V\VWHP DUH GLIIHUHQW +HQFH W\SLFDO LQWHUFRQQHFW SHUIRUPDQFH PHWULFV KDYH WR EH FRQYHUWHG WR PHWULFV IRU D ZLUHOHVV FRPPXQLFDWLRQV V\VWHP ,Q WKLV ZRUN WKH ZLUHOHVV LQWHUFRQQHFW V\Vn WHP LV EHLQJ XWLOL]HG IRU FORFN GLVWULEXWLRQ WKXV D VHW RI UDGLRIUHTXHQF\ 5)f PHWULFV LQFOXGLQJ JDLQ QRLVH ILJXUH OLQHDULW\ ,,3f DQG PDWFKLQJ KDV WR EH GHYHORSHG WR PHHW WKH PDLQ FORFN UHTXLUHPHQWV RI VNHZ DQG MLWWHU )LUVW WR PD[LPL]H WKH SRZHU WUDQVIHU IURP WKH FORFN VRXUFH WR WKH ORFDO FORFN V\Vn WHP PDWFKLQJ JDLQ DQG DQWHQQD UHTXLUHPHQWV DUH GLVFXVVHG 6HFRQG FORFN VNHZ DQG MLWn WHU DUH GHILQHG 7KLUG FORFN VNHZ DQG MLWWHU DUH XVHG WR VHW UHTXLUHPHQWV IRU WKH SRZHUOHYHO DW WKH LQSXW RI WKH IUHTXHQF\ GLYLGHU WKH VLJQDOWRQRLVH UDWLR 615f DW WKH LQSXW RI WKH IUHTXHQF\ GLYLGHU WKH QRLVH ILJXUH IRU WKH /1$ ZLWK VRXUFHIROORZHU EXIIHUV DQG DQ ,,3 IRU WKH /1$ DQG VRXUFHIROORZHU EXIIHUV )LQDOO\ WKH RYHUDOO V\VWHP UHTXLUHn PHQWV DUH VXPPDUL]HG DQG WDEXODWHG

PAGE 137

3RZHU 7UDQVIHU IURP &ORFN 6RXUFH WR /RFDO &ORFN 6\VWHP $Q LOOXVWUDWLRQ RI WKH ZLUHOHVV LQWHUFRQQHFW LV VKRZQ LQ )LJXUH 7KH ZLUHOHVV LQWHUFRQQHFW V\VWHP KDV WR JHQHUDWH D JOREDO FORFN VLJQDO *&/.f EURDGFDVW WKLV VLJQDO DFURVV D GLVWDQFH FRQWDLQLQJ PHWDO LQWHUIHUHQFH VWUXFWXUHV UHFHLYH *&/. ZLWK WKH UHFHLYn LQJ DQWHQQD DPSOLI\ DQG GLYLGH WKLV GRZQ WR WKH ORFDO FORFN VLJQDO /&/.f DQG WKHQ GLVn WULEXWH WKLV VLJQDO WR DGMDFHQW FLUFXLWV 7KH VLJQDO OHYHOV LQ )LJXUH DUH UHSUHVHQWHG DV YROWDJH DPSOLWXGHV LQ G%9 GHFLEHOV ZLWK UHVSHFW WR 9 ZKLFK LV RJ^DPSOLWXGH`f 9ROWDJH VKRXOG EH XVHG DW SRLQWV LQ WKH V\VWHP ZKHUH WKH ORDG LV FDSDFLWLYH $W SRLQWV LQ WKH V\VWHP ZKHUH WKH LPSHGDQFH LV IL[HG KDYLQJ D UHDO FRPSRQHQW WKH VLJQDO FDQ EH H[SUHVVHG LQ SRZHU LH G%Pf UDWKHU WKDQ YROWDJH 7KH FRQYHUVLRQ IURP G%P WR G%9 GHSHQGV RQ WKH UHVLVWDQFH DW WKDW SRLQW 5f DQG LV JLYHQ DV IROORZV f )RU D 4 VLQJOHHQGHGf LPSHGDQFH WKH FRQYHUVLRQ LV G%9 G%P G% 7KH WRWDO JDLQ RI WKH LQWHUFRQQHFW LV WKH FDVFDGHG JDLQV RI HDFK FRPSRQHQW LQFOXGn LQJ DQ\ PLVPDWFK ORVV 7KH RXWSXW /&/. YROWDJH VZLQJ LV WKHQ WKH LQSXW YROWDJH VZLQJ WLPHV WKH WRWDO V\VWHP YROWDJH JDLQ OLPLWHG E\ WKH VXSSO\ YROWDJH 7KXV WKH PDLQ WKLQJV ZKLFK HIIHFW WKH /&/. VZLQJ DUH DV IROORZV RXWSXW SRZHU RI WKH WUDQVPLWWHU DQWHQQDWR DQWHQQD JDLQ LQFOXGLQJ LQWHUIHUHQFH HIIHFWV WRWDO UHFHLYHU JDLQ DQG PLVPDWFK ORVV DW WKH DQWHQQDVFLUFXLW LQWHUIDFH (DFK FRPSRQHQW ZLOO EH DGGUHVVHG LQ WKLV VHFWLRQ DQG QRPLQDO YDOXHV ZLOO EH UHFRPPHQGHG

PAGE 138

U )UHTXHQF\ *+]f FR R J nWR )UHTXHQF\ *+]f 'LYLGHU%XIIHU Y &RQYHUVLRQ *DLQ )UHTXHQF\ *+]f $YRLG 0LVPDWFK ? $YRLG L 0LVPDWFK )UHTXHQF\ 'LYLGHU f 5HTXLUHG A -LWWHU6NHZ /RFDO &ORFN 2XWSXWV 5HTXLUHG 61 5DWLR 2XWSXW %XIIHUV W )LJXUH G%9f G%9f G%9f :LUHOHVV LQWHUFRQQHFW RSWLPL]DWLRQ RG; 7UDQVPLWWHU 3RZHU 7KH DPRXQW RI SRZHU WKDW WKH SRZHU DPSOLILHU 3$f LQ WKH WUDQVPLWWHU FDQ GHOLYHU GHSHQGV RQ WKH VXSSO\ YROWDJH WKH &026 WHFKQRORJ\ HJ JDWHR[LGH WKLFNQHVVf WKH SRZHU DPSOLILHU FLUFXLW WRSRORJ\ WKH IUHTXHQF\ DQG WKH DQWHQQD LPSHGDQFH 'HSHQGLQJ RQ WKH 3$ WRSRORJ\ FODVV $% ( ) HWFf WKH SHDNWRSHDN YROWDJH VZLQJ DFURVV WKH RXWSXW WUDQVLVWRU FDQ EH EHWZHHQ 9GG DQG 9GG >/HH@ :KLOH UHFHQW UHVXOWV KDYH VKRZQ &026 3$V FDSDEOH RI GHOLYHULQJ G%P RI SRZHU DW 0+] XVLQJ _LP WHFKQRORJLHV >)DO2O .XR2O 6KL2O@ WR WKH DXWKRUfV NQRZOHGJH WKHUH DUH QR &026 3$ UHVXOWV LQ WKH WR *+] UHJLRQ RWKHU WKDQ WKDW ZKLFK ZLOO EH SUHVHQWHG LQ FKDSWHU 7KXV D G%P RXWSXW SRZHU LV VHW DV D JRDO IRU WKH 3$ DW *+] DQG DERYH 7KLV FRUUHVSRQGV WR D SHDNWRSHDN YROWDJH VZLQJ RI 9 IRU D 4 VLQJOHHQGHG ORDG )XUWKHU ZRUN LV UHTXLUHG WR GHYHORS &026 3$fV RSHUDWLQJ DERYH *+] DQG WKH G%P VSHFLILFDWLRQ VKRXOG EH PRGLILHG DFFRUGLQJO\

PAGE 139

$QWHQQD 6SHFLILFDWLRQV 3UREDEO\ WKH PRVW FULWLFDO FRPSRQHQW IRU WKH ZLUHOHVV LQWHUFRQQHFW V\VWHP LV WKH DQWHQQD $V WKH ORVV WKURXJK WKH DQWHQQDV LQFUHDVHV WKH WUDQVPLVVLRQ SRZHU DQGRU WKH UHFHLYHU JDLQ KDYH WR EH LQFUHDVHG $OVR DQ\ LPSHGDQFH PLVPDWFK EHWZHHQ WKH DQWHQQD DQG WKH FLUFXLW WUDQVPLWWHU RU UHFHLYHUf FDXVHV UHIOHFWLRQ RU PLVPDWFK ORVV )LQDOO\ WKH VLJn QDOV ZLOO EH GLVWULEXWHG DFURVV D GLJLWDO FKLS ZKLFK KDV PXOWLSOH PHWDO LQWHUIHUHQFH DQG SDFNDJH VWUXFWXUHV EHWZHHQ WKH DQWHQQDV 7KLV ZLOO IXUWKHU GHJUDGH WKH DQWHQQD JDLQ 'XH WR LWV LPSRUWDQFH WR WKH V\VWHP WKH DQWHQQD JDLQ DQG LPSHGDQFH VKRXOG EH DFFXUDWHO\ SUHn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f )RU D VTXDUH GLH ZLWK WKH WUDQVPLWWHU DW WKH FHQWHU DQG DQ HYHQO\VSDFHG JULG RI FORFN UHFHLYHUV WKH DQWHQQD SURSDJDWLRQ GLVWDQFHV ZLOO EH HLWKHU RU FP %DVHG RQ WKH H[SHULPHQWDO ZRUN SUHVHQWHG LQ >.LP22E@ DQWHQQDWRDQWHQQD WUDQVPLVVLRQ JDLQV RI G% DQG G% DUH DVVXPHG IRU DQG FP VHSDUDWLRQ GLVWDQFHV DW *+] ZLWKRXW LQWHUIHUHQFH VWUXFWXUHV 7KLV VKRZV D G% GHFUHDVH LQ JDLQ SHU RFWDYH RI GLVn WDQFH 8VLQJ WKHVH UHVXOWV WKH JDLQV DW DQG FP DUH H[WUDSRODWHG WR EH 7KLV JDLQ LV IRU D PP ORQJ ]LJ]DJ GLSROH DQWHQQD SP DERYH D ILFP VLOLFRQ VXEVWUDWH ZLWK D r EHQGDQJOH DQG D JP WUDFH ZLGWK

PAGE 140

DQG G% UHVSHFWLYHO\ 0HWDO LQWHUIHUHQFH VWUXFWXUHV KDYH EHHQ VKRZQ H[SHULPHQWDOO\ DQG LQ &KDSWHU f WR GHJUDGH WKH JDLQ E\ WR G% >
PAGE 141

LQSXW VHQVLWLYLW\ LV HYLGHQW 7R PD[LPL]H WKH UHFHLYHU JDLQ WKH /1$ SHDN JDLQ VKRXOG RFFXU DW D IUHTXHQF\ FORVH WR WKH LQSXW VHOIUHVRQDQW IUHTXHQF\ RI WKH GLYLGHU 7KLV PLQLn PL]HV WKH PLQLPXP GHWHFWDEOH VLJQDO RI WKH UHFHLYHU E\ DOORZLQJ IRU D VPDOOOHYHO VLJQDO WR LQMHFWLRQORFN WKH GLYLGHU ,I QHHGHG IRU WHVWLQJ SXUSRVHV WKH VXSSO\ YROWDJH RI WKH GLYLGHU FDQ EH DGMXVWHG WR WXQH WKH GLYLGHU VHOIUHVRQDQW IUHTXHQF\ $V FDQ EH VHHQ LI D GLYLGHU ZLWKRXW FRQYHUVLRQ JDLQ LV XVHG WKHQ HLWKHU WKH /1$ JDLQ ZRXOG KDYH WR EH LQFUHDVHG E\ G% RU WKH WUDQVPLWWHU ZRXOG KDYH WR EURDGFDVW G% PRUH SRZHU ,QFUHDVLQJ WKH /1$ JDLQ WR a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n 0DQWn ZKHUH 3GHO LV WKH SRZHU GHOLYHUHG IURP WKH DQWHQQD WR WKH FLUFXLW 3$96 LV WKH DYDLODEOH SRZHU IURP WKH DQWHQQD DQG 0DQW LV WKH PLVPDWFK IDFWRU 7KH PLVPDWFK IDFWRU LV D SRZHU WUDQVPLVVLRQ FRHIILFLHQW WUDQVPLWWDQFHf EHWZHHQ WZR PHGLXPV 7KH DQWHQQD PLVPDWFK IDFWRU LV JLYHQ E\ >*RQ@ )RU WKH UHVW RI WKLV FKDSWHU ZKHQ UHIHUULQJ WR WKH /1$ ERWK WKH /1$ DQG VRXUFHIROORZn HUV DUH DVVXPHG

PAGE 142

0 B L_UFOfL_UD-f f§ U U $ F$ DQW? f ZKHUH UF LV WKH UHIOHFWLRQ FRHIILFLHQW RI WKH FLUFXLW DQG UDQW LV WKH UHIOHFWLRQ FRHIILFLHQW RI WKH DQWHQQD ERWK ZLWK UHVSHFW WR e GLIIHUHQWLDOO\f ,Q WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP WKHUH DUH WZR PLVPDWFK IDFWRUVf§RQH DW WKH WUDQVPLWWHU DQG RQH DW WKH UHFHLYHU WKXV WKHVH IDFWRUV VKRXOG EH PD[LPL]HG 7KH VSHFLILFDWLRQ IRU WKH WRWDO PLVPDWFK ORVV /PPf LQ WKH V\VWHP LV G% JLYHQ E\ WKH IROORZLQJ RJ/ [ff Yr DQLO OUO DQW 7\SLFDOO\ LPSHGDQFH PLVPDWFK LV VSHFLILHG DV HLWKHU D YROWDJHVWDQGLQJZDYH UDWLR 96:5f RU D UHIOHFWLRQ FRHIILFLHQW 7KHVH SDUDPHWHUV DUH QRW VWUDLJKWIRUZDUG LQ WKLV V\Vn WHP GXH WR WKH FRPSOH[ LPSHGDQFHV RI WKH VRXUFH FLUFXLWf DQG WKH ORDG DQWHQQDf $V ZDV GRQH LQ >*RQ@ HTXDWLRQ f FDQ EH UHZULWWHQ DV IROORZV 30 3D\V n 3D\V n _Umf f ZKHUH UHT LV DQ HTXLYDOHQW UHIOHFWLRQ FRHIILFLHQW GHILQHG E\ ,Um_ f 1RZ D 96:5 FDQ EH GHILQHG DV 96:5 OU+UfO f )RU D G% WRWDO PLVPDWFK ORVV WKH HTXLYDOHQW VSHFLILFDWLRQV IRU 7HT DQG 96:5 DW HDFK DQWHQQD DUH G% DQG UHVSHFWLYHO\ *LYHQ DQ DQWHQQD LPSHGDQFH RU UHIOHFWLRQ FRHIILFLHQWf WKH UDQJH RI DOORZDEOH FLUn FXLW LPSHGDQFHV RU UHIOHFWLRQ FRHIILFLHQWVf LV UHTXLUHG 7A LV H[SUHVVHG LQ WHUPV RI UF DQG UDQW E\ SOXJJLQJ f LQWR f UHVXOWLQJ LQ

PAGE 143

0 f r F DQW U U FO DQW f $V FDQ EH VHHQ 7HT LV D ELOLQHDU WUDQVIRUP RI UF $ FLUFOH ZLWK UDGLXV ,)HTO FRUUHVSRQGLQJ WR 0DQW /PP RU 96:5f LQ WKH UHT SODQH ZLOO WKHUHIRUH PDS LQWR D FLUFOH LQ WKH UF SODQH DOORZLQJ WKH FLUFXLW LPSHGDQFH WR EH FKRVHQ IRU D JLYHQ DQWHQQD LPSHGDQFH $SSHQGL[ $ RI >*RQ@ FRQWDLQV HTXDWLRQV IRU FLUFOH PDSSLQJ GXH WR ELOLQHDU WUDQVIRUPV WKXV WKH GHUn LYDWLRQV DUH QRW UHSHDWHG KHUH 7KH HTXDWLRQ IRU WKH FLUFOH LQ WKH 7F SODQH IRU D JLYHQ 7DQW LV _UFF_ U f ZKHUH F DQG U DUH WKH FHQWHU DQG UDGLXV RI WKH FLUFOH LQ WKH 7F SODQH JLYHQ DV IROORZV F Ur F DQW I L f§ LU A U HJ? +UHOUM U Oe-+7 DQWL L f§ LU U L HT_ _[ DQW? f f $V FDQ EH VHHQ IRU SHUIHFW PDWFKLQJ LH 0 RU 7HT f WKH FLUFOH LQ WKH 7F SODQH LV FHQWHUHG DW 7rDQW ZLWK D UDGLXV RI FRUUHVSRQGLQJ WR FRQMXJDWH PDWFKLQJ )LJXUH VKRZV D IDPLO\ RI FLUFOHV LQ WKH UF SODQH FRUUHVSRQGLQJ WR PLVPDWFK ORVVHV RI ^ DQG ` G% 96:5V RI ^ DQG ` DQG 7HTV RI ^ DQG ` G% ZKHQ UHDGLQJ IURP WKH RXWVLGH FLUFOH WR WKH LQVLGH FLUFOH 7KHVH FLUFOHV DUH IRU DQ DQWHQQD LPSHGDQFH RI M e SORWWHG RQ D 6PLWK FKDUW QRUPDOL]HG E\ e 7KH VHFRQG FLUFOH IURP WKH RXWVLGH FRUUHVSRQGV WR 96:5 WKHUHIRUH DOO FLUFXLW LPSHGDQFHV ORFDWHG ZLWKLQ WKLV FLUFOH ZLOO PHHW WKH V\VWHP VSHFLILFDWLRQ UHVXOWLQJ LQ OHVV WKDQ G% PLVPDWFK ORVV SHU DQWHQQD

PAGE 144

M )LJXUH &RQVWDQW 7HT FLUFOHV LQ WKH UF SODQH 7KLV VKRZV WKH DOORZDEOH FLUFXLW LPSHGDQFHV IRU D JLYHQ DQWHQQD UHIOHFWLRQ FRHIILFLHQW 7DQW IRU ]DQW Mff 'HILQLWLRQ RI &ORFN 6NHZ DQG -LWWHU $ FULWLFDO UHTXLUHPHQW IRU FORFN GLVWULEXWLRQ V\VWHPV LV HQVXULQJ WKDW DOO RI WKH GLVn WULEXWHG FORFN VLJQDOV DUH V\QFKURQL]HG WKURXJKRXW WKH FKLS 6LQFH WKHVH FORFN VLJQDOV DUH XVHG WR PDLQWDLQ V\VWHP WLPLQJ DQ\ XQFHUWDLQW\ LQ WKH HGJH RI WKH FORFN VLJQDO UHGXFHV WKH DPRXQW RI WLPH DYDLODEOH WR SHUIRUP XVHIXO RSHUDWLRQV $V D UHVXOW WKH PLQLPXP FORFN F\FOH WLPH KDV WR EH LQFUHDVHG E\ WKH XQFHUWDLQW\ VORZLQJ WKH V\VWHP GRZQ 7KXV LQWHUn FRQQHFW V\VWHPV ZKLFK FDQ GHOLYHU KLJK IUHTXHQF\ FORFN VLJQDOV ZLWK PLQLPDO WLPLQJ XQFHUWDLQW\ W\SLFDOO\ b RI D FORFN F\FOHf DUH UHTXLUHG 7KH RULJLQ RI FORFN WLPLQJ XQFHUWDLQW\ LV PLVPDWFK RU GLIIHUHQFHV LQ WKH SURSDJDn WLRQ GHOD\V RI WKH VLJQDOV 7KH GLIIHUHQFH KDV ERWK WLPHLQYDULDQW VWDWLFf DQG WLPHYDULDQW G\QDPLFf FRPSRQHQWV &ORFN VNHZ LV GHILQHG DV WKH VWDWLF GLIIHUHQFH EHWZHHQ FORFN HGJHV WKURXJKRXW WKH FKLS ZKLOH FORFN MLWWHU LV GHILQHG DV WKH G\QDPLF GLIIHUHQFH EHWZHHQ WKH

PAGE 145

6WDWLF VNHZf '\QDPLF MLWWHUf )LJXUH 3UREDELOLW\ GHQVLW\ IXQFWLRQ RI SURSDJDWLRQ GHOD\ LQFOXGLQJ ERWK VWDWLF DQG G\QDPLF FRPSRQHQWV UHVXOWLQJ LQ FORFN VNHZ DQG MLWWHU DIWHU ('1 -XO\ SS f FORFN HGJHV WKURXJKRXW WKH FKLS 7KHVH TXDQWLWLHV DUH UHODWHG WR WKH YDULDQFHV Df RI WKH VWDWLF DQG G\QDPLF GHOD\ SUREDELOLW\ GHQVLW\ IXQFWLRQV 3')fVf UHVSHFWLYHO\ )LJXUH LOOXVWUDWHV HDFK W\SH RI FRPSRQHQW DV LW UHODWHV WR WKH WRWDO GHOD\ 3') +HUH WKH WRWDO VWDWLF YDULDQFH LV WKH VXP RI LQGLYLGXDO VWDWLF YDULDQFHV HJ SURFHVV YDULDn WLRQ FDSDFLWLYH ORDG PLVPDWFK HWFf \LHOGLQJ WKH FORFN VNHZ /LNHZLVH WKH WRWDO G\QDPLF YDULDQFH LV WKH VXP RI LQGLYLGXDO G\QDPLF YDULDQFHV HJ WKHUPDO QRLVH VRXUFH MLWWHU HWFf \LHOGLQJ WKH FORFN MLWWHU %RWK VWDWLF DQG G\QDPLF YDULDQFHV DGG WR JLYH WKH WRWDO GHOD\ YDULDQFH 7KH YDULDQFHV FDQ EH TXDQWLILHG DV HLWKHU D SHDNWRSHDN YDULDQFH DVVXPHG WR EH a Df RU D URRWPHDQVTXDUHG 506f YDULDQFH Df ZKHUH W\SLFDOO\ SHDNWRSHDN LV XVHG WR TXDQWLI\ FORFN VNHZ DQG HLWKHU FDQ EH XVHG WR TXDQWLI\ FORFN MLWWHU &ORFN 6NHZ 9HUVXV $PSOLWXGH 0LVPDWFK ,Q WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP WKH VLJQDOV ZLOO DUULYH DW WKH DQWHQQDV WKURXJKRXW WKH FKLS DW GLIIHUHQW SRZHU OHYHOV GHSHQGLQJ RQ WKH GLVWDQFH EHWZHHQ WKH

PAGE 146

WUDQVPLWWHU DQG UHFHLYHU $VVXPLQJ WKDW WKH UHFHLYHUV GR QRW FRUUHFW IRU WKLV GLIIHUHQFH LQ VLJQDO VWUHQJWK WKHQ WKH YROWDJH VZLQJ XVHG WR WULJJHU RU ORFN WKH GLYLGHUV ZLOO EH PXFK ORZHU LQ UHFHLYHUV IDUWKHU IURP WKH WUDQVPLWWHU 7KLV FDQ SRWHQWLDOO\ OHDG WR D FORFN VNHZ GXH WR D FRQYHUVLRQ IURP DPSOLWXGH PRGXODWLRQ WR SKDVH PRGXODWLRQ $0 WR )0f 7KH FRQYHUVLRQ PHFKDQLVP GHSHQGV RQ ZKHWKHU WKH GLYLGHU LV RSHUDWLQJ LQ D ODWFKHG PRGH RU DQ LQMHFWLRQORFNHG PRGH DQG HDFK FDVH ZLOO EH H[DPLQHG /DWFKHG 0RGH )RU D GLYLGHU RSHUDWLQJ LQ WKH ODWFKHG PRGH $0WR)0 FRQYHUVLRQ ZLOO UHVXOW LI WKH GLYLGHU VZLWFKHV DW D SRLQW RWKHU WKDQ WKH ]HURFURVVLQJ RI WKH LQSXW VLJQDO 7KLV LV LOOXVn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f RU QRW ULJKWf

PAGE 147

WULJJHU OHYHO DV ZHOO DV WKH GLIIHUHQFH EHWZHHQ WKH ORZOHYHO DQG KLJKOHYHO LQSXW VLJQDOV UHIHUUHG WR DV DPSOLWXGH PLVPDWFK 7KH FORFN VNHZ DV D IXQFWLRQ RI WULJJHU OHYHO DQG LQSXW VLJQDO OHYHOV FDQ UHDGLO\ EH REWDLQHG 7KH SKDVH DQJOHV c DQG DW ZKLFK WKH PD[LPXP DQG PLQLPXP LQSXW VLJQDOV DUH HTXDO WR WKH WULJJHU OHYHO 9UHI DUH DV IROORZV 7KH FORFN VNHZ LQ WHUPV RI SKDVH RI WKH LQSXW VLJQDO LV WKHQ f f $If VLQ ? \ Y PLQn !‘ PP PD[n 9 ‘ 9 PLQ PD[n f 6ROYLQJ IRU WKH DPSOLWXGH PLVPDWFK UDWLR 9PLQ9PD[ \LHOGV WKH DOORZDEOH VLJQDO OHYHO GLIn IHUHQFH IRU D JLYHQ LQSXW DQJXODU FORFN VNHZ $cQ DV IROORZV 9 f PLQ 9 PD[ VLQ VLQ ?6f$4LQf@ f ZKHUH UHI 9 ‘ PLQ f 7KH PD[LPXP FORFN VNHZ ZLOO RFFXU ZKHQ )RU WKLV FDVH f FDQ EH UHGXFHG WR 9 ‘ PLQ 9 PD[ FRV$Qf FRV:WF A N RXWn f ZKHUH WKH LQSXW DQJXODU FORFN VNHZ KDV EHHQ FRQYHUWHG WR D FORFN VNHZ $W ZLWK UHVSHFW WR WKH RXWSXW ORFDO FORFN SHULRG XVLQJ WKH GLYLVLRQ UDWLR RI WKH IUHTXHQF\ GLYLGHU 1

PAGE 148

)LJXUH $PSOLWXGH PLVPDWFK DOORZHG DW GLYLGHU LQSXW YHUVXV ORFDO FORFN VNHZ SHUFHQWDJH ZLWK YDU\LQJ IURP WR LQ LQFUHPHQWV )LJXUH VKRZV D SORW RI WKH DPSOLWXGH PLVPDWFK LQ GHFLEHOV YHUVXV ORFDO FORFN VNHZ LQ SHUFHQWDJH IRU 1 7KHVH SORWV DUH RI f ZLWK VFDQQLQJ IURP WR $V GHFUHDVHV D ODUJHU DPSOLWXGH PLVPDWFK FDQ EH WROHUDWHG PHDQLQJ WKDW WKH GLYLGHUfV $0 UHMHFWLRQ LPSURYHV DV WKH WULJJHU OHYHO DSSURDFKHV ]HUR $V FDQ EH VHHQ IRU WKH DPSOLWXGH PLVPDWFK FDQ EH fLQILQLWHf ZLWKRXW JHQHUDWLQJ PRUH WKDQ b RI FORFN VNHZ 7KH PD[LPXP FORFN VNHZ DOORZHG IRU DPSOLWXGH PLVPDWFK LV VHW WR b )URP f DQG )LJXUH WKH DPSOLWXGHV DW WKH LQSXW RI WKH IUHTXHQF\ GLYLGHU VKRXOG EH PDWFKHG WR ZLWKLQ G% IRU b VNHZ WR DFFRPPRGDWH WKH ZRUVWFDVH WULJJHULQJ FRQGLWLRQ f 7KHVH UHVXOWV VKRZ FORFN VNHZ DV D IXQFWLRQ RI WULJJHU OHYHO IRU GLIIHUHQW LQSXW VLJn QDO VZLQJV WR WKH GLYLGHU 'LYLGHUV XWLOL]LQJ VRXUFHFRXSOHG ORJLF 6&/f ZLOO LGHDOO\ WULJn JHU DW WKH ]HUR FURVVLQJV RI WKH LQSXW VLJQDO GXH WR WKH GLIIHUHQWLDO VWUXFWXUH RI WKH 6&/ 'IOLSIORS +RZHYHU FLUFXLW PLVPDWFKHV ZLOO FDXVH WKH WULJJHU OHYHO WR RFFXU VOLJKWO\ DERYH RU EHORZ WKH ]HUR FURVVLQJV UHVXOWLQJ LQ D QRQ]HUR DQG FORFN VNHZ 7KH VL]H RI

PAGE 149

ZLOO EH VPDOO KRZHYHU VLQFH WKH RIIVHW YROWDJH GXH WR FLUFXLW PLVPDWFKHV VKRXOG EH OHVV WKDQ P9 ,Q WKH ODWFKHG PRGH 9PLQ LV RQ WKH RUGHU RI 9 WKXV VKRXOG EH OHVV WKDQ PHDQLQJ WKDW DPSOLWXGH PLVPDWFK ZLOO QRW FDXVH PRUH WKDQ b RI FORFN VNHZ $V D UHVXOW DXWRPDWLF JDLQ FRQWURO $*&f VKRXOG QRW EH UHTXLUHG IRU 6&/ GLYLGHUV RSHUDWn LQJ LQ WKH ODWFKHG PRGH KRZHYHU DV ZLOO EH VKRZQ $*& LV UHTXLUHG IRU GLYLGHUV RSHUDWn LQJ LQ WKH LQMHFWLRQORFNHG PRGH ,QMHFWLRQ/RFNHG 0RGH :KHQ WKH GLYLGHU LV RSHUDWLQJ DV D VXSHUKDUPRQLF LQMHFWLRQORFNHG RVFLOODWRU ,/2f D SKDVH GLIIHUHQFH H[LVWV EHWZHHQ WKH ORFNLQJ VLJQDO DQG RVFLOODWRU RXWSXW VLJQDO 7KLV SKDVH GLIIHUHQFH LV D IXQFWLRQ RI WKH LQSXW VLJQDO OHYHOV DV ZHOO DV WKH GLIIHUHQFH EHWZHHQ WKH ORFNLQJ IUHTXHQF\ DQG WKH QDWXUDO IUHTXHQF\ RI WKH ,/2 $SSHQGL[ ( UHYLHZV WKH EDVLF WKHRU\ EHKLQG ,/2V ,Q SDUWLFXODU WKH VWHDG\VWDWH SKDVH UHODWLRQVKLS EHWZHHQ WKH LQSXW DQG RXWSXW VLJQDOV RI WKH ,/2 ZDV VKRZQ WR EH >3DF@ I ? 46V VLQ $FG$ : VP O $P$ fD $Fr $f f R LQ HTXDWLRQ (f +HUH $FR LV WKH GLIIHUHQFH EHWZHHQ WKH QDWXUDO DQG ORFNLQJ IUHTXHQFLHV FRFR/f $ LV WKH GHULYDWLYH RI WKH SKDVH UHVSRQVH RI WKH ,/2 LH GLYLGHUf ZLWK UHVSHFW WR IUHTXHQF\ $ Af IURP ZKLFK DQ HTXLYDOHQW TXDOLW\ IDFWRU FDQ EH GHILQHG DQG 9/ DQG 9 DUH WKH ORFNLQJ DQG RVFLOODWRU RXWSXW YROWDJH VZLQJV UHVSHFWLYHO\ $ SKDVH GLIIHUHQFH H[LVWV EHFDXVH DQ RVFLOODWRU JHQHUDWHV r RI SKDVHVKLIW DURXQG WKH ORRS DW LWV QDWXUDO RVFLOODWLRQ IUHTXHQF\ :KHQ DQ LQSXW VLJQDO LV LQMHFWHG LQWR WKH ORRS D SKDVH GLIIHUn HQFH UHVXOWV DW WKDW LQMHFWLRQ SRLQW 7KHUHIRUH WKH ORRS DGMXVWV WKH RVFLOODWLRQ IUHTXHQF\ WR FRXQWHUDFW WKLV SKDVH GLIIHUHQFH 7KLV LV GLVFXVVHG LQ PRUH GHWDLO LQ $SSHQGL[ (

PAGE 150

2QFH DJDLQ D FORFN VNHZ FDQ EH GHILQHG LQ WHUPV RI WKH LQSXW SKDVH GLIIHUHQFH LQ UDGLDQV 7DNLQJ WKH GLIIHUHQFH EHWZHHQ WKH VWHDG\VWDWH SKDVH GLIIHUHQFH JLYHQ LQ f IRU WKH PD[LPXP DQG PLQLPXP LQSXW VZLQJ ORFNLQJ VLJQDOV 9/PD[ 9/PLQf DVVXPLQJ WKDW WKH RXWSXW VZLQJV DUH WKH VDPH IRU HDFK FDVH DQG PDNLQJ WKH IROORZLQJ VXEVWLWXWLRQ R $FR $ f,/2 9/PLQ $ Y f -O $FR D$< UHVXOWV LQ f 7KXV WKH UHODWLRQVKLS EHWZHHQ FORFN VNHZ DQG DPSOLWXGH PLVPDWFK DW WKH GLYLGHU LQSXW LV WKH VDPH IRU GLYLGHUV RSHUDWLQJ LQ WKH ODWFKHG RU LQMHFWLRQORFNHG PRGH ,Q IDFW DQ HTXLYDOHQW WULJJHU OHYHO 9A" FDQ EH GHILQHG IRU WKH ,/2 GLYLGHU DV IROORZV 9,/" 9 f Y UHI Y2 $FR$ AOI$FR $f f $V WKH ORFNLQJ IUHTXHQF\ DSSURDFKHV WKH QDWXUDO IUHTXHQF\ RI WKH ,/2 WKH HTXLYDOHQW WULJn JHU OHYHO DSSURDFKHV ]HUR 7KH DPSOLWXGH PLVPDWFK IRU WKH ZRUVW FDVH FORFN VNHZ FDQ DJDLQ EH TXDQWLILHG )RU DQ LQMHFWLRQORFNHG IUHTXHQF\ GLYLGHU ,/)'f WKH PLQLPXP LQSXW VLJQDO OHYHO ZLOO EH WKDW IRU MXVW ORFNLQJ WKH ,/2 7KLV ORFNLQJ FRQGLWLRQ LV JLYHQ LQ DSSHQGL[ ( LQ (f ZKHUH WKH PLQLPXP 9/ LV 9/PLQ 9 ‘ $£R$ f 3OXJJLQJ WKLV LQWR f DQG DVVXPLQJ WKDW $FR$ m ZKLFK LV YDOLG LQ WKLV FDVH VLQFH 9/PLQ m 9f UHVXOWV LQ :LWK WKHVH VXEVWLWXWLRQV f EHFRPHV A f >VLQVLQBOf$-f@ VLQA$RFRV$ff f DV LQ f DQG )LJXUH UHPDLQV YDOLG 7R REWDLQ OHVV WKDQ b RI FORFN VNHZ IURP DPSOLWXGH PLVPDWFK WKH LQSXWV WR DOO RI WKH IUHTXHQF\ GLYLGHUV VKRXOG EH PDWFKHG WR

PAGE 151

ZLWKLQ G% 7KLV ZRUVWFDVH VLWXDWLRQ RFFXUV ZKHQ DW OHDVW RQH GLYLGHU LV EHLQJ ORFNHG DW LWV PLQLPXP GHWHFWDEOH VLJQDO 0'6f +RZHYHU LQ WKH DFWXDO V\VWHP WKH VLJQDOV DW WKH GLYLGHU LQSXWV VKRXOG DOO EH DERYH WKH 0'6 WR SURYLGH UREXVWQHVV DJDLQVW SURFHVV YDULDn WLRQ 7KXV 9/PLQ LQFUHDVHV DQG GHFUHDVHV $V D UHVXOW PRUH DPSOLWXGH PLVPDWFK FDQ EH WROHUDWHG ,I WKH PLQLPXP LQSXW VLJQDO OHYHO DW WKH GLYLGHU LQSXW 9/PMQf LV G% DERYH WKH GLYLGHUfV 0'6 WKHQ ZRXOG EH HTXDO WR DQG WKH DPSOLWXGH PLVPDWFK FDQ EH XS WR G% IRU OHVV WKDQ b RI FORFN VNHZ 7KHUHIRUH $*& LV UHTXLUHG IRU GLYLGHUV RSHUDWLQJ LQ WKH LQMHFWLRQORFNHG PRGH 7KLV PRGH DV RSSRVHG WR WKH ODWFKHG PRGH ZLOO EH WKH PRVW FRPPRQ RQH IRU WKH GLYLGHUV LQ WKH FORFN UHFHLYHU VLQFH LQMHFWLRQ ORFNLQJ LQFUHDVHV WKH FRQYHUVLRQ JDLQ 7KH ZRUVWFDVH V\VWHP VSHFLILFDWLRQ IRU WKH UHFHLYHU ZKRVH GLYLGHU LV LQMHFWLRQORFNHGf LV WKDW WKH DPSOLn WXGHV DW WKH LQSXW RI WKH IUHTXHQF\ GLYLGHU KDYH WR EH PDWFKHG WR ZLWKLQ G% IRU b FORFN VNHZ )LQDOO\ LI WKH LQSXWV WR WKH GLYLGHUV DUH JXDUDQWHHG WR EH DW OHDVW G% DERYH WKH GLYLGHUfV 0'6 WKHQ WKH DPSOLWXGH PLVPDWFK FDQ EH DV ODUJH DV G% &ORFN -LWWHU 9HUVXV 6LJQDOWR1RLVH 5DWLR ,Q WUDGLWLRQDO FRPPXQLFDWLRQV V\VWHPV D KLJK VLJQDOWRQRLVH UDWLR 615f LV UHTXLUHG LQ RUGHU IRU WKH UHFHLYHU WR PDNH FRUUHFW HVWLPDWLRQV RI WKH PRGXODWHG GDWD TXDQn WLILHG LQ WHUPV RI D ELWHUURU UDWH %(5f $V WKH 615 UDWLR LQFUHDVHV WKH %(5 GHFUHDVHV WKHUHIRUH D PLQLPXP 615 LV UHTXLUHG VXFK WKDW WKH %(5 LV ORZHU WKDQ WKDW ZKLFK WKH PRGXODWLRQ FRGH FDQ FRUUHFW ,Q WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP %(5 LV QRW UHDOO\ PHDQLQJIXO VLQFH WKH V\VWHP LV RQO\ VHQGLQJ D FDUULHU VLJQDO ,QVWHDG QRLVH ZLOO FRUUXSW WKH ]HURFURVVLQJV RI WKH VLJQDOV UHVXOWLQJ LQ MLWWHUSKDVH QRLVH DW WKH RXWSXW RI WKH IUHn TXHQF\ GLYLGHU 7KHUHIRUH D UHODWLRQVKLS EHWZHHQ WKH 615 DW WKH LQSXW RI WKH IUHTXHQF\

PAGE 152

GLYLGHU DQG WKH RXWSXW FORFN MLWWHU LV UHTXLUHG 7R DFFRPSOLVK WKLV WKH SKDVH QRLVH RI IUHn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ff ,Q DGGLWLRQ WR WKHVH WKH GLYLGHU ZLOO JHQHUDWH SKDVH QRLVH LWVHOI WKXV WKH WRWDO RXWSXW SKDVH QRLVH LV WKH VXP RI WKH GLYLGHG DQG ILOWHUHG IRU ,/)'Vf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f %RWK RI WKHVH FRPSRQHQWV DUH GLYLGHG E\ WKH GLYLVLRQ UDWLR VTXDUHG ZKHQ UHIHUUHG WR WKH RXWSXW 7KH WKLUG VRXUFH LV MLWWHU LQ WKH GLYLGHU GXH WR YDULDWLRQV LQ WKH GHOD\ LW WDNHV D VLJQDO WR SURSDJDWH WKURXJK WKH FLUn FXLW 7KH IRXUWK VRXUFH LV RXWSXW SKDVH QRLVH ZKLFK LV GXH WR DGGLWLYH QRLVH DW WKH RXWSXW RI

PAGE 153

)LJXUH 9Q&26<_ 3KDVRU GLDJUDP IRU DGGLWLYH QRLVH VKRZLQJ KRZ SKDVH QRLVH LV JHQHUDWHG WKH GLYLGHU 6LQFH D IUHTXHQF\ GLYLGHU EDVLFDOO\ VDPSOHV WKH LQSXW SKDVH GXULQJ D ]HUR FURVVLQJ RI WKH VLJQDO WKH SKDVH QRLVH VSHFWUXPV ZLOO EH UHSOLFDWHG DQG DOLDVHG DW PXOWLSOHV RI WKH VDPSOLQJ IUHTXHQF\ >(JD@ 6LQFH WKHUH DUH WZR ]HURFURVVLQJV SHU F\FOH WKH VDPn SOLQJ UDWH LV WZLFH WKH RXWSXW IUHTXHQF\ RI WKH GLYLGHU 7KH PHDVXUHG UHVXOWV LQ >(JD@ VKRZ WKDW WKH GRPLQDQW VRXUFHV DUH WKH RXWSXW DGGLWLYH QRLVH DQG WKH VDPSOHG LQSXW DGGLn WLYH QRLVH 7KHUHIRUH WKHVH VRXUFHV ZLOO EH XVHG WR GHULYH WKH RXWSXW SKDVH QRLVH DV D IXQFn WLRQ RI WKH 615 DW WKH GLYLGHU LQSXW &RQYHUVLRQ )URP $GGLWLYH 1RLVH WR 3KDVH 1RLVH $GGLWLYH QRLVH DV LWV QDPH LPSOLHV DGGV WR WKH VLJQDOV SUHVHQW LQ WKH V\VWHP 7KLV UHVXOWV LQ ERWK DQ DPSOLWXGH DQG SKDVH IOXFWXDWLRQ LQ WKH VLJQDO $OWKRXJK WKLV VHHPV VWUDLJKWIRUZDUG HQRXJK D TXLFN GHULYDWLRQ IRU WKLV FDVH ZLOO EH SUHVHQWHG ,Q DGGLWLRQ WKH GLIIHUHQFH EHWZHHQ VLQJOHVLGHEDQG SKDVHQRLVH GHQVLW\ / DQG SKDVH SRZHUVSHFWUDOGHQn VLW\ 36'f 6A ZLOO EH H[SODLQHG WR FOHDU XS DQ\ PLVFRQFHSWLRQV )LJXUH VKRZV D SKDVRU GLDJUDP UHSUHVHQWLQJ DQ H[DPSOH ZKHUH QRLVH LV DGGHG WR D VLQXVRLGDO VLJQDO ZLWK D URRWPHDQVTXDUHG 506f YROWDJH DPSOLWXGH $ 7KH DGGLWLYH QRLVH LV UHSUHVHQWHG DV YQH-nWf ZKHUH YQ LV DQ 506 YROWDJH 7KLV QRLVH FDQ EH GLYLGHG LQWR RUWKRJRQDO FRPSRQHQWV ZKHUH WKH FRPSRQHQW SDUDOOHO WR $ LV GHILQHG DV DPSOLWXGH QRLVH DQG WKH FRPSRQHQW SHUSHQGLFXODU WR $ LV GHILQHG DV SKDVH QRLVH 7KH SKDVH VKLIW UHVXOWLQJ IURP WKH DGGLWLYH QRLVH LV DV IROORZV

PAGE 154

-f WDQ LI 9QVLQg A WDQO U96LQA 9QVLQ9 $ YfFRV?M$ fn f ZKHUH Mf LV LQ UDGLDQV 7KH ILUVW DSSUR[LPDWLRQ LV D UHVXOW RI YQ m $ ZKLOH WKH VHFRQG DSSUR[LPDWLRQ LV GXH WR D VPDOODQJOH DVVXPSWLRQ 7KLV VKRZV WKDW WKH DPRXQW RI SKDVHVKLIW LV DSSUR[LPDWHO\ HTXDO WR WKH SHUSHQGLFXODU FRPSRQHQW LQ YROWVf GLYLGHG E\ WKH VLJQDO OHYHO LQ YROWVf 'XH WR WKH VPDOODQJOH DSSUR[LPDWLRQ WKH SKDVH QRLVH FDQ HLWKHU UHIHU WR _f GLUHFWO\ RU WR WKH YQ VLQ?ff FRPSRQHQW 7KH PHDQ VTXDUHG YDOXH RI WKH SKDVH VKLIW LV W! e H ? YQ VLQ ?_ f97LI VLQY_fIL_ f§9 f $ N-T $ ZKHUH ( LV WKH H[SHFWHGYDOXH RSHUDWRU YQ DQG ?_ DUH LQGHSHQGHQW DQG ??I KDV D XQLIRUP SUREDELOLW\ GHQVLW\ IXQFWLRQ 7KH IDFWRU RI LQ f PHDQV WKDW KDOI RI WKH DGGLWLYH QRLVH SRZHU LV LQ WKH SKDVH ZKHUHDV WKH RWKHU KDOI LV LQ WKH DPSOLWXGH 7KH DYDLODEOH SRZHU IURP RR D UHVLVWRU LV YQ N7 ‘ $ ZKHUH N LV %ROW]PDQQfV FRQVWDQW [ -.f DQG 7 LV DEVROXWH WHPSHUDWXUH RI WKH V\VWHP 7KLV 36' LV D GRXEOHVLGHEDQG GHQVLW\ PHDQLQJ WKDW LW LV IRU SRVLWLYH IUHTXHQFLHV RQO\ 7KHUHIRUH WKH FRQYHUVLRQ IURP DQ DGGLWLYH QRLVH VRXUFH WR D SKDVH 36' 6AIf LV F Q B  B 1DGG f9n$ $\ 3 N7 VLJ $I f 3 f 6OJ ZKHUH 3VLJ LV WKH SRZHU RI WKH VLJQDO $Af DQG “LV WKH DGGLWLYH QRLVH VSHFWUDO GHQn VLW\ 7KH XQLWV RI f DUH UDG+] 3KDVH QRLVH LV IUHTXHQWO\ GHILQHG DV D VLQJOHVLGHEDQG GHQVLW\ eIf 6LQFH WKH SKDVH 36' 6AIf LV D GRXEOHVLGHEDQG GHQVLW\ WKHQ /LIf LV VLPSO\ RQH KDOI RI 6AIf >(JD@ RU

PAGE 155

/If LV G% ORZHU WKDQ 6AIf 8QLWV RI G%U+] GHFLEHOV DERYH UDG+]f DUH W\SLFDOO\ XVHG IRU 6I\If DQG G%F+] GHFLEHOV DERYH WKH SRZHU RI WKH FDUULHUf IRU /LIf 2XWSXW 3KDVH 1RLVH RI &ORFN 5HFHLYHU 6LQFH WKH IUHTXHQF\ GLYLGHU LQ WKH FORFN UHFHLYHU LV SUHFHGHG E\ DQ /1$ ZLWK VRXUFHIROORZHU EXIIHUVf WKH DGGLWLYH QRLVH VSHFWUXP DW WKH LQSXW RI WKH GLYLGHU LV “GLYDGG N7 ‘ )If ‘ *^If ‘ $I f ZKHUH )If DQG *If DUH WKH QRLVH IDFWRU DQG SRZHU JDLQ RI WKH /1$ ,QWHJUDWLQJ WKLV 36' RYHU DOO IUHTXHQFLHV \LHOGV WKH WRWDO QRLVH SRZHU 1GLY7 DV IROORZV 1GLY7 e>mf )If ‘ *If@GI f $Q HTXLYDOHQW QRLVH EDQGZLGWK LQ +HUW]f FDQ EH GHILQHG DV %:Q N7I4)If*IfGI N7 ‘ )I f*I'f f f ZKHUH I LV WKH UHVRQDQW IUHTXHQF\ RI WKH /1$ 7KH DGGLWLYH 36' LQ f LV FRQYHUWHG WR D SKDVH 36' XVLQJ f WR \LHOG WKH SKDVH QRLVH DW WKH LQSXW RI WKH GLYLGHU DV IROORZV N7 f )If ‘ *^If f 3 VLJ f 7KH SKDVH QRLVH DW WKH RXWSXW RI D GLYLGHU RSHUDWLQJ LQ WKH ODWFKHG PRGH ZLWK GLYLn VLRQ UDWLR 1 GXH WR WKH LQSXW DGGLWLYH QRLVH LV >(JD@ VRXWIf N7 f )If f *If r 1 ‘ 3 ‘ 6OJ f :KHQ WKH GLYLGHU LV DQ ,/)' WKH QRLVH WUDQVIHU IXQFWLRQ IURP WKH LQSXW WR WKH RXWSXW LV ORZSDVV ILOWHUHG OLNH D ILUVWRUGHU SKDVHORFNHG ORRS ZLWK DQ HTXLYDOHQW ORRS EDQGZLGWK

PAGE 156

RI % +]f JLYHQ LQ (O f 7KH RXWSXW SKDVHQRLVH LV WKHQ UHIHU WR DSSHQGL[ (f AK ?f f f ZKHUH 6A LV JLYHQ LQ f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n TXHQF\ 2XWSXW DGGLWLYH SKDVH QRLVH ZLOO VLPSO\ DGG WR f -LWWHU LQ &ORFN 5HFHLYHU $V GHULYHG LQ $SSHQGL[ DQG JLYHQ LQ >(JD@ WKH FORFN MLWWHU DW D QRGH LV UHODWHG WR WKH SKDVH QRLVH DW WKDW QRGH WKURXJK WKH IROORZLQJ IRUPXOD ‘eVWfVLQMFU: f ZKHUH 7 LV WKH REVHUYDWLRQ WLPH 7KH XQLWV RI f DUH LQ VHF 7KH 506 F\FOHWRF\FOH MLWWHU ZRXOG EH WKH VTXDUH URRW RI f ZLWK 7 HTXDO WR WKH SHULRG RI WKH RXWSXW FORFN VLJn QDO 7KH VLQHVTXDUHG IXQFWLRQ KDV QXOOV DW A RU KDUPRQLFV RI WKH RXWSXW IUHTXHQF\ ,I WKH SKDVH 36' LV QRW YDU\LQJ YHU\ IDVW YHUVXV IUHTXHQF\ ZLWK UHVSHFW WR VLQ WI7f WKHQ WKH DYHUDJH YDOXH RI VLQWI7f FDQ EH XVHG 7KLV UHVXOWV LQ WKH IROORZLQJ MLWWHU YDULDQFH f

PAGE 157

1RZ DOO WKH WRROV DUH DYDLODEOH WR GHULYH WKH UHODWLRQVKLS EHWZHHQ VLJQDOWRQRLVH UDWLR 615f DQG FORFN MLWWHU GXH WR LQSXW DGGLWLYH QRLVH 7KH RXWSXW MLWWHU GXH WR DQ LQMHFn WLRQORFNHG GLYLGHU ZLOO EH OHVV WKDQ WKDW IRU D ODWFKHG GLYLGHU KHQFH f LV XVHG IRU WKH RXWSXW SKDVH QRLVH RI WKH GLYLGHU UHSUHVHQWLQJ WKH ZRUVW FDVH 3OXJJLQJ f LQWR f WKH WRWDO RXWSXW MLWWHU RI WKH FORFN UHFHLYHU GXH WR LQSXW DGGLWLYH QRLVH LV DV IROORZV f 6XEVWLWXWLQJ f UHVXOWV LQ f 7KH WLPLQJ MLWWHU FDQ EH FRQYHUWHG WR D SKDVH MLWWHU &F\FLH LQ F\FOHV RU HTXLYDOHQWO\ DV D SHUFHQWDJH RI WKH SHULRGf E\ VLPSO\ PXOWLSO\LQJ f E\ WKH IUHTXHQF\ VTXDUHG \LHOGLQJ f )LJXUH VKRZV D SORW RI F\FOHWRF\FOH 506 FORFN MLWWHU YHUVXV 615 LQ G%f IRU D GLYLVLRQ UDWLR RI 1 $V FDQ EH VHHQ DV WKH 615 LPSURYHV WKH FORFN MLWWHU LPSURYHV DV ZHOO 7KH VORSH RI WKH OLQH LV OLQHDU ZKHQ SORWWLQJ WKH MLWWHU RQ D ORJDULWKPLF VFDOH ZLWK D VORSH RI G% 615f SHU RFWDYH GHFUHDVH LQ MLWWHU 7R REWDLQ DQ 506 MLWWHU RI b RI WKH ORFDO FORFN SHULRG D G% 615 LV UHTXLUHG DW WKH LQSXW RI WKH GLYLGHU :KHQ WKH 615 LV DQG G% WKH ORFDO FORFN MLWWHU LV b DQG b UHVSHFWLYHO\ 7KH IDFW WKDW G% 615 \LHOGV RQO\ b 506 MLWWHU LV GXH WR WKH GLYLGHU UHGXFLQJ WKH MLWWHU LQ SHUFHQWDJHf E\ WKH GLYLn VLRQ UDWLR 7KXV WKH 506 MLWWHU DW WKH LQSXW RI WKH GLYLGHU LV b 7KH SHDNWRSHDN MLWWHU DW WKH LQSXW RI WKH GLYLGHU IRU D G% 615 LV WKHQ DLQ b

PAGE 158

)LJXUH /RFDO FORFN MLWWHU YHUVXV VLJQDOWRQRLVH UDWLR DW LQSXW WR GLYLGHU $ VXUYH\ RI MLWWHU UHVXOWV REWDLQHG LQ WKH OLWHUDWXUH IRU KLJKSHUIRUPDQFH PLFURSURn FHVVRUV >7DP22 %RH .DH 0DL
PAGE 159

7R YHULI\ WKH GHSHQGHQFH RI MLWWHU RQ 615 IRU D IUHTXHQF\ GLYLGHU D ILP IUHn TXHQF\ GLYLGHU ZDV VLPXODWHG DW GLIIHUHQW 615V 7R JHQHUDWH WUDQVLHQW QRLVH DQ DUUD\ RI GDWD SRLQWV UDQGRPO\ GLVWULEXWHG DFFRUGLQJ WR D JDXVVLDQ SUREDELOLW\ GHQVLW\ IXQFn WLRQ ZDV JHQHUDWHG XVLQJ 0DWODE 7KH PHDQ ZDV VHW WR ]HUR DQG WKH VWDQGDUG GHYLDWLRQ ZDV VHW WR 7KLV DUUD\ ZDV LPSRUWHG LQWR 63,&( DW SV WLPH LQWHUYDOV JHQHUDWLQJ D QRLVH WUDQVLHQW YROWDJH QV ORQJ 7KH SV WLPH LQWHUYDO VHWV WKH PD[LPXP IUHTXHQF\ RI WKH QRLVH ZKLOH WKH QV WRWDO OHQJWK RI WLPH VHWV WKH QXPEHU RI GDWD SRLQWV ZKLFK ZLOO EH REWDLQHG IURP WKH VLPXODWLRQ ,Q RWKHU ZRUGV WKHVH WZR QXPEHUV EDVLFDOO\ VHW WKH OLPLWV RI LQWHJUDWLRQ IRU f ZLWK WKH ORZHU OLPLW VHW E\ QVf DQG WKH XSSHU OLPLW VHW E\ SVf %\ DPSOLI\LQJ WKLV QRLVH DQG DGGLQJ LW LQ VHULHV ZLWK WKH LQSXW VLJQDO D JLYHQ 615 FDQ EH REWDLQHG 7KH 615 LV VLPSO\ HTXDO WR WKH UDWLR RI WKH VLJQDO WR WKH VWDQGDUG GHYLD VWDQGDUG GHYLDWLRQ LV HTXDO WR WKH DPSOLILFDWLRQ IDFWRU DSSOLHG WR WKH QRLVH GDWD 7KH SHULRGV RI WKH DQG RXWSXW VLJQDOV ZHUH WKHQ REWDLQHG DQG WKH VWDQGDUG GHYLDWLRQ ZDV FDOFXODWHG 7KH UHVXOW LV VKRZQ LQ )LJXUH ZKLFK SORWV WKH VLPXODWHG MLWWHU YHUVXV 615 FRPSDUHG ZLWK WKH MLWWHU SUHGLFWHG E\ f $V FDQ EH VHHQ WKH VLPXODWHG MLWWHU IRU WKH GLYLGHU DJUHHV YHU\ ZHOO ZLWK WKH SUHGLFWHG MLWWHU &OHDUO\ WKH ORJDULWKP RI WKH MLWWHU GHFUHDVHV OLQHDUO\ ZLWK 615 LQ G%f $OVR )LJXUH VKRZV WKDW WKH MLWWHU LV GHFUHDVLQJ DV 1 EHFRPHV ODUJHU %RWK RI WKHVH UHVXOWV YHULI\ f +RZHYHU WKH VLPXODWHG MLWWHU IRU WKH GLYLGHU LV VOLJKWO\ OHVV WKDQ WKH SUHGLFWHG MLWWHU 7KLV LV WKRXJKW WR EH GXH WR RQO\ KDYLQJ KDOI DV PDQ\ GDWD SRLQWV IRU WKH SHULRGV DV FRPSDUHG WR WKH SHULRGV 6LQFH MLWWHU LQFUHDVHV DV PRUH GDWD DUH LQFOXGHG 7KLV VLPXODWLRQ SURFHGXUH ZDV GHYHORSHG E\ -LP &DVHUWD RI WKH 8QLYHUVLW\ RI )ORULGD

PAGE 160

)LJXUH 6LPXODWLRQ UHVXOWV RI 506 MLWWHU YHUVXV LQSXW 615 IURP _LP IUHTXHQF\ GLYLGHU DIWHU D GLYLVLRQ E\ DQG GLYLVLRQ E\ 7KH VLPXODWHG MLWWHU SRLQWVf LV FRPSDUHG WR WKH SUHGLFWHG MLWWHU OLQHf IURP f FRUUHVSRQGLQJ WR LQWHJUDWLQJ WKH SRZHU VSHFWUDO GHQVLW\ WR ORZHU IUHTXHQFLHVf WKH GLIIHUn HQFH EHWZHHQ WKH H[SHFWHG DQG PHDVXUHG MLWWHU LV DQWLFLSDWHG IRU VKRUW REVHUYDWLRQ WLPHV 6HQVLWLYLW\ DQG 1RLVH 5HTXLUHPHQWV 5HFHLYHU 6HQVLWLYLW\ +DYLQJ GHILQHG WKH 615 UHTXLUHG DW WKH LQSXW RI WKH GLYLGHU WKH UHFHLYHU VHQVLWLYLW\ FDQ QRZ EH GHILQHG 6HQVLWLYLW\ LV GHILQHG DV WKH PLQLPXP VLJQDO OHYHO ZKLFK FDQ EH GHWHFWHG IRU WKH UHTXLUHG 615 $W WKH GLYLGHU LQSXW WKLV VLJQDO OHYHO LV 3VLJ 615 ‘ 1GLY7 615 ‘ N7I4 )If f *IfGI f 615 ‘ N7 ‘ )IDf f *If ‘ %:1 ZKHUH f DQG f KDYH EHHQ XVHG 7KH WHPSHUDWXUH LQ WKLV HTXDWLRQ UHIHUV WR WKH HTXLYDOHQW WHPSHUDWXUH RI WKH DQWHQQD UHSUHVHQWLQJ WKH WRWDO QRLVH DW WKH LQSXW RI WKH

PAGE 161

UHFHLYHU 1RWH WKDW LI WKH LQSXW LQWHUIHUHQFH LV YHU\ ODUJH WKHQ WKH HTXLYDOHQW DQWHQQD WHPn SHUDWXUH FDQ EH PXFK ODUJHU WKDQ >6PL@ (TXLYDOHQWO\ WKH QRLVH ILJXUH FDQ EH FRQVLGHUHG WKDW RI WKH DQWHQQD DQG UHFHLYHU ZKHUH D QRLVH ILJXUH IRU WKH DQWHQQD LV GHILQHG E\ WKH WRWDO QRLVH DQG LQWHUIHUHQFHf DW WKH LQSXW RI WKH UHFHLYHU GLYLGHG E\ WKH DYDLODEOH WKHUPDO QRLVH N7f 1RWH WKDW IRU YHU\ ODUJH DQWHQQD QRLVH ILJXUHV WKH QRLVH ILJXUH RI WKH UHFHLYHU EHFRPHV XQLPSRUWDQW PHDQLQJ WKDW WKH QRLVH DGGHG E\ WKH UHFHLYHU LV D VPDOO IUDFWLRQ RI WKH QRLVH DOUHDG\ SUHVHQW DW WKH LQSXW &RQYHUWLQJ f WR G% DQG G%P \LHOGV WKH IROORZLQJ IRU 7 r & ZKLFK LV DVVXPHG WR EH WKH WHPSHUDWXUH RI WKH PLFURSURFHVVRUf 3fr?G%P G%P f Ae 1)If *If?G%PRJO4%:1f 615PLQ?G% 6XEWUDFWLQJ WKH /1$ JDLQ *If?%f IURP WKH VLJQDO SRZHU DW WKH GLYLGHU LQSXW DQG VXEVWLn WXWLQJ 615PLQ G% \LHOGV WKH LQSXW VHQVLWLYLW\ IRU WKH FORFN UHFHLYHU DV IROORZV G%P 1)IDfORJ -%:Qf f G%P +] 5HIHUULQJ EDFN WR VHFWLRQ WKH WUDQVPLWWHU SRZHU OHYHO LV VSHFLILHG WR EH G%P WKH DQWHQQDWRDQWHQQD JDLQ LV VSHFLILHG WR EH G% DW D FP GLVWDQFH DQG WKH PLVPDWFK ORVV LV VSHFLILHG WR EH G% 7KHUHIRUH WKH GHVLUHG VHQVLWLYLW\ LV G%P G%P G% G% f 5HFHLYHU 1RLVH 5HTXLUHPHQWV 7KH QRLVH SHUIRUPDQFH RI WKH UHFHLYHU FDQ QRZ EH VSHFLILHG E\ SOXJJLQJ LQ WKH G%P VHQVLWLYLW\ LQWR f 7KLV \LHOGV WKH QRLVH ILJXUH RI WKH UHFHLYHU DW WKH FHQWHU IUHTXHQF\ DQG WKH QRLVH EDQGZLGWK DV IROORZV

PAGE 162

1)IDf O2ORJ :1f G% f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f 5HIHUULQJ WR WKH UHVXOWV LQ &KDSWHU LV DSSUR[LPDWHO\ DW *+] IRU WKH SP /1$f DQG DW *+] IRU WKH _LP /1$ 7KXV %:1 *+] DQG RJM%:1f G% :LWK WKLV DVVXPSWLRQ WKH PD[LPXP QRLVH ILJXUH FDQ EH VSHFLILHG DV G% ZLWK WKH H[SOLFLW HTXDWLRQ EHLQJ IN7 ? 1)IRfPD[ 3VHQV?G%PAORJ>A\615PLQ?G%O2ORJOR%:1f f 7KXV WKH V\VWHP VSHFLILFDWLRQ IRU PD[LPXP QRLVH ILJXUH RI WKH UHFHLYHU LQFOXGLQJ WKH DQWHQQD QRLVH ILJXUH DWI4 LV VHW WR G% 7KLV QRLVH ILJXUH VSHFLILFDWLRQ LV IRU D G%P VHQVLWLYLW\ +RZHYHU DV PHQWLRQHG HDUOLHU LI WKH LQSXW LQWHUIHUHQFH WR WKH DQWHQQD LV VLJQLILFDQW WKHQ WKH 1) LQ f LV WKH WRWDO QRLVH ILJXUH RI WKH DQWHQQD SOXV WKH UHFHLYHU LH )A )UFYU f $V WKH WRWDO QRLVH DQG LQWHUIHUHQFH DW WKH UHFHLYHU LQSXW LV TXDQWLILHG WKH G% VSHFLILFDWLRQ IRU 1) PD\ EHFRPH LQIHDVLEOH $V VXFK WKH VHQVLWLYLW\ RI WKH UHFHLYHU ZLOO KDYH WR EH PRGLILHG WR D ODUJHU SRZHU WR PDLQWDLQ WKH G% 615 DW WKH GLYLGHU LQSXW ,Q RWKHU ZRUGV FXUUHQWO\ WKH VHQVLWLYLW\ VSHFLILFDWLRQ LV GULYLQJ WKH QRLVH ILJXUH VSHFLILFDWLRQ +RZHYHU ZLWK WKH WRWDO

PAGE 163

QRLVH DQG LQWHUIHUHQFH DW WKH LQSXW RI WKH UHFHLYHU IXOO\ TXDQWLILHG WKLV QRLVH ZLOO GULYH WKH VHQVLWLYLW\ VSHFLILFDWLRQ (VWLPDWLRQ RI 7RWDO 1RLVH 615 DQG -LWWHU IRU SP 5HFHLYHU 7KH FRQFHSWV LQ WKH SDVW WZR VHFWLRQV KDYH EHHQ DSSOLHG WR WKH _LP /1$ DQG VRXUFHIROORZHU FLUFXLWV )LJXUH Df VKRZV WKH VLPXODWHG JDLQ DQG 1) IRU WKH /1$ 7KH PD[LPXP JDLQ LV G% DQG WKH PLQLPXP 1) LV G% DW *+] )LJXUH Ef VKRZV WKH SRZHU VSHFWUDO GHQVLW\ RI WKH QRLVH DW WKH LQSXW RI WKH IUHTXHQF\ GLYLGHU ZKLFK ZDV JLYHQ LQ f ,QWHJUDWLQJ WKLV 36' RYHU DOO IUHTXHQFLHV \LHOGV WKH WRWDO WKHUPDO QRLVH SRZHUf§1GLY7 IURP fZKLFK LV G%P 7R SHUIRUP DQ LQWHJUDO LQ 63,&( RYHU DOO IUHTXHQFLHV WKH IUHTXHQF\ OLPLWV DUH LQFUHDVHG XQWLO WKH WRWDO LQWHJUDO GRHV QRW FKDQJH a *+] IRU WKLV FDVHf 8VLQJ WKH YDOXHV DQG G% IRU *If DQG 1)If UHVSHFWLYHO\ DQG 1GLY; RI G%P %:1 LV FDOFXODWHG WR EH *+] 7KLV LV VKRZQ LQ )LJXUH Ef )LJXUH Df 6LPXODWHG JDLQ DQG QRLVH ILJXUH IRU _OP /1$ ZLWK VRXUFHIROORZHUV Ef 6LPXODWHG SRZHU VSHFWUDO GHQVLW\ RI WKHUPDO QRLVH DW LQSXW RI IUHTXHQF\ GLYLGHU DORQJ ZLWK HTXLYDOHQW QRLVH EDQGZLGWK %:1f

PAGE 164

ZKHUH WKH 36' LV IL[HG DW WKH FHQWHUIUHTXHQF\ YDOXH DFURVV WKH QRLVH EDQGZLGWK 7KH QRLVH EDQGZLGWK DJUHHV YHU\ ZHOO ZLWK WKH HVWLPDWHG YDOXH RI *+] JLYHQ LQ f 7KH 615 FDQ QRZ EH FDOFXODWHG DVVXPLQJ WKDW WKH LQSXW VLJQDO WR WKH UHFHLYHU LV G%P DQG WKDW WKH WRWDO LQSXW QRLVH LV GRPLQDWHG E\ WKHUPDO QRLVH DV ZDV VLPLODUO\ GRQH LQ >%UDY22E@f 7KH G%P LQSXW SRZHU LV WKH GHVLUHG VHQVLWLYLW\ ZKLFK DJDLQ LV D UHVXOW RI G%P WUDQVPLWWHG SRZHU G% DQWHQQDWRDQWHQQD JDLQ DW FP G% RI LQWHUIHUHQFH HIIHFWV DQG G% RI PLVPDWFK ORVV 7KH VLJQDO SRZHU OHYHO DW WKH LQSXW RI WKH GLYLGHU LV G%P SOXV WKH JDLQ RI WKH /1$ G%f RU G%P 7KHUHIRUH WKH VLPn XODWHG 615 LV G%P G%Pf G% $SSO\LQJ WKH G% 615 WR f WKH H[SHFWHG 506 F\FOHWRF\FOH MLWWHU LV SV DW WKH RXWSXW FORFN IUHTXHQF\ RI *+] FRUUHn VSRQGLQJ WR b 506 MLWWHU 7KH SHDNWRSHDN MLWWHU LV SV ZKLFK LV ab RI WKH FORFN SHULRG 7KLV HDVLO\ PHHWV WKH VSHFLILFDWLRQ RI OHVV WKDQ b RI SHDNWRSHDN MLWWHU /LQHDULW\ 6SHFLILFDWLRQ 1RQOLQHDULW\ LQ WKH /1$ ZLWK VRXUFHIROORZHUVf ZLOO UHVXOW LQ WKH IROORZLQJ PDLQ HIIHFWV JHQHUDWLRQ RI KDUPRQLFV JDLQ FRPSUHVVLRQ VSHFLILHG LQ WHUPV RI D G% FRPSUHVn VLRQ SRLQW 3cG%f! DQG LQWHUPRGXODWLRQ VSHFLILHG LQ WHUPV RI D WKLUGRUGHU LQWHUFHSW SRLQW ,3f $ GLIILFXOW\ HQFRXQWHUHG LQ WKH FORFN UHFHLYHU LV WKH IDFW WKDW WKH VLJQDOV RXWVLGH RI WKH fEDQG RI LQWHUHVWf DUH QRW KHDYLO\ DWWHQXDWHG ,Q RWKHU ZRUGV WKH UHVSRQVH RI WKH /1$ LV YHU\ EURDGEDQG GXH WR WKH OLPLWHG LQGXFWRU 4 LQ WKH /1$ 7KHUHIRUH LQWHUIHULQJ VLJQDOV WKHPVHOYHV SRVH PRUH RI D SUREOHP WKDQ WKH KDUPRQLFV RU LQWHUPRGXODWLRQ SURGXFWV WKHVH VLJQDOV JHQHUDWH ,Q IDFW LI WKH LQWHUIHULQJ VLJQDOV EHFRPH WRR ODUJH WKH\ FDQ ORFN WKH GLYLGHU DW D GLIIHUHQW IUHTXHQF\ HYHQ ZLWK WKH UHGXFHG JDLQ RI WKH /1$ DW WKH LQWHUIHUHUfV IUHTXHQF\ $W WKH YHU\ OHDVW WKHVH LQWHUIHULQJ VLJQDOV ZLOO GHJUDGH WKH 615 LQFUHDVLQJ WKH

PAGE 165

FORFN MLWWHU 7KRXJK QRW WKH PDLQ FRQFHUQ WKH KDUPRQLFV DQG LQWHUPRGXODWLRQ SURGXFWV ZLOO VWLOO GHJUDGH WKH 615 2QH RI WKH PDLQ UHDVRQV WKH JOREDO FORFN LV GLVWULEXWHG DW D KLJKHU IUHTXHQF\ WKDQ WKH ORFDO FORFN LV WR WU\ WR RSHUDWH LQ D EDQG ZKLFK LV QRW DV QRLV\ 7KH V\VWHP ZLOO EH JHQn HUDWLQJ D VLJQLILFDQW DPRXQW RI QRLVH DW DQG EHORZ WKH ORFDO FORFN IUHTXHQF\ DQG DW IUHn TXHQFLHV GXH WR WKH HGJHUDWHV RI WKH VLJQDOV 2QH SRWHQWLDO SUREOHPDWLF IUHTXHQF\ LV DURXQG RQHWKLUG RI WKH JOREDO FORFN IUHTXHQF\ *&/.f 'XH WR QRQOLQHDULW\ WKH /1$ ZLOO JHQHUDWH D KDUPRQLF QHDU *&/. :KHQ WKLV VLJQDO LV DGGHG WR WKH GHVLUHG JOREDO FORFN WKH 615 ZLOO GHFUHDVH DQG MLWWHU ZLOO LQFUHDVH 7KLV VFHQDULR FDQ WKHQ EH XVHG WR VSHFLI\ WKH ,,3 RI WKH /1$ 1RWH WKDW D VFHQDULR H[LVWV IRU WKH FDVH RI WZR LQWHUIHUHUV JHQn HUDWLQJ DQ LQWHUPRGXODWLRQ SURGXFW +RZHYHU VLQFH WKHVH LQWHUIHUHUV PXVW EH IDU HQRXJK DZD\ IURP *&/. WR SUHYHQW GLYLGHU ORFNLQJ WKH +3 ZRXOG EH YHU\ VLPLODU WR WKH FDVH GLVFXVVHG KHUH )LJXUH VKRZV DQ LOOXVWUDWLRQ RI WKH FDVH EHLQJ FRQVLGHUHG 7ZR VLQXVRLGDO LQSXW VLJQDOV DUH SUHVHQW DW WKH LQSXW RI WKH UHFHLYHU 2QH LV DW WKH JOREDO FORFN IUHTXHQF\ If KDYLQJ D YROWDJH DPSOLWXGH $VHQV FRUUHVSRQGLQJ WR WKH VHQVLWLYLW\ 7KH LQWHUIHULQJ IR VLJQDO LV DW D IUHTXHQF\ ZLWK DQ DPSOLWXGH $LQW 7KH UHVSRQVH RI D QRQOLQHDU /1$ FDQ EH PRGHOHG ZLWK D SRZHU VHULHV DV \If W;M f [Wf D f [Wf D f [Wf f )RU WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP WKH ORFDO FORFN IUHTXHQF\ ZLOO EH DW RI *&/. $VVXPLQJ WKDW WKLV ORFDO FORFN VLJQDO LV VTXDUH LWV WKLUG KDUPRQLF ZLOO EH TXLWH VWURQJ RFFXUULQJ DW D IUHTXHQF\ RI RU f RI WKH *&/. 7KLV VXEVHTXHQWO\ FDQ JHQHUDWH D WKLUGRUGHU KDUPRQLF DW RU f RI WKH *&/. GXH WR /1$ QRQOLQHDULWLHV

PAGE 166

'HVLUHG ,QWHUIHUHU $c QW 'HVLUHG *&/. £ APGV I IR 6LJQDOV ,QWR 5HFHLYHU ,QWHUIHUHU D![. m LIRf$PGV +DUPRQLF IR?$ m, ,$LQW  IR 6LJQDOV 2XW RI /1$ ,QWR 'LYLGHU )LJXUH ,OOXVWUDWLRQ RI VFHQDULR XVHG WR VSHFLI\ /1$ ,,3 ZKHUH [ DQG \ DUH WKH LQSXW DQG RXWSXW RI WKH /1$ 2&@ LV WKH JDLQ RI WKH /1$ DQG D DQG D DUH WKH VHFRQG DQG WKLUGRUGHU QRQOLQHDULW\ YROWDJH FRHIILFLHQWV UHVSHFWLYHO\ )RU D IXOO\GLIIHUHQWLDO FLUFXLW WKH HYHQRUGHU QRQOLQHDULW\ FRHIILFLHQWV DUH FORVH WR ]HUR >5D]@ 7KHUHIRUH WKH RXWSXW IURP WKH /1$ IRU WKH LQWHUIHULQJ VLJQDO DW LV DV IROORZV \ >D\rf$Of _DMf$MFRVrf LDAf$AFR6,W} f ZKHUH WKH IUHTXHQF\ GHSHQGHQFH RI WKH D SDUDPHWHUV KDV EHHQ LQFOXGHG 7KH WKLUGKDUn PRQLF WHUP LQ WKLV H[SUHVVLRQ ZLOO LQWHUIHUH ZLWK WKH GHVLUHG *&/. VLJQDO GHJUDGLQJ WKH 615 DQG HYHQWXDOO\ RYHUWDNLQJ WKH GHVLUHG VLJQDO 7KH VFHQDULR XVHG WR VSHFLI\ -,3 LV QRZ GHVFULEHG ,QWHJUDO WR DQ\ ,,3 VSHFLILFDn WLRQ LV WKH FRQVWUDLQW RQ WKH SRZHU RI WKH LQWHUIHUHU ,Q WKH FDVH RI WKH FORFN UHFHLYHU WKH SRZHU RI WKH LQWHUIHUHU LV FRQVWUDLQHG E\ VHWWLQJ WKH UDWLR EHWZHHQ WKH RXWSXW SRZHU RI WKH IR GHVLUHG VLJQDO DW ZKHQ LW LV DW LWV VHQVLWLYLW\ WR WKH RXWSXW SRZHU RI WKH LQWHUIHUHU DW WR WKH UHTXLUHG 615 WR PHHW D JLYHQ MLWWHU VSHFLILFDWLRQ ,Q HTXDWLRQ IRUP WKLV LV ZULWWHQ DV

PAGE 167

615JWf f >Df$QA ZKHUH 615D7f LV JLYHQ LQ f 1RQOLQHDULW\ ZLOO JHQHUDWH D WKLUGRUGHU KDUPRQLF RI WKH LQWHUIHUHU ZKLFK ZLOO FRUUXSW WKH 615 7KH '3 LV VSHFLILHG IRU WKH FDVH ZKHQ WKH 615 LV GHJUDGHG E\ G% GXH WR WKLV WKLUGRUGHU KDUPRQLF 7KXV WKH RXWSXW SRZHU RI WKH IR WKLUGRUGHU KDUPRQLF DW ID VKRXOG EH HTXDO WR WKH RXWSXW SRZHU RI WKH LQWHUIHUHU DW M 7KLV UHVXOWV LQ :W!-f ‘ >+Wn.@f 7KH UDWLR RI mM WR RF FDQ EH UHODWHG WR WKH ,,3 DW WKDW IUHTXHQF\ >5D]@ 6LQFH RFM LV QRW D UHDGLO\NQRZQ TXDQWLW\ WKH IROORZLQJ DVVXPSWLRQ LV PDGH f R9 m IRf Df a $3 f f 7KLV DVVXPSWLRQ VWDWHV WKDW WKH WKLUGRUGHU QRQOLQHDULW\ KDV WKH VDPH IUHTXHQF\ UHVSRQVH IR DV WKH JDLQ RU HTXLYDOHQWO\ WKDW ,,3f LV HTXDO WR ,,3Af 8VLQJ f DQG f LQ f UHVXOWV LQ WKH IROORZLQJ H[SUHVVLRQ IRU $,,3 WKH LQSXW DPSOLWXGH LQ 9ROWV FRUUHVSRQGLQJ WR ,,3 LQ :DWWVf $,,3 a m f -615D7f f 7KLV FDQ EH H[SUHVVHG LQ G%P DV IROORZV 3, 3? nG%P VHQV?G%P */1$IRf?G%*/1$^Mf G% 615FWf f§ f

PAGE 168

7KH /1$ JDLQ DW L-! FDQ EH IRXQG IURP VLPXODWLRQV 7KHVH VLPXODWLRQV VKRZ WKDW WKH DWWHQXDWLRQ DW LV DSSUR[LPDWHO\ G% OHVV WKDQ WKH JDLQ DW I ,I KDOI RI WKH JDLQ LQ WKH UHFHLYHU LV WKH JDLQ LQ WKH /1$ WKHQ IRU D G%P VHQVLWLYLW\ D G% UHFHLYHU JDLQ DQG D G% 615 IRU b 506 MLWWHUf WKH VSHFLILHG +3 LV f G%P $OWKRXJK WKLV b 506 MLWWHU LV PRUH WKDQ WKH b WDUJHW WKLV UHVXOWV LQ D PRUH DJJUHVVLYH ,,3 VSHFLILFDWLRQ 7KXV WKH V\VWHP VSHFLILFDWLRQ IRU ,,3 LV G%P 6XPPDU\ ,Q WKLV FKDSWHU V\VWHP VSHFLILFDWLRQV KDYH EHHQ GHYHORSHG IRU D ZLUHOHVV FORFN GLVn WULEXWLRQ V\VWHP FRQYHUWLQJ VWDQGDUG FORFN PHWULFV LQWR 5) PHWULFV 7R PD[LPL]H WKH SRZHU WUDQVIHU IURP WKH FORFN VRXUFH WR WKH ORFDO FORFN V\VWHP PDWFKLQJ JDLQ DQG DQWHQQD UHTXLUHPHQWV KDYH EHHQ VSHFLILHG ,Q SDUWLFXODU WKH UHFHLYHU JDLQ VKRXOG DSSUR[Ln PDWHO\ HTXDO WKH ORVV WKURXJK WKH DQWHQQDV &ORFN VNHZ DQG MLWWHU KDYH EHHQ GHILQHG DQG DUH XVHG WR VHW WKH UHPDLQLQJ SHUIRUPDQFH PHWULFV RI WKH V\VWHP $PSOLWXGH PLVPDWFK DW WKH LQSXW RI WKH IUHTXHQF\ GLYLGHU UHVXOWV LQ FORFN VNHZ WKXV WKH VLJQDOOHYHOV DW WKH LQSXW RI WKH GLYLGHU VKRXOG EH HTXDO 7KHUPDO QRLVH DW WKH LQSXW RI WKH GLYLGHU ZLOO UHVXOW LQ FORFN MLWWHU GXH WR LWV FRQYHUVLRQ WR SKDVH QRLVH $V WKH VLJQDOWRQRLVH UDWLR 615f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

PAGE 169

GLIIHUHQW JOREDO FORFN IUHTXHQF\ LV WR EH XVHG WKHQ WKHVH VSHFLILFDWLRQV FDQ EH PRGLILHG XVLQJ WKH IRUPXODV SUHVHQWHG LQ WKLV FKDSWHU 6RPH JHQHUDO UHTXLUHPHQWV IRU WKH V\VWHP ZKLFK KDYH QRW EHHQ GLVFXVVHG LQFOXGH SRZHU FRQVXPSWLRQ EDODQFHG WRSRORJ\ DQG SDVVLYH FRPSRQHQWV )LUVW WKH SRZHU FRQn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fV RI a DW *+] DUH UHFRPPHQGHG +RZHYHU DV KDV EHHQ VKRZQ LQ &KDSWHU DQG ZLOO EH VKRZQ LQ &KDSWHU KLJK4 FLUFXLWV DUH PRUH VXVFHSWLEOH WR SURFHVV YDULDWLRQV WKXV WKHUH DUH WUDGHRIIV LQYROYHG LQ FKRRVLQJ WKH LQGXFWRU 4 0RUH ZRUN LV UHTXLUHG WR GHYHORS WKH RSWLPDO 4 DQG RU WR GHYHORS KLJK4 FLUFXLWV ZKLFK DUH QRW DV VXVFHSWLEOH WR SURFHVV YDULDWLRQV

PAGE 170

7DEOH 6\VWHP VSHFLILFDWLRQV IRU ZLUHOHVV FORFN GLVWULEXWLRQ DW *+] 3HUIRUPDQFH 0HWULF 6SHFLILFDWLRQ *OREDO FORFN IUHTXHQF\ *+] /RFDO &ORFN )UHTXHQF\ *+] 2XWSXW &ORFN 6NHZ b RI ORFDO FORFN SHULRGf 2XWSXW &ORFN -LWWHU 3HDNWRSHDN F\FOHWRF\FOH b 506 F\FOHWRF\FOH b 7UDQVPLWWHU 3RZHU G%P $QWHQQDWR$QWHQQD *DLQ ZLWK ,QWHUIHUHQFH G% DW FP VHSDUDWLRQ G% DW FP VHSDUDWLRQ ERWK LQFOXGLQJ G% LQWHUIHUHQFH ORVVf 5HFHLYHU *DLQ G% *DLQ DQWHQQDSDLU ORVVf /1$ UHVRQDQW IUHT IL62 RI GLYLGHUf 0DWFKLQJ %HWZHHQ $QWHQQD DQG &LUFXLW 7RWDO PLVPDWFK ORVV G% UHT G% 96:5 5HFHLYHU 6HQVLWLYLW\ G%P 1RLVH )LJXUH G% ,,3 G%P $PSOLWXGH 0DWFKLQJ DW 'LYLGHU ,QSXW IRU b FORFN VNHZf ZLWKLQ G% ZRUVWFDVH DW OHDVW GLYLGHU DW 0'6f ZLWKLQ G% DOO GLYLGHUV DW OHDVW 0'6 G%f 6LJQDOWR1RLVH 5DWLR DW 'LYLGHU ,QSXW G% IRU b 506 MLWWHUf 5HFHLYHU 3RZHU &RQVXPSWLRQ P: SHU UHFHLYHU DW 9AG 1RWH 7KHVH VSHFLILFDWLRQV DUH GHSHQGHQW RQ JOREDO FORFN IUHTXHQF\

PAGE 171

&+$37(5 :,5(/(66 ,17(5&211(&76 )25 &/2&. ',675,%87,21 2YHUYLHZ 7KH FORFN UHFHLYHU VKRZQ LQ )LJXUH OEf FRQWDLQV DQ LQWHJUDWHG UHFHLYLQJ DQWHQQD D IXOO\GLIIHUHQWLDO /1$ D SDLU RI VRXUFHIROORZHU EXIIHUV D IUHTXHQF\ GLYLGHU ZKLFK WUDQVODWHV WKH JOREDO FORFN IUHTXHQF\ WR WKH ORFDO FORFN IUHTXHQF\ DQG RXWSXW EXIIn HUV ZKLFK GULYH WKH ORFDOFORFN FDSDFLWLYH ORDG 7KH GHVLJQ DQG LPSOHPHQWDWLRQ RI &026 /1$V DQG IUHTXHQF\ GLYLGHUV KDV EHHQ SUHVHQWHG LQ FKDSWHUV ZKLOH FKDSWHU SUHn VHQWHG WKH V\VWHP UHTXLUHPHQWV IRU ZLUHOHVV FORFN GLVWULEXWLRQ 7KLV FKDSWHU XVHV WKHVH FRQFHSWV DQG FLUFXLWV WR LPSOHPHQW ZLUHOHVV LQWHUFRQQHFW V\VWHPV DQG WR GHPRQVWUDWH V\Vn WHP RSHUDWLRQ $ FULWLFDO FRPSRQHQW RI D ZLUHOHVV LQWHUFRQQHFW V\VWHP LV WKH DQWHQQD ZKLFK DFWV DV D WUDQVGXFHU EHWZHHQ WKH HOHFWURPDJQHWLF ZDYH DQG WKH FXUUHQW DQG YROWDJH ZLWKLQ WKH FLUFXLWU\ 7KLV FKDSWHU SUHVHQWV D EULHI UHYLHZ RI DQWHQQD IXQGDPHQWDOV 7KHQ WKH PHDVXUHG FKDUDFWHULVWLFV RI RQFKLS DQWHQQD SDLUV DQG DQWHQQDV FRQQHFWHG WR DQ /1$ DUH SUHVHQWHG 7KHVH UHVXOWV DUH IURP D _LP &026 WHVW FKLS 8VLQJ WKLV FKLS D *+] VLQJOHUHFHLYHU ZLUHOHVV LQWHUFRQQHFW DFURVV D PP GLVWDQFH LV SUHVHQWHGf§WKH ILUVW RQFKLS &026 ZLUHOHVV LQWHUFRQQHFW HYHU GHPRQVWUDWHG $ VHFRQG WHVW FKLS ZDV LPSOHPHQWHG LQ D _LP &026 SURFHVV ZLWK FRSSHU LQWHUFRQQHFWV ,Q WKLV FKLS PXOWLn SOH DQWHQQD WHVW VWUXFWXUHV DUH LQFOXGHG DQG FKDUDFWHUL]HG $OVR LQ WKH SP WHVW FKLS VLQJOHUHFHLYHU DQG GRXEOHUHFHLYHU ZLUHOHVV LQWHUFRQQHFWV DUH GHPRQVWUDWHG DFURVV DQG PP LQWHUFRQQHFWLRQ GLVWDQFHV DW *+] 7KHVH UHFHLYHUV KDYH LQLWLDOL]DWLRQ

PAGE 172

FLUFXLWU\ WR UHGXFH WKH FORFN VNHZ )LQDOO\ D ZLUHOHVV FORFN WUDQVPLWWHU QRW LQ D 3//f KDV EHHQ LPSOHPHQWHG FKDUDFWHUL]HG DQG GHPRQVWUDWHG IRU D PP LQWHUFRQQHFWLRQ GLVn WDQFH 7KHVH UHVXOWV FRQFOXVLYHO\ GHPRQVWUDWH V\VWHP RSHUDWLRQ DQG SODXVLELOLW\ 2Q&KLS $QWHQQDV $QWHQQD )XQGDPHQWDOV 7R HYDOXDWH RQFKLS DQWHQQDV D UHYLHZ RI EDVLF DQWHQQD IXQGDPHQWDOV LV UHTXLUHG VXFK DV UDGLDWLRQ UHVLVWDQFH DQWHQQD LPSHGDQFH GLUHFWLYLW\ DQWHQQD JDLQ DQG QHDU IDUILHOG GHILQLWLRQV 5DGLDWLRQ UHVLVWDQFH 5Uf LV HTXDO WR WKH WRWDO SRZHU UDGLDWHG E\ WKH DQWHQQD GLYLGHG E\ WKH URRWPHDQVTXDUHG UPVf FXUUHQW LQ WKH DQWHQQD +HQFH 5U UHSUHn VHQWV WKH UDGLDWLRQ RI DQ DQWHQQD $Q DQWHQQD ZLWK D VPDOO UDGLDWLRQ UHVLVWDQFH LQGLFDWHV WKDW D ODUJH FXUUHQW ZRXOG EH UHTXLUHG WR UDGLDWH SRZHU .HHS LQ PLQG WKDW DQWHQQDV DUH UHFLSURFDO ZKLFK PHDQV WKDW WKH UDGLDWLRQ SURSHUWLHV DSSO\ ERWK WR DQ DQWHQQDfV WUDQVPLWn WLQJ DQG UHFHLYLQJ FDSDELOLWLHV 7KH DQWHQQD LPSHGDQFH LV WKH LPSHGDQFH fVHHQf ORRNLQJ LQWR WKH DQWHQQD GHILQHG DV WKH UDWLR RI YROWDJH WR FXUUHQW DW WKH DQWHQQD WHUPLQDOV 7KH DQWHQQD UHVLVWDQFH LV FRPn SRVHG RI WKH UDGLDWLRQ UHVLVWDQFH DQG ORVV UHVLVWDQFH RI WKH DQWHQQD DV ZHOO DV VXEVWUDWH FRPSRQHQWV 7KH DQWHQQD UHDFWDQFH LV FRPSRVHG RI WKH FDSDFLWDQFH DQG LQGXFWDQFH RI WKH DQWHQQD OLQHV DV ZHOO DV WKH VXEVWUDWH FRPSRQHQWV 7R PD[LPL]H WKH SRZHU WUDQVIHU EHWZHHQ WKH DQWHQQD DQG WKH FLUFXLWU\ WKH FLUFXLW LPSHGDQFH VKRXOG EH WKH FRQMXJDWH RI WKH DQWHQQD LPSHGDQFH 'LUHFWLYLW\ 'f LV D ILJXUHRIPHULW XVHG WR TXDQWLI\ WKH GLUHFWLRQDO SURSHUWLHV RI DQ DQWHQQD 7KH GLUHFWLYLW\ RI DQ DQWHQQD LV WKH UDGLDWLRQ LQWHQVLW\ UDGLDWHG SRZHU SHU XQLW VROLG DQJOHf IRU WKDW DQWHQQD LQ D JLYHQ GLUHFWLRQ GLYLGHG E\ WKH UDGLDWLRQ LQWHQVLW\ RI D

PAGE 173

QRQGLUHFWLYH RU LVRWURSLF DQWHQQD +HQFH WKH GLUHFWLYLW\ LV PD[LPXP LQ WKH GLUHFWLRQ RI PD[LPXP UDGLDWLRQ 7KH KLJKHU WKH GLUHFWLYLW\ WKH VWURQJHU WKH DQWHQQDfV UDGLDWHG HQHUJ\ LV IRU WKDW GLUHFWLRQ $OO SUDFWLFDO DQWHQQDV DUH GLUHFWLRQDO '!f LQ WKDW WKH GLUHFWLYLW\ LV QRW FRQVWDQW IRU DOO GLUHFWLRQV $ VXEFDWHJRU\ RI GLUHFWLRQDO DQWHQQDV LV WKH RPQLGLUHFn WLRQDO DQWHQQD ZKRVH GLUHFWLYLW\ LV HVVHQWLDOO\ FRQVWDQW DFURVV D FHUWDLQ SODQH 7KH JDLQ EHWZHHQ D WUDQVPLWWLQJUHFHLYLQJ DQWHQQD SDLU FDQ EH GHWHUPLQHG XVLQJ )ULLV WUDQVPLVVLRQ HTXDWLRQ >%DO@ ZKLFK GHVFULEHV WKH SRZHU UHFHLYHG WR WKH SRZHU WUDQVPLWWHG EHWZHHQ WZR DQWHQQDV DV IROORZV f ZKHUH Hc LV DQ HIILFLHQF\ UHSUHVHQWLQJ ORVV LQ WKH FRQGXFWRUV DQG GLHOHFWULFV HTXDO WR WKH UDGLDWLRQ UHVLVWDQFH GLYLGHG E\ WKH VXP RI WKH UDGLDWLRQ UHVLVWDQFH DQG WKH ORVV UHVLVWDQFH >%DO@f Uc LV WKH UHIOHFWLRQ FRHIILFLHQW DW WKH DQWHQQD WHUPLQDOV 'c LV WKH GLUHFWLYLW\ DQG $ DQG U DUH WKH ZDYHOHQJWK DQG VHSDUDWLRQ GLVWDQFH UHVSHFWLYHO\ :KHQ FKDUDFWHUL]LQJ DQ DQWHQQD SDLU XVLQJ 6SDUDPHWHUV WKLV )ULLV WUDQVPLVVLRQ HTXDWLRQ LV HTXDO WR ,6LO >.LP22D@ ZKLFK LV HTXDO WR WKH WUDQVGXFHU JDLQ *7f ZKHQ WKH DQWHQQDV DUH PHDVXUHG ZLWK 4 WUDQVPLVVLRQ OLQHV DQG WHUPLQDWLRQ LPSHGDQFHV >*RQ@ 7R FKDUDFWHUL]H WKH DQWHQQD SHUIRUPDQFH XQGHU PDWFKHG FRQGLWLRQV D WUDQVPLVVLRQ JDLQ *Df LV GHILQHG DV f O_6Q_fO_6_f 5HIHUULQJ WR f *D LV HTXDO WR WKH TXDQWLW\ HWHU'W'U$WUf 7KLV JDLQ LV HTXDO WR WKH SRZHU DYDLODEOH DW WKH RXWSXW GLYLGHG E\ WKH SRZHU GHOLYHUHG WR WKH LQSXW UHSUHVHQWLQJ WKH EHVW SRVVLEOH JDLQ IRU WKH DQWHQQDV ZKHUH ERWK DQWHQQDV DUH FRQMXJDWHO\ PDWFKHG

PAGE 174

7KHUHIRUH *D LV XVHG WR FKDUDFWHUL]H RQFKLS DQWHQQD SHUIRUPDQFH >.LP22D@ 1RWH WKDW LQ DFWXDO V\VWHP LPSOHPHQWDWLRQV WKH DQWHQQDV ZRXOG RQO\ EH PDWFKHG IRU D OLPLWHG IUHn TXHQF\ UDQJH WKXV *D ZRXOG RQO\ EH HTXDO WR WKH DFWXDO JDLQ RYHU WKRVH IUHTXHQFLHV )ULLV WUDQVPLVVLRQ IRUPXOD DVVXPHV WKDW WKH WUDQVPLWWLQJ DQG UHFHLYLQJ DQWHQQDV DUH IDU HQRXJK DSDUW VXFK WKDW WKH\ DUH LQ WKH IDUILHOG UHJLRQ ,Q WKH IDUILHOG UHJLRQ RQO\ UDGLDWLRQ WHUPV Uf RI WKH HOHFWURPDJQHWLF ILHOG GRPLQDWH SRZHU GHQVLW\ a Uf DQG WKH DQJXODU GLVWULEXWLRQ RI WKHVH WHUPV LV LQGHSHQGHQW RI GLVWDQFH >:DQ@ $ QHDUILHOG UHJLRQ DOVR H[LVWV ZKHUH WKH UHDFWLYH WHUPV RI WKH HOHFWURPDJQHWLF ILHOG GRPLQDWH LH WKH SRZHU GHQVLW\ LV FRPSOH[f 7KLV UHDFWLYH SRZHU UHIHUV WR HQHUJ\ VWRUDJH LQ WKH HOHFWULF DQG PDJn QHWLF ILHOGV WKXV ZDYH SURSDJDWLRQ LV QRW RFFXUULQJ )LQDOO\ DQ LQWHUPHGLDWHILHOG UHJLRQ H[LVWV ZKHUH UDGLDWLRQ WHUPV GRPLQDWH KRZHYHU WKHLU DQJXODU GLVWULEXWLRQ LV D IXQFWLRQ RI GLVWDQFH 0XOWLSOH GHILQLWLRQV DERXQG IRU WKH ERXQGDULHV EHWZHHQ WKHVH UHJLRQV )RU DQ HOHFn WULFDOO\ VKRUW GLSROH DQWHQQD ZKRVH OHQJWK /f LV OHVV WKDQ $ WKH IDUILHOG FULWHULRQ LV >%DO .LP22D@ DV IROORZV 7KHVH HOHFWULFDOO\ VKRUW DQWHQQDV DUH XVHG LQ ZLUHOHVV LQWHUFRQQHFWV GXH WR UHVWULFWHG GLH DUHD IRU WKH DQWHQQDV ,I WKH } V\PERO LV LQWHUSUHWHG DV D VLQJOH RUGHU RI PDJQLWXGH WKH ERXQGDU\ EHWZHHQ WKH LQWHUPHGLDWH DQG IDUILHOG UHJLRQV EHFRPHV DV IROORZV U!$ f ZKLFK DJUHHV H[DFWO\ ZLWK >%DQ@ 1RWH WKDW WKLV FULWHULRQ FKDQJH DV WKH DQWHQQD EHFRPHV HOHFWULFDOO\ ORQJHU LH / JRHV XS RU IUHTXHQF\ JRHV XSf ZLWK WKH ERXQGDU\ VKLIWLQJ WR / IRU ; / ? DQG /N IRU / $ >%DQ@

PAGE 175

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f DUH YHU\ FRPPRQ GXH WR WKHLU UDGLDWLRQ UHVLVWDQFH EHLQJ a e ZKLFK LV D WUDGLWLRQDO WUDQVPLVVLRQOLQH FKDUDFWHULVWLF LPSHGDQFH 7KH FRUUHVSRQGLQJ IUHTXHQF\ DW ZKLFK WKH DQWHQQD LV D KDOIZDYHOHQJWK ORQJ LV WKH UHVRQDQW IUHTXHQF\ RI WKH DQWHQQD )RU RQFKLS DQWHQQDV WKH DQWHQQD VL]H LV OLPLWHG E\ WKH VL]H RI WKH FKLS 7KHUHIRUH WR PD[LPL]H WKH DQWHQQDfV UDGLDWLRQ ZKLOH OLPLWLQJ WKH SK\VLFDO VL]H RI WKH DQWHQQD UHTXLUHV RSHUDWLQJ DW KLJKHU IUHTXHQFLHV HJ *+]f FRUUHVSRQGLQJ WR VPDOOHU $ )RU UHIHUHQFH SXUSRVHV 7DEOH VKRZV WKH ZDYHOHQJWK YHUVXV IUHTXHQF\ IRU WKH IROORZLQJ PHGLXPV IUHHVSDFH VLOLFRQ GLR[LGH DQG VLOLFRQ VXEVWUDWH 7KH GLSROH DQWHQQD OHQJWK KDV EHHQ OLPLWHG WR PP FRUUHVSRQGLQJ WR $ DQG $ DW DQG *+] UHVSHFWLYHO\ LQ

PAGE 176

7DEOH :DYHOHQJWK LQ PLOOLPHWHUV YHUVXV IUHTXHQF\ IRU GLIIHUHQW PHGLXPV )UHTXHQF\ *+] *+] *+] *+] *+] AfIUHHVSDFH AR[LGH 7O AVLOLFRQ Df Ef )LJXUH 'LH SKRWRJUDSK RI Df OLQHDU DQG Ef ]LJ]DJ GLSROH DQWHQQDV VLOLFRQ GLR[LGH 6LQFH WKH DQWHQQD LV QRW RSHUDWLQJ DW UHVRQDQFH LH WKH OHQJWK LV QRW ;f WKHQ WKH DQWHQQD LPSHGDQFH KDV D UHDFWLYH FRPSRQHQW 7KHUHIRUH WKLV UHDFWDQFH KDV WR EH FRQMXJDWHO\ PDWFKHG ZLWK WKH LPSHGDQFHV RI WKH WUDQVPLWWHU DQG UHFHLYHU %RWK OLQHDU DQG ]LJ]DJ GLSROH DQWHQQDV KDYH EHHQ LPSOHPHQWHG 'LH SKRWRJUDSKV RI H[DPSOH DQWHQQDV ZLWK SDGV DUH VKRZQ LQ )LJXUH >.LP22D@ 7KH D[LDO OHQJWK RI WKH DQWHQQDV LV PP DQG WKH OLQH WUDFH ZLGWK LV _LP )RU WKH ]LJ]DJ DQWHQQD VKRZQ LQ )LJn XUH OEf WKH DUP HOHPHQW OHQJWK LV _LP DQG WKH EHQG DQJOH LV r ZKLOH WKH ]LJ]DJ DQWHQQDV LPSOHPHQWHG ZLWK FORFN UHFHLYHUV DQG WUDQVPLWWHUV KDYH DQ DUP HOHPHQW OHQJWK RI _LP DQG D EHQG DQJOH RI r $OWKRXJK WKH ODWHUDO ILHOG FRPSRQHQWV FDQFHO HDFK RWKHU LQ D ]LJ]DJ VWUXFWXUH WKH ORQJLWXGLQDO FRPSRQHQWV UHLQIRUFH HDFK RWKHU 6LQFH WKH SK\VLFDO OHQJWK RI D ]LJ]DJ DQWHQQD LV ORQJHU WKDQ WKH OLQHDU GLSROH DQWHQQD D ]LJ]DJ GLSROH DQWHQQD

PAGE 177

FDQ KDYH WKH VDPH JDLQ DV D ORQJHU OLQHDU GLSROH DQWHQQD ZKLOH EHLQJ PRUH DUHD HIILFLHQW DW WKH FRVW RI DGGHG FDSDFLWDQFH >.LP22D@ ,Q RWKHU ZRUGV IRU WKH VDPH D[LDO OHQJWK WKH ]LJn ]DJ GLSROH DQWHQQD ZLOO KDYH KLJKHU JDLQ WKDQ D OLQHDU GLSROH DQWHQQD /RRS DQWHQQDV DUH DOVR EHLQJ FRQVLGHUHG IRU ZLUHOHVV LQWHUFRQQHFWV $ ORRS DQWHQQD ZKRVH WRWDO FLUFXPIHUHQFH LQFOXGLQJ PXOWLSOH WXUQVf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a_LP GLDPHWHU PHWDO EDOOV ZLOO LQWHUIHUH ZLWK WKH JOREDO FORFN VLJQDOV )LQDOO\ ZLWKLQ WKH 3& ERDUG WKHUH ZLOO EH JURXQG SODQHV ZKLFK ZLOO DJDLQ UHVXOW LQ UHGXFHG DQWHQQD JDLQ

PAGE 178

S VXEVWUDWH 'LHOHFWULF 0HGLXP 7UDQVPLWWLQJ $QWHQQD 0HWDO +HDWVLQNf 3DWK $ 5HFHLYLQJ $QWHQQD )LJXUH &URVVVHFWLRQDO LOOXVWUDWLRQ RI FKLS IRU PHWDO DQWHQQDV ZLWK LQWHUIHUHQFH VWUXFWXUHV 7KUHH SRVVLEOH VLJQDO SDWKV EHWZHHQ DQWHQQDVf§SDWK $ LV GLUHFW SDWK SDWK % LV VXEVWUDWH VXUIDFH ZDYH DQG SDWK & LV VXEVWUDWH UHIOHFWLRQ %HQHDWK WKH R[LGH LV WKH VXEVWUDWH 7KLV VXEVWUDWH WKHQ LV FRQQHFWHG WR D KHDWVLQN ZKLFK LV XVHG WR UHPRYH WKH KHDW IURP WKH PLFURSURFHVVRU ([SHULPHQWDO UHVXOWV WKRXJK VKRZ WKDW WKLV VLWXDWLRQ LV QRW YHU\ FRQGXFLYH WR RQFKLS VLJQDO WUDQVPLVVLRQ IRU VXEVWUDWH UHVLVWLYLW\ OHVV WKDQ a 4FP ,QVWHDG LI D GLHOHFWULF OD\HU LV SODFHG EHWZHHQ WKH VXEVWUDWH DQG KHDWVLQN WKH DQWHQQD JDLQ FDQ LQFUHDVH E\ DV PXFK DV G% >.LP22D *XR2O@ GHSHQGLQJ RQ WKH GLHOHFWULF PDWHULDO DQG LWV WKLFNQHVV 6LQFH KHDW VWLOO QHHGV WR EH UHPRYHG IURP WKH FKLS WKLV GLHOHFWULF OD\HU KDV WR EH WKHUPDOO\ FRQGXFWLYH $ SRVVLEOH PDWHULDO ZKLFK PHHWV ERWK FULWHULD LV DOXPLQXP QLWULGH $,1f >*XR2O@ $OXPLQXP QLWULGH KDV D WKHUPDO FRQGXFWLYLW\ DSSUR[LPDWHO\ WKH VDPH DV WKDW RI VLOLFRQ DQG DERXW WLPHV WKDW RI JODVV DQG LWV GLHOHFWULF FRQVWDQW LV 'XH WR WKH LQWHUIHUHQFH VWUXFWXUHV WKH GLUHFW SDWK IURP RQH DQWHQQD WR DQRWKHU ODEHOHG SDWK $ LQ )LJXUH ZLOO EH YHU\ DWWHQXDWHG IRU RQFKLS ZLUHOHVV LQWHUFRQQHFWLRQ 7KLV GLUHFW SDWK LV DOVR UHIHUUHG WR DV D VXUIDFH ZDYH 7KH JDLQ RI WKLV SDWK ZLOO GHSHQG RQ

PAGE 179

WKH LQWHUIHUHQFH VWUXFWXUHV EHWZHHQ WKH DQWHQQDV 8QIRUWXQDWHO\ D ZLUHOHVV LQWHUFRQQHFW GHVLJQHU KDV QR FRQWURO RYHU WKHVH OD\HUV DQG WKXV FDQf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n HUDO ZDYHf ODEHOHG SDWK % 7KLV ZDYH LV JHQHUDWHG IURP DQWHQQDV ZKLFK KDYH ILQLWH EHDP ZLGWKV >.LP22D@ LH LW GRHV QRW H[LVW IRU SODQH ZDYHVf $QDO\VHV RI WKLV SDWK DUH FRQn WDLQHG LQ >%RD@ DQG >.LP22D@ ([SHULPHQWV VKRZ WKDW IRU WKH FDVH ZKHQ QR LQWHUIHUHQFH VWUXFWXUHV DUH SUHVHQW WKH JDLQ WKURXJK WKH VXEVWUDWH SDWKV % DQG &f FDQ EH FRPSDUDEOH WR WKH JDLQ WKURXJK WKH GLUHFW SDWK $ 7KH UHODWLYH VWUHQJWK RI SDWKV % DQG & GHSHQGV RQ WKH W\SH DQG VL]H RI WKH XQGHUOD\LQJ GLHOHFWULF PHGLXP >*XR2O@ 0XFK JUHDWHU GHWDLO RQ WKH PRGHOLQJ RI HDFK RI WKHVH SDWKV LV FRQWDLQHG LQ >.LP22D *XR2O@ $QWHQQD &KDUDFWHULVWLFV LQ >LP &026 )LJXUH VKRZV D SKRWRLOOXVWUDWLRQ RI D WHVW FKLS XVHG WR FKDUDFWHUL]H ERWK DQWHQn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

PAGE 180

a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n QDOV IURP WKH QHWZRUN DQDO\]HU WR EDODQFHG VLJQDOV XVHG WR H[FLWH WKH DQWHQQDV 6HPLULJLG FDEOHV DUH XVHG WR LQFUHDVH PHDVXUHPHQW UHOLDELOLW\ $OVR VKRZQ LV D FURVVVHFWLRQ RI WKH RQZDIHU PHDVXUHPHQW VHWXS 7KH GLHV DUH PRXQWHG RQ D JODVV VOLGH ZKLFK LV WKHQ SODFHG RQ YDULRXV LQVXODWRUV ZLWK NfV UDQJLQJ IURP ZRRGf WR $,1f )LJXUH Df VKRZV WKH PHDVXUHG WUDQVGXFHU DQG WUDQVPLVVLRQ JDLQ IURP WR *+] EHWZHHQ WZR RQFKLS OLQHDU GLSROH DQWHQQDV ZKLFK DUH VHSDUDWHG E\ PP 7KH

PAGE 181

3RUW 1HWZRUN $QDO\]HU 3RUW %DLXQ ? %DODQFHG 3UREH ‘I A $QWHQQD G R/L Yf§ 6HPL5LJLG &DEOHV %DODQFHG f§0f§ I 'LHOHFWULF %DLXQ ? $QWHQQD r :DIHU *ODVV PP )LJXUH 0HDVXUHPHQW VHWXS XVHG WR FKDUDFWHUL]H RQFKLS DQWHQQD SDLU 2QZDIHU PHDVXUHPHQWV DUH SHUIRUPHG ZLWK WKH GLH PRXQWHG RQ D JODVV VOLGH DERYH D GLHOHFWULF LQVXODWRU )LJXUH Df 7UDQVGXFHU DQG WUDQVPLVVLRQ JDLQ RI OLQHDU GLSROH DQWHQQD SDLU Ef 3KDVH YHUVXV IUHTXHQF\ IRU OLQHDU GLSROH DQWHQQD SDLU DQWHQQDV DUH LPSOHPHQWHG LQ PHWDO DQG VHSDUDWHG IURP WKH VXEVWUDWH E\ -LP RI R[LGH 7KHVH UHVXOWV DUH IRU D VXEVWUDWH UHVLVWLYLW\ RI a 4FP $V FDQ EH VHHQ WKH DQWHQQD LV PRUH HIILFLHQW DW KLJKHU IUHTXHQFLHV RU ZKHQ VLJQDO ZDYHOHQJWKV DUH VPDOOHU $ WUDQVGXFHU fJDLQf RI G% LV DWWDLQHG DW *+] ZKLOH WKH WUDQVPLVVLRQ JDLQ LV G% 7KH ODUJH GLIIHUHQFH EHWZHHQ WKH WUDQVGXFHU DQG WUDQVPLVVLRQ JDLQ LV GXH WR WKH DQWHQQD LPSHGDQFH EHLQJ VLJQLILFDQWO\ GLIIHUHQW WKDQ WKH 4 GLIIHUHQWLDO LPSHGDQFH LH GLIIHUHQWLDO YROWn DJH RYHU GLIIHUHQWLDO FXUUHQWf 6LQFH WKH _LP /1$ SUHVHQWHG LQ VHFWLRQ LV

PAGE 182

)LJXUH $QWHQQD LPSHGDQFH YHUVXV IUHTXHQF\ IRU OLQHDU GLSROH DQWHQQD PDWFKHG GLIIHUHQWLDOO\ WR e WKH PLVPDWFK IRU WKH DQWHQQDV LQ WKH PHDVXUHPHQW V\VWHP LV YLUWXDOO\ WKH VDPH DV ZKHQ WKH DQWHQQD LV LQWHJUDWHG ZLWK WKH /1$ 7KH PHDVXUHG DQWHQQD LPSHGDQFH VKRZQ LQ )LJXUH DW *+] LV DSSUR[LPDWHO\ M e &RPn SDULQJ WKLV LPSHGDQFH WR 4 WKH WRWDO PLVPDWFK ORVV IRU WKH DQWHQQD SDLU FDQ EH FDOFXn ODWHG WR EH G% XVLQJ HTXDWLRQV f DQG f 7KH UHDVRQ IRU VXFK D ORZ DQWHQQD LPSHGDQFH LV VWLOO QRW NQRZQ IRU FHUWDLQ 3UHYLn RXV DQWHQQD LPSOHPHQWDWLRQV LQGLFDWH WKDW WKLV VDPH DQWHQQD VKRXOG KDYH DQ LPSHGDQFH FORVHU WR IW >.LP22D@ $ SRWHQWLDO UHDVRQ IRU WKLV ORZ LPSHGDQFH LV WKH HIIHFW RI LQWHUn IHUHQFH VWUXFWXUHV LQFOXGLQJ PHWDO GXPP\ SDWWHUQV IRU FKHPLFDOPHFKDQLFDO SROLVKLQJ RQ DQWHQQD LPSHGDQFH :KDW LV NQRZQ IRU FHUWDLQ LV WKDW D G% PLVPDWFK ORVV LV XQDFFHSWn DEOH &OHDUO\ EHWWHU WHFKQLTXHV DUH UHTXLUHG WR SUHGLFW WKH DQWHQQD LPSHGDQFH VR WKDW SURSHU PDWFKLQJ FLUFXLWU\ FDQ EH GHVLJQHG $OWHUQDWLYHO\ DGMXVWDEOH PDWFKLQJ QHWZRUNV FDQ EH GHVLJQHG WR DOORZ IRU WKH UHFHLYHU LQSXW PDWFK WR EH WXQHG WR DFFRXQW IRU DQWHQQD LPSHGDQFH YDULDWLRQV

PAGE 183

)LJXUH Ef VKRZV WKH PHDVXUHG SKDVH YHUVXV IUHTXHQF\ IRU WKH YROWDJH UDWLR EHWZHHQ D UHFHLYLQJ DQG WUDQVPLWWLQJ DQWHQQD SDLU :LWK 4 WUDQVPLVVLRQ OLQHV DW ERWK SRUWV RI WKH QHWZRUN DQDO\]HU WKLV SKDVH LV GHILQHG DV IROORZV ZKHUH Dc DQG Ec DUH WKH QRUPDOL]HG LQFLGHQW DQG UHIOHFWHG YROWDJH ZDYHV DW HDFK SRUW 7KH SKDVH GHOD\ GHFUHDVHV OLQHDUO\ ZLWK IUHTXHQF\ f FROFHIIf LQGLFDWLQJ ZDYH SURSDJDWLRQ UDWKHU WKDQ VRPH W\SH RI OXPSHG HOHPHQW 5& FRXSOLQJ >.LP22D@ )LQDOO\ WKH WUDQVGXFHU JDLQ EHWZHHQ WZR VHWV RI SDGV ZDV PHDVXUHG WR EH DSSUR[LPDWHO\ G% OHVV WKDQ WKH PHDn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f VKRZV WKH PHDVXUHG WUDQVGXFHU JDLQ EHWZHHQ WKH DQWHQQD SDLU DQG WKH DQWHQQD ZLWK /1$ IRU D VHSDUDWLRQ GLVWDQFH RI a PP $V FDQ EH VHHQ WKH /1$ LV SURYLGLQJ JDLQ IRU IUHTXHQFLHV EHORZ a *+] 7KH WUDQVGXFHU JDLQ RI WKH DQWHQQD SDLU FDQ WKHQ EH GHHPEHGGHG IURP WKH DQWHQQD ZLWK /1$ PHDVXUHn PHQW WR H[WUDFW WKH /1$ JDLQ 7KLV UHVXOW LV VKRZQ LQ )LJXUH Ef DORQJ ZLWK WKH

PAGE 184

Df Ef )LJXUH Df 0HDVXUHG WUDQVGXFHU JDLQ EHWZHHQ DQWHQQD SDLU DQG DQWHQQD SDLU ZLWK /1$ Ef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fXQXVHGf SRUWLRQ RQ HLWKHU VLGH RI WKH UHFHLYHU LV [ PP ZKLOH WKH VL]H RI WKH UHFHLYHU DORQH H[FOXGLQJ SDG DUHDf LV [ PP 6LQFH WKH FORFN UHFHLYHU LV IXOO\ GLIIHUHQWLDO WKH UHFHLYHU OD\RXW VKRXOG EH V\PPHWULF IRU HDFK KDOIFLUFXLW RI WKH UHFHLYHU LQFOXGLQJ DOO FRPSRQHQWV VXSSO\ YROWDJH OLQHV DQG SDGV 7KLV DYRLGV DQ\ V\VWHPDWLF HUURUV EHLQJ LQWURGXFHG IURP OD\RXW

PAGE 185

'LSROH $QWHQQD /RZ 1RLVH $PSOLILHU ZLWK VRXUFHIROORZHUVf )LJXUH )UHTXHQF\ 'LYLGHU DQG 2XWSXW %XIIHUV 'LH SKRWRJUDSK RI _[P FORFN UHFHLYHU ZLWK OLQHDU GLSROH DQWHQQD PLVPDWFK $GGLWLRQDOO\ PXOWLSOH VXEVWUDWH FRQQHFWLRQV VKRXOG EH LQFOXGHG IRU ERWK WKH /1$ DQG GLYLGHU ,Q SDUWLFXODU WR PLQLPL]H DQ\ VXEVWUDWH FURVVWDON D JXDUGULQJ VWUXFWXUH VKRXOG VHSDUDWH WKH /1$ IURP WKH GLYLGHU )LQDOO\ WKH /1$ DQG EXIIHU VKRXOG KDYH D VHSn DUDWH VXSSO\ YROWDJH IURP WKH IUHTXHQF\ GLYLGHU )LJXUH Df VKRZV WKH PHDVXUHPHQW VHWXS XVHG WR GHPRQVWUDWH D VLQJOHUHFHLYHU ZLUHOHVV LQWHUFRQQHFW $ *+] JOREDO FORFN VLJQDO LV H[WHUQDOO\ JHQHUDWHG DQG WKHQ DPSOLILHG WR G%P WR RYHUFRPH WKH ORZ DQWHQQD WUDQVGXFHU JDLQ 7KLV VLJQDO LV FRQn YHUWHG WR D EDODQFHG VLJQDO DQG LQMHFWHG WR WKH RQFKLS WUDQVPLWWLQJ DQWHQQD 7KH JOREDO FORFN VLJQDO SURSDJDWHV PP DFURVV WKH FKLS WR WKH UHFHLYHU ZKRVH RXWSXW LV PHDVXUHG XVLQJ DQ RVFLOORVFRSH )LJXUH Ef VKRZV SORWV RI WKH LQSXW YROWDJH EHIRUH WKH H[WHUQDO DPSOLILHUf WR WKH WUDQVPLWWLQJ DQWHQQD DQG WKH RXWSXW YROWDJH IRU WKH ZLUHOHVV FORFN UHFHLYHU $V FDQ EH VHHQ D 0+] *+] GLYLGHG E\ HLJKWf ORFDO FORFN VLJQDO LV JHQn HUDWHG GHPRQVWUDWLQJ WKH RSHUDWLRQ RI WKH VLQJOHUHFHLYHU RQFKLS ZLUHOHVV LQWHUFRQQHFW IRU FORFN GLVWULEXWLRQ 7KH UHGXFHG YROWDJH VZLQJ DW WKH RXWSXW LV IURP GULYLQJ D  ORDG $V LQIHUUHG IURP WKH PHDVXUHG DQWHQQDSDLU WUDQVGXFHU JDLQ RI G% WKH UHFHLYHG VLJQDO OHYHO LV G%P DW WKH /1$ LQSXW 6LQFH WKH UHFHLYHU ZRXOG QRW ORFN WR WKH

PAGE 186

Df Ef WLPH QVf )LJXUH Df 0HDVXUHPHQW VHWXS XVHG WR FKDUDFWHUL]H FORFN UHFHLYHU Ef ,QSXW DQG RXWSXW ZDYHIRUPV IRU WKH FORFN UHFHLYHU ILQ *+] DQG IRXW *+] 0+]f IRU D LQWHUFRQQHFWLRQ GLVWDQFH RI PP WUDQVPLWWHG VLJQDO IRU ORZHU LQSXW SRZHU OHYHOV WKH PLQLPXP GHWHFWDEOH VLJQDO RI WKH UHFHLYHU LV G%P 7KH UHFHLYHU FRQVXPHV P: IURP D 9 VXSSO\ 7KLV UHVXOW LV WKH ILUVW GHPRQVWUDWLRQ RI DQ RQFKLS ZLUHOHVV LQWHUFRQQHFW $ *+] JOREDO FORFN VLJQDO LV VXFFHVVIXOO\ UHFHLYHG RYHU D PP GLVWDQFH ZLWK LQWHUIHUn HQFH VWUXFWXUHV ORFDWHG LQ EHWZHHQ WKH DQWHQQDV DQG D 0+] ORFDO FORFN VLJQDO LV JHQn HUDWHG $OVR WKLV UHVXOW LV WKH ILUVW WR LQWHJUDWH DQWHQQDV DQG &026 FLUFXLWU\ RQ D VLQJOH FKLS 7KHUHIRUH SODXVLELOLW\ RI D ZLUHOHVV V\VWHP KDV EHHQ GHPRQVWUDWHG XVLQJ D VWDQGDUG _LP &026 WHFKQRORJ\ DW a *+] %DVHG RQ WKLV UHVXOW VXSHULRU SHUIRUPDQFH LV H[SHFWHG E\ XVLQJ D PRUH DGYDQFHG &026 WHFKQRORJ\ 7KLV DOORZV WKH RSHUDWLQJ IUHn TXHQF\ WR LQFUHDVH LPSURYLQJ WKH HIILFLHQF\ RI WKH DQWHQQDV ZKLFK LQ WXUQ DOORZV ORQJHU LQWHUFRQQHFWLRQ GLVWDQFHV

PAGE 187

PP )LJXUH /D\RXW RI 80& WHVWFKLS VKRZLQJ ORFDWLRQV RI UHOHYDQW ]LJ]DJ DQWHQQDV =[f ORRS DQWHQQD /3Mf FORFN WUDQVPLWWHUV 7[f DQG FORFN UHFHLYHUV 5[f $QWHQQD &KDUDFWHULVWLFV LQ IOP &026 &KLS ,PSOHPHQWDWLRQ %DVHG RQ WKH UHVXOWV DFKLHYHG LQ WKH _LP &026 WHVW FKLS D WHVW FKLS ZDV LPSOHPHQWHG LQ D SP &026 WHFKQRORJ\ ZLWK FRSSHU LQWHUFRQQHFWV 7KLV WHFKQRORJ\ ZDV REWDLQHG IURP 80& DV SDUW RI WKH 65& &RSSHU 'HVLJQ &KDOOHQJH VSRQVRUHG E\ WKH 6HPLFRQGXFWRU 5HVHDUFK &RUSRUDWLRQ 65&f 1RYHOOXV 6SHHG)DP,3(& DQG 80& 7KH VXEVWUDWH UHVLVWLYLW\ LV 4FP 6L[ OD\HUV RI FRSSHU LQWHUFRQQHFW ZHUH SURYLGHG 0XOWLSOH DQWHQQD WHVW VWUXFWXUHV /1$V IUHTXHQF\ GLYLGHUV FORFN UHFHLYHUV DQG FORFN WUDQVPLWWHUV KDYH EHHQ LQFOXGHG LQ WKLV WHVW FKLS )LJXUH VKRZV WKH OD\RXW RI WKH

PAGE 188

R 80& WHVWFKLS ZKRVH DUHD LV [ PP 7KH ORFDWLRQV RI UHOHYDQW ]LJ]DJ DQWHQQDV =[f DQG WKH ORRS DQWHQQD /3Mf KDYH EHHQ QRWHG &ORFN WUDQVPLWWHUV 7[O 7[f DQG UHFHLYHUV 5[O5[f KDYH EHHQ ODEHOHG DV ZHOO /RFDWHG LQ EHWZHHQ WKH DQWHQQDV DUH WHVW VWUXFWXUHV ZKLFK LQWHUIHUH ZLWK WKH FORFN WUDQVPLVVLRQ DQG UHFHSWLRQ 7KHVH WHVW VWUXFWXUHV FRQWDLQ PXOWLSOH PHWDO LQWHUFRQQHFWV YLDV VXEVWUDWH FRQQHFWLRQV SDVVLYDWLRQ RSHQLQJV DQG PHWDOILOO SDWWHUQV QRW VKRZQf IRU PHWDO OD\HUV WKURXJK 7KHUHIRUH WKH GHQVLW\ RI VWUXFWXUHV EHWZHHQ WKH DQWHQQDV LV KLJK $QWHQQD 'HVFULSWLRQV 7KH PDMRULW\ RI WKH DQWHQQDV LPSOHPHQWHG DUH PP ORQJ ]LJ]DJ GLSROH DQWHQQDV ODEHOHG =[ LQ )LJXUH 7KH ]LJ]DJ DQWHQQDV LOOXVWUDWHG LQ )LJXUH KDYH D _LP WUDFHZLGWK DQ _LP DUP HOHPHQW OHQJWK DQG D r EHQG DQJOH 7KHVH YDOXHV ZHUH EDVHG RQ WKH EHVW UHVXOWV FXUUHQWO\ DYDLODEOH IURP DQWHQQD GHVLJQ H[SHULPHQWV >.LP22E@ $QWHQn QDV =D WR =I DUH LPSOHPHQWHG LQ PHWDO DW YDULRXV ORFDWLRQV WKURXJKRXW WKH FKLS ZLWK VSDFLQJV RI G ^ ` PP IRU DQWHQQD SDLUV =^D E EF UHVSHFWLYHO\ 7KH GLVWDQFH IURP PHWDO WR WKH VXEVWUDWH LV a _LP $OVR GLIIHUHQW PHWDO OD\HUV ZHUH XVHG WR IDEULFDWH WKH DQWHQQDV WR H[DPLQH WKHLU GHSHQGHQFH RQ DQWHQQD JDLQ 7KHVH DQWHQQDV DUH ODEHOHG =c DQG = DQG DUH VHSDUDWHG E\ PP 7ZR DQWHQQDV DUH HYDOXDWHG IRU XVH LQ WKH WUDQVPLWWHUf§D ]LJ]DJ GLSROH DQWHQQD DQG D ORRS DQWHQQD /3Mf ZKLFK KDV DQ RPQLGLUHFWLRQDO LQSODQH UDGLDWLRQ SDWWHUQ >%DO@ 7KH

PAGE 189

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n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k m R Z R R F 2f F )LJXUH &URVVVHFWLRQDO LOOXVWUDWLRQ RI DQWHQQDV LQ GLUHFW FRQWDFW ZLWK VXEVWUDWH

PAGE 190

)LJXUH Df =LJ]DJ]LJ]DJ WUDQVPLVVLRQ JDLQV IRU G PP =Af DQG PP =FBIf $OVR VKRZQ LV WKH SDG WUDQVGXFHU JDLQ ,6LOfDW G PP IRU UHIHUHQFH Ef 3KDVH YHUVXV IUHTXHQF\ RI YROWDJH ZDYH IRU G DQG PP 0HDVXUHG $QWHQQD &KDUDFWHULVWLFV 7KH EDODQFHG PHDVXUHPHQW VHWXS VKRZQ LQ )LJXUH ZDV XVHG WR FKDUDFWHUL]H WKH DQWHQQD SDLUV )LJXUH Df VKRZV WKH PHDVXUHG ]LJ]DJ]LJ]DJ WUDQVPLVVLRQ JDLQ *Df YHUVXV IUHTXHQF\ IRU DQG PP VHSDUDWLRQV FRUUHVSRQGLQJ WR DQWHQQD SDLUV =E DQG =EI! UHVSHFWLYHO\ 7KH JDLQ LQFUHDVHV ZLWK IUHTXHQF\ DQG GHFUHDVLQJ VHSDUDWLRQ $W *+] *D LV DQG G% IRU DQG PP VHSDUDWLRQV UHVSHFWLYHO\ $OVR VKRZQ LV WKH JDLQ ,6LOf EHWZHHQ WZR VHWV RI SDGV VHSDUDWHG E\ PP ZKLFK LV DERXW G% EHORZ WKH WUDQVPLVVLRQ JDLQ DW *+] )RU WKLV PHDVXUHPHQW VLWXDWLRQ WKH SDGWRSDG JDLQ LV FORVH WR WKH LQVWUXPHQW QRLVH IORRU 7KH SKDVH GHOD\V EHWZHHQ WKH YROWDJHV DW WKH UHFHLYLQJ DQG WUDQVPLWWLQJ DQWHQQDV DUH VKRZQ LQ )LJXUH Ef 7KH SKDVH GHOD\ GHFUHDVHV OLQHDUO\ ZLWK IUHTXHQF\ DJDLQ LQGLFDWLQJ ZDYH SURSDJDWLRQ UDWKHU WKDQ VRPH W\SH RI OXPSHGHOHPHQW 5& FRXSOLQJ 7KLV IDFW DORQJ ZLWK WKH ORZ SDGWRSDG JDLQ VKRZ

PAGE 191

)LJXUH Df /RRS]LJ]DJ WUDQVPLVVLRQ JDLQV IRU G /3S=If DQG PP /3M=Df $OVR VKRZQ LV WKH SDG WUDQVGXFHU JDLQ ,6LOf DW G PP IRU UHIHUHQFH Ef =LJ]DJ WUDQVPLVVLRQ JDLQV IRU PHWDO DQWHQQDV =JDEFf DW G PP GHPRQVWUDWLQJ WKH HIIHFW RI LQWHUIHUHQFH VWUXFWXUHV RQ JDLQ FRQFOXVLYHO\ WKDW WKH VLJQDO LV SURSDJDWLQJ IURP RQH DQWHQQD WR WKH RWKHU DQG WKDW WKHVH ZDYHV DUH ODXQFKHG PXFK PRUH HIILFLHQWO\ IURP WKH DQWHQQDV WKDQ IURP WKH SDGV DORQH )LJXUH Df VKRZV *D YHUVXV IUHTXHQF\ IRU D ORRS]LJ]DJ DQWHQQD SDLU DW DQG PP VHSDUDWLRQV FRUUHVSRQGLQJ WR DQWHQQD SDLUV /3U=I DQG /3U=D UHVSHFn WLYHO\f $W *+] *D LV DQG G% UHVSHFWLYHO\ ZKLFK LV OHVV WKDQ WKH ]LJ]DJ]LJ]DJ JDLQV 7KHVH UHVXOWV DUH SURPLVLQJ WKRXJK DQG VKRZ WKDW DW KLJK HQRXJK IUHTXHQFLHV ORRS DQWHQQDV DUH XVHIXO IRU WKH WUDQVPLWWLQJ DQWHQQDV )LJXUH Ef VKRZV *D YHUVXV IUHTXHQF\ IRU WKUHH SDLUV RI ]LJ]DJ DQWHQQDV (DFK SDLU LV VHSDUDWHG E\ PP FRUUHVSRQGLQJ WR DQWHQQDV =D EE EF 7KHVH UHVXOWV GHPRQn VWUDWH WKH HIIHFW RI LQWHUIHUHQFH VWUXFWXUHV RQ DQWHQQD JDLQ $V FDQ EH VHHQ IURP )LJXUH HDFK SDLU KDV GLIIHUHQW W\SHV DQG GHQVLWLHV RI PHWDO DQG DFWLYH VWUXFWXUHV ORFDWHG LQ

PAGE 192

EHWZHHQ 7KH WUDQVPLVVLRQ JDLQ FKDQJHV E\ WR G% GHSHQGLQJ RQ WKH DQWHQQD DQG IUHn TXHQF\ $OVR WKH DQWHQQDV KDYH VOLJKWO\ GLIIHUHQW LPSHGDQFHV QRW VKRZQf VLQFH WKH UHIOHFWLRQ FRHIILFLHQW LV D IXQFWLRQ RI WKH PHWDO VWUXFWXUHV UHIOHFWLQJ WKH WUDQVPLWWHG VLJQDO EDFN LQWR WKH DQWHQQD 7KLV VKRZV WKDW WKH VWUXFWXUHV ORFDWHG DURXQG WKH DQWHQQD DIIHFW WKH DQWHQQD SHUIRUPDQFH 7KH LPSOLFDWLRQ RQ V\VWHP GHVLJQ LV WKDW WKLV GHJUDGDWLRQ LQ VLJQDO OHYHO KDV WR EH DFFRXQWHG IRU ZKHQ HVWLPDWLQJ WKH ZRUVWFDVH VLJQDOWRQRLVH UDWLR $OVR WKLV VKRZV WKH QHHG IRU HOHFWURPDJQHWLF VLPXODWLRQ WRROV ZKLFK FDQ EH XVHG WR WU\ WR HVWLn PDWH WKH HIIHFWV RI WKHVH VWUXFWXUHV EHIRUHKDQG DOORZLQJ GHVLJQ UXOHV WR EH GHYHORSHG 0XFK JUHDWHU DQDO\VLV RI WKLV SUREOHP LV FRQWDLQHG LQ >
PAGE 193

)LJXUH Df ,PSHGDQFH VPLWK FKDUW VKRZLQJ UHIOHFWLRQ FRHIILFLHQWV RI ]LJ]DJ DQG ORRS DQWHQQDV ,PSHGDQFHV DUH QRUPDOL]HG WR 4f Ef $QWHQQD LPSHGDQFH IRU ]LJ]DJ DQG ORRS DQWHQQDV =E DQG /3Mf f f§f§f§f§f§!f§f§f§f§,f§f§f§ PHWDO G PP ‘ WQ R SQZHOO G PP f R &2 ‘ DS f G PP L L L L L L L L L L L )UHTXHQF\ *+]f )UHTXHQF\ *+]f Df Ef )LJXUH Df =LJ]DJ WUDQVPLVVLRQ JDLQV DW G PP IRU DQWHQQDV LQ PHWDOV RU Ef =LJ]DJ WUDQVPLVVLRQ JDLQV DW G PP =SS =QSf IRU DQWHQQDV LPSOHPHQWHG LQ GLUHFW FRQWDFW ZLWK VXEVWUDWH ZLWK QSOXV LQ SVXE DQG SSOXV LQ QZHOO 0HWDO DQWHQQD DW G PP LV VKRZQ IRU FRPSDULVRQ SXUSRVHV

PAGE 194

7DEOH &RPSDULVRQ EHWZHHQ PHDVXUHG DQG H[SHFWHG DQWHQQD FKDUDFWHULVWLFV SP $QWHQQD 3DLU 3DUDPHWHU DW *+] ([SHFWHG 0HDVXUHG =LJ]DJ WR =LJ]DJ *DLQ G PPf G% G% ,PSHGDQFH =LJ]DJf M 4 M 4 /RRS WR =LJ]DJ *DLQ G PPf G% G% ,PSHGDQFH /RRSf M4 MOO e )LJXUH Ef VKRZV WKH PHDVXUHG JDLQ YHUVXV IUHTXHQF\ IRU WKH VXEVWUDWH DQWHQQDV 7KHVH VXEVWUDWH DQWHQQDV DUH LPSOHPHQWHG LQ Q LQ SVXEVWUDWH DQG S LQ QZHOO RYHUODLG ZLWK PHWDO DQG PXOWLSOH FRQWDFWV $W *+] IRU D PP VHSDUDWLRQ WKH JDLQV IRU D QSOXV DQWHQQD LPSOHPHQWHG LQ SVXEVWUDWH DQG IRU D SSOXV DQWHQQD LPSOHn PHQWHG LQ QZHOO DUH DQG G% UHVSHFWLYHO\ 7KH JDLQV DUH EHWZHHQ WR G% OHVV WKDQ WKDW IRU WKH PHWDO ]LJ]DJ G PPf IRU IUHTXHQFLHV EHWZHHQ WR *+] UHVSHFn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f XQLIRUPLW\ 2QH GLIIHUHQFH EHWZHHQ WKH FKLSV LV WKDW WKH _LP FKLS DOORZHG IRU

PAGE 195

7UDQVPLWWLQJ $QWHQQD *+]f )LJXUH %ORFN GLDJUDP RI FORFN WUDQVPLWWHU IRU SP WHVW FKLS WKHVH ILOO SDWWHUQV WR EH EORFNHG DURXQG WKH DQWHQQD ZKLOH QR EORFNLQJ ZDV XVHG LQ WKH _LP FKLS &XUUHQWO\ LW LV QRW NQRZQ ZKHWKHU WKLV GLIIHUHQFH FDQ DFFRXQW IRU WKH GLIIHUn HQFH LQ WKH LPSHGDQFH 0RUH H[SHULPHQWDO DQG PRGHOLQJ ZRUN LV UHTXLUHG WR XQGHUVWDQG WKHVH GLVFUHSDQFLHV VLQFH SUHGLFWLQJ WKH DQWHQQD SHUIRUPDQFH LV FULWLFDO WR D ZRUNLQJ ZLUHn OHVV LQWHUFRQQHFW V\VWHP :LUHOHVV ,QWHUFRQQHFWV LQ ILP &026 &ORFN 7UDQVPLWWHU )LJXUH VKRZV D VLPSOLILHG EORFN GLDJUDP IRU WKH FORFN WUDQVPLWWHU 7KH *+] VLJQDO LV JHQHUDWHG XVLQJ D GLIIHUHQWLDO YROWDJHFRQWUROOHG RVFLOODWRU 9&2f 7KH 9&2 LV UHTXLUHG WR KDYH ORZ SKDVH QRLVH WR GHFUHDVH WKH ORFDO FORFN MLWWHU 7KH VLJQDO IURP WKH 9&2 LV WKHQ DPSOLILHG E\ D WZRVWDJH RXWSXW SRZHU DPSOLILHU DQG GHOLYHUHG WR WKH WUDQVPLWWLQJ DQWHQQD ,Q WKH ILQDO FORFN GLVWULEXWLRQ V\VWHP LPSOHPHQWDWLRQ WKH 9&2 ZLOO EH SKDVHORFNHG WR DQ H[WHUQDO UHIHUHQFH +RZHYHU WR HDVH WKH LPSOHPHQWDWLRQ UHTXLUHn PHQWV IRU WKLV FKLS WKH WUDQVPLWWHU ZDV RSHUDWHG RSHQORRS ZKHUH WKH IUHTXHQF\ RI WKH 9&2 ZDV FRQWUROOHG GLUHFWO\ ZLWK LWV GF LQSXW 7KLV FORFN WUDQVPLWWHU ZDV SULPDULO\ GHVLJQHG E\ &KLK0LQJ +XQJ DQG FKDUDFWHUL]HG E\ WKH DXWKRU $FFRUGLQJO\ WKH EDVLF FLUn FXLW GHVLJQ ZLOO RQO\ EH EULHIO\ UHYLHZHG VSHFLILFDOO\ IRU WKH 9&2 ZKHUH WKH LQWHUHVWHG

PAGE 196

)LJXUH 'LH SKRWRJUDSK RI JP FORFN WUDQVPLWWHU ZLWK ]LJ]DJ GLSROH DQWHQQD UHDGHU FDQ UHIHU WR >+XQ22D@ ,QVWHDG WKH GLVFXVVLRQ ZLOO IRFXV RQ WKH PHDVXUHG FKDUDFWHUn LVWLFV DQG WKHLU LPSOLFDWLRQ IRU ZLUHOHVV LQWHUFRQQHFW GHVLJQ )LJXUH VKRZV D GLH SKRWRJUDSK RI WKH FORFN WUDQVPLWWHU ZLWK D ]LJ]DJ DQWHQQD 7KH VL]H RI WKH WUDQVPLWWHU LQFOXGLQJ WKH fXQXVHGf SRUWLRQV RQ HLWKHU VLGH RI WKH FLUFXLW LV [ PP ZKLOH WKH DFWLYH DUHDV H[FOXGLQJ SDGVf LV [ PP 7KH OD\RXW LV V\Pn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

PAGE 197

ORZHU PRELOLW\ 7KLV LPSOLHV WKDW WKH WRWDO JDWH DUHD LV ODUJHU LI 3026 LV XVHG IRU WKH RVFLOODWRU 6LQFH I QRLVH LV DSSUR[LPDWHO\ LQYHUVHO\ SURSRUWLRQDO WR WKH JDWH DUHD WKH FORVHLQ SKDVH QRLVH FDQ EH UHGXFHG )XUWKHUPRUH 3026 WUDQVLVWRUV KDYH PXFK ORZHU KRWFDUULHU QRLVH DFFRUGLQJ WR >&KH@ WKXV WKH IDURXW SKDVH QRLVH ZRXOG DOVR EH LPSURYHG +HQFH 3026 WUDQVLVWRUV DUH H[FOXVLYHO\ XVHG LQ WKLV 9&2 GHVLJQ >+XQ22D@ 7KH LQGXFWDQFH RI ERWK /M DQG / ZDV GHVLJQHG WR EH Q+ ZKLOH PHDVXUHn PHQWV UHYHDO DQ LQGXFWDQFH RI Q+ 7KLV LQFUHDVH LQ LQGXFWDQFH DV ZHOO DV LQFUHDVHG FDSDFLWDQFH IURP WKH WUDQVLVWRUV DV SUHVHQWHG LQ VHFWLRQ GHFUHDVHG WKH 9&2 RSHUDWn LQJ IUHTXHQF\ IURP a *+] WR *+] 7KH LQGXFWRUV ZHUH LPSOHPHQWHG ZLWK PHWDO DQG OD\HUV VKXQWHG WRJHWKHU DERYH D SDWWHUQHG JURXQG VKLHOG $V PHQWLRQHG LQ &KDSWHU WKLV LPSURYHV WKH LQGXFWRU 4 E\ SURYLGLQJ D ORZ UHVLVWDQFH SDWK WR JURXQG IRU FDSDFLWLYHO\ U? FRXSOHG FXUUHQW 7KH DUHD RI WKH LQGXFWRU LV [ SP DQG WKH PHWDO VSDFLQJ ZLGWK DQG QXPEHU RI WXUQV DUH _LP SP DQG UHVSHFWLYHO\ 7KH H[SHFWHG LQGXFWRU 4

PAGE 198

ZDV ZKLOH WKH PHDVXUHG 4 ZDV a 7KLV UHGXFWLRQ LQ 4 ZDV GXH WR WKH WKLQQLQJ RI WKH WRSOD\HU PHWDOV RQFH DJDLQ GLVFXVVHG LQ VHFWLRQ 7KH YDUDFWRU LV LPSOHPHQWHG XVLQJ DQ DFFXPXODWLRQPRGH KLJK4 026 FDSDFLWRU >+XQ@ 0HDVXUHPHQWV VKRZ WKDW WKH YDUDFWRU 4 LV DSSUR[LPDWHO\ DW *+] $ 9&2 WHVW VWUXFWXUH FRQVLVWLQJ RI WKH 9&2 FRUH SOXV D VLQJOHVWDJH RXWSXW EXIIHU ZDV FRQQHFWHG WR D VSHFWUXP DQDO\]HU \LHOGLQJ WKH RXWSXW VLQJOHVLGHEDQG SKDVHQRLVH SORW DQG VSHFWUXP VKRZQ LQ )LJXUH :LWK 9FWL 9 WKH PHDVXUHG FHQn WHU IUHTXHQF\ LV *+] ZLWK DQ RXWSXW SRZHU OHYHO RI G%P 7KH 9&2 FRUH FRQn VXPHV P: IURP D 9 VXSSO\ 7KH SKDVH QRLVHV DW 0+] DQG 0+] RIIVHWV DUH DQG G%F+] UHVSHFWLYHO\ )LQDOO\ DQ RVFLOODWLRQ IUHTXHQF\ YHUVXV FRQWURO YROWn DJH SORW LV VKRZQ LQ )LJXUH 7KH WXQLQJ UDQJH LV 0+] IRU 9FWO EHWZHHQ DQG 9 7KH OLQHDU SRUWLRQ RI WKLV FXUYH UHYHDOV D 9&2 JDLQ RI DSSUR[LPDWHO\ 0+]9

PAGE 199

7DEOH VKRZV D FRPSDULVRQ EHWZHHQ WKH H[SHFWHG DQG PHDVXUHG 9&2 UHVXOWV 7KH RYHUDOO 4 RI WKH /& WDQN LV WKH PDMRU OLPLWLQJ IDFWRU RI WKH SKDVH QRLVH SHUIRUPDQFH ZKHUH WKH WDQN FRQVLVWV RI /M / &YO &Y WKH JDWH DQG GUDLQ FDSDFLWDQFHV RI 0M DQG 0 DQG WKH JDWH FDSDFLWDQFHV RI 0 DQG 0 6LQFH WKH PHDVXUHG 4 RI WKH LQGXFWRU LV a WLPHV VPDOOHU WKDQ WKDW RI WKH YDUDFWRU WKH 4 RI WKH RYHUDOO /& WDQN LV OLPLWHG E\ WKH LQGXFWRU 7KH SKDVH QRLVH RI D 9&2 LQ I UHJLRQ UHIHU WR )LJXUH f FDQ EH DSSUR[LPDWHG E\ /HHVRQfV IRUPXOD >/HH@ DV IROORZV /$FRf f ORJ )N74OUVf >.Vt-@ f ZKHUH ) LV WKH QRLVH IDFWRU RI WKH DPSOLILHU N LV %ROW]PDQQfV FRQVWDQW 7 LV WKH WHPSHUDn WXUH 9 LV WKH YROWDJH DFURVV WKH WDQN 4/ LV WKH 4 RI / UV LV WKH VHULHV UHVLVWDQFH RI WKH LQGXFWRU DQG $FR LV WKH IUHTXHQF\ RIIVHW IURP WKH FDUULHU $VVXPLQJ WKDW WKH ODVW WHUP LQ WKH HTXDWLRQ LV PXFK JUHDWHU WKDQ DQG DOO RWKHU SDUDPHWHUV H[FHSW UV LQ WKH HTXDWLRQ DUH FRQn VWDQW D b 4 GHJUDGDWLRQ LQFUHDVHV WKH SKDVH QRLVH E\ a G% ORJff ZKLFK LV

PAGE 200

VLJQLILFDQW 7KXV GRXEOLQJ WKH WKLFNQHVVHV RI PHWDOV DQG VKRXOG LPSURYH WKH SKDVH QRLVH WR DQG G%F+] DW 0+] DQG 0+] RIIVHWV UHVSHFWLYHO\ 7KH SRZHU FRQn VXPSWLRQ VKRXOG DOVR GHFUHDVH VLQFH WKH YROWDJH DFURVV WKH WDQN ZRXOG UHPDLQ DSSUR[Ln PDWHO\ FRQVWDQW ZKLOH WKH HIIHFWLYH UHVLVWDQFH RI WKH WDQN 4UVf ZRXOG LQFUHDVH 7KLV H[SODLQV ZK\ WR RYHUFRPH WKH ORZHU 4 WKH FXUUHQW OHYHO ZDV LQFUHDVHG H[SODLQLQJ WKH DGGHG SRZHU FRQVXPSWLRQ OLVWHG LQ 7DEOH )LQDOO\ WR XQGHUVWDQG WKH FRPSHWLWLYHQHVV RI WKLV 9&2 UHVXOW ZLWK SUHYLRXVO\ SXEn OLVKHG UHVXOWV WKH SKDVH QRLVH FDQ EH VFDOHG WR D *+] UHJLPH 7KXV IRU D JLYHQ LQGXFWRU 4 WKH *+] 9&2 ZRXOG FRUUHVSRQG WR D *+] 9&2 DFKLHYLQJ D SKDVH QRLVH RI G%F+] DW D 0+] RIIVHW 7KLV UHVXOW LV a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n LFDQWO\ ORDGLQJ WKH 9&2 RXWSXW $W WKH WXQHG IUHTXHQF\ WKH VLQJOHVLGHG RXWSXW KDV DQ DPSOLWXGH DSSUR[LPDWHO\ HTXDO WR WKH VXSSO\ YROWDJH ZLWK DQ RIIVHW YROWDJH RI 9GG 3$ 7KH

PAGE 201

VHFRQG VWDJH DFWV DV D FODVV( DPSOLILHU ZLWKRXW D EDQGSDVV ILOWHU WUDGLWLRQDOO\ XVHG WR VHOHFW WKH IXQGDPHQWDOf DQG LV WXQHG WRJHWKHU ZLWK WKH DQWHQQD LPSHGDQFH +HQFH DV WKH DPSOLILHU VZLWFKHV ERWK WKH IXQGDPHQWDO DQG KLJKHURUGHU KDUPRQLFV DUH WUDQVPLWWHG 7KH VL]H RI WKH WUDQVLVWRUV LQ VWDJH LV WLPHV ODUJHU WKDQ WKDW RI WKH ILUVW VWDJH 'XH WR WKH WKLQ JDWH R[LGH DQG WKXV ORZ JDWHR[LGH EUHDNGRZQ YROWDJH WKH PD[LPXP RXWSXW SRZHU LV GHVLJQHG WR EH a G%P WKH YROWDJH DPSOLWXGH RI WKH GLIIHUHQWLDO RXWSXW LV 9 DQG WKH UHDO SDUW RI WKH DQWHQQD LPSHGDQFH LV f 7KH SRZHU HIILFLHQF\ LV b 7KH RXWSXW SRZHU FDQ EH VLPSO\ FRQWUROOHG E\ WKH VXSSO\ YROWDJH )LJXUH VKRZV D VLPXODWHG GLIn IHUHQWLDO RXWSXW ZDYHIRUP DFURVV WKH DQWHQQD 7KH RXWSXW ZDYHIRUP LV QRW VLQXVRLGDO DW WKH ]HURFURVVLQJ SRLQW GXH WR WKH SDVVLQJ RI KLJKHURUGHU KDUPRQLFV LQ WKH 3$ LQWURGXFLQJ GLVWRUWLRQ WR WKH VLJQDO )LJXUH Df VKRZV WKH PHDVXUHG VPDOOVLJQDO 6SDUDPHWHUV IRU D GLIIHUHQWLDO 3$ UHIHUHQFHG WR 4f 7KH LQSXW RI WKLV VWDQGDORQH 3$ KDG 4 UHVLVWRUV WR JURXQG DW HDFK LQSXW QRGH 7R REWDLQ WKH PHDVXUHPHQW EDOXQV DUH SODFHG DW WKH LQSXW DQG RXWSXW RI WKH 3$ WR WUDQVIRUP WKH XQEDODQFHG VLJQDOV IURP WKH QHWZRUN DQDO\]HU WR EDODQFHG VLJQDOV )LJXUH $ VLPXODWHG RXWSXW ZDYHIRUP DFURVV WKH DQWHQQD IURP WKH 3$

PAGE 202

)UHTXHQF\ *+]f Df )UHTXHQF\ *+]f Ef )LJXUH Df 6SDUDPHWHUV RI SRZHU DPSOLILHU GULYLQJ  GLIIHUHQWLDOO\f Ef 2SHUDWLQJ SRZHU JDLQ RI WKH 3$ ZKHQ FRQQHFWHG WR D ]LJ]DJ DQWHQQD IRU WKH 3$ 7KH IUHTXHQF\ UDQJH RI WKH EDOXQV LV OLPLWHG WR *+] WKXV WKH PHDVXUHPHQW U? LV DOVR OLPLWHG WR WKLV IUHTXHQF\ UDQJH 7KH 3$ GHPRQVWUDWHV DQ ,6LO RI G% DW *+] ZKLOH FRQVXPLQJ P: IURP D 9 VXSSO\ 7KH 6SDUDPHWHUV DUH REWDLQHG ZKHQ GULYn LQJ D 4 ORDG +RZHYHU VLQFH WKH DQWHQQD LPSHGDQFH LV QRW HTXDO WR 4 DQG KDV D UHDFWLYH FRPSRQHQW WKH JDLQ DQG UHVRQDQW IUHTXHQF\ IRU WKH 3$ GULYLQJ WKH DQWHQQD DUH GLIIHUHQW WKDQ WKDW IURP WKH 6SDUDPHWHU PHDVXUHPHQWV 7KH RSHUDWLQJ SRZHU JDLQ SORWWHG LQ )LJXUH Ef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

PAGE 203

7DEOH 6XPPDU\ RI PHDVXUHG FKDUDFWHULVWLFV IRU JP &026 9&2 DQG 3$ 9&2 ([SHFWHG 0HDVXUHG 3$ ([SHFWHG 0HDVXUHG )UHTXHQF\ 5DQJH *+] *+] IR *+] *+] 3RZHU &RQVXPSWLRQ P: P: 3RZHU &RQVXPSWLRQ P: 9GG 9f P: 9GG 9f 3KDVH 1RLVH # 0+] G%F+] 2XWSXW 3RZHU WR $QWHQQD G%P G%P 3KDVH 1RLVH # 0+] G%F+] 2SHUDWLQJ 3RZHU *DLQ If G% G% PHDVXUHPHQW RI WKH IXOO WUDQVPLWWHU ZLWK DQ LQWHJUDWHG DQWHQQD GLVFXVVHG LQ WKH QH[W VHFn WLRQf UHYHDOLQJ DQ DYDLODEOH SRZHU RI G%P DW 9 WR WKH DQWHQQD DW *+] 7DEOH VKRZV D FRPSDULVRQ EHWZHHQ WKH H[SHFWHG DQG PHDVXUHG 3$ UHVXOWV $V FDQ EH VHHQ WKH SRZHU JDLQ RI WKH DPSOLILHU LV OHVV WKDQ H[SHFWHG DQG WKH UHVRQDQW IUHn TXHQF\ KDV EHHQ UHGXFHG 'HFUHDVLQJ WKH LQGXFWRU 4 E\ KDOI VKRXOG UHGXFH WKH VPDOOVLJn QDO 3$ JDLQ E\ DSSUR[LPDWHO\ G% 7KH ILUVW G% FRPHV IURP WKH SUHDPSOLILHU VWDJH ZKRVH RXWSXW LV D SDUDOOHO UHVRQDQW FLUFXLW FRPSRVHG RI / DQG VKXQW FDSDFLWDQFHV 7KXV WKH ORDG LPSHGDQFH DW UHVRQDQFH LV SURSRUWLRQDO WR WKH 4 RI / $QRWKHU G% GHJUDGDn WLRQ FRPHV IURP WKH VHFRQG VWDJH ZKHUH LWV RXWSXW LV ORDGHG E\ WKH DQWHQQD +RZHYHU G% RI WKH JDLQ UHGXFWLRQ LV VWLOO XQDFFRXQWHG IRU )LQDOO\ WKH UHGXFWLRQ LQ UHVRQDQW IUHn TXHQF\ LV GXH WR LQFUHDVHG WUDQVLVWRU FDSDFLWDQFHV &ORFN 7UDQVPLWWHU ZLWK ,QWHJUDWHG $QWHQQD 5HIHUULQJ EDFN WR )LJXUH WKH FORFN WUDQVPLWWHU FRQVLVWV RI D 9&2 FRQQHFWHG WR D 3$ ZKLFK WKHQ GULYHV WKH WUDQVPLWWLQJ DQWHQQD 2SHUDWLRQ RI WKH FORFN WUDQVPLWWHU LV GHPRQVWUDWHG E\ RQFKLS JHQHUDWLRQ RI D JOREDO FORFN VLJQDO DQG UHFHSWLRQ RI WKLV VLJQDO

PAGE 204

2XWSXW WR 6SHFWUXP $QDO\]HU Df L 2O R f§ Ua f§ G PP G PP R &' G%P Q Q G%P ; ; ( ( P &4 ‘r P f n !f§ n ORRW BB L )UHTXHQF\ *+]f Ef )UHTXHQF\ *+]f Ff )LJXUH Df 0HDVXUHPHQW VHWXS XVHG WR FKDUDFWHUL]H ZLUHOHVV FORFN WUDQVPLWWHU Ef t Ff 2XWSXW VSHFWUD IURP UHFHLYLQJ DQWHQQD GHWHFWHG IURP FORFN WUDQVPLWWHU DFURVV D t PP GLVWDQFH XVLQJ UHFHLYLQJ DQWHQQDV )LJXUH Df VKRZV D EORFN GLDJUDP RI WKH PHDVXUHPHQW VHWXS XVHG WR WHVW WKH FORFN WUDQVPLWWHU 7KH '& FRQWURO DQG VXSSO\ YROWDJHV ZHUH VXSSOLHG WR D WUDQVPLWWHU GULYLQJ D ]LJ]DJ GLSROH DQWHQQD 7KH RXWSXW VSHFWUXP ZDV WKHQ REWDLQHG E\ SURELQJ UHFHLYLQJ DQWHQQDV ORFDWHG DW PP DQG PP VHSDUDWLRQV IURP WKH WUDQVPLWWLQJ DQWHQQD )LJXUHV Ef DQG Ff VKRZ WKH UHVXOWDQW RXWSXW VSHFWUD PHDVXUHG DW PP DQG PP GLVWDQFHV UHVSHFWLYHO\ KDYLQJ SHDN RXWSXW SRZHU OHYHOV RI DQG G%P DW a *+] 7KH WUDQVPLWWHU FRQVXPHV a P: RI SRZHU 7KXV DQ RQFKLS FORFN WUDQVPLWWHU ZLWK DQ LQWHJUDWHG DQWHQQD LV GHPRQVWUDWHG IRU WKH ILUVW WLPH

PAGE 205

:KLOH WKLV ILUVW RQFKLS ZLUHOHVV WUDQVPLWWHU GHPRQVWUDWLRQ LV VLJQLILFDQW WKH UHFHLYHG SRZHU OHYHO LV YHU\ ORZ 7R XQGHUVWDQG WKLV WKH SRZHU GHOLYHUHG WR WKH ]LJ]DJ WUDQVPLWWLQJ DQWHQQD E\ WKH 3$ LV REWDLQHG XVLQJ WKH PHDVXUHG *D IRU WKH PP DQWHQQD VHSDUDWLRQ DQG FDOFXODWLQJ WKH PLVPDWFK ORVV LQ WKH V\VWHP $W *+] WKH PLVPDWFK ORVV DW WKH 3$DQWHQQD LQWHUIDFH LV G% DQG WKH PLVPDWFK ORVV DW WKH DQWHQQDVSHFWUXP DQDn O\]HU LQWHUIDFH LV G% 7KHVH QXPEHUV DUH REWDLQHG XVLQJ f 5HIHUULQJ WR )LJXUH Df WKH DQWHQQD WUDQVPLVVLRQ JDLQ LV G% IRU D PP VHSDUDWLRQ:KLOH WKLV LV QRW WKH PHDVXUHG JDLQ IRU D PP VHSDUDWLRQ LW LV ZLWKLQ WKH GHYLDWLRQ H[SHFWHG IURP YDU\n LQJ PHWDO LQWHUIHUHQFH VWUXFWXUHV ZKLFK LV G% DV GHPRQVWUDWHG LQ )LJXUH Ef 7KHUHIRUH ZLWK DQ DQWHQQD WUDQVPLVVLRQ JDLQ RI G% WKH H[WUDFWHG SRZHU DYDLODEOH IURP WKH 3$ LV G%P f ZKLOH WKH SRZHU GHOLYHUHG WR WKH ]LJ]DJ DQWHQQD LV G%P f 7KH 3$ ZDV RULJLQDOO\ GHVLJQHG WR GHOLYHU G%P WR WKH DQWHQQD WKXV WKHUH LV D G% GLIIHUHQFH EHWZHHQ WKH H[SHFWHG DQG PHDVXUHG UHVXOWV 5HIHUULQJ EDFN WR VHFWLRQ WKH UHGXFWLRQ LQ LQGXFWRU 4 GXH WR WKLQ PHWDOV DQG DFFRXQWV IRU G% RI WKH 3$ JDLQ UHGXFWLRQ 6HFRQG RSHUDWLQJ WKH 3$ DW *+] DV RSSRVHG WR DW LWV UHVRQDQW IUHn TXHQF\ *+]f DFFRXQWV IRU DQRWKHU G% 7KLUG WKH GFELDV OHYHO SURYLGHG E\ WKH 9&2 DW WKH LQSXW RI WKH 3$ LV OHVV WKDQ WKDW XVHG WR FKDUDFWHUL]H WKH VWDQGDORQH 3$ 7KLV UHGXFHG ELDV SRLQW LV GXH WR WKH 9&2 FRUH FXUUHQW EHLQJ LQFUHDVHG WR RYHUFRPH WKH 4 UHGXFWLRQ 7KXV WKH WUDQVFRQGXFWDQFH RI WKH 3$ LV GHFUHDVHG UHVXOWLQJ LQ DQRWKHU aG% GHJUDGDWLRQ 7DNLQJ DOO RI WKHVH HIIHFWV LQWR DFFRXQW WKH H[SHFWHG SRZHU GHOLYHUHG E\ WKH 3$ VKRXOG EH G%P 7KLV LV FORVH WR WKH PHDVXUHG 3$fV DYDLODEOH SRZHU RI G%P ZKLFK DVVXPHV WKDW WKH 3$ DQG WKH DQWHQQD DUH FRQMXJDWHO\ PDWFKHG DV ZDV WKH FDVH LQ WKH

PAGE 206

)LJXUH 'LH SKRWRJUDSK RI SP FORFN UHFHLYHU ZLWK ]LJ]DJ GLSROH DQWHQQD VLPXODWLRQV ,Q VXPPDU\ DOWKRXJK RSHUDWLRQ RI WKH ZLUHOHVV WUDQVPLWWHU KDV EHHQ GHPRQn VWUDWHG WKH SRZHU WKDW WKH WUDQVPLWWHU FDQ SURYLGH WR WKH UHFHLYHUV WKURXJK WKH DQWHQQD SDLU LV YHU\ ORZ DQG DV ZLOO EH VKRZQ LQ WKH QH[W VHFWLRQ LV EHORZ WKDW ZKLFK WKH UHFHLYHU FDQ GHWHFW $V FDQ EH VHHQ WKH 3$ DSSHDUV WR EH WKH ZHDNHVW OLQN LQ WKH WUDQVPLWWHU DQG UHTXLUHV FLUFXLW RSWLPL]DWLRQ 6LQJOH &ORFN 5HFHLYHU ZLWK ,QWHJUDWHG $QWHQQD )LJXUH VKRZV D GLH SKRWRJUDSK RI WKH _LP FORFN UHFHLYHU ZLWK DQ LQWHn JUDWHG ]LJ]DJ GLSROH DQWHQQD 7KH VL]H RI WKH UHFHLYHU LQFOXGLQJ WKH DQWHQQD LV [ PP ZKLOH WKH DFWLYH DUHD LV [ PP 7KH UHFHLYHU FRQVLVWV RI D ]LJ]DJ GLSROH DQWHQQD D GLIIHUHQWLDO /1$ ZLWK VRXUFHIROORZHU EXIIHUV DQ IUHTXHQF\ GLYLGHU DQG RXWSXW EXIIHUV 7KH SHUIRUPDQFH RI WKH _LP /1$ DQG VRXUFHIROORZHUV ZDV SUHVHQWHG LQ &KDSWHU ZKLOH WKH GLYLGHU SHUIRUPDQFH ZDV SUHVHQWHG LQ &KDSWHU $ VXPPDU\ RI WKH UHVXOWV LV VKRZQ LQ 7DEOH IRU HDV\ UHIHUHQFH 7R GHFUHDVH WKH PLQLPXP GHWHFWDEOH VLJn QDO RI WKH UHFHLYHU WKH SHDN /1$EXIIHU JDLQ VKRXOG FRLQFLGH ZLWK WKH VHOIUHVRQDQFH RI WKH IUHTXHQF\ GLYLGHU 7R DFKLHYH WKLV WKH VXSSO\ YROWDJH RI WKH IUHTXHQF\ GLYLGHU ZDV

PAGE 207

7DEOH 6XPPDU\ RI PHDVXUHG FKDUDFWHULVWLFV IRU SP &026 /1$ DQG GLYLGHU /1$ DQG 6) %XIIHU ([SHFWHG 0HDVXUHG )UHTXHQF\ 'LYLGHU ([SHFWHG 0HDVXUHG I *+] *+] 'LYLVLRQ 5DWLR *DLQ G% G% IPD[9GG 9f 9GG 9f *+] *+] *+] *+] 1) G% G% IL62 9GG 9f 9GG 9f *+] *+] *+] *+] 3RZHU P: P: 3RZHU 9GG 9f 9GG 9f P: P: P: P: LQFUHDVHG IURP WR 9 LQFUHDVLQJ ILVR IURP WR *+] 2SHUDWLRQ RI WKH FORFN UHFHLYHU LV GHPRQVWUDWHG E\ WUDQVPLVVLRQ RI D JOREDO FORFN VLJQDO DFURVV WKH FKLS GHWHFWLRQ RI WKLV VLJQDO DQG JHQHUDWLRQ RI D ORFDO FORFN VLJQDO E\ D VLQJOH FORFN UHFHLYHU 7KH VDPH PHDVXUHPHQW VHWXS VKRZQ LQ )LJXUH Df DQG DJDLQ LQ )LJXUH Df ZDV XVHG WR WHVW WKH FORFN UHFHLYHU 7KH LQSXW VLJQDO LV JHQHUDWHG RIIFKLS DQG H[WHUQDOO\ DPSOLILHG 7KLV VLJQDO LV WKHQ WUDQVIRUPHG LQWR D EDODQFHG VLJQDO DQG LQMHFWHG LQWR WKH WUDQVPLWWLQJ DQWHQQD RQFKLS 7KH RXWSXW RI WKH UHFHLYHU ORFDWHG DFURVV WKH FKLS LV WKHQ SUREHG DQG H[DPLQHG XVLQJ DQ RVFLOORVFRSH 7ZR FORFN UHFHLYHUV 5[O DQG 5[ IURP )LJn XUH f ZHUH WHVWHG KDYLQJ DQWHQQD VHSDUDWLRQV RI DQG PP UHVSHFWLYHO\ )LJXUH Ef VKRZV SORWV RI WKH LQSXW YROWDJH WR WKH WUDQVPLWWLQJ DQWHQQD DQG WKH RXWSXW YROWDJH IRU WKH ZLUHOHVV FORFN UHFHLYHU IRU D PP WUDQVPLVVLRQ GLVWDQFH 5[f GHPRQVWUDWLQJ RSHUDWLRQ RI WKH FORFN UHFHLYHU ZLWK LQWHJUDWHG DQWHQQD 7KH LQSXW JOREDO FORFN IUHTXHQF\ LV *+] DQG WKH RXWSXW ORFDO FORFN IUHTXHQF\ LV *+] *+] GLYLGHG E\ f 7KH PLQLPXP SRZHU OHYHO QHHGHG DW WKH WUDQVPLWWHU IRU WKH FORFN UHFHLYHU WR ORFN WR WKH LQSXW VLJQDO LV G%P 1RWH WKDW WKH LQSXW VLJQDO SORWWHG LQ )LJXUH Ef LV WDNHQ EHIRUH WKH

PAGE 208

)LJXUH Df 0HDVXUHPHQW VHWXS XVHG WR FKDUDFWHUL]H ZLUHOHVV FORFN UHFHLYHU Ef ,QSXW DQG RXWSXW ZDYHIRUPV IRU FORFN UHFHLYHU GHPRQVWUDWLQJ RSHUDWLRQ RI UHFHLYHU IRU PP VHSDUDWLRQ DW *+] H[WHUQDO DPSOLILHU 7KH UHGXFHG YROWDJH VZLQJ DW WKH RXWSXW UHVXOWV IURP GULYLQJ D  ORDG 7KH WRWDO SRZHU FRQVXPSWLRQ IRU WKH UHFHLYHU LV P: ,QIHUULQJ IURP WKH PHDVXUHG ]LJ]DJ]LJ]DJ DQWHQQD SDLU WUDQVPLVVLRQ JDLQ a G%f DQG DFFRXQWLQJ IRU PLVPDWFK ORVV LQ WKH V\VWHP WKH SRZHU GHOLYHUHG WR WKH LQSXW RI WKH UHFHLYHU LV G%P $V H[SHFWHG IRU WKH PP VHSDUDWLRQ WKH LQSXW SRZHU OHYHO WR WKH WUDQVPLWWLQJ DQWHQQD FDQ EH G% ORZHU VLQFH WKH DQWHQQD WUDQVPLVVLRQ JDLQ LV G% KLJKHU 6LPXOWDQHRXV 7UDQVPLWWHU DQG 5HFHLYHU 2SHUDWLRQ %\ FRPSDULQJ WKH UHVXOWV IURP ERWK WKH WUDQVPLWWHU DQG UHFHLYHU FLUFXLWV ZLWK LQWHn JUDWHG DQWHQQDV LW FDQ EH VHHQ WKDW VLPXOWDQHRXV RSHUDWLRQ RI WKHVH WZR FLUFXLWV LH LQWHn JUDWHG WUDQVPLWWHU FRPPXQLFDWLQJ DFURVV WKH FKLS WR LQWHJUDWHG UHFHLYHUf LV FXUUHQWO\ QRW SRVVLEOH 7KH LQSXW SRZHU OHYHO IRU WKH UHFHLYHU LV G%P ZKLOH WKH SRZHU UHFHLYHG IURP WKH WUDQVPLWWHU LV DQG G%P IRU KDOIFKLS DQG IXOOFKLS WUDQVPLVVLRQ

PAGE 209

UHVSHFWLYHO\ 7KXV WKHUH LV D G% DQG G% GHILFLW LQ fV\VWHP JDLQf IRU DFKLHYLQJ D IXOO\ LQWHJUDWHG KDOI DQG IXOOFKLS ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP UHVSHFWLYHO\ $ SULn PDU\ UHDVRQ IRU WKLV GHILFLW LV WKH UHGXFHG WKLFNQHVV RI PHWDOV DQG $V KDV EHHQ VKRZQ WKH HIIHFW RI ORZHU WKDQ H[SHFWHG LQGXFWRU 4 DQG UHGXFHG RSHUDWLRQ IUHTXHQF\ KDV UHGXFHG WKH RYHUDOO V\VWHP JDLQ )LUVW WKH b 4 GHJUDGDWLRQ UHGXFHG WKH 3$ JDLQ E\ DW OHDVW G% 4 DQG JP HIIHFWf DQG KDV UHGXFHG WKH /1$ JDLQ E\ DSSUR[LPDWHO\ G% 6HFRQG UHIHUn ULQJ WR )LJXUH Df RSHUDWLQJ WKH DQWHQQDV DW *+] DV RSSRVHG WR DW *+] UHVXOWV LQ DSSUR[LPDWHO\ G% OHVV WUDQVPLVVLRQ JDLQ IRU WKH IXOOFKLS WUDQVPLVVLRQ GLVWDQFH $OVR WKH PLVPDWFK LQ WKH 9&2 IUHTXHQF\ UDQJH DQG SHDN JDLQ IUHTXHQF\ IRU WKH 3$ OHG WR DQRWKHU G% GHJUDGDWLRQ RI WKH V\VWHP JDLQ WKH 3$ DSSHDUV WR FXUUHQWO\ EH D VHQVLWLYH VSRW LQ WKH V\VWHPf 7KHUHIRUH D G% GLIIHUHQFH FDQ EH DFFRXQWHG IRU LQGLFDWLQJ WKDW WKH IXOO\LQWHJUDWHG V\VWHP ZRXOG KDYH ZRUNHG LI WKH FRSSHU WKLFNQHVV ZDV GRXEOHG DQG WKH RSHUDWLQJ IUHTXHQF\ KDG EHHQ DERYH *+] $V LW LV E\ MXVW GRXEOLQJ WKH PHWDO WKLFNQHVV RI PHWDO DQG OD\HUV WKH V\VWHP JDLQ ZLOO LQFUHDVH E\ DSSUR[LPDWHO\ G% ZKLFK VKRXOG EH HQRXJK WR DOORZ RSHUDWLRQ RI WKH IXOO\LQWHJUDWHG KDOIFKLS GLVWULEXWLRQ V\VWHP $V PHQWLRQHG HDUOLHU WKHVH GHVLJQV ZHUH LPSOHPHQWHG DV SDUW RI WKH 65& &RSSHU 'HVLJQ &KDOOHQJH 'XH WR FLUFXLW IDEULFDWLRQ SUREOHPV DW 80& DQG WKH WKLQQLQJ RI WKH WRSOD\HU PHWDOV WKH PDVNV DUH EHLQJ UHJHQHUDWHG DQG DQRWKHU UXQ LV IRUWKFRPLQJ :LWK WKLV QHZ UXQ LW LV H[SHFWHG WKDW DOO RI WKH FLUFXLW SHUIRUPDQFHV ZLOO GUDPDWLFDOO\ LPSURYH DQG WKDW VLPXOWDQHRXV RSHUDWLRQ RI D FORFN UHFHLYHU DQG WUDQVPLWWHU ZLOO EH DFKLHYHG 'RXEOH5HFHLYHU :LUHOHVV ,QWHUFRQQHFW $ FULWLFDO WHVW IRU WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP LV WR GHPRQVWUDWH WZR FORFN UHFHLYHUV RSHUDWLQJ VLPXOWDQHRXVO\ DQG V\QFKURQRXVO\ IURP WKH VDPH WUDQVPLWWHG

PAGE 210

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r LQFUHPHQWV DV SUHVHQWHG LQ VHFWLRQ 7KLV SURJUDPPLQJ LV XVHG WR WXQH RXW DQ\ WLPHLQYDULDQW GHOD\ PLVPDWFK EHWZHHQ WKH UHFHLYHUV OLPLWLQJ WKH FORFN VNHZ WR XQGHU b IRU DQ GLYLGHU 7KH WUDQVPLWWLQJ DQWHQQDV XVHG IRU WKH GRXn EOHUHFHLYHU WHVW DUH =E DQG =F SRVLWLRQHG RQ WKH OHIWKDQG VLGH RI WKH FKLS $QWHQQD =E LV ORFDWHG PP IURP ERWK 5[ DQG 5[ ZKLOH =F LV ORFDWHG PP IURP 5[ DQG PP IURP 5[ 0HDVXUHPHQW 6HWXS 7KH GRXEOHUHFHLYHU PHDVXUHPHQW LV LOOXVWUDWHG LQ )LJXUH 7KH LQSXW JOREDO FORFN VLJQDO *&/.f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

PAGE 211

)LJXUH 'LDJUDP RI PHDVXUHPHQW VHWXS GHYHORSHG IRU GRXEOHUHFHLYHU LQWHUFRQQHFW 0HDVXULQJ WKH FORFN VNHZ EHWZHHQ WZR FORFN UHFHLYHUV XVLQJ ZDIHU SURELQJ ZRXOG UHTXLUH PDWFKLQJ WKH GHOD\ WKURXJK WKH H[WHUQDO PHDVXUHPHQW V\VWHP VXFK WKDW WKH PHDn VXUHG FORFN VNHZ LV GXH RQO\ WR WKH VNHZ LQ WKH FLUFXLWV WKHPVHOYHV 6LQFH WKLV LV GLIILFXOW DW a *+] WKH FORFN VNHZ LV GHWHFWHG RQ FKLS XVLQJ D SKDVH GHWHFWRU 3'f 7KH 3' FRQYHUWV RQFKLS VNHZ WR YROWDJH DV SUHVHQWHG LQ &KDSWHU VHFWLRQ 7KH 3' LV V\PPHWULF ZLWK UHVSHFW WR ERWK LQSXW VLJQDOV SUHYHQWLQJ D V\VWHPDWLF VNHZ RU SKDVH GLIIHUHQFH IURP DULVLQJ LQ WKH 3' LWVHOI $V VKRZQ LQ )LJXUH Ef WKH RXWSXW YROWDJH UHVSRQVH IURP WKH 3' LV DSSUR[LPDWHO\ OLQHDU YHUVXV LQSXW VNHZ DQG LV V\PPHWULF ZLWK UHVSHFW WR WKH \D[LV $ FXVWRPL]HG SUREH ZDV UHTXLUHG WR PHDVXUH WKH WZR UHFHLYHUV DQG WKH SKDVH GHWHFWRU RXWSXW VLPXOWDQHRXVO\ 6LQFH HDFK FORFN UHFHLYHU KDV SUREHV RQ WKUHH VLGHV WKH RXWSXWV IURP WZR VLGHV RI HDFK UHFHLYHU ZHUH URXWHG VXFK WKDW RQH SUREH FRXOG EH XVHG LQ

PAGE 212

SODFH RI IRXU RULJLQDO SUREHV 7KHUHIRUH D SLQ SUREH ZDV UHTXLUHG REWDLQHG IURP **% ,QGXVWULHVf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f ZKLFK UHVWHG RQ WKH PHWDO FKXFN RI WKH SUREH VWDWLRQ )LJXUH VKRZV WKH PHDVXUHG UHVXOWV IRU WKH GRXEOHUHFHLYHU ZLUHOHVV LQWHUFRQQHFW 7KH LQSXW IUHn TXHQF\ LV *+] DQG WKH RXWSXW IUHTXHQFLHV IRU ERWK FORFN UHFHLYHUV DUH *+] 7KH LQSXW SRZHU DYDLODEOH WR WKH WUDQVPLWWLQJ DQWHQQD LV VHW WR G%P $V FDQ EH VHHQ WKLV UHVXOW GHPRQVWUDWHV WZR UHFHLYHUV RSHUDWLQJ VLPXOWDQHRXVO\ )URP WKH ILJXUH WKH WZR FORFN VLJQDOV DSSHDU WR KDYH YHU\ ORZ VNHZ +RZHYHU WKLV FDQ EH PLVOHDGLQJ EHFDXVH WKHVH ZDYHIRUPV ZHUH REWDLQHG WKURXJK WKH H[WHUQDO FDEOHV LQ WKH PHDVXUHPHQW V\VWHP 7KXV WKHUH LV VNHZ LQGXFHG E\ PLVPDWFKHV EHWZHHQ HDFK FDEOH DQG LWV FRQQHFWRUV 7KH H[WUDFWHG RQFKLS VNHZ IRU WKLV FDVH REWDLQHG ZLWK WKH SKDVH GHWHFWRU ZLOO EH SUHVHQWHG LQ WKH QH[W VHFWLRQ 7R REWDLQ WKH SRZHU GHOLYHUHG WR HDFK UHFHLYHU WKH DQWHQQD WUDQVGXFHU JDLQ ZDV PHDVXUHG IRU HDFK DQWHQQD SDLU ZLWK D ZRRG GLHOHFWULF OD\HU EHQHDWK WKH FKLS 7KH PHDVXUHG DQWHQQD WUDQVGXFHU JDLQV DW *+] DUH G% IRU ERWK DQWHQQD SDLUV =EBD DQG =EBF )URP WKH DQWHQQD LPSHGDQFH WKH LQSXW LPSHGDQFH RI WKH

PAGE 213

' 2f &2 ff§} e r 4B & ILQ *+]A ? ,' L DV WKH WUDQVPLWWLQJ DQWHQQD ZLWK D ZRRG DV WKH EDFN GLHOHFWULF ,QSXW IUHTXHQF\ LV *+] RXWSXW IUHTXHQFLHV DUH *+] GLVWDQFHV DUH PP HDFK UHFHLYHU DQG WKH LPSHGDQFH RI WKH LQSXW EDOXQ WKH PLVPDWFK ORVV LQ WKH V\VWHP LV FDOFXn ODWHG WR EH G% :LWK D G%P DYDLODEOH LQSXW SRZHU WKH SRZHU GHOLYHUHG WR HDFK UHFHLYHU LV G%P FRUUHVSRQGLQJ WR WKH PLQLPXP GHWHFWDEOH VLJQDO IRU WKH UHFHLYHU 7KH VHFRQG GRXEOHUHFHLYHU LQWHUFRQQHFW ZDV GHPRQVWUDWHG XVLQJ =JF DV WKH WUDQVn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

PAGE 214

)LJXUH 0HDVXUHG LQSXW DQG RXWSXW ZDYHIRUPV RI GRXEOHUHFHLYHU LQWHUFRQQHFW XVLQJ =F DV WKH WUDQVPLWWLQJ DQWHQQD ZLWK D $,1L DV WKH EDFN GLHOHFWULF ,QSXW IUHTXHQF\ LV *+] RXWSXW IUHTXHQFLHV DUH *+] GLVWDQFHV DUH PP IRU 5[ DQG PP IRU 5[ 7LPHRIIOLJKW GHOD\ PLVPDWFK DQG DPSOLWXGH PLVPDWFK DUH SUHVHQW )RU WKLV VHFRQG PHDVXUHPHQW WKH 80& WHVWFKLS ZDV PRXQWHG RQ D JODVV VOLGH DQG SODFHG RQ D PP WKLFN VODE RI DOXPLQXP QLWULGH $,1f 7KLV WKLFNQHVV ZDV FKRVHQ EDVHG RQ WKH UHVXOWV SUHVHQWHG LQ >*XR2O@ 5HFDOO WKDW $,1 KDV D KLJK WKHUPDO FRQGXFWLYLW\ DSSUR[LPDWHO\ WKH VDPH DV VLOLFRQf ZKLOH VWLOO EHLQJ D JRRG GLHOHFWULF NDQ f 7KXV LW LV D VXLWDEOH PHGLXP IRU WKH EDFNVLGH GLHOHFWULF OD\HU ZKLFK ZLOO EH FRQQHFWHG WR WKH KHDWn VLQN )LJXUH VKRZV WKH PHDVXUHG ZDYHIRUPV IRU WKH GRXEOHUHFHLYHU LQWHUFRQQHFW ZLWK =F DV WKH WUDQVPLWWLQJ DQWHQQD 2QFH DJDLQ WKH LQSXW IUHTXHQF\ LV *+] DQG WKH RXWSXW IUHTXHQFLHV DUH *+] GHPRQVWUDWLQJ WZR UHFHLYHUV RSHUDWLQJ VLPXOWDQHRXVO\ 7KH FORFN VNHZ IRU WKLV FDVH ZLOO EH SUHVHQWHG LQ WKH IROORZLQJ VHFWLRQ KRZHYHU E\ LQVSHFWLRQ WKH VNHZ DSSHDUV WR EH VLPLODU WR WKH UHVXOW VKRZQ LQ )LJXUH

PAGE 215

7KH PHDVXUHG WUDQVGXFHU JDLQV IRU WKH DQWHQQDV RQ WKH $,1 VXEVWUDWH DW *+] DUH G% IRU DQWHQQD SDLU =FBD GLVWDQFH PPf DQG G% IRU DQWHQQD SDLU =IF6F GLVWDQFH f§ PPf ,QWHUHVWLQJO\ WKH JDLQ IRU WKH GLDJRQDO SDLU ]FD! LV ODU6HU! HYHQ WKRXJK WKH GLVWDQFH LV JUHDWHU 7KLV VHHPV WR LQGLFDWH WKDW WKH VLJQDO GHJUDGDWLRQ GXH WR LQWHUIHUHQFH VWUXFWXUHV LV JUHDWHU IRU DQWHQQD SDLU =FBF :LWK D G%P DYDLODEOH LQSXW SRZHU WR WKH WUDQVPLWWLQJ DQWHQQD DQG D G% PLVPDWFK ORVV WKH SRZHU GHOLYHUHG WR HDFK UHFHLYHU LV DV IROORZV G%P IRU 5[ DQG G%P IRU 5[ 7KH G% GLIIHUHQFH LQ LQSXW SRZHU OHYHOV FDQ UHVXOW LQ DV PXFK DV b RI FORFN VNHZ DFFRUGLQJ WR VHFWLRQ HTXDWLRQ f LH VNHZ f§FRVB^ n`f 7KXV DW OHDVW D WRWDO RI EHWZHHQ •& b RI V\VWHPDWLF FORFN VNHZ GXH WR WLPHRIIOLJKW GHOD\ PLVPDWFK DQG DPSOLWXGH PLVPDWFK VKRXOG EH SUHVHQW LQ WKLV PHDVXUHPHQW 0HDVXUHG 6NHZ %HWZHHQ 7ZR &ORFN 5HFHLYHUV 7KH VNHZ EHWZHHQ WKH WZR UHFHLYHUV IRU WKH FDVH ZLWK =E DV WKH WUDQVPLWWLQJ DQWHQQDf ZDV H[WUDSRODWHG IURP WKH UHVSRQVH RI WKH SKDVH GHWHFWRU ZKHQ WKH VWDUWXS VWDWHV RI WKH GLYLGHUV ZHUH YDULHG DFURVV WKHLU SRVVLEOH YDOXHV 6LQFH WKH VNHZ LQGXFHG E\ WKH PHDVXUHPHQW V\VWHP ZLOO EH FRQVWDQW DW D VLQJOH IUHTXHQF\ WKH RXWSXW RI WKH SKDVH GHWHFn WRU FDQ EH FRPSDUHG WR WKH VNHZ PHDVXUHG ZLWK WKH RVFLOORVFRSH 7KLV UHVXOWV LQ D YROW DJHYHUVXVVNHZ UHVSRQVH IRU WKH SKDVH GHWHFWRU ZKLFK VKRXOG EH PD[LPXP DW WKH SRLQW RI ]HUR VNHZ EHWZHHQ WKH WZR LQSXW VLJQDOV DQG ZKLFK VKRXOG EH VLPLODU WR WKH UHVSRQVH VKRZQ LQ )LJXUH Ef $V GLVFXVVHG LQ &KDSWHU WKH SURJUDPPDEOH GLYLGHU KDV D VWDWHGHSHQGHQW LQLWLDOn L]DWLRQ IDLOXUH ZKLFK FDXVHV WKH GLYLGHU WR LQLWLDOL]H WR DQ\ RI WKH SRVVLEOH VWDWHV IRU D JLYHQ LQSXW GHVLUHG VWDWH 7KH SURJUDPPHG VWDWH GHSHQGV RQ WKH VWDWH RI WKH GLYLGHU ZKHQ

PAGE 216

7LPH 'LIIHUHQFH 0HDVXUHG ZLWK 2VFLOORVFRSH RU 2II&KLS 6NHZ SVf EHWZHHQ ZDYHIRUPV IURP 5[ DQG 5[ XVLQJ RVFLOORVFRSH /LQHV ZLWK HTXDO DQG RSSRVLWH VORSHV DUH ILW WR ERWK WKH SRVLWLYH DQG QHJDWLYH SRUWLRQV RI WKH FXUYH ZKHUH WKH OLQHV LQWHUVHFW DW WKH SRLQW RI ]HUR VNHZ WKH LQLWLDOL]DWLRQ VLJQDO LV HQJDJHG (YHQ ZLWK WKLV IDLOXUH WKRXJK WKH GLYLGHU ZLOO WDNH RQ GLVFUHWL]HG VWDWHV RU SKDVHV WKXV WKH PLQLPXP DGMXVWHG VNHZ FDQ VWLOO EH H[WUDFWHG )LJXUH VKRZV WKH RXWSXW YROWDJH RI WKH SKDVH GHWHFWRU YHUVXV WKH VNHZ EHWZHHQ WKH WZR ZDYHIRUPV PHDVXUHG ZLWK WKH RVFLOORVFRSH WKURXJK WKH H[WHUQDO PHDVXUHn PHQW V\VWHP 7KH WRS [D[LV LV IRU VNHZ PHDVXUHG GLUHFWO\ ZLWK WKH RVFLOORVFRSH RU RIIFKLS VNHZ ZKLFK ZRXOG LQFOXGH VNHZ DGGHG E\ WKH PHDVXUHPHQW V\VWHP 8VLQJ WKHVH YDOXHV OLQHV DUH ILW XVLQJ 0DWODEf WR ERWK WKH LQFUHDVLQJ DQG GHFUHDVLQJ SRUWLRQ RI WKH YROWDJH UHVSRQVH 7KH VORSHV RI WKHVH OLQHV DUH VHW WR EH HTXDO DQG RSSRVLWH VLQFH WKH SKDVH GHWHFWRU KDV D UHVSRQVH ZKLFK KDV HYHQRUGHU V\PPHWU\ ZLWK UHJDUGV WR LQSXW VNHZ 7KH SKDVH GHWHFWRU ZDV GHVLJQHG VXFK WKDW ERWK LQSXWV KDYH LGHQWLFDO ORDGV HOLPLQDWLQJ DQ\ V\VWHPDWLF SKDVH HUURU 8QGHU WKLV DVVXPSWLRQ WKH WZR OLQHV LQ )LJXUH LQWHUVHFW DW WKH

PAGE 217

SRLQW RI ]HUR LQSXW VNHZ )URP WKLV LQWHUVHFWLRQ WKH VNHZ DGGHG E\ WKH PHDVXUHPHQW V\Vn WHP FDQ EH GHWHUPLQHG WR EH SV 7KLV WKHQ DOORZV WKH RQFKLS VNHZ WR EH GHWHUPLQHG E\ VXEWUDFWLQJ SV IURP WKH PHDVXUHG RIIFKLS VNHZ UHVXOWLQJ LQ WKH VHFRQG [D[LV LQ )LJn XUH 7KXV WKH PLQLPXP RQFKLS VNHZ LV SV ZKLFK LV b RI WKH ORFDO FORFN SHULRG %\ DGMXVWLQJ WKH VWDUWXS VWDWHV RI WKH GLYLGHUV WKH VNHZ FDQ EH UHGXFHG WR XQGHU b ZKLFK LV WKH UHTXLUHPHQW VHW IRU JOREDO FORFN GLVWULEXWLRQ V\VWHPV 7KLV VDPH VNHZ H[WUDFWLRQ PHWKRGRORJ\ KDV EHHQ DSSOLHG WR WKH FDVH ZKHUH =JF LV XVHG DV WKH WUDQVPLWWLQJ DQWHQQD VKRZLQJ SV RI VNHZ RU b 7KLV LV VLJQLILFDQW VLQFH WKH PHDVXUHPHQW FDVH KDV DW OHDVW b RI V\VWHPDWLF VNHZ GXH WR WLPHRIIOLJKW GHOD\ PLVPDWFK DQG DPSOLn WXGH PLVPDWFK LQ DGGLWLRQ WR SURFHVV YDULDWLRQ 8QIRUWXQDWHO\ WKH XQFRUUHFWHG RU XQSURJUDPPHG VNHZ FDQ QRW EH GHWHUPLQHG IURP WKLV PHDVXUHPHQW 7KLV LV GXH WR WKH VWDWHGHSHQGHQW LQLWLDOL]DWLRQ IDLOXUH RI WKH GLYLGHUV ,I WKH GLYLGHUV LQLWLDOL]H SURSHUO\ WKHQ WKH XQFRUUHFWHG VNHZ ZRXOG EH WKH VNHZ PHDVXUHG ZKHQ ERWK GLYLGHUV DUH LQLWLDOL]HG WR WKH VDPH VWDWH +RZHYHU VLQFH HDFK GLYLGHU FXUUHQWO\ LQLWLDOL]HV WR D UDQGRP VWDWH LW LV LPSRVVLEOH WR NQRZ LI WKH GLYLGHUV DUH LQLWLDOn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

PAGE 218

, r r r r r 7LPH QVf )LJXUH ,OOXVWUDWLRQ RI WKH PHDVXUHG MLWWHU RI WKH ZLUHOHVV FORFN UHFHLYHU 7KLV MLWWHU ZDV PHDVXUHG RII RI WKH *+] LQSXW VLJQDO \LHOGLQJ WKH MLWWHU RI WKH *+] RXWSXW VLJQDO XVHG DV WKH WULJJHU 7KH PHDVXUHG SHDNWRSHDN MLWWHU ZDV SV VLJQDO ZKLFK VKRXOG KDYH YHU\ VPDOO MLWWHU RU SKDVH QRLVH \LHOGV WKH MLWWHU RI WKH WULJJHUHG *+] ZDYHIRUP $Q LOOXVWUDWLRQ RI WKH UHVXOWDQW SORW ZLWK LQILQLWH SHUVLVWHQFH LV VKRZQ LQ )LJXUH 7KLV LV DQ LOOXVWUDWLRQ UDWKHU WKDQ WKH UHDO GDWD VLQFH D VFUHHQ FDSn WXUH RI WKH RVFLOORVFRSH FRXOG QRW EH REWDLQHG 7KH PHDVXUHG SHDNWRSHDN MLWWHU LV SV DW *+] FRUUHVSRQGLQJ WR b RI WKH ORFDO FORFN SHULRG 7KLV LV XQGHU WKH V\VWHP UHTXLUHPHQW RI b SHDNWRSHDN MLWWHU 1RWH WKDW WKLV MLWWHU LV IRU D fTXLHWf V\VWHP PHDQLQJ WKDW WKH VXSSO\ YROWDJHV DUH QRW EHLQJ PRGLILHG E\ GLJLWDO FLUFXLWV ,Q WKH OLWHUDWXUH WKLV LV NQRZQ DV WKH TXLHWVXSSO\ MLWWHU ,Q DQ RSHUDWLQJ PLFURSURFHVVRU WKH MLWWHU ZLOO EH LQFUHDVHG GXH WR ODUJHU VZLWFKLQJ QRLVH DQG QRLV\ VXSSOLHV 7KH VLJQDOWRQRLVH UDWLR DW WKH LQSXW RI WKH GLYLGHU FDQ EH GHWHUPLQHG XVLQJ f DQG WKH PHDVXUHG MLWWHU DQG DVVXPLQJ WKDW WKH MLWWHU LV GRPLQDWHG E\ LQSXW DGGLWLYH QRLVH $VVXPLQJ WKDW WKH 506 MLWWHU LV RQHVL[WK RI WKH SHDNWRSHDN MLWWHU RU SV bf WKHQ WKH 615 DW WKH GLYLGHU LQSXW LV FDOFXODWHG WR EH G% 7KLV DJUHHV UHPDUNDEO\ ZHOO

PAGE 219

ZLWK WKH 615 RI G% HVWLPDWHG XVLQJ 63,&( ZKLFK ZDV SUHVHQWHG LQ VHFWLRQ RI &KDSWHU 6LQFH WKH V\VWHP UHTXLUHPHQW IRU WKH 615 WR REWDLQ b 506 MLWWHU LV G% D G% PDUJLQ LV SUHVHQW +RZHYHU WKLV PDUJLQ FRXOG RI FRXUVH EH FRQVXPHG DQG SHUn KDSV H[FHHGHGf E\ WKH LQWHUIHUHQFH GXH WR GLJLWDO VZLWFKLQJ QRLVH 6XPPDU\ ,Q WKLV FKDSWHU WKH LPSOHPHQWDWLRQ RI RQFKLS ZLUHOHVV LQWHUFRQQHFWV ZLWK LQWHn JUDWHG DQWHQQDV KDV EHHQ SUHVHQWHG 7ZR GLIIHUHQW WHVW FKLSV KDYH EHHQ LPSOHPHQWHGf§RQH LQ D _LP &026 SURFHVV DQG RQH LQ D _LP &026 SURFHVV 0HDVXUHG DQWHQQD FKDUDFWHULVWLFV IRU ERWK FKLSV VKRZ WKDW DQWHQQD JDLQ LQFUHDVHV ZLWK IUHTXHQF\ ZKHUH WKH DQWHQQD EHFRPHV HOHFWULFDOO\ ORQJHU $OVR WKH SUHVHQFH RI LQWHUIHUHQFH VWUXFWXUHV EHWZHHQ WKH DQWHQQDV FDQ DOWHU WKH JDLQ E\ G% /RRS DQWHQQDV DQG DQWHQQDV LQ GLUHFW FRQWDFW ZLWK WKH VXEVWUDWH KDYH EHHQ FKDUDFWHUL]HG DQG VKRZ SURPLVLQJ UHVXOWV IRU KLJKHU IUHTXHQn FLHV $GGLWLRQDOO\ WKH SKDVH YHUVXV IUHTXHQF\ UHVSRQVH IRU WKH DQWHQQDV LV OLQHDU LQGLFDWn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n PHQWHG $ *+] JOREDO FORFN VLJQDO ZDV VXFFHVVIXOO\ UHFHLYHG RYHU D PP GLVWDQFH ZLWK LQWHUIHUHQFH VWUXFWXUHV ORFDWHG LQ EHWZHHQ WKH DQWHQQDV DQG D 0+] ORFDO FORFN VLJQDO ZDV JHQHUDWHG 7KLV UHVXOW LV WKH ILUVW GHPRQVWUDWLRQ RI DQ RQFKLS ZLUHOHVV LQWHUFRQn QHFW $OVR WKLV UHVXOW LV WKH ILUVW WR LQWHJUDWH DQWHQQDV DQG &026 FLUFXLWU\ RQ D VLQJOH FKLS

PAGE 220

,Q D _LP FRSSHU WHFKQRORJ\ D *+] ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP FRPSRVHG RI LQWHJUDWHG WUDQVPLWWHUV UHFHLYHUV DQG DQWHQQDV KDV EHHQ LPSOHPHQWHG $ ZLUHOHVV WUDQVn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n WDQFH VFDOH URXJKO\ ZLWK 0RRUHfV /DZ XVLQJ WZR GDWD SRLQWVf )LQDOO\ D ZLUHOHVV LQWHUFRQQHFW FRPSRVHG RI WZR FORFN UHFHLYHUV ORFNHG WR WKH VDPH JOREDO FORFN VLJQDO KDV EHHQ GHPRQVWUDWHG DW *+] 7KH GLVWDQFHV EHWZHHQ WKH WUDQVPLWWLQJ DQWHQQD DQG WKH WZR FORFN UHFHLYHU DUH PP HDFK LQ RQH FDVH DQG PP DQG PP IRU WKH RWKHU FDVH 7KH PP LQWHUFRQQHFWLRQ GLVWDQFH LV WKH ORQJHVW WR GDWH 7KH UHVXOWV VKRZ WKDW WKH WZR UHFHLYHUV DUH V\QFKURQL]HG KDYLQJ DQ RXWSXW FORFN VNHZ RI SV DW *+] RU bf 7KH PHDVXUHG SHDNWRSHDN FORFN MLWWHU ZLWK TXLHW VXSSOLHV LV SV DW *+] RU bf 7KHVH UHVXOWV VXJJHVW WKDW D ZLUHOHVV GLVWULEXWLRQ V\VWHP FDQ PHHW WKH VNHZ DQG MLWWHU UHTXLUHPHQWV IRU FORFN GLVWULEXWLRQ V\VWHPV $OO RI WKHVH UHVXOWV GHPRQVWUDWH WKH SODXVLELOLW\ RI DQ LQWUDFKLS ZLUHOHVV LQWHUFRQQHFW V\VWHP FRQWDLQn LQJ LQWHJUDWHG DQWHQQDV DQG &026 WUDQVPLWWHU DQG UHFHLYHU FLUFXLWU\

PAGE 221

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n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

PAGE 222

UHTXLUHV RU PRUH FORFN UHFHLYHUV RSHUDWLQJ DW *+] RU KLJKHU WR EH LPSOHPHQWHG RQFKLS WKLV DGGHG SRZHU GLVVLSDWLRQ E\ WKH UHFHLYHUV LV D FRQFHUQ 7R HYDOXDWH WKH LPSDFW WKHVH UHFHLYHUV KDYH RQ WKH WRWDO SRZHU FRQVXPSWLRQ RI PLFURSURFHVVRUV WKH SRZHU UHTXLUHPHQWV IRU D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP DUH HVWLPDWHG IRU WKH SP JHQHUDn WLRQ DQG WKHQ FRPSDUHG WR WKRVH IRU FRQYHQWLRQDO JOREDO FORFN GLVWULEXWLRQ V\VWHPV 3RZHU &RQVXPSWLRQ &RPSDULVRQ 0HWKRGRORJ\ 7KH HQWLUH FORFN GLVWULEXWLRQ V\VWHP FDQ EH GLYLGHG LQWR D JOREDO V\VWHP DQG D ORFDO V\VWHP 7KH JOREDO GLVWULEXWLRQ V\VWHP LV DVVXPHG WR EH WKH QHWZRUN ZKLFK GHOLYHUV WKH FORFN IURP LWV VRXUFH RIIFKLS RU RQFKLSf WR YDULRXV ORFDWLRQV WKURXJKRXW WKH FKLS 7KH ORFDO GLVWULEXWLRQ V\VWHP LV DVVXPHG WR EH WKDW ZKLFK WDNHV WKH JOREDO FORFN JHQHUDWHV GLIn IHUHQW FORFN SKDVHV DV QHHGHG DQG GLVWULEXWHV WKHVH VLJQDOV WR DOO RI WKH ORFDO FLUFXLWV 7KH SRZHU GLVVLSDWLRQ RI WKH JOREDO GLVWULEXWLRQ V\VWHP FDQ EH PRGHOHG DV G\QDPLF DFFRUGLQJ WR WKH IROORZLQJ IRUPXOD 3' &79GGI f ZKHUH &7 LV WKH WRWDO HTXLYDOHQWf JOREDO FDSDFLWDQFH EHLQJ VZLWFKHG LQFOXGLQJ WKH FDSDFLn WDQFH RI EXIIHUV 9GG LV WKH V\VWHP VXSSO\ YROWDJH DQG I LV WKH FORFN IUHTXHQF\ 7R FRPn SDUH WKH SRZHU UHTXLUHPHQWV EHWZHHQ GLIIHUHQW W\SHV RI JOREDO FORFN GLVWULEXWLRQ V\VWHPV WKH V\VWHP YROWDJH DQG IUHTXHQFLHV DUH DVVXPHG WR EH HTXDO $OVR DQ HTXDO FDSDFLWLYH ORDG UHSUHVHQWLQJ WKH ORFDO FORFN JHQHUDWRUV RU GLVWULEXWLRQ V\VWHP LV DVVXPHG IRU HDFK W\SH RI JOREDO GLVWULEXWLRQ V\VWHP 8QGHU WKHVH DVVXPSWLRQV WKH SRZHU GLVVLSDWLRQ FDQ EH FRQYHUWHG WR FDSDFLWDQFHV DQG WKHVH FDQ EH XVHG WR FRPSDUH WKH SRZHU GLVVLSDWLRQ RI GLIn IHUHQW JOREDO GLVWULEXWLRQ V\VWHPV VLPLODU WR DQ DSSURDFK WDNHQ LQ >5HV@

PAGE 223

)LJXUH VL]H IDFWRU D VWDJH QXPEHU Q Q n/RDG )LJXUH 6FKHPDWLF RI WDSHUHG EXIIHU XVHG LQ JOREDO FORFN GLVWULEXWLRQ (DFK VWDJH LV D WLPHV DV ODUJH DV WKH SUHYLRXV EXIIHU VWDJH UHVXOWLQJ LQ PLQLPXP GHOD\ 7KH WRWDO JOREDO FDSDFLWDQFH FDQ EH DOORFDWHG DPRQJ WKUHH FRPSRQHQWV DV IROORZV &R &Z DQG &/ >)ORE@ 7KHVH FRPSRQHQWV DUH LOOXVWUDWHG LQ )LJXUH &* LV WKH HTXLYn DOHQW FDSDFLWDQFH RI WKH KLJKHVW OHYHO QHWZRUN ZKLFK GHOLYHUV WKH FORFN IURP LWV VRXUFH WR YDULRXV VHFWRUV RU VSLQH ORFDWLRQV GLVWULEXWHG WKURXJKRXW WKH FKLS &* LQFOXGHV WKH WRWDO FDSDFLWDQFH RI WKH ILQDO GULYHU VWDJH KHUHLQ WHUPHG WKH VHFWRU EXIIHUf SOXV DQ\ EXIIHUV OHDGn LQJ XS WR WKH VHFWRU EXIIHU 7KH PXOWLVWDJH EXIIHUV VKRZQ LQ )LJXUH DUH DVVXPHG WR EH WDSHUHG ZKHUH HDFK EXIIHU VWDJH LV D WLPHV DV ODUJH DV WKH SUHYLRXV EXIIHU VWDJH ,W FDQ EH VKRZQ WKDW WKLV UHVXOWV LQ PLQLPXP GHOD\ WKURXJK WKH EXIIHU FKDLQ >0HD@ 7KH WRWDO FDSDFLWDQFH RI DQ 1VWDJH EXIIHU GULYLQJ D FDSDFLWLYH ORDG &MAA FDQ EH HVWLPDWHG DV

PAGE 224

^&/RDG\ f F EXIIHU Q $ VHFRQG FRPSRQHQW WR &7 LV &Z ZKLFK LV WKH FDSDFLWDQFH RI WKH LQWHUFRQQHFWLQJ ZLUHV IRU GHOLYHULQJ WKH FORFN IURP WKH RXWSXW RI WKH VHFWRU EXIIHUV WR WKH ORFDO GLVWULEXWLRQ V\Vn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f ,QLWLDO SP $O N 9 0+] *ULG +WUHH F* FZ &/ &*OREDO &T&Z&O b 7RWDO 3RZHU

PAGE 225

QHWZRUNV QHHG WR EH DGMXVWHG ,Q WKLV ZRUN WKH IRUPHU ZDV FKRVHQ WR NHHS WKH DQDO\VLV VWUDLJKWIRUZDUG 7KLV UHVXOWV LQ VKLIWLQJ VRPH SRZHU IURP WKH JOREDO GLVWULEXWLRQ V\VWHP WR WKH ORFDO GLVWULEXWLRQ V\VWHP UHVXOWLQJ LQ DQ XQGHUHVWLPDWLRQ RI WKH WRWDO JOREDO FDSDFLn WDQFH +RZHYHU VLQFH HDFK FORFN GLVWULEXWLRQ V\VWHP GULYHV DQ LGHQWLFDO FORFN ORDG PHDQn LQJIXO FRPSDULVRQV DUH VWLOO REWDLQHG 7ZR FDVHV DUH XVHG LQ FRPSDULQJ WKHVH FORFN GLVWULEXWLRQ V\VWHPV IRU _LP JHQn HUDWLRQ PLFURSURFHVVRUV &DVH LV IRU DOXPLQXP $Of LQWHUFRQQHFWV DQG FRQYHQWLRQDO GLHOHFWULFV FDVH LV IRU FRSSHU &Xf LQWHUFRQQHFWV DQG ORZ. GLHOHFWULFV 7R REWDLQ FDVH &Z DQG WKH FRPSRQHQWV RI &* DULVLQJ IURP ZLUHV ZHUH VFDOHG E\ DGGLWLRQDO IDFWRUV RI DQG UHSUHVHQWLQJ WKH GHFUHDVHG ZLGWK RI &XOLQHV DV FRPSDUHG WR $OOLQHV IRU FRQVWDQW UHVLVWDQFH OLQHV DQG WKH UHGXFWLRQ LQ IURP WR UHVSHFWLYHO\ &ORFN 'LVWULEXWLRQ 6\VWHPV *ULG%DVHG 6\VWHP 7KH JULGEDVHG V\VWHP VKRZQ LQ )LJXUH Df DQG EDVHG RQ '(&fV >*LH@ FRQVLVWV RI D JOREDO WUHH VXSSO\LQJ WKH FORFN WR GLIIHUHQW VSLQH ORFDn WLRQV DQG WKH FDSDFLWDQFH RI WKLV QHWZRUN LV &* 7KHVH VSLQHV GULYH WKH FORFN JULG ZKLFK KDV D FDSDFLWDQFH &Z IURP WKH IRXU HGJHV RI WKH JULG 7KH ORFDO FORFN JHQHUDWRUV WKHQ WDS RII RI WKH JULG 'XH WR WKH DPRXQW RI ZLULQJ XVHG WR IRUP WKH JULG &Z LV ODUJH &RQVHn TXHQWO\ ODUJH VHFWRU EXIIHUV DUH QHHGHG WKHUHE\ LQFUHDVLQJ &* +7UHH %DVHG 6\VWHP 7KH +WUHH V\VWHP VKRZQ LQ )LJXUH Ef DQG EDVHG RQ ,%0fV 6 >:HE@ FRQVLVWV RI D JOREDO WUHH VXSSO\LQJ WKH FORFN WR GLIIHUHQW VHFWRU EXIIHU ORFDWLRQV (DFK EXIIHU GULYHV D EDODQFHG +WUHH ZKLFK GULYHV WKH ORFDO FORFN JHQHUn DWRUV &Z LV VPDOOHU IRU WKH +WUHH WKHUHIRUH D VPDOOHU VHFWRU EXIIHU LV UHTXLUHG

PAGE 226

)LJXUH ,OOXVWUDWLRQ RI VLQJOH TXDGUDQW RI JOREDO FORFN GLVWULEXWLRQ V\VWHPV IRU Df JULG Ef +WUHH DQG Ff ZLUHOHVV :LUHOHVV 6\VWHP 7KH ZLUHOHVV V\VWHP VKRZQ LQ )LJXUH Ff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n QHQW RI &* LV ]HUR +RZHYHU VLQFH WKH ZLUHOHVV V\VWHP FRQWDLQV FRPSRQHQWV ZLWK VWDWLF SRZHU GLVVLSDWLRQ DQ HTXLYDOHQW JOREDO FDSDFLWDQFH >)ORE@ UHSUHVHQWLQJ WKLV SRZHU GLVn VLSDWLRQ LV QHHGHG 7KH HTXLYDOHQW FDSDFLWDQFH LV GHILQHG DV IROORZV f B 37; 1 f 35; n*HT 9I f ZKHUH 37; DQG 3U[ DUH WKH SRZHU FRQVXPSWLRQV RI WKH WUDQVPLWWHU DQG UHFHLYHU DQG 1 LV WKH WRWDO QXPEHU RI UHFHLYHUV %DVHG RQ PHDVXUHG UHVXOWV IRU WKH SP *+] ZLUHOHVV LQWHUFRQQHFW DQG DVVXPLQJ D 9 VXSSO\ WKH WUDQVPLWWHU FRQVXPHV P: ZKLOH HDFK

PAGE 227

7DEOH 7RWDO JOREDO FDSDFLWDQFH IRU OLP WHFKQRORJ\ &DSDFLWDQFH S)f &DVH SP $O N 9 *+] &DVH SP &X N 9 *+] *ULG +WUHH :LUHOHVV *ULG +WUHH :LUHOHVV F* FZ &/ &*OREDO &J&?Y&O b 7RWDO 3RZHU UHFHLYHU FRQVXPHV P: 7KH HTXLYDOHQW FDSDFLWDQFH FDQ WKHQ EH XVHG WR PDNH D FRPn SDULVRQ WR WKH JULG DQG +WUHH V\VWHPV 5HVXOWV DQG &RQFOXVLRQV IRU 3RZHU &RQVXPSWLRQ 7DEOH VKRZV D EUHDNGRZQ RI WKH JOREDO FDSDFLWLYH ORDGLQJ IRU WKH WKUHH GLVWULn EXWLRQ V\VWHPV IRU FDVHV DQG $OO FDSDFLWDQFH XQLWV DUH LQ SLFRIDUDGV 7KH ILQDO URZ JLYHV WKH SRZHU FRQVXPHG E\ WKH JOREDO FORFN GLVWULEXWLRQ V\VWHPV DV D SHUFHQWDJH RI WRWDO PLFURSURFHVVRU SRZHU 7KLV SHUFHQWDJH UHSUHVHQWV WKH UHODWLYH DPRXQW RI SRZHU GLVVLSDWHG LQ WKH JOREDO V\VWHP WR GULYH D ORDG RI &/ 7R REWDLQ WKHVH QXPEHUV WKH WRWDO SRZHU FRQn VXPSWLRQ SURMHFWHG LQ WKH ,756 LV FRQYHUWHG LQWR DQ HTXLYDOHQW FDSDFLWDQFH XVLQJ f DQG WKHQ FRPSDUHG ZLWK &JOREDL )RU FODULILFDWLRQ WKLV SHUFHQWDJH LV QRW WKH UHODWLYH DPRXQW RI SRZHU GLVVLSDWHG LQ WKH HQWLUH FORFN V\VWHP ZKLFK LV HVWLPDWHG WR EH EHWZHHQ DQG b >*LH@ 7KLV LV EHFDXVH WKH SHUFHQWDJH GRHV QRW LQFOXGH WKH GLVVLSDWLRQ LQ WKH ORFDO FORFN GLVWULEXWLRQ V\VWHP ZKLFK ZLOO EH PXFK JUHDWHU WKDQ WKH GLVVLSDWLRQ LQ WKH JOREDO V\VWHP 7KH UHVXOWV VKRZ WKDW IRU ERWK FDVHV WKH ZLUHOHVV V\VWHP LV FRPSDUDEOH LQ SHUIRUn PDQFH WR WKH +WUHH V\VWHP DQG EHWWHU LQ SHUIRUPDQFH WKDQ WKH JULGEDVHG V\VWHP LQ WHUPV

PAGE 228

RI SRZHU GLVVLSDWLRQ 7KH UHVXOWV DOVR VKRZ WKDW WHFKQRORJ\ GHYHORSPHQWV VXFK DV &X RU ORZ. ZLOO KDYH WKH JUHDWHVW SRVLWLYH LPSDFW RQ V\VWHPV ZKRVH WRWDO HTXLYDOHQW FDSDFLWDQFH LV GRPLQDWHG E\ &Z VXFK DV WKH JULG )LQDOO\ WKH UHVXOWV VKRZ WKDW WKH SRZHU GLVVLSDWHG LQ WKH FORFN UHFHLYHUV JLYHQ E\ &T VKRXOG EH D VPDOO IUDFWLRQ bf RI WKH WRWDO SRZHU GLVn VLSDWHG LQ WKH PLFURSURFHVVRU 7KHVH UHVXOWV VKRZ WKDW SRZHU GLVVLSDWLRQ GRHV QRW LPSRVH OLPLWDWLRQV IRU ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHPV 3URFHVV 9DULDWLRQ 7KH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP FRQVLVWV RI PXOWLSOH FORFN UHFHLYHUV GLVWULEn XWHG WKURXJKRXW WKH PLFURSURFHVVRU 7KHVH UHFHLYHUV ZLOO KDYH WR EH PDWFKHG DFURVV WKH FKLS LQ WHUPV RI JDLQ DQG SKDVH VR WKDW WKH RXWSXW ORFDO FORFN VLJQDOV DUH V\QFKURQL]HG +RZHYHU SURFHVV YDULDWLRQ DQG WHPSHUDWXUH JUDGLHQWV PDNH WKLV PDWFKLQJ GLIILFXOW WKXV WKHUH ZLOO EH D QRQ]HUR FORFN VNHZ GXH WR SURFHVV YDULDWLRQ 7KH HIIHFWV RI SURFHVV YDULDn WLRQ RQ D ORZ QRLVH DPSOLILHU IUHTXHQF\ GLYLGHU DQG FORFN UHFHLYHU ZLOO EH VLPXODWHG LQ WKLV VHFWLRQ XVLQJ 0RQWH &DUOR DQDO\VHV DQG WKH UHVXOWDQW ORFDO FORFN VNHZ ZLOO EH VKRZQ 6LPXODWLRQ 0HWKRGRORJ\ 7ZR ZD\V SURFHVV YDULDWLRQ FDQ EH VLPXODWHG DUH E\ XVLQJ HLWKHU FRPHU PRGHOV RU 0RQWH &DUOR DQDO\VHV 7KH FRPHU PRGHOV SURYLGH D ZRUVWFDVH YDULDWLRQ WKDW FDQ EH H[SHFWHG ORWWRORW UHSUHVHQWLQJ DSSUR[LPDWHO\ D D YDULDWLRQ 7KHVH PRGHOV FDQ EDVLn FDOO\ EH XVHG WR FRQILUP FLUFXLW RSHUDWLRQ DFURVV DOO ZDIHUV :KLOH WKLV LV LPSRUWDQW IRU WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP HYDOXDWLQJ WKH YDULDWLRQ ZLWKLQ D VLQJOH GLH LV LPSRUWDQW DV ZHOO 7KH ZLWKLQGLH YDULDWLRQ LV VLPXODWHG XVLQJ 0RQWH &DUOR DQDO\VHV LQ 63,&( :LWK WKHVH DQDO\VHV VHOHFW PRGHO SDUDPHWHUV DUH GHILQHG E\ D SUREDELOLW\ GHQVLW\ IXQFWLRQ

PAGE 229

3')f DQG IRU HDFK 0RQWH &DUOR LWHUDWLRQ 63,&( UDQGRPO\ VHOHFWV YDOXHV IRU WKH SDUDPn HWHUV DFFRUGLQJ WR WKH 3')V LH WKH YDOXHV DUH PRUH OLNHO\ WR EH FORVHU WR WKH PHDQ WKDQ WR WKH PHDQ Df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f )LQDOO\ WKH UHVLVWDQFH DVVRFLDWHG ZLWK WKH PHWDO VSLn UDOLQGXFWRUV ZDV YDULHG DVVXPLQJ D D YDULDWLRQ RI b ,Q +63,&( WKH SDUDPHWHU VWDWHPHQW LV HLWKHU YDULDEOH $*$866PHDQ YDULDWLRQ VLJf RU *$866PHDQ YDULDWLRQ VLJf 7KH PHDQ DQG YDULDWLRQ DUH VSHFLILHG ZLWK VLJ EHLQJ WKH QXPEHU RI VWDQGDUG GHYLDWLRQV WKH YDULDWLRQ UHSUHVHQWV +HUH $*$866 GHQRWHV DQ DEVRn OXWH JDXVVLDQ ZKHUH WKH YDULDWLRQ LV DQ DEVROXWH TXDQWLW\ DQG *$866 GHQRWHV D UHODWLYH JDXVVLDQ ZKHUH WKH YDULDWLRQ LV ZLWK UHVSHFW WR WKH PHDQ )RU WKH R[LGHWKLFNQHVV 3') H[DPSOH DERYH HTXLYDOHQW VWDWHPHQWV ZRXOG EH WR[ $*$866Q Q f RU WR[ *$866Q f ZKHUH WKH D YDULDWLRQ RI Q LV b RI WKH PHDQYDOXH Q

PAGE 230

)LJXUH 3URFHVV YDULDWLRQ RI VLQJOHHQGHG /1$fV JDLQ 6Mf QRLVH ILJXUH 1)f DQG LQSXW PDWFKLQJ 6Xf REWDLQHG ZLWK 0RQWH &DUOR DQDO\VLV LWHUDWLRQVf /1$ DQG )UHTXHQF\ 'LYLGHU 9DULDWLRQ )LJXUH VKRZV WKH VLPXODWHG JDLQ QRLVH ILJXUH 1)f DQG LQSXWPDWFKLQJ 6Xf YDULDWLRQ IRU D VLQJOHHQGHG ORZ QRLVH DPSOLILHU /1$f IURP D 0RQWH &DUOR VLPXODWLRQ ZLWK LWHUDWLRQV $W *+] WKH JDLQ 6Mf YDULHV E\ G% WKH 1) YDULHV E\ G% DQG WKH 6X YDULHV E\ G% 7KH YDULDWLRQ LQ WKH SKDVH QRW VKRZQf RI 6L LV r UHVXOWLQJ LQ D b VNHZ UHIHUHQFHG WR WKH RXWSXW ORFDO FORFN VLJQDO )LJXUH VKRZV WKH YDULDWLRQ REWDLQHG ZLWK 0RQWH &DUOR VLPXODWLRQV RI WKH IUHTXHQF\ GLYLGHU IRU D P9 DPSOLWXGH *+] LQSXW VLJQDO DW 9GG 9 $W 9 WKH VNHZ LV SV DW WKH ULVLQJ HGJH DQG SV DW WKH IDOOLQJ HGJH 6LQFH WKH RXWSXW FORFN SHULRG LV SV *+]f WKH VNHZ GXH WR SURFHVV YDULDWLRQ LQ WKH IUHTXHQF\ GLYLGHU LV b 7R REWDLQ WKH YDULDELOLW\ LQ WKH VHOIUHVRQDQW IUHTXHQFLHV RI WKH GLYLGHUV D IDVWIRXULHU WUDQVIRUP ZDV WDNHQ RQ WKH RXWSXW VLJQDOV RI WKH ILUVW GLYLGHU ZKHQ LW ZDV

PAGE 231

7LPH QVf )LJXUH 3URFHVV YDULDWLRQ RI IUHTXHQF\ GLYLGHU REWDLQHG ZLWK 0RQWH &DUOR DQDO\VLV LWHUDWLRQVf IRU DQ *+] P9 DPSOLWXGH LQSXW VLJQDO 9GG 9 VHOIRVFLOODWLQJ $ 0+] YDULDWLRQ LQ WKH RXWSXW IUHTXHQF\ RI WKH GLYLGHU ZDV REVHUYHG FRUUHVSRQGLQJ WR D *+] YDULDWLRQ LQ WKH LQSXWUHIHUUHG VHOIRVFLOODWLRQ IUHn TXHQF\ ZKLFK LV VLJQLILFDQW &ORFN 5HFHLYHU 9DULDWLRQ 7KH YDULDELOLW\ RI WKH HQWLUH FORFN UHFHLYHU ZDV DQDO\]HG XVLQJ ERWK VPDOOVLJQDO $&f DQDO\VLV DQG WUDQVLHQW DQDO\VLV 5HIHUULQJ WR )LJXUH OEf WKH UHFHLYHU FRQVLVWV RI D IXOO\GLIIHUHQWLDO /1$ D SDLU RI VRXUFHIROORZHU EXIIHUV DQ IUHTXHQF\ GLYLGHU DQG RXWSXW EXIIHUV )LJXUH Df VKRZV WKH VLPXODWHG JDLQ YDULDWLRQ IURP D 0RQWH &DUOR DQDOn \VLV LWHUDWLRQVf HTXDO WR WKH YROWDJH JDLQ YHUVXV IUHTXHQF\ IURP WKH LQSXW RI WKH UHFHLYHU WR WKH LQSXW RI WKH IUHTXHQF\ GLYLGHU 'XH WR ERWK WKH /1$ GULYLQJ D FDSDFLWLYH ORDG DQG WKH QHJDWLYH LQSXW FRQGXFWDQFH RI WKH VRXUFHIROORZHUV WKH JDLQ LV PXFK ODUJHU WKDQ WKDW RI WKH /1$ GULYLQJ D UHVLVWLYH ORDG )XUWKHUPRUH VLQFH WKH UHVRQDQW IUHTXHQF\ RI WKH /1$ LQ WKH UHFHLYHU LV VHW E\ WKH VPDOO LQSXW FDSDFLWDQFH RI WKH VRXUFHIROORZHU WKH

PAGE 232

)LJXUH 6LPXODWHG SURFHVV YDULDWLRQ RI Df YROWDJH JDLQ DQG Ef SKDVHVKLIW RI UHFHLYHU IURP DQWHQQD LQSXW WR LQSXW RI IUHTXHQF\ GLYLGHU YDULDWLRQ LQ WKH UHVRQDQW IUHTXHQF\ FDQ EH ODUJH 7KH PD[LPXP JDLQ DW *+] YDULHV IURP WR G% ZLWK WKH UHVRQDQW IUHTXHQF\ YDU\LQJ E\ 0+] $V FDQ EH VHHQ WKH JDLQ YDULDWLRQ LV SULPDULO\ GXH WR WKH UHVRQDQW IUHTXHQF\ VKLIWLQJ UDWKHU WKDQ WKH SHDN JDLQ

PAGE 233

VKLIWLQJ 7KLV G% JDLQ YDULDWLRQ LV VHYHUH DQG FDQ UHVXOW LQ FORFN VNHZ GXH WR DPSOLWXGH PLVPDWFK 5HIHUULQJ WR VHFWLRQ )LJXUH D G% DPSOLWXGH YDULDWLRQ FDQ UHVXOW LQ DV PXFK DV SV RI VNHZ DW *+] GLYLGHG E\ RU b RXWSXW FORFN VNHZ $GGLWLRQDOO\ WKH YDULDWLRQ LQ WKH UHVRQDQW IUHTXHQF\ ZLOO FDXVH GLIIHUHQW SKDVHVKLIWV WKURXJK WKH DPSOLILHU DW D JLYHQ LQSXW IUHTXHQF\ )LJXUH Ef VKRZV WKH SKDVHVKLIW WR WKH LQSXW RI WKH GLYLGHU YHUVXV IUHTXHQF\ $W *+] WKH YDULDWLRQ LQ SKDVHVKLIW LV r \LHOGLQJ SV RI RXWSXW VNHZ RU ab RI WKH ORFDO FORFN SHULRG 7KH SKDVH YDULDWLRQ LQFUHDVHV DURXQG WKH UHVRQDQW IUHTXHQF\ WKRXJK $W *+] FRUUHn VSRQGLQJ WR WKH ZRUVWFDVH YDULDWLRQ WKH SKDVH YDULDWLRQ LV r ZKLFK LV b RI WKH LQSXW SHULRG DQG b RI WKH RXWSXW SHULRG LH b GLYLGHG E\ f (YHQ PRUH VHYHUH WKDQ WKH VNHZ GXH WR DPSOLWXGH PLVPDWFK DQG UHVRQDQW IUHn TXHQF\ YDULDWLRQV LV WKH IDFW WKDW WKLV JDLQ YDULDWLRQ FDQ FDXVH WKH VLJQDO DW WKH LQSXW RI WKH GLYLGHU WR EH WRR ORZ IRU WKH GLYLGHU WR ORFN ,Q WKDW FDVH WKH GLYLGHU ZRXOG VHOIRVFLOODWH DQG WKH V\VWHP ZRXOG QRW EH V\QFKURQL]HG 7KH JDLQ YDULDWLRQ LQ )LJXUH Df LV FRPn SRXQGHG E\ WKH YDULDWLRQ LQ WKH LQSXW VHQVLWLYLW\ RI WKH GLYLGHU $V ZDV MXVW PHQWLRQHG WKH LQSXWUHIHUUHG VHOIRVFLOODWLRQ IUHTXHQF\ IL6f FDQ YDU\ E\ *+] 7KXV WKH FRQYHUn VLRQ JDLQ RI WKH GLYLGHU DQG KHQFH WKH JDLQ RI WKH UHFHLYHU ZLOO YDU\ FRQVLGHUDEO\ )LJXUH VKRZV WKH RXWSXW FORFN ZDYHIRUPV IURP WKH 0RQWH &DUOR VLPXODWLRQV LWHUDWLRQVf IRU WKUHH GLIIHUHQW LQSXW SRZHU OHYHOV WR WKH UHFHLYHU 5HDGLQJ WRS WR ERWWRP WKH LQSXW SRZHU OHYHOV DUH G%P G%P DQG G%P $W D G%P LQSXW SRZHUf§)LJXUH Dff§WKH RXWSXW FORFN ZDYHIRUPV DUH DOO DW WKH VDPH IUHTXHQF\ ZKLFK LV *+] GLYLGHG E\ *+] 7KH VNHZ DW WKH ULVLQJ DQG IDOOLQJ HGJHV DUH DQG SV UHVSHFWLYHO\ ZKLFK FRUUHVSRQGV WR D b VNHZ $W D G%P LQSXW SRZHUf§)LJXUH

PAGE 234

EfaWKH PDMRULW\ RI WKH RXWSXW FORFN ZDYHIRUPV DUH DW *+] ZLWK VLPLODU VNHZ WR WKDW MXVW PHQWLRQHG +RZHYHU WKHUH DUH D IHZ LWHUDWLRQV ZKHUH WKH RXWSXW ZDYHIRUPV DUH DW D GLIIHUHQW IUHTXHQF\ VKRZLQJ WKDW WKH GLYLGHU LV VHOIRVFLOODWLQJ 7KLV PHDQV WKDW WKH LQSXW VLJQDO WR WKH GLYLGHU LV QRW ODUJH HQRXJK WR ORFN WKH GLYLGHU )LQDOO\ DW D G%P LQSXW SRZHUf§)LJXUH Fff§YLUWXDO FKDRV LV HYLGHQW ZLWK PRVW RI WKH FORFN ZDYHIRUPV DW GLIIHUHQW IUHTXHQFLHV $W WKLV SRZHU OHYHO WKH V\VWHP KDV IDLOHG &RQFOXVLRQV IRU 3URFHVV 9DULDWLRQ 7DEOH VXPPDUL]HV WKH SURFHVV YDULDWLRQ UHVXOWV 0RVW VLJQLILFDQW DPRQJ WKHVH UHVXOWV DUH WKH SKDVHVKLIWV GXH WR SURFHVV YDULDWLRQ UHVXOWLQJ LQ RXWSXW FORFN VNHZ DQG WKH ODUJH YDULDWLRQ LQ IL62 $OVR WKHVH UHVXOWV VKRZ WKDW DV WKH LQSXW VLJQDO OHYHO GHFUHDVHV WKH UHFHLYHU EHFRPHV PRUH VXVFHSWLEOH WR SURFHVV YDULDWLRQ DQG HYHQWXDOO\ IDLOV DW ORZ LQSXW OHYHOV /DUJHO\ WKLV SUREOHP LV GXH WR WKH LQSXW VLJQDO OHYHO WR WKH GLYLGHU QRW EHLQJ ODUJH HQRXJK WR ORFN WKH GLYLGHU ,Q RWKHU ZRUGV WKLV PHDQV WKDW WKH RYHUDOO UHFHLYHU JDLQ FRQn VLVWLQJ RI WKH DPSOLILHU DQG GLYLGHU FRQYHUVLRQ JDLQ KDV LWV SHDN DW D GLIIHUHQW IUHTXHQF\ 7KHVH UHVXOWV ILUVW PRWLYDWH WKH QHHG IRU VRPH W\SH RI DXWRPDWLF JDLQ FRQWURO $*&f LQ WKH UHFHLYHU 7KH $*& FLUFXLWV FRXOG WXQH WKH UHVRQDQW IUHTXHQF\ RI WKH DPSOLn ILHU DQGRU DGMXVW WKH SHDN JDLQ OHYHO ,Q VR GRLQJ WKH YDULDWLRQ LQ SKDVHVKLIW WKURXJK WKH DPSOLILHU ZRXOG EH FRQWUROOHG DV ZHOO 0RVW LPSRUWDQWO\ WKH IL62 RI WKH LQMHFWLRQORFNHG IUHTXHQF\ GLYLGHU VKRXOG EH WXQDEOH VXFK WKDW WKH LQSXW VLJQDO ZLOO DOZD\V EH DEOH WR ORFN WKH GLYLGHU 7KLV WXQDELOLW\ FRXOG EH REWDLQHG E\ GHVLJQLQJ WKH ILUVW LQMHFWLRQORFNHG GLYLGHU DV D GLIIHUHQWLDO YROWDJHFRQWUROOHG RVFLOODWRU HJ >5DW@f 6HFRQG WKHVH UHVXOWV KDYH EHHQ REWDLQHG ZLWK YHU\ SULPLWLYH ELDVLQJ WHFKQLTXHV 8VLQJ PRUH DGYDQFHG ELDVLQJ WHFKQLTXHV VKRXOG LPSURYH WKH FLUFXLWVf LPPXQLW\ WR SURFHVV YDULDWLRQ

PAGE 235

Df Z R Ef Z R &f Z ff§r R 3LQ G%P 7LPH QVf )LJXUH 2XWSXW FORFN VLJQDOV IURP IUHTXHQF\ GLYLGHU ZLWK SURFHVV YDULDWLRQ ,QSXW SRZHU OHYHOV DUH Df G%P Ef G%P Ff G%P 6\QFKURQL]DWLRQ 3UREDEO\ WKH VLQJOHPRVW LPSRUWDQW FULWHULRQ IRU D JOREDO FORFN GLVWULEXWLRQ V\VWHP LV WKDW DOO RI WKH RXWSXW FORFN VLJQDOV VKRXOG EH V\QFKURQL]HG KDYLQJ ORZ FORFN VNHZ DQG ORZ FORFN MLWWHU 7KH FORFN VNHZ VKRXOG EH OHVV WKDQ b RI WKH FORFN SHULRG ZKLOH WKH SHDNWRSHDN FORFN MLWWHU VKRXOG EH OHVV WKDQ b RI WKH FORFN SHULRG 7KLV VHFWLRQ ZLOO ORRN

PAGE 236

7DEOH 6XPPDU\ RI VLPXODWHG SURFHVV YDULDWLRQ IRU SP UHFHLYHU FLUFXLWV 3DUDPHWHU $EVROXWH 9DULDWLRQ 3HUFHQWDJH /1$ *DLQ G% b ,QSXW 0DWFKLQJ 6MMf G% b 1RLVH )LJXUH G% b 'LYLGHU 2XWSXW 6LJQDOV :KHQ /RFNHG SV b IRXW *+] 'LYLGHU ILVR *+] b LQSXWf 5HFHLYHU *DLQ WR GLYLGHU LQSXWf G% ILQ *+] G% ILQ *+] b ILQ *+] b ILQ *+] 3KDVH6KLIW 'XH 7R $PSOLWXGH 0LVPDWFK r ILQ *+] r ILQ *+] b IRXW *+] b IRXW *+] 5HFHLYHU 3KDVH6KLIW WR GLYLGHU LQSXWf r ILQ *+] r ILQ *+] b IRXW *+] b IRXW *+] 5HFHLYHU 2XWSXW 6LJQDOV SV 3LQ G%P b IRXW *+] 'LYLGHU )DLOV DW 3LQ G%Pf DW WKH PDLQ FRPSRQHQWV ZKLFK FRQWULEXWH WR VNHZ DQG MLWWHU IRU D ZLUHOHVV FORFN GLVWULEXn WLRQ V\VWHP HVWLPDWH WKH FXUUHQWf ZRUVWFDVH VNHZ DQG MLWWHU DQG WKHQ WKHVH HVWLPDWHV FRPSDUH ZLWK WKH PHDVXUHG UHVXOWV IURP &KDSWHU &ORFN 6NHZ )RU D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP WKH PDLQ HOHPHQWV ZKLFK ZLOO FRQWULEXWH WR FORFN VNHZ DUH DV IROORZV SURFHVV YDULDWLRQ DIIHFWLQJ UHFHLYHU JDLQ DQG SKDVHVKLIW WLPHRIIOLJKW GHOD\ PLVPDWFKHV GXH WR WKH UHFHLYHUV EHLQJ GLIIHUHQW GLVWDQFHV DZD\ IURP WKH WUDQVPLWWHU DPSOLWXGH PLVPDWFKHV DW WKH LQSXW RI WKH GLYLGHU FDXVLQJ $0WR30 FRQn YHUVLRQ VHFWLRQ f LQWHUIHUHQFH VWUXFWXUHV ORFDWHG EHWZHHQ WKH WUDQVPLWWLQJ DQG UHFHLYn LQJ DQWHQQDV DIIHFWLQJ DQWHQQD WUDQVPLVVLRQ JDLQ LPSHGDQFH DQG SKDVHVKLIW DQG ILQDOO\ PLVPDWFKHV EHWZHHQ WKH ORDGV DW WKH FORFN UHFHLYHU RXWSXWV 7KH ZRUVWFDVH VNHZ FDQ EH HVWLPDWHG DV WKH VXP RI WKH PHDQ VNHZ DQG VWDQGDUG GHYLDWLRQV RI WKH UDQGRP VNHZ ZKHUH WKH YDULDQFHV Df RI DOO RI UDQGRP VNHZV ZLOO DGG WR \LHOG WKH WRWDO VNHZ YDULDQFH

PAGE 237

7KH QRQUDQGRP RU V\VWHPDWLF VNHZ FRPSRQHQW ZLOO EH GXH WR WLPHRIIOLJKW 72)f GHOD\ PLVPDWFKHV 6LQFH WKH UHFHLYHU ORFDWLRQV FDQQRW EH SUHFLVHO\ NQRZQ WKH HYHQO\VSDFHG JULG RI UHFHLYHUV LV PRUH FRQFHSWXDO WKDQ SUDFWLFDOf WKHUH FDQ EH D VLWXDn WLRQ ZKHUH WKH 72) PLVPDWFK LV KDOI RI WKH LQSXW FORFN SHULRG )RU H[DPSOH WKH VSHHG RI OLJKW LQ D VLOLFRQVXEVWUDWH SURSDJDWLRQ PHGLXP LV [O PV ,I WKH GLVWDQFH EHWZHHQ WKH WUDQVPLWWHU DQG WZR UHFHLYHUV GLIIHUV E\ PP SOXV DQ\ LQWHJHU ZDYHOHQJWK Vc PP DW *+]f WKHQ WKH 72) GHOD\ PLVPDWFK LV SV ZKLFK LV KDOI RI WKH SHULRG RI D *+] LQSXW VLJQDO 7KHUHIRUH WKH ZRUVWFDVH 72) PLVPDWFK LV b ZLWK UHVSHFW WR WKH LQSXW VLJQDO UHVXOWLQJ LQ b RI VNHZ ZLWK UHVSHFW WR WKH RXWSXW VLJQDO 5DQGRP VNHZ FRPSRQHQWV ZLOO EH GXH WR SURFHVV YDULDWLRQ DPSOLWXGH PLVPDWFKHV DQG ORDG PLVPDWFKHV 7KH HIIHFWV RI SURFHVV YDULDWLRQ RQ VNHZ ZHUH SUHVHQWHG LQ WKH SUHn YLRXV VHFWLRQ 7KH UHVXOWV VKRZ WKDW IRU WKH _OP UHFHLYHU WKH VLPXODWHG VNHZ GXH WR SURFHVV YDULDWLRQ LV b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a G% 7KXV WKHUH LV D PD[LPXP RI G% RI DPSOLWXGH PLVPDWFK DW WKH LQSXW RI WKH GLYLGHU $SSO\LQJ f WKLV UHVXOWV LQ b RI VNHZ 6LQFH PXOWLSOH SDWKV H[LVW IURP WKH WUDQVPLWWLQJ WR WKH UHFHLYLQJ DQWHQQD WKH HIIHFWLYH GLHOHFWULF FRQVWDQW DQG KHQFH VSHHG RI OLJKW ZLOO EH EHWZHHQ WKDW RI VLOLFRQ DQG 62

PAGE 238

ZLWK UHVSHFW WR WKH RXWSXW VLJQDO )LQDOO\ PLVPDWFKHV LQ WKH ORDG FDSDFLWDQFH RI WKH FORFN UHFHLYHUV ZLOO JHQHUDWH FORFN VNHZ 7KHVH FDSDFLWDQFHV DUH DVVXPHG WR EH PDWFKHG WR ZLWKLQ b WKXV WKH VNHZ GXH WR WKLV PLVPDWFK LV DOVR b 7KH WRWDO ZRUVWFDVH HVWLPDWHG VNHZ FDQ QRZ EH REWDLQHG 6LQFH HDFK RI WKH UDQn GRP VNHZ FRPSRQHQWV DUH LQGHSHQGHQW WKH LQGLYLGXDO YDULDQFHV FDQ EH VXPPHG $OVR VLQFH WKH QXPEHUV OLVWHG LQ WKH SUHYLRXV SDUDJUDSK UHSUHVHQW J YDULDWLRQV WKH WRWDO J YDULDWLRQ LV HTXDO WR D b 7KHUHIRUH WKH ZRUVWFDVH XQFRUUHFWHG VNHZ IRU D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP LV HVWLPDWHG WR EH WKH PHDQ VNHZ bf SOXV WKH J UDQGRP VNHZ bf UHVXOWLQJ LQ b VNHZ ZLWK UHVSHFW WR WKH RXWSXW VLJQDO 7KLV VNHZ HVWLPDWLRQ DVVXPHV WKDW WKHUH LV QR $*& LQ WKH UHFHLYHU :LWK $*& WKH VNHZ GXH WR DPSOLWXGH PLVPDWFK ZRXOG EH HOLPLQDWHG DQG WKH VNHZ GXH WR SURFHVV YDULDWLRQ ZRXOG EH UHGXFHG $VVXPLQJ WKDW $*& LV XVHG WKH ZRUVWFDVH VNHZ ZRXOG EH OHVV WKDQ b )LQDOO\ WKH XVH RI PRUH VRSKLVWLFDWHG ELDVLQJ WHFKQLTXHV VKRXOG UHGXFH WKH SUR FHVVLQGXFHG VNHZ IXUWKHU $V ZDV VKRZQ LQ &KDSWHU SURJUDPPDEOH GLYLGHUV FDQ EH XVHG WR VWDUWXS WKH UHFHLYHU LQ D GLIIHUHQW VWDWH DQG WR UHGXFH FORFN VNHZ 7KH W\SHV RI FORFN VNHZ WKDW FDQ EH FRUUHFWHG IRU LQFOXGH ERWK V\VWHPDWLF DQG UDQGRP VNHZ VLQFH ERWK DUH WLPHLQYDULDQW )RU DQ GLYLGHU WKH VNHZ ZLOO EH OLPLWHG WR XQGHU saa RU b 7KXV WKH ZRUVWFDVH VNHZ XVLQJ D SURJUDPPDEOH GLYLGHU LV b ZKLFK PHHWV WKH V\VWHP VSHFLILFDWLRQ RI OHVV WKDQ b RI VNHZ 1RWH KRZHYHU WKDW WR FRUUHFW IRU WKH VNHZ XVLQJ SURJUDPPDEOH GLYLGHUV UHTXLUHV D VWDUWXS PHWKRGRORJ\ IRU WKH FORFN UHFHLYHUV ZKLFK ZRXOG HVWLPDWH WKH XQFRUUHFWHG VNHZ XVLQJ D SKDVH GHWHFWRU PRGLI\ WKH VWDWHV RI WKH UHFHLYHU DFFRUGLQJO\ DQG WKHQ V\QFKURQRXVO\ UHOHDVH WKH FORFN UHFHLYHUV IURP WKH LQLWLDOL]HG VWDWH UHIHU WR VHFWLRQ

PAGE 239

f $GGLWLRQDOO\ WKLV VNHZ FRUUHFWLRQ WHFKQLTXH LV VWDWLF KHQFH WKH UHFHLYHUV FRXOG DFFXPXODWH SKDVH HUURUV LI IRU H[DPSOH D VLQJOH FORFN SXOVH LV PLVVHG GXH WR LQWHUIHUHQFH 0RUH ZRUN LV UHTXLUHG WR DQDO\]H WKLV SKDVHHUURU DFFXPXODWLRQ ,Q &KDSWHU WKH PHDVXUHG VNHZ EHWZHHQ WZR UHFHLYHUV ZLWK SURJUDPPDEOH GLYLGn HUV ZDV b 7KLV VNHZ LQFOXGHG SURFHVV YDULDWLRQ 72) GHOD\ PLVPDWFKHV DPSOLWXGH PLVPDWFK DQG GLIIHUHQW LQWHUIHUHQFH VWUXFWXUHV $V H[SHFWHG WKH PHDVXUHG VNHZ LV OHVV WKDQ b 7KHVH UHVXOWV VKRZ WKDW WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP VKRXOG EH DEOH WR PHHW WKH VNHZ UHTXLUHPHQWV RI JOREDO FORFN GLVWULEXWLRQ V\VWHPV +RZHYHU DGGLWLRQDO ZRUN LV QHHGHG WR SHUIHFW WKH VWDUWXS WHFKQLTXH IRU WKH SURJUDPPDEOH GLYLGHUV DQG FRUn UHFW IRU WKH VWDWHGHSHQGHQW LQLWLDOL]DWLRQ IDLOXUH RI WKH GLYLGHUV GHVFULEHG LQ f 7KLV VWDUWXS PHWKRGRORJ\ DSSHDUV WR EH WKH XOWLPDWH NH\ IRU GHPRQVWUDWLQJ IHDVLELOLW\ LQ WHUPV RI FORFN VNHZ IRU D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP &ORFN -LWWHU 7KH PDLQ HOHPHQWV FRQWULEXWLQJ WR FORFN MLWWHU IRU D ZLUHOHVV FORFN GLVWULEXWLRQ V\Vn WHP DUH DV IROORZV WKHUPDO QRLVH UHFHLYHG E\ WKH DQWHQQD DQG JHQHUDWHG LQ WKH UHFHLYHU DQG DQWHQQD VZLWFKLQJ QRLVH RU GLJLWDO LQWHUIHUHQFH UHFHLYHG E\ WKH DQWHQQD SRZHUVXSSO\ QRLVH DQG SKDVH QRLVH IURP WKH SKDVHORFNHG ORRS 3//f ZKLFK V\QWKHVL]HV WKH JOREDO FORFN VLJQDO LQ WKH WUDQVPLWWHU 6LQFH WKHVH VRXUFHV RI MLWWHU DUH DOO LQGHSHQGHQW WKHLU YDULn DQFHV DGG $V ZDV VKRZQ LQ &KDSWHU WKHUPDO QRLVH DQG LQWHUIHUHQFH GHJUDGH WKH VLJ QDOWRQRLVH UDWLR 615f DW WKH LQSXW RI WKH IUHTXHQF\ GLYLGHU UHVXOWLQJ LQ MLWWHU DW WKH RXWSXW RI WKH GLYLGHU JLYHQ E\ ff 1RLVH RQ WKH SRZHU VXSSOLHV ZLOO WUDQVODWH LQWR QRLVH RQ WKH VLJQDOV WKHPVHOYHV GHSHQGLQJ RQ WKH SRZHUVXSSO\UHMHFWLRQUDWLR RI WKH

PAGE 240

FLUFXLW )LQDOO\ MLWWHU LQ WKH VRXUFH RU 3// ZLOO FRUUHVSRQG WR MLWWHU RQ WKH RXWSXW VLJQDO ZKHUH WKH MLWWHU DV D SHUFHQWDJH RI WKH SHULRG LV GLYLGHG E\ WKURXJK WKH FORFN UHFHLYHU (VWLPDWLRQV IRU HDFK RI WKHVH MLWWHU FRPSRQHQWV ZLOO QRZ EH SUHVHQWHG )LUVW WKH MLWWHU GXH WR WKHUPDO QRLVH KDV EHHQ VLPXODWHG ZLWK 63,&( $V ZDV VKRZQ LQ VHFWLRQ WKH VLPXODWHG 615 GXH WR WKHUPDO QRLVH ZDV G% UHVXOWLQJ LQ b RI SHDN MLWWHU 7KLV VLPXODWLRQ DVVXPHG DQ LQSXW WKHUPDO QRLVH RI N7 6HFRQG WR HVWLPDWH WKH DPRXQW RI LQWHUn IHUHQFH DW WKH UHFHLYHU LQSXW GLJLWDO WHVW FLUFXLWV RU QRLVH JHQHUDWRUV KDYH EHHQ SODFHG YHU\ FORVH WR DQWHQQDV >%UDY22D %UDY22E@ DOORZLQJ WKH SRZHU VSHFWUDO GHQVLW\ WR EH REWDLQHG 0HDVXUHPHQWV DQG VLPXODWLRQV LQ >%UDY22D %UDY22E@ VKRZ WKDW WKH WRWDO SRZHU GXH WR GLJLWDO VZLWFKLQJ QRLVH LV G%P ,I WKH LQSXW VLJQDO OHYHO LV DW WKH GHVLUHG UHFHLYHU VHQVLn WLYLW\ RI G%P WKHQ WKH VLJQDOWRLQWHUIHUHQFH UDWLR 6,5f LV G% UHVXOWLQJ LQ b RI SHDN MLWWHU 7KLUG VLQFH D &026 3// KDV QRW EHHQ EXLOW DW *+] WKH SHDN MLWWHU RI WKH 3// DW *+] LV DVVXPHG WR EH b &RPSDUHG WR UHVXOWV IRXQG LQ OLWHUDWXUH >
PAGE 241

GLVWULEXWLRQ V\VWHP VKRXOG EH LQGHSHQGHQW RI WKH GLJLWDO VXSSOLHV (YHQ VWLOO QRLVH ZLOO FRXSOH WR WKHVH VXSSOLHV IURP ERWK WKH GLJLWDO FLUFXLWV DQG WKH VZLWFKLQJ RI WKH UHFHLYHUV 6HFRQG WKH DPRXQW RI VZLWFKLQJ QRLVH ZLOO PRVW OLNHO\ EH ODUJHU WKDQ WKH G%P PHDn VXUHG LQ >%UDY22D %UDY22E@ ZKLFK ZDV IRU RQO\ RQH QRLVH JHQHUDWLQJ FLUFXLWf ,I WKH DPRXQW RI QRLVH LV TXDGUXSOHGf§FRUUHVSRQGLQJ WR QRLVH JHQHUDWRUV RQ DOO IRXU VLGHV RI WKH UHFHLYHUf§WKHQ WKH SRZHU ZRXOG LQFUHDVH WR G%P WKH 6,5 ZRXOG GHFUHDVH WR G% WKH SHDN MLWWHU GXH WR LQWHUIHUHQFH ZRXOG LQFUHDVH WR b DQG WKH WRWDO SHDN MLWWHU ZRXOG LQFUHDVH WR b 7KLV ZRXOG RQO\ OHDYH b RI SHDNMLWWHU PDUJLQ IRU WKH QRLV\ VXSSOLHV ,Q &KDSWHU WKH PHDVXUHG SHDNWRSHDN MLWWHU ZLWK D TXLHW VXSSO\ DQG ZLWKRXW DQ\ GLJLWDO VZLWFKLQJ QRLVH ZDV b 7KLV MLWWHU LV GXH WR WKHUPDO QRLVH FRUUHVSRQGLQJ WR D G% 615 $V FDQ EH VHHQ WKLV DJUHHV YHU\ ZHOO ZLWK WKH HVWLPDWHG SHDN MLWWHU IURP WKHUPDO QRLVH RI b YDOLGDWLQJ ERWK WKH VLPXODWHG 615 DQG WKH 615MLWWHU UHODWLRQVKLS 7KHUHIRUH WKHVH UHVXOWV VXJJHVW WKDW WKH SHDNWRSHDN MLWWHU GXH WR WKHUPDO QRLVH LQWHUIHUn HQFH DQG VRXUFH MLWWHU VKRXOG EH DSSUR[LPDWHO\ b PHHWLQJ WKH V\VWHP VSHFLILFDWLRQ RI b SHDNWRSHDN MLWWHU +RZHYHU PRUH ZRUN LV UHTXLUHG WR TXDQWLI\ KRZ PXFK MLWWHU LV LQWURGXFHG E\ QRLV\ VXSSOLHV $OVR WKH MLWWHU IURP D 3// DW *+] VKRXOG EH TXDQWLILHG DQG FKHFNHG DJDLQVW WKH DVVXPHG YDOXH ,I WKH MLWWHU VSHFLILFDWLRQ LV H[FHHGHG E\ WKH QRLV\ VXSSOLHV DQG WKH 3// WKHQ WKH UHFHLYHU FLUFXLWU\ FRXOG EH RSWLPL]HG WR GHFUHDVH WKH MLWWHU GXH WR WKHUPDO QRLVH DQG LQWHUIHUHQFH 2QH SRWHQWLDO LPSURYHPHQW ZRXOG EH WR XVH D GLIIHUn HQWLDO /&RVFLOODWRU DV WKH ILUVW LQMHFWLRQORFNHG GLYLGHU >5DW@ 'XH WR WKH KLJKHU 4 RI WKLV FLUFXLW DV FRPSDUHG WR WKH ORZ 4 RI D ULQJ RVFLOODWRU WKH RXWSXW SKDVH QRLVH DQG KHQFH MLWWHU ZLOO EH GHFUHDVHG 7KLV GHFUHDVH LV GXH WR WKH ILOWHULQJ SURYLGHG E\ WKH LQMHFWLRQORFNLQJ ZKHUH WKH RXWSXW SKDVH FDQ RQO\ IROORZ WKH LQSXW SKDVH RYHU D OLPLWHG ORFNLQJ UDQJH DV VKRZQ LQ $SSHQGL[ (

PAGE 242

&RQFOXVLRQV IRU 6\QFKURQL]DWLRQ 7KLV VHFWLRQ KDV HVWLPDWHG WKH FXUUHQWf ZRUVWFDVH FORFN VNHZ DQG MLWWHU WKDW FDQ EH REWDLQHG ZLWK D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP 7KH ZRUVWFDVH XQFRUUHFWHG VNHZ GXH WR SURFHVV YDULDWLRQ 72) PLVPDWFK DPSOLWXGH PLVPDWFK IURP GLIIHUHQW DQWHQQD JDLQVf DQG PLVPDWFKHV LQ WKH RXWSXW ORDG LV HVWLPDWHG WR EH b 8VLQJ D SURJUDPPDn EOH GLYLGHU WKRXJK ZLOO UHGXFH WKH VNHZ WR XQGHU b 7KLV QHFHVVLWDWHV D VWDUWXS PHWKRGRORJ\ WR SURJUDP DQG UHOHDVH WKH GLYLGHUV V\QFKURQRXVO\ 7KH ZRUVWFDVH SHDNWRSHDN MLWWHU GXH WR WKHUPDO QRLVH VZLWFKLQJ QRLVH GLJLWDO LQWHUIHUHQFHf DQG SKDVH QRLVH RQ WKH JOREDO FORFN VLJQDO LV HVWLPDWHG WR EH b +RZHYHU PRUH ZRUN LV UHTXLUHG WR HYDOXDWH WKH MLWWHU LQWURGXFHG E\ SRZHUVXSSO\ QRLVH DQG WKH MLWWHU RI D *+] 3// 0HDVXUHG UHVXOWV RI D *+] GRXEOHUHFHLYHU ZLUHOHVV LQWHUFRQQHFW KDYH VKRZQ D FORFN VNHZ RI b DQG D SHDNWRSHDN FORFN MLWWHU RI b 7KHVH UHVXOWV KHOS WR FRQILUP WKH HVWLPDWLRQV PDGH LQ WKLV VHFWLRQ VKRZLQJ WKDW D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP VKRXOG EH DEOH WR PHHW WKH VNHZ DQG MLWWHU UHTXLUHPHQWV IRU PLFURSURFHVVRUV +RZHYHU PRUH ZRUN LV UHTXLUHG WR YDOLGDWH WKHVH HVWLPDWLRQV DQG WR GHPRQVWUDWH WKH FORFN VNHZ DQG MLWWHU LQ D ZRUNLQJ PLFURSURFHVVRU RU FORVH SUR[LPLW\ WKHUHRIf /DWHQF\ RI SP :LUHOHVV ,QWHUFRQQHFW 7KH LQFUHDVLQJ ODWHQF\ RI FRQYHQWLRQDO LQWHUFRQQHFWV LV D SULPDU\ PRWLYDWLRQ IRU LQYHVWLJDWLQJ DOWHUQDWLYH LQWHUFRQQHFW V\VWHPV 7KHUHIRUH WKH ODWHQF\ RI WKH LP ZLUHn OHVV LQWHUFRQQHFW KDV EHHQ FDOFXODWHG 7KH ODWHQF\ ZLOO LQFOXGH 72) GHOD\ DQG GHOD\ WKURXJK WKH FORFN UHFHLYHU ,I WKH VLOLFRQVXEVWUDWH LV DVVXPHG DV WKH SURSDJDWLRQ PHGLXP ZKLFK ZRXOG KDYH WKH ZRUVWFDVH ODWHQF\ GXH WR KLJKHU GLHOHFWULF FRQVWDQWf WKHQ WKH 72) GHOD\ IRU D FP SURSDJDWLRQ GLVWDQFH LV SV WKLV GLVWDQFH LV DVVXPHG IRU WKH QP

PAGE 243

WHFKQRORJ\ JHQHUDWLRQ ZKLFK KDV D *+] JOREDO FORFN DQG D PP DUHD >6,$@f§IRU DQ HYHQO\VSDFHG JULG RI UHFHLYHUV WKH ORQJHVW GLVWDQFH IURP WKH FHQWHU WR WKH FRPHU UHFHLYHU LV FPf 7KH ODWHQF\ WKURXJK WKH FORFN UHFHLYHU ZDV VLPXODWHG ZLWK 63,&( KDYLQJ WKH IROORZLQJ FRPSRQHQWV /1$ DQG VRXUFHIROORZHU GHOD\ SV ILUVW GLYLGHU GHOD\ SV VHFRQG GLYLGHU GHOD\ SV DQG WKLUG GLYLGHU GHOD\ SV 7KXV WKH WRWDO ODWHQF\ WKURXJK WKH _LP UHFHLYHU LV SV DQG WKH ZLUHOHVV LQWHUFRQn QHFW KDV D WRWDO ODWHQF\ IURP WKH WUDQVPLWWHU WR WKH XQEXIIHUHGf RXWSXW RI WKH UHFHLYHU RI DSSUR[LPDWHO\ SV 7KLV FRUUHVSRQGV WR b RI D *+] RXWSXW ZDYHIRUP 7KH ODWHQF\ WKURXJK WKH RXWSXW EXIIHU ZDV QRW LQFOXGHG VLQFH WKLV ZRXOG GHSHQG RQ WKH ORDG WKDW WKH UHFHLYHU LV GULYLQJ $OVR WKH GHOD\ WKURXJK WKLV EXIIHU ZRXOG EH WKH VDPH IRU DQ\ LQWHUFRQQHFW V\VWHP VLQFH WKH ORFDO FORFN ORDGV DUH HTXDO $ URXJK HVWLPDWH RI WKH FORFN VNHZ FRXOG EH DSSUR[LPDWHO\ b RI WKH ODWHQF\ ZKLFK LV SV RU b RI WKH RXWSXW FORFN SHULRG 7KLV LV OHVV WKDQ WKH HVWLPDWHG ZRUVWFDVH VNHZ RI b DQG LV ZLWKLQ WKH DOORZDEOH VNHZ UDQJH RI b 7KH ODWHQF\ WKURXJK WKH LQWHUFRQQHFW LV VWLOO IDLUO\ ODUJH :KDW PDNHV ZLUHOHVV LQWHUFRQQHFWV DWWUDFWLYH WKRXJK LV WKDW WKH ODWHQF\ WKURXJK WKH UHFHLYHU ZLOO GHFUHDVH ZLWK VXEVHTXHQW WHFKQRORJ\ JHQHUDWLRQV ZKHUHDV WKH ODWHQF\ WKURXJK FRQYHQWLRQDO LQWHUFRQn QHFW FRQWLQXHV WR LQFUHDVH UHIHU WR )LJXUH f )RU H[DPSOH VFDOLQJ IURP WKH QP WR WKH QP JHQHUDWLRQ WKH JDWH GHOD\ GHFUHDVHV E\ b ZKHUHDV WKH JOREDO LQWHUFRQQHFW GHOD\ LQFUHDVHV GUDPDWLFDOO\ >6,$ %RK@ 7KH PD[LPXP LQWHUFRQQHFW GLVWDQFH LQFUHDVHV DFFRUGLQJ WR WKH VTXDUHURRW RI WKH DUHD LQFUHDVH RU IURP FP WR FP bf 7KH 72) GHOD\ WKHUHIRUH LQFUHDVHV E\ b ZKLOH WKH GHOD\ WKURXJK WKH UHFHLYHU GHFUHDVHV E\ b 7KXV WKH ZLUHOHVV LQWHUFRQQHFW ODWHQF\ ZLOO GHFUHDVH IURP WR SVf§D b

PAGE 244

GHFUHDVH 7KLV GHFUHDVH ZLOO EH OHVVHQHG ZKHQ WKH RXWSXW EXIIHU ODWHQF\ LV LQFOXGHG ZKLFK ZLOO EH LQFUHDVLQJ ZLWK HDFK WHFKQRORJ\ JHQHUDWLRQ +RZHYHU WKLV H[DPSOH VKRZV WKDW WKH ZLUHOHVV LQWHUFRQQHFW ODWHQF\ LV GHFUHDVLQJ ZLWK WHFKQRORJ\ JHQHUDWLRQ ZKHUHDV WKH FRQn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f\ r\ DQG PP UHVSHFWLYHO\ (DFK DQWHQQD RFFXSLHV PP )RU D ZLUHOHVV FORFN GLVn WULEXWLRQ V\VWHP FRQVLVWLQJ RI RQH WUDQVPLWWHU UHFHLYHUV DQG DQWHQQDV WKH WRWDO DUHD FRQVXPHG LV PP 7KH ,756 SURMHFWV WKH GLH VL]H RI D SP JHQHUDWLRQ PLFURSURFHV VRU WR EH PP 7KXV ab RI WKH PLFURSURFHVVRU DUHD ZLOO EH FRQVXPHG E\ WKH ZLUHn OHVV FORFN GLVWULEXWLRQ V\VWHP $SSUR[LPDWHO\ KDOI RI WKLV DUHD LV GXH WR WKH PP ORQJ DQWHQQDV WKXV PRUH DUHDHIILFLHQW DQWHQQDV ZRXOG EH GHVLUDEOH 7KH DUHD FRQVXPHG E\ WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP KDV WR EH ZHLJKHG DJDLQVW WKH PHWDO DUHD VDYHG 6LQFH WKH JOREDO FORFN WUHH ZLOO EH HOLPLQDWHG WKHUH ZLOO EH OHVV DUHD FRQVXPHG E\ URXWLQJ JOREDO VLJQDOV 7KLV ZLOO WKHQ IUHH XS PRUH RI WKH WRSOHYHO PHWDO IRU WKH IXQFWLRQDO EORFNV DFURVV WKH PLFURSURFHVVRU $V D UHVXOW WKHLU DUHD FRXOG

PAGE 245

SRWHQWLDOO\ EH GHFUHDVHG 7KXV FRVWIXQFWLRQV IRU ERWK WKH DFWLYH DUHD DQG WKH PHWDO DUHD VKRXOG EH GHYHORSHG WR HYDOXDWH WKH WRWDO DUHD FRVW RI D ZLUHOHVV LQWHUFRQQHFW V\VWHP $QRWKHU LQWDQJLEOH IHDVLELOLW\ LWHP LV WKH DELOLW\ WR DFFXUDWHO\ YHULI\ LI WKH V\VWHP ZLOO IXQFWLRQ SURSHUO\ NQRZQ DV GHVLJQ YHULILFDWLRQ :LWK DQ\ JOREDO FORFN GLVWULEXWLRQ V\VWHP LW LV LPSHUDWLYH WR PHHW WKH FORFN UHTXLUHPHQWV ZLWK fILUVW VLOLFRQf )RU FRQYHQn WLRQDO LQWHUFRQQHFW V\VWHPV VRSKLVWLFDWHG FRPSXWHU DLGHG GHVLJQ &$'f WRROV KDYH EHHQ GHYHORSHG WR PRGHO WKH WUDQVPLVVLRQOLQH HIIHFWV H[WUDFW WKH SDUDVLWLFV DVVRFLDWHG ZLWK D OLQH DQG PRGHO FXUUHQW UHWXUQSDWKV )RU D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP &$' WRROV ZRXOG DOVR KDYH WR EH GHYHORSHG RU DW OHDVW LPSURYHG WR PRGHO WKH DQWHQQD FKDUDFWHULVn WLFV WKH HIIHFWV RI PHWDO LQWHUIHUHQFH VWUXFWXUHV EHWZHHQ WKH DQWHQQDV WKH GLJLWDO VZLWFKLQJ QRLVH DQG WKH 5) SHUIRUPDQFH RI WKH FLUFXLWV UHTXLULQJ DFFXUDWH LQGXFWRU DQG WUDQVLVWRU PRGHOOLQJf :KLOH WKLV KXUGOH LV QRW LQVXUPRXQWDEOH LW LV TXLWH ODUJH SRLQWLQJ RXW WKH QHHG IRU DGGLWLRQDO UHVHDUFK &RQFOXVLRQV DQG )XWXUH :RUN )HDVLELOLW\ 6XPPDU\ 7KH IHDVLELOLW\ RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP KDV EHHQ HYDOXDWHG \LHOGLQJ WKH IROORZLQJ UHVXOWV )LUVW WKH SRZHU FRQVXPSWLRQ RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP LV VLPLODU WR WKDW RI FRQYHQWLRQDO V\VWHPV ZKHUH WKH FORFN UHFHLYHUV DQG WUDQVPLWWHU ZLOO FRQVXPH OHVV WKDQ b RI WKH V\VWHP SRZHU 6HFRQG WKH FORFN VNHZ RI WKH ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP ZLOO EH OHVV WKDQ b ZKLOH WKH SHDNWRSHDN FORFN MLWWHU VKRXOG EH OHVV WKDQ b DQG ERWK RI WKHVH QXPEHUV KDYH EHHQ YHULILHG ZLWK PHDVXUHPHQWV 7KXV D ZLUHOHVV V\VWHP VKRXOG EH DEOH WR PHHW WKH VNHZ DQG MLWWHU UHTXLUHPHQWV IRU FORFN GLVWULEXn WLRQ V\VWHPV 7KLUG SURFHVV YDULDWLRQ KDV EHHQ VKRZQ WR DIIHFW ERWK WKH VNHZ DQG WKH EDVLF

PAGE 246

RSHUDWLRQ ZKHUH DW ORZ LQSXW SRZHU OHYHOV WKH V\VWHP GRHV QRW RSHUDWH DFURVV DOO SURFHVV YDULDWLRQV 7KLV QHFHVVLWDWHV WKH QHHG WR FRQWURO WKH JDLQ LQ WKH UHFHLYHU WKURXJK DGMXVWLQJ WKH UHVRQDQW IUHTXHQF\ RI WKH DPSOLILHU DQG DGMXVWLQJ WKH VHOIUHVRQDQW IUHTXHQF\ RI WKH LQMHFWLRQORFNHG GLYLGHU )RXUWK WKH ODWHQF\ RI WKH V\VWHP LV SV IRU D *+] SP ZLUHOHVV LQWHUFRQQHFW DQG IXWXUH WHFKQRORJ\ VFDOLQJ ZLOO EHQHILW ZLUHOHVV LQWHUn FRQQHFW ODWHQF\ RYHU FRQYHQWLRQDO LQWHUFRQQHFW ODWHQF\ )LIWK WKH ZLUHOHVV FORFN GLVWULEXn WLRQ V\VWHP ZLOO FRQVXPH DSSUR[LPDWHO\ b RI WKH SURMHFWHG PLFURSURFHVVRU DUHD KRZHYHU WKLV LV RIIVHW E\ WKH WRSOHYHO PHWDO DUHD WKDW ZLOO EH FRQVHUYHG 6L[WK GHVLJQ YHUn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n YHQWLRQDO V\VWHPV ZLWK SRWHQWLDO FRVWV RI DGGHG DUHD DQG PRUH GLIILFXOW GHVLJQ YHULILFDn WLRQ 7KH EHQHILWV RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP DV GLVFXVVHG LQ &KDSWHU FDQ EH VXPPDUL]HG DV EHLQJ DEOH WR SURYLGH KLJK IUHTXHQF\ FORFN VLJQDOV ZLWK OLWWOH WR QR GLVSHUn VLRQ RYHU ODUJH GLVWDQFHV ZKHUH WKH ODWHQF\ RI WKH LQWHUFRQQHFW LV LPSURYLQJ ZLWK WHFKQROn RJ\ 7KLV VKRXOG LQFUHDVH WKH PD[LPXP FORFN IUHTXHQF\ RI PLFURSURFHVVRUV $GGLWLRQDOO\ WKHVH EHQHILWV FDQ EH REWDLQHG XVLQJ FRQYHQWLRQDO &026 WHFKQRORJ\ 7KXV ZKLOH WKHUH LV D FLUFXLW RYHUKHDG IRU D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP VXFK D V\VWHP FRXOG EH XVHG WR SURYLGH JOREDO LQWHUFRQQHFWV IRU PLFURSURFHVVRUV RSHUDWLQJ ZHOO DERYH 0+] ,Q

PAGE 247

DGGLWLRQ WKH XSSHU IUHTXHQF\ OLPLW LV VHW E\ KRZ IDVW WKH 5) FLUFXLWV FDQ RSHUDWH UDWKHU WKDQ KRZ IDVW WKH LQWHUFRQQHFWV FDQ RSHUDWH 7KLV LV D KLGGHQ EHQHILW RI ZLUHOHVV LQWHUFRQn QHFWV LQ WKDW WKH SHUIRUPDQFH LPSURYHV DV WKH IUHTXHQF\ LQFUHDVHV WKH LPSURYHPHQWV DUH ODUJHO\ GXH WR WKH DQWHQQDV DQG LQGXFWRUV EHFRPLQJ PRUH HIILFLHQWf 7KXV ZLUHOHVV LQWHUn FRQQHFWV VKLIW WKH PLFURSURFHVVRU SHUIRUPDQFH IURP EHLQJ EDFNHQG LH LQWHUFRQQHFWf OLPLWHG WR RQFH DJDLQ EHLQJ IURQWHQG LH WUDQVLVWRUf OLPLWHG 7KHUHIRUH WKLV ZRUN KDV GHPRQVWUDWHG ERWK WKH SODXVLELOLW\ DQG IHDVLELOLW\ RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP )LUVW WKH GHVLJQ DQG LPSOHPHQWDWLRQ RI /1$ DQG IUHn TXHQF\ GLYLGHU FLUFXLWV RSHUDWLQJ EHWZHHQ DQG *+] ZDV SUHVHQWHG 7KH UHVXOWV GHPn RQVWUDWH WKH FRPSHWLWLYHQHVV RI &026 IRU ERWK LQWHUFRQQHFW DQG JHQHUDO ZLUHOHVV FRPPXQLFDWLRQ V\VWHPV 6HFRQG WKH V\VWHP UHTXLUHPHQWV RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP ZHUH GHULYHG WUDQVODWLQJ WKH FORFN UHTXLUHPHQWV RI VNHZ DQG MLWWHU LQWR VWDQGDUG UDGLR UHTXLUHPHQWV 7KLUG WKH SODXVLELOLW\ RI D ZLUHOHVV LQWHUFRQQHFW V\VWHP ZDV GHPRQn VWUDWHG E\ LPSOHPHQWLQJ RQFKLS DQWHQQD SDLUV FORFN UHFHLYHUV DQG FORFN WUDQVPLWWHUV 7KHVH DUH WKH ILUVW GHPRQVWUDWLRQV RI ZLUHOHVV LQWHUFRQQHFWV DQG WKH ILUVW WLPH DQWHQQDV DQG FLUFXLWU\ KDYH EHHQ LQFRUSRUDWHG RQ WKH VDPH VLOLFRQ FKLS )LQDOO\ WKH IHDVLELOLW\ RI D ZLUHn OHVV LQWHUFRQQHFW V\VWHP ZDV HYDOXDWHG LQ WHUPV RI SRZHU GLVVLSDWLRQ V\QFKURQL]DWLRQ SURFHVV YDULDWLRQ ODWHQF\ DUHD DQG GHVLJQ YHULILFDWLRQ $OO RI WKHVH UHVXOWV LQGLFDWH WKH SRWHQWLDO RI DQ RQFKLS ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP %URDGHU $SSOLFDELOLW\ 7KH UHVXOWV IURP D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP FDQ EH DSSOLHG WRn JHQHUDO ZLUHOHVV LQWHUFRQQHFW V\VWHPV DV ZHOO DV 5) V\VWHPV ZLWK RQFKLS DQWHQQDV %DVLFDOO\ D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP FRQVLVWV RI FRPPXQLFDWLRQ RI MXVW D FDUULHU VLJQDO

PAGE 248

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n JOHFKLS UDGLR LQFOXGLQJ DQWHQQDf FRXOG EH LPSOHPHQWHG IRU ORZFRVW DSSOLFDWLRQV 7KHUHn IRUH LQ VXPPDU\ WKLV ZRUN LV DSSOLFDEOH WR ERWK ZLUHOHVV LQWHUFRQQHFW V\VWHPV DQG JHQHUDO ZLUHOHVV FRPPXQLFDWLRQ V\VWHPV 6XJJHVWHG )XWXUH :RUN &LUFXLW 'HVLJQ DQG ,PSOHPHQWDWLRQ )LUVW WKH VWDWHGHSHQGHQW LQLWLDOL]DWLRQ IDLOXUH LQ WKH SURJUDPPDEOH GLYLGHU KDV WR EH FRUUHFWHG XVLQJ WKH QHZ ODWFKHV SUHVHQWHG LQ )LJXUH 6HFRQG DXWRPDWLF JDLQ DQG IUHTXHQF\f FRQWURO FLUFXLWV VKRXOG EH LPSOHPHQWHG LQ WKH UHFHLYHU WR WXQH WKH JDLQ LQ WKH DPSOLILHU DQG WKH VHOIRVFLOODWLRQ IUHTXHQF\ RI WKH LQMHFn WLRQORFNHG GLYLGHU 3RWHQWLDOO\ WKH ILUVW GLYLGHU FRXOG EH FKDQJHG WR DQ /& LQMHFn WLRQORFNHG RVFLOODWRU 7KLUG WKH V\VWHPf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n ODWLRQ WRROV 7KLV LV YHU\ LPSRUWDQW WR WKH RYHUDOO V\VWHP RSHUDWLRQ VLQFH ERWK WKH DQWHQQD

PAGE 249

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n PHQWHG ZKLFK FRXOG LQFOXGH SKDVH GHWHFWRUV WR PHDVXUH WKH VWDWLF SKDVH HUURU DQG DQD ORJWRGLJLWDO FRQYHUWHUV WR WUDQVODWH WKLV SKDVHHUURU LQWR D VSHFLILF FRXQW IRU WKH UHFHLYHUf 6HFRQG WKH DFFXPXODWLRQ RI SKDVHHUURUV LQ WKH VWDWLF SKDVHFRUUHFWLRQ V\VWHP VKRXOG EH TXDQWLILHG 7KLUG WKH MLWWHU GXH WR QRLV\ VXSSOLHV DQG UHDOLVWLF VZLWFKLQJ QRLVH KDV WR EH TXDQWLILHG )LQDOO\ DQ LQWUDFKLS ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP FRQVLVWLQJ RI D WUDQVn PLWWHU DQG PXOWLSOH UHFHLYHUV VKRXOG EH XVHG WR SURYLGH D FORFN VLJQDO WR DQ DFWXDO PLFURn SURFHVVRU FRUH 8VLQJ WKLV SODWIRUP WKH FORFN VNHZ DQG MLWWHU RI WKH ZLUHOHVV FORFN GLVWULEXWLRQ FDQ EH REWDLQHG DQG UHDOLVWLF V\VWHP RSHUDWLRQ FDQ EH GHPRQVWUDWHG

PAGE 250

$33(1',; $ 7+(25< )25 &20021*$7( /2: 12,6( $03/,),(5 $ ,QSXW ,PSHGDQFH )LJXUH $O VKRZV WKH VPDOOVLJQDO PRGHO RI D FRPPRQJDWH ORZ QRLVH DPSOLILHU /1$f %\ LQVSHFWLRQ WKH LQSXW DGPLWWDQFH IRU WKH FRPPRQJDWH DPSOLILHU LV
PAGE 251

U B OGO B 9 I O6P fB 6P 96 Ym +5M6PmV $f fn6 YV $V FDQ EH VHHQ 5V GHJHQHUDWHV WKH WUDQVLVWRU 7KHUHIRUH WKH FRPPRQJDWH /1$ LV W\SLn FDOO\ PRUH OLQHDU WKDQ DQ LQGXFWLYHO\GHJHQHUDWHG FRPPRQVRXUFH /1$ )RU D WXQHG ORDG RXWSXW WKH 6L LV HTXDO WR 1 .8FR 6P?=/HTRf? P56 $f ZKHUH =/HTFRf LV HTXDO WR WKH WRWDO ORDG LPSHGDQFH DW WKH DPSOLILHU RXWSXW DW UHVRQDQFH $VVXPLQJ WKDW WKH RXWSXW PDWFKLQJ QHWZRUN LV WKH VDPH DV WKDW XVHG LQ WKH VRXUFHGHJHQn HUDWHG /1$ LH D PDWFKLQJ QHWZRUN FRQWDLQLQJ D GUDLQ LQGXFWRU DQG D FDSDFLWLYH WUDQVn IRUPHUf WKHQ L B P4/GAR5G P56 $f ZKHUH DQG /G DUH WKH TXDOLW\ IDFWRU DQG LQGXFWDQFH RI WKH GUDLQ LQGXFWRU DQG K*Rf LV D FRPSOH[ FDSDFLWLYHWUDQVIRUPHU UDWLR GHILQHG LQ f $ 1RLVH )DFWRU 7R GHULYH WKH QRLVH IDFWRU RI WKLV DPSOLILHU WKH VKRUWFLUFXLW RXWSXW FXUUHQW LV FDOFXn ODWHG IRU WKH HQWLUH FLUFXLW LQFOXGLQJ WKH VRXUFH UHVLVWDQFH &RQYHUWLQJ WKLV WR D SRZHU VSHFWUDO GHQVLW\ 36'f DQG WDNLQJ WKH UDWLR RI WKH WRWDO 36' WR WKH 36' IURP WKH VRXUFH UHVLVWDQFH \LHOGV WKH QRLVH IDFWRU 7KH WKHUPDO QRLVH JHQHUDWHG E\ WKH VRXUFH UHVLVWDQFH LV UHSUHVHQWHG E\ D YROWDJH VRXUFH ZLWK D 36' JLYHQ E\ WKH IROORZLQJ Y? $N7 5V $f

PAGE 252

1HJOHFWLQJ DOO RWKHU VRXUFHV RI QRLVH H[FHSW WKH WKHUPDO QRLVH LQ WKH 026 WUDQVLVWRU OXPSHG RQO\ DW WKH GUDLQf WKH WRWDO VKRUWFLUFXLW RXWSXW FXUUHQW FDQ EH VKRZQ WR EH rVF AP95V L B S } P.6 $f ZKHUH *P LV GHILQHG LQ $f DQG LG LV WKH FKDQQHO WKHUPDO QRLVH OXPSHG DW WKH GUDLQ 7KLV GUDLQ WKHUPDO QRLVH KDV D 36' RI N7\JGR $f ZKHUH \ LV D ELDVGHSHQGHQW SDUDPHWHU IRU ORQJFKDQQHO GHYLFHVf N LV %ROW]PDQQfV FRQVWDQW [ n -.f 7 LV DEVROXWH WHPSHUDWXUH DQG JGR LV WKH VKRUWFLUFXLW GUDLQ FRQGXFWDQFH 6LQFH WKHVH WZR QRLVH VRXUFHV DUH XQFRUUHODWHG WKH RXWSXW 36' LV a OVF a OG Y5V f§ P $f ZKHUH $f KDV EHHQ VXEVWLWXWHG 7KHUHIRUH WKH QRLVH IDFWRU LV OG N7
PAGE 253

$33(1',; % 287387 0$7&+,1* 86,1* &$3$&,7,9( 75$16)250(5 % 7ZR(OHPHQW 0DWFKLQJ 7HFKQLTXH 0XOWLSOH PHWKRGV ERWK JUDSKLFDO DQG DQDO\WLFDO H[LVW WR TXLFNO\ PDWFK RQH LPSHGDQFH WR DQRWKHU XVLQJ WZR UHDFWLYH HOHPHQWV $Q DQDO\WLFDO PHWKRG IRXQG LQ >%RZ@ ZKLFK LV ERWK HDV\ WR XQGHUVWDQG XVH DQG FRGH LH LQWR 0DWODEf LV VXPPDn UL]HG KHUH IRU FRQYHQLHQFH 7KH VLPSOHVW PDWFKLQJ VFHQDULR LV PDWFKLQJ WZR UHVLVWDQFHV XVLQJ DQ f/f RU WZRHOHPHQW QHWZRUN ,Q WKLV FDVH WKH ODUJHU UHVLVWDQFH WHUPHG 5S UHTXLUHV D VKXQW UHDFWLYH FRPSRQHQW GHFUHDVLQJ WKH LPSHGDQFHf ZKLOH WKH VPDOOHU UHVLVn WDQFH WHUPHG 5VH UHTXLUHV D VHULHV UHDFWLYH FRPSRQHQW 7KH VLJQ RI WKHVH UHDFWDQFHV LH LQGXFWLYH RU FDSDFLWLYHf DUH RSSRVLWH RI RQH DQRWKHU IRU WKH QHW UHDFWDQFH WR EH ]HUR 7KH IRUPXODV IRU WKLV PHWKRG DUH DV IROORZV H -IHn sHrf 5f %Of %f %f ZKHUH ;V DQG ;S DUH WKH UHTXLUHG VHULHV DQG VKXQW UHDFWDQFHV RI WKH / QHWZRUN $Q LOOXVWUDn WLRQ RI WKLV SURFHGXUH LV JLYHQ LQ )LJXUH %O 7KH YDOXH RI WKH LQGXFWDQFH DQG FDSDFLWDQFH RI HDFK FRPSRQHQW FDQ HDVLO\ EH REWDLQHG IURP WKH UHDFWDQFH IRU D JLYHQ DQJXODU IUHn TXHQF\ 7R KDQGOH FRPSOH[ VRXUFHV DQG ORDGV WKH UHDFWDQFH RI WKH VRXUFHORDG LV DEVRUEHG

PAGE 254

U Q ;V 45VH ‘, b L B/ L c [S UST c c 7 / /0DWFKLQJ 1HWZRUN )LJXUH % f/f QHWZRUN PDWFKLQJ GHVLJQ LQWR WKH PDWFKLQJ QHWZRUN +HQFH WKH WRWDO UHDFWDQFHV LQ WKH PDWFKLQJ QHWZRUN DQG WKH VRXUFHORDG VKRXOG EH HTXDO WR WKH YDOXHV ;V RU ;S DV JLYHQ E\ %f DQG %f % $SSOLFDWLRQ WR &DSDFLWLYH 7UDQVIRUPHU LQ /1$ )LJXUH % VKRZV WKH RXWSXW PDWFKLQJ QHWZRUN IRU WKH /1$ LQFOXGLQJ WKH FDSDFLn WLYH WUDQVIRUPHU +HUH WKH RXWSXW LPSHGDQFH RI WKH FDVFRGH KDV EHHQ PRGHOHG E\ 5SOO&S ZKLOH WKH GUDLQ LQGXFWRU LV PRGHOHG DV D SDUDOOHO QHWZRUN FRPSRVHG RI /G 53/G DQG &MAM $V FDQ EH VHHQ &@ LV WKH VKXQW UHDFWDQFH ZKLOH & SOXV WKH HTXLYDOHQW VHULHV UHDFWDQFH IURP /G DQG &/G&3 LV WKH WRWDO VHULHV UHDFWDQFH +HQFH WKH WRWDO LPSHGDQFH DW 5_B 4 = VH / f§ B B n 9 /0DWFKLQJ 1HWZRUN ) )LJXUH % 6PDOOVLJQDO PRGHO RI FRPPRQJDWH ORZ QRLVH DPSOLILHU

PAGE 255

WKH RXWSXW RI WKH FDVFRGH =VH LQFOXGLQJ WKH GUDLQ LQGXFWRU VKRXOG ILUVW EH FDOFXODWHG DV IROORZV ]f A:c0&f&SfL5fL;ff 5f ZKHUH 5VH DQG ;VH DUH WKH UHDO DQG LPDJLQDU\ SDUWV RI =VH UHVSHFWLYHO\ 6HWWLQJ 5S 5/  DQG WKHQ XVLQJ %Of \LHOGV WKH TXDOLW\ IDFWRU 4 RI WKH / PDWFKLQJ QHWZRUN (TXDn WLRQV %f DQG %f DUH WKHQ XVHG WR ILQG & DQG &M UHVSHFWLYHO\ DV IROORZV & Q;f45LHf %f &O 4 WRn %f ,I HLWKHU RU ERWKf RI WKHVH YDOXHV LV QHJDWLYH LW VLJQLILHV WKDW DQ LQGXFWRU VKRXOG EH XVHG LQ SODFH RI WKH FDSDFLWRU ,I WKLV LV XQDFFHSWDEOH IRU WKH /1$ GHVLJQ WKHQ WKH YDOXH RI /G DQG LWV DVVRFLDWHG SDUDVLWLFVf VKRXOG EH PRGLILHG

PAGE 256

$33(1',; & '(5,9$7,21 2) 12,6( 3$5$0(7(56 & (TXLYDOHQW ,QSXW 1RLVH *HQHUDWRUV $Q\ fQRLV\f WZRSRUW QHWZRUN FDQ EH UHSUHVHQWHG DV D fQRLVHOHVVf WZRSRUW QHWn ZRUN ZLWK HTXLYDOHQW YROWDJH YQf DQG FXUUHQW LQf QRLVH JHQHUDWRUV DW WKH LQSXW RI WKH QHWn ZRUN >*UD@ DV VKRZQ LQ )LJXUH &O 7KH WZR QRLVH JHQHUDWRUV DOORZ WKH QRLVH SURSHUWLHV RI WKH QHWZRUN WR EH FRUUHFWO\ UHSUHVHQWHG IRU DQ\ LQSXW LPSHGDQFH DGPLWn WDQFHf 7R FDOFXODWH WKHVH QRLVH JHQHUDWRUV WKH RXWSXW VKRUWFLUFXLW LVFf FXUUHQW FDQ EH HTXDWHG IRU ERWK QHWZRUNV ZLWK WKH LQSXW ILUVW VKRUWFLUFXLWHG DQG WKHQ RSHQFLUFXLWHG \LHOGLQJ YQ DQG LQ UHVSHFWLYHO\ 7UDGLWLRQDOO\ WKH QRLVH IDFWRU RI WKH QHWZRUN LV FKDUDFWHUL]HG LQ WHUPV RI IRXU QRLVH SDUDPHWHUV DV IROORZV Df f1RLV\f 7ZR3RUW 2 2 f1RLVHOHVVf 7ZR3RUW 2 & )LJXUH &O 5HSUHVHQWDWLRQ RI D fQRLV\f WZRSRUW QHWZRUN DV D fQRLVHOHVVf WZRSRUW ZLWK HTXLYDOHQW LQSXW QRLVH JHQHUDWRUV

PAGE 257

&Of ZKHUH )PLQ LV WKH PLQLPXP REWDLQDEOH QRLVH IDFWRU 5Q LV WKH QRLVH FRQGXFWDQFH
PAGE 258

7R DFFRXQW IRU FRUUHODWLRQ EHWZHHQ YQ DQG LQ YQ FDQ EH GLYLGHG LQWR FRPSRQHQWV ZKLFK DUH FRUUHODWHG YFf DQG XQFRUUHODWHG YXf ZLWK LQ DV IROORZV 9Q 9X 9F 9X =FLQ &f 7KH FRUUHODWLRQ LPSHGDQFH =F LV B rQr9Q &f Q 7KH SRZHU VSHFWUDO GHQVLWLHV RI HDFK RI WKHVH QRLVH VRXUFHV LQ YXf FDQ EH UHSUHn VHQWHG E\ HLWKHU D UHVLVWDQFH IRU YXf RU D FRQGXFWDQFH IRU LQf DV IROORZV *f Q N7$I f &f 5 f N7$In ZKHUH 5X FDQ EH REWDLQHG XVLQJ WKH HTXLYDOHQW LQSXW QRLVH JHQHUDWRUV DV IROORZV &f 5 A?=F?*Q 5f_=*f f N7$I 6XEVWLWXWLQJ &f &f &f DQG &f LQWR &f \LHOGV &f _Y n=L =RfI n 5V=V =HO n &f 7DNLQJ WKH SDUWLDO GHULYDWLYHV RI &f ZLWK UHVSHFW WR WKH VRXUFH UHVLVWDQFH DQG UHDFWDQFH DQG VHWWLQJ WKH UHVXOWV HTXDO WR ]HUR JLYHV WKH RSWLPXP VRXUFH LPSHGDQFH UHVXOWLQJ LQ PLQn LPXP QRLVH IDFWRU DV IROORZV &f

PAGE 259

;RSW ;F &OOf 7KH PLQLPXP QRLVH ILJXUH )PLQf FDQ QRZ EH REWDLQHG E\ SOXJJLQJ LQ &f DQG &O f LQWR &f DV IROORZV A! *Q5RSW 5Ff &f 7KXV WKH WRWDO QRLVH IDFWRU FDQ EH UHSUHVHQWHG DV )}Q )PLQ *Q 5RSW ;RSW@ >)PLQ 5Q *RSW %RSW@ HWFf 6XUYH\LQJ WKH UHVXOWV LQ WKH SUHYLRXV VHFWLRQ DQ DOWHUQDWLYH VHW RI SDUDPHWHUV ZKLFK FRPH GLUHFWO\ IURP WKH HTXLYDOHQW LQSXW QRLVH JHQHUDWRUV FDQ EH GHILQHG IURP ZKLFK DQ\ RWKHU QRLVH SDUDPHn WHU VHW FDQ EH HDVLO\ REWDLQHG 7KHVH DOWHUQDWLYH SDUDPHWHUV DUH 5Q *Q DQG 3Q ZKHUH *Q LV JLYHQ E\ &f ZKLOH 5 Q N7 &f 3 Q N7 n &f +HUH 3Q LV D GLPHQVLRQOHVV FRUUHODWHG fSRZHU UDWLRf 7KHVH SDUDPHWHUV QDWXUDOO\ DUH QHHGHG ZKHQ WU\LQJ WR REWDLQ DQ\ RI WKH WUDGLWLRQDO QRLVH SDUDPHWHUV WKXV WKH\ DUH FRQYHn QLHQW IRU QXPHULFDO WHFKQLTXHV ([DPLQLQJ WKHVH QHZ SDUDPHWHUV LW FDQ HDVLO\ EH VKRZQ WKDW

PAGE 260

RU DOWHUQDWLYHO\ &f Y rL Q Q S r Q a5a )LQDOO\ WR REWDLQ )PcQ WKH IROORZLQJ UHODWLRQVKLSV DUH XVHG &f )fWQ *-5 5&f L*f5F M*f5X *; &f 5H3Qf -J-A =2f IWH3ff 5H3Qf _If_ 5H?SQf 5H3 Qf A*QLQ P3Qf 7KXV 5Q *Q DQG 3Q FDQ EH XVHG WR REWDLQ DQ\ RI WKH RWKHU QRLVH SDUDPHWHUV 7KURXJK WKH DXWKRUfV H[SHULHQFH WKHVH DOWHUQDWLYH SDUDPHWHUV DUH PRUH FRQYHQLHQW IRU GHULYDWLRQV DQG QXPHULFDO WHFKQLTXHV SDUWLFXODUO\ IRU GHULYDWLRQV RI )PLQ

PAGE 261

$33(1',; 12,6( 3$5$0(7(56 2) 026)(7 7UDQVLVWRU 0RGHO DQG (TXLYDOHQW ,QSXW 1RLVH *HQHUDWRUV ,Q WKLV DSSHQGL[ WKH QRLVH SDUDPHWHUV RI D VLQJOH LQWULQVLF 026)(7 DUH GHULYHG LQFOXGLQJ WKH HIIHFWV RI WKHUPDO QRLVH LQ WKH FKDQQHO OXPSHG ERWK DW WKH JDWH JDWH LQGXFHG QRLVHf§*,1f DQG WKH GUDLQ VXEVWUDWH UHVLVWDQFH DQG DOO RI WKH SDUDVLWLF FDSDFLn WDQFHV )LJXUH 'O VKRZV D KLJKIUHTXHQF\ TXDVLVWDWLF PRGHO RI WKH 026)(7 >7VL@ LQFOXGLQJ QRLVH VRXUFHV 7KH QRLVH VRXUFHV KDYH WKH IROORZLQJ SRZHU VSHFWUDO GHQVLWLHV neI 87\6GR 'Of $ P $ 'f LIWE $N7 $ 5E f 'f 9EV pn5E )LJXUH 'O +LJKIUHTXHQF\ TXDVLVWDWLF PRGHO RI LQWULQVLF 026)(7 >7VL@

PAGE 262

ZKHUH \ DQG DUH ELDVGHSHQGHQW SDUDPHWHUV N LV %ROW]PDQQfV FRQVWDQW 7 LV DEVROXWH WHPSHUDWXUH DQG JGR LV WKH VKRUWFLUFXLW GUDLQ FRQGXFWDQFH 7KH SRODULW\ RI WKH QRLVH VRXUFHV LG DQG L5E LV DUELWUDU\ KRZHYHU WKH SRODULW\ RI LJ VKRXOG EH WDNHQ RSSRVLWH WR LG WR \LHOG D FRUUHODWLRQ FRHIILFLHQW ZLWK WKH FRUUHFW SRODULW\ HTXDO WR 'f F )RU ORQJFKDQQHO GHYLFHV F LV SXUHO\ LPDJLQDU\ WDNLQJ D YDOXH F M $V GLVFXVVHG LQ $SSHQGL[ & DQ\ QRLV\ WZRSRUW QHWZRUN FDQ EH UHSUHVHQWHG DV D QRLVHOHVV WZRSRUW QHWZRUN ZKLFK KDV HTXLYDOHQW YROWDJH YQf DQG FXUUHQW LQf QRLVH JHQHUn DWRUV DW WKH LQSXW RI WKH QHWZRUN 2QFH WKHVH QRLVH JHQHUDWRUV DUH REWDLQHG WKH QRLVH SDUDPHWHUV FDQ EH GHULYHG DV GLVFXVVHG LQ $SSHQGL[ & 7R REWDLQ YQ WKH RXWSXW VKRUWFLUn FXLW FXUUHQW LVFf LV HTXDWHG IRU ERWK D QRLV\ )LJXUH 'O DV VKRZQf DQG QRLVHOHVV QHWZRUN )LJXUH 'O FRQWDLQLQJ YQ DQG LQ DW WKH LQSXW DQG H[FOXGLQJ LJ LG DQG L5Ef ZKHQ WKH LQSXW LV VKRUWFLUFXLWHG 7R REWDLQ LQ WKH VDPH SURFHGXUH LV IROORZHG H[FHSW WKH LQSXW LV QRZ RSHQFLUFXLWHG 7KH UHVXOWV DUH DV IROORZV 'f 'f $ WUDQVDGPLWWDQFH \M UHIHUULQJ RXWSXW FXUUHQW WR LQSXW YROWDJH LV IRXQG WR EH \L J-L0f0&JG$&GEf 'f r 9 YRXW Z

PAGE 263

,Q WKHVH HTXDWLRQV $ LV WKH YROWDJH GLYLVLRQ UDWLR EHWZHHQ YJV DQG YEV 5EL&JE &VE 4IFf@ &*7 a &JV&JG&JE ‘ 'f \E P&VE &OW &-Ef -> MR!D!%f@ 'f 'f 'O f 'f 'f 1RLVH 3DUDPHWHUV IRU &RPSOHWH 0RGHO 8VLQJ 'f DQG 'f QRLVH SDUDPHWHUV IRU WKH 026)(7 LQFOXGLQJ DOO VHFRQG RUGHU HIIHFWV FDQ EH GHULYHG DV RXWOLQHG LQ $SSHQGL[ & 7KH DOWHUQDWLYH QRLVH SDUDPHWHUV 5Q *Q DQG 3Qf GHILQHG LQ &f &f DQG &f DUH JLYHQ LQ 7DEOH 'O ZKHUH L >FRFARAM 'f 2f&2Jf $V FDQ EH VHHQ E\ H[DPLQLQJ 'f 'f DQG 'f WKH JDWHWRERG\ FDSDFLWDQFH JUHDWO\ FRPSOLFDWHV DOO RI WKH HTXDWLRQV 7KHVH DOWHUQDWLYH QRLVH SDUDPHWHUV FDQ WKHQ EH XVHG WR ILQG PRUH FRQYHQWLRQDO QRLVH SDUDPHWHUV XVLQJ &f IRU =F &f IRU 5X &f IRU 5RSW &OOf IRU ;RSW DQG

PAGE 264

&f IRU )PLQ 1RZ WKDW H[SOLFLW HTXDWLRQV DUH LQ KDQG IRU WKH QRLVH SDUDPHWHUV IRU WKH FRPSOHWH PRGHO RI D 026)(7 D QXPHULFDO VLPXODWRU HJ 0DWODEf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f DV ZHOO DV WKH HIIHFW RI &JG 6XEVWUDWH UHVLVn WDQFH DQG DOO RWKHU FDSDFLWDQFHV H[FHSW &JV DUH QHJOHFWHG 8QGHU WKHVH DVVXPSWLRQV \L JP MRR&JG &*7 &JV &JG DQG WKH LQSXW QRLVH JHQHUDWRUV EHFRPH B MGB OG 9QO "O PaMp&JG 'f

PAGE 265

 f§ f&*Uf A A -&AM?!7f 'f ZKHUH WKH FXWRII IUHTXHQF\ RI WKH 026)(7 LV HTXDO WR 6Q &2QQ A U FfU Ff 'f nJV JG 7KH DOWHUQDWLYH DQG LPSHGDQFHEDVHG QRLVH SDUDPHWHUV DUH HDVLO\ REWDLQHG DQG DUH VKRZQ LQ WKH IROORZLQJ WDEOH 7R REWDLQ WKH QRLVH SDUDPHWHUV ZLWKRXW &JG VLPSO\ VHW &JG 7DEOH 1RLVH SDUDPHWHUV FRQVLGHULQJ &JG $OWHUQDWLYH 1RLVH 3DUDPHWHUV ,PSHGDQFH%DVHG 1RLVH 3DUDPHWHUV r‘n U Df D+8 -)PLQO 'f *}! ‘ r1ref U 'f / Y P 3QL a 'f ]fnf ‘ 8F*7 'f /RRNLQJ DW WKHVH QRLVH SDUDPHWHUV LW FDQ ILUVW EH VHHQ WKDW WKH RSWLPDO VRXUFH LPSHGDQFH LV DQ LQGXFWRU ZKLFK VHULHV UHVRQDWHV ZLWK &JV 6HFRQG WKH RSWLPDO VRXUFH UHVLVWDQFH LV D VKRUW PHDQLQJ WKDW ZKHQ GULYLQJ WKH QHWZRUN ZLWK D UHVLVWLYH ORDG D QRLVH PLVPDWFK RFFXUV )XUWKHUPRUH )PLQ PHDQLQJ WKDW ZLWK SHUIHFW QRLVH PDWFKLQJ LH 5V f WKH 026)(7 GRHV QRW DGG QRLVH 7KLV LV GXH WR WKHUH EHLQJ RQO\ RQH QRLVH VRXUFH ZKLFK LV IXOO\ FRUUHODWHG WR LWVHOI KHQFH 5X ZKLOH 5F DV ZHOO +RZHYHU )PcQ LV SK\VLFDOO\ XQUHDVRQDEOH PHDQLQJ WKDW VHFRQGRUGHU HIIHFWV ZLOO GHWHUPLQH )PMQ

PAGE 266

' &DVH (IIHFW RI *,1 7KH HIIHFW RI *,1 LV QRZ DQDO\]HG ZKLOH QHJOHFWLQJ VXEVWUDWH UHVLVWDQFH DQG DOO RWKHU FDSDFLWDQFHV H[FHSW &JV 8QGHU WKHVH DVVXPSWLRQV \? JP! &JW &JV DQG WKH LQSXW QRLVH JHQHUDWRUV EHFRPH 9Q tQ 'f 'f 7KH QRLVH SDUDPHWHUV FDQ WKHQ EH GHULYHG DQG DUH JLYHQ LQ WKH IROORZLQJ WDEOH 7DEOH 1RLVH SDUDPHWHUV FRQVLGHULQJ *,1 $OWHUQDWLYH 1RLVH 3DUDPHWHUV ,PSHGDQFH%DVHG 1RLVH 3DUDPHWHUV 5f I 'f 'f mf r!&fE <6fAfI! 'f 3Q 5QMp&JVfEL 'f 7KH IROORZLQJ YDULDEOHV KDYH EHHQ LQWURGXFHG E? $_F_ 'f E $_F_ $ 'f $ i 9 \ 'f ([DPLQLQJ WKLV VHW RI QRLVH SDUDPHWHUV LW FDQ ILUVW EH VHHQ WKDW WKH LQFOXVLRQ RI *,1 FDXVHV WKH RSWLPDO VRXUFH UHDFWDQFH WR VOLJKWO\ GHFUHDVH 7KLV PHDQV WKDW VHULHV

PAGE 267

UHVRQDQFH VKRXOG RFFXU DW D KLJKHU IUHTXHQF\ DV LQGLFDWHG LQ >6KD@ 6HFRQG DQ RSWLPDO VRXUFH UHVLVWDQFH LV QRZ SUHVHQW FDXVLQJ )PLQ WR EH JUHDWHU WKDQ RQH 7KLV LV GXH WR WKH GUDLQ DQG JDWH QRLVHV QRW EHLQJ IXOO\ FRUUHODWHG 7KLUG )PLQ LV SURSRUWLRQDO WR FR2S PHDQn LQJ WKDW DV WHFKQRORJLHV VFDOH WR VKRUWHU FKDQQHO OHQJWKV ORZHU QRLVH ILJXUHV ZLOO RFFXU IRU D JLYHQ RSHUDWLQJ IUHTXHQF\ )RXUWK WKH RSWLPDO VRXUFH LPSHGDQFH LV GLUHFWO\ SURSRUWLRQDO WR 4JV ZKHUH 4JV LV GHILQHG LQ f 7KLV PHDQV WKDW DV WKH GHYLFH EHFRPHV VPDOOHU RU IUHTXHQF\ GHFUHDVHVf 4JV DQG =RSW ERWK LQFUHDVH &DVH (IIHFW RI 5E ([FOXGLQJ 6XEVWUDWH UHVLVWDQFH JHQHUDWHV WKHUPDO QRLVH ZKLFK LV FRXSOHG LQWR WKH GHYLFH WKURXJK EXON WUDQVFRQGXFWDQFH DQG EXON FDSDFLWDQFHV &JE &GE &VEf 7KH LQFOXVLRQ RI &JE FRPSOLFDWHV WKH HTXDWLRQV FRQVLGHUDEO\ WKHUHIRUH WKH HTXDWLRQV IRU WKH IXOO PRGHO VKRXOG EH XVHG WR H[DPLQH WKH HIIHFW RI &JE RQ QRLVH 1HJOHFWLQJ &JE DQG &JG WKHQ \@ JP $ DQG &*7 &JV 7KH EXON DGPLWWDQFH
PAGE 268

7DEOH 1RLVH SDUDPHWHUV FRQVLGHULQJ VXEVWUDWH UHVLVWDQFH DQG H[FOXGLQJ &JE $OWHUQDWLYH 1RLVH 3DUDPHWHUV ,PSHGDQFH%DVHG 1RLVH 3DUDPHWHUV ILf '! f6GR Af O '! *f 'nf 3 'f m mGR = 'f WKH RWKHU KDQG IRU YHU\ ODUJH 5E )PLQ DSSURDFKHV DV ZHOO WKH OLPLW RI %5E DV 5E DSSURDFKHV LQILQLW\ LV ]HUR VLQFH !% DSSURDFKHV ]HUR IDVWHU WKDQ 5E DSSURDFKHV LQILQLW\f )PLQ LV ODUJHU WKDQ IRU 5E YDOXHV LQ EHWZHHQ DQG LQILQLW\ 7KLV VKRZV WKDW VKRUW RU RSHQFLUFXLWLQJ WKH VXEVWUDWH LV YDOXDEOH IRU PLQLPL]LQJ QRLVH ILJXUH 7KLV LV WKH FDVH IRU PRVW RWKHU 5) SHUIRUPDQFH PHWULFV DV D IXQFWLRQ RI VXEVWUDWH UHVLVWDQFH )RU &GE a &JV U_ DQG FR2S a )PLQ SHDNV DW DURXQG  )RU WKLV FDVH WKH VXEVWUDWH UHVLVn WDQFH UDQJH EHWZHHQ DQG Ne VKRXOG EH DYRLGHG LI SRVVLEOH WR PLQLPL]H )PLQ &DUH PXVW EH WDNHQ LQ WU\LQJ WR REWDLQ WKH QRLVH SDUDPHWHUV DV D IXQFWLRQ RI RQO\ 5E DQG GUDLQ QRLVH LH QHJOHFWLQJ DOO EXON FDSDFLWDQFHVf 2Q ILUVW JODQFH LW VHHPV WKDW WKLV FDQ EH DFKLHYHG E\ VHWWLQJ IL% +RZHYHU WKLV ZLOO RQO\ PDWFK IRU VPDOO YDOXHV RI 5E 7KH FKDUDFWHULVWLF RI GHFUHDVLQJ )PLQ IRU YHU\ ODUJH YDOXHV RI 5E ZLOO QRW EH REWDLQHG VLQFH 2E VKRXOG DSSURDFK ]HUR DV 5E DSSURDFKHV LQILQLW\ 7KLV PHDQV WKDW IRU ODUJH YDOXHV RI 5E WKH EXON FDSDFLWDQFHV &GE DQG &VEf FDQQRW EH QHJOHFWHG

PAGE 269

$33(1',; ( ,1-(&7,21 /2&.,1* 2) 26&,//$7256 (O 2YHUYLHZ $Q\ RVFLOODWRU FDQ EH ORFNHG RU V\QFKURQL]HG WR DQ H[WHUQDO VLJQDO ZKRVH IUHn TXHQF\ RU KDUPRQLFf LV FORVH WR WKH QDWXUDO IUHTXHQF\ RI WKH RVFLOODWRU 7KLV W\SH RI FLUFXLW LV NQRZQ DV DQ LQMHFWLRQORFNHG RVFLOODWRU ,/2f ZKHUH WKH RVFLOODWRU LV ORFNHG WR HLWKHU WKH IXQGDPHQWDO VXEKDUPRQLF RU VXSHUKDUPRQLF RI WKH QDWXUDO RVFLOODWLQJ IUHTXHQF\ 7KH ILUVW SUHVHQWDWLRQ RI ,/2V ZDV E\ $GOHU LQ >$GO@ ZKHUH KH GHYHORSHG D GLIIHUHQWLDO HTXDWLRQ UHODWLQJ WKH SKDVH GLIIHUHQFH EHWZHHQ WKH ORFNLQJ LQSXWf DQG RVFLOODWRU RXWSXWf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f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

PAGE 270

LQSXW DQG RXWSXW VLJQDOV 7KLV WKHRU\ LV D VXPPDU\ RI WKH UHOHYDQW UHVXOWV SUHVHQWHG LQ >$GO@ DQG >3DF@ )LQDOO\ WKH SKDVH QRLVH RI WKH ,/2 ZLOO EH SUHVHQWHG ( 7KHRU\ IRU ,QMHFWLRQ /RFNLQJ ( %DVLF 0RGHO )LJXUH (ODf VKRZV WKH EDVLF PRGHO IRU DQ ,/2 +HUH WKH RVFLOODWRU LV PRGHOHG VLPSO\ DV D IHHGEDFN QHWZRUN ZLWK D IRUZDUG SDWK RI +Vf 7KLV PRGHO FDQ WKHUHIRUH EH DSSOLHG WR HLWKHU /& RVFLOODWRUV RU ULQJ RVFLOODWRUV ,Q WKH DEVHQFH RI DQ LQSXW VLJQDO WKH FLUFXLW ZLOO RVFLOODWH DW LWV QDWXUDO IUHTXHQF\ WR $W WKLV UHVRQDQW IUHTXHQF\ DQG RQO\ DW WKLV IUHTXHQF\ WKH RXWSXW RI WKH FLUFXLW 9 DQG WKH HUURU VLJQDO ( DUH LQ SKDVH :KHQ D ORFNn LQJ VLJQDO ZLWK DPSOLWXGH 9/ LV LQMHFWHG LQWR WKH ORRS DW D IUHTXHQF\ FR/ FR $JR D GLIn IHUHQW HUURU VLJQDO LV SURGXFHG ZKLFK LV SKDVHVKLIWHG DQG PDJQLWXGHVKLIWHGf ZLWK UHVSHFW WR 9 7KH ORRS ZLOO WKHQ RVFLOODWH DW D IUHTXHQF\ ZKHUH WKH SKDVHVKLIW WKURXJK +Vf LV WKH RSSRVLWH RI WKH SKDVHVKLIW LQGXFHG EHWZHHQ 9R DQG ( 3KDVRU UHSUHVHQWDWLRQV FDQ EH XVHG WR LOOXVWUDWH WKH FLUFXLW RSHUDWLRQ ZKHUH ( LV WKH YHFWRU VXP RI WKH SKDVRUV 9/ DQG 9 DV VKRZQ LQ )LJXUH (OEf +HUH 9/ LV DVVXPHG WR ‘GHGW )LJXUH (O Df %DVLF IHHGEDFN PRGHO RI DQ LQMHFWLRQORFNHG RVFLOODWRU Ef 3KDVRU UHSUHVHQWDWLRQ RI WKH V\VWHP

PAGE 271

EH DW UHVW ZLWK UHVSHFW WR WKH REVHUYHU WKXV WKH SODQH LWVHOI LV URWDWLQJ DW DQ DQJXODU IUHn TXHQF\ RI FR/ $OWHUQDWHO\ WKH DFWXDO VLQXVRLGDO YROWDJH VLJQDOV ZLOO EH WKH SURMHFWLRQ RI WKHVH SKDVRUV RQWR D OLQH ZKLFK LV URWDWLQJ DW FR/ %RWK 94 DQG ( ZLOO URWDWH LQ WKLV UHIHUn HQFH SODQH DW D YHORFLW\ RI A FRUUHVSRQGLQJ WR DQ DQJXODU IUHTXHQF\ RI FR/A ( 'LIIHUHQWLDO /RFNLQJ (TXDWLRQ )URP )LJXUH (OEf ff FDQ EH UHODWHG WR 9/ DQG 9 E\ WKH IROORZLQJ WDQMf 9AVLQ* 9 T 9OFRV4 (Of 7KH RSHQORRS RVFLOODWRU FLUFXLWf§WKDW ZKLFK LV UHSUHVHQWHG E\ +Vff§LV DVVXPHG WR EH JHQn HUDWLQJ D SKDVH VKLIW ZKLFK GHFUHDVHV OLQHDUO\ YHUVXV IUHTXHQF\ DURXQG WKH EDQG RI LQWHUn HVW 7KH VORSH RI WKLV SKDVH YHUVXV IUHTXHQF\ UHVSRQVH LV $ GMf Ge! (f )RU D WXQHG FLUFXLW ZLWK TXDOLW\ IDFWRU 4 $ fFGR 7KHUHIRUH DW DQ LQSXW IUHTXHQF\ G% RI WR FR/f§ +Vf JHQHUDWHV D SKDVH VKLIW ZLWK UHVSHFW WR WKH SKDVH DW FRf RI f! $WR FRf DmO MW a f $LRff (ff ZKHUH $R LV WKH GLIIHUHQFH EHWZHHQ WKH QDWXUDO DQG ORFNLQJ IUHTXHQFLHV JFROf 3OXJn JLQJ WKLV LQWR (Of DQG PDNLQJ D VPDOODQJOH DSSUR[LPDWLRQ RI WDQ f! f! UHVXOWV LQ WKH ORFNLQJ HTXDWLRQ IRU DQ ,/2 DV IROORZV >$GO 3DF@ G4 B BL?B VLQ GWa f 8n 9-n O 9/\RfFRVf (f

PAGE 272

( /RFNLQJ 5DQJH DQG /RFNLQJ 6LJQDO /HYHO 7KH ,/2 LV ORFNHG ZKHQ LV FRQVWDQW ZKLFK LV ZKHQ A 7KHUHIRUH WKH ORFN GW LQJ UDQJH $FR FDQ EH REWDLQHG IURP (f DV IROORZV VLQ $ 9-? 9O9fFRV4n )RU VPDOO LQSXW ORFNLQJ VLJQDOV 9/m9Rf WKLV ORFNLQJ UDQJH UHGXFHV WR (f LnAfVLQf (nf $V FDQ EH VHHQ WKH ORFNLQJ UDQJH LQFUHDVHV DV 9/ LQFUHDVHV $OVR DV WKH VORSH RI WKH SKDVH VKLIW YHUVXV IUHTXHQF\ GHFUHDVHV WKH ORFNLQJ UDQJH LQFUHDVHV 7KLV FRUUHVSRQGV WR D ORZHU HIIHFWLYH 4 RI WKH FLUFXLW 7KH UHTXLUHG LQSXW ORFNLQJ VLJQDO OHYHO FDQ EH REWDLQHG IURP (f E\ NQRZLQJ WKDW ,VLQ LV DOZD\V OHVV WKDQ UHVXOWLQJ LQ WKH IROORZLQJ A $FRf$ f Y R (f ( 6WHDG\6WDWH 3KDVH (UURU 7KH VWHDG\VWDWH SKDVH HUURU EHWZHHQ WKH LQSXW ORFNLQJ VLJQDO DQG WKH ,/2 RXWSXW VLJQDO LV JLYHQ DV IROORZV >3DF@ ? 4VV VLQ O $FR$ -O $DR$f 99 VLQ O $FR$ DL $p R$< (f R $W WKH QDWXUDO IUHTXHQF\ RI WKH ,/2 66 LV ]HUR 7KLV VWHDG\VWDWH SKDVH HUURU ZLOO UHVXOW LQ D VNHZ IRU WKH ,/2 ZKHQ LW LV XVHG DV D IUHTXHQF\ GLYLGHU LQ WKH FORFN UHFHLYHU 7KLV VNHZ ZLOO GHSHQG RQ WKH VLJQDO OHYHO RI WKH LQMHFWHG VLJQDO 9/ )RU D VXSHUKDUPRQLF ,/2 XVHG DV

PAGE 273

D IUHTXHQF\ GLYLGHU WKLV SKDVH GLIIHUHQFH ZLOO EH ZLWK UHVSHFW WR WKH LQSXW IUHTXHQF\ WKXV WKH RXWSXW VWHDG\VWDWH SKDVH HUURU LV WKDW JLYHQ LQ (f GLYLGHG E\ 1 ( 3KDVH 1RLVH RI ,QMHFWLRQ/RFNHG 2VFLOODWRUV 7KH SKDVH WUDQVIHUIXQFWLRQ RI DQ (/2 ZLWK D QRLV\ VLJQDO ZLWK DPSOLWXGH 9Q DGGHG DW WKH LQSXW FDQ EH VKRZQ WR EH >/OR 5DW .XU@ 2 RXW FRVH $9 VV YQ pLQ A 9O !On ƒ7YcFRVr (f 7KH ODVW WHUP 9Q9/ UHSUHVHQWV WKH LQSXW SKDVH QRLVH 7KH GLYLVLRQ E\ WKH LQSXW LQMHFWHG SRZHU OHYHO LV GXH WR WKH FRQYHUVLRQ IURP LQSXW DGGLWLYH QRLVH WR SKDVH QRLVH >(JD@ 7DNn LQJ WKH PDJQLWXGH VTXDUHG RI (f WKH RXWSXW SKDVH QRLVH RI WKH ,/2 LV fmr! AA f (f ZKHUH 6RXW DQG 6cQ DUH WKH RXWSXW DQG LQSXW SRZHU VSHFWUDO GHQVLWLHV 36'fVf RI WKH SKDVH DQG % FDQ EH VKRZQ WR EH V MAFRVf DAfDf}f (f ,I WKH ,/2 LV EHLQJ XVHG DV D IUHTXHQF\ GLYLGHU WKH RXWSXW SKDVH 36' LQ (f LV GLYLGHG E\ 1 ZKHUH 1 LV WKH GLYLVLRQ UDWLR >5DW@ ([DPLQLQJ (f LW FDQ EH VHHQ WKDW WKH ,/2 EHKDYHV OLNH D ILUVWRUGHU SKDVHORFNHG ORRS ZLWK UHVSHFW WR LQSXW SKDVH QRLVH >5DW@ ILOn WHULQJ WKH QRLVH ZLWK D ORZSDVV UHVSRQVH +HUH WKH EDQGZLGWK RI WKH HTXLYDOHQW ORZSDVV ILOWHU LV D IXQFWLRQ RI WKH LQSXW LQMHFWLRQ OHYHO

PAGE 274

$33(1',; ) 48$/,7< )$&725 2) 5,1* 26&,//$725 ,W LV XVHIXO WR PRGHO WKH TXDOLW\ IDFWRU 4f RI D ULQJ RVFLOODWRU 7KLV 4 FDQ WKHQ EH DSSOLHG WR WKH UHVXOWV SUHVHQWHG LQ $SSHQGL[ ( RQ LQMHFWLRQORFNHG RVFLOODWRUV 8VLQJ WKH PRGHO SURYLGHG LQ >(JD@ D ULQJ RVFLOODWRU FDQ EH PRGHOHG DV DQ LGHDO LQYHUWHU ZLWK ]HUR GHOD\ SURYLGLQJ D r SKDVH VKLIWf LQ D ORRS ZLWK DQ HOHPHQW ZKLFK KDV D GHOD\ RI 7 DV VKRZQ LQ )LJXUH )O 7KH GHOD\ HOHPHQW FDXVHV D SKDVHVKLIW LQ WKH ORRS HTXDO WR FR? )Of 7KH ORRS ZLOO RVFLOODWH DW WKH IUHTXHQF\ ZKHUH LV HTXDO WR Q VLQFH WKH LQYHUWHU SURYLGHV WKH RWKHU r SKDVH VKLIW 7KXV WKH RVFLOODWLRQ IUHTXHQF\ LV HTXDO WR 7KH TXDOLW\ IDFWRU 4f RI DQ RVFLOODWRU LV UHODWHG WR WKH RVFLOODWLRQ IUHTXHQF\ DQG SKDVH VKLIW WKURXJK WKH RVFLOODWRU DV IROORZV >5D]@ )f ,GHDO ,QYHUWHU QR GHOD\f 'HOD\ 7 )LJXUH )O 0RGHO RI ULQJ RVFLOODWRU DV DQ LGHDO LQYHUWHU IROORZHG E\ D GHOD\ HOHPHQW XVHG WR ILQG WKH 4 RI D ULQJ RVFLOODWRU

PAGE 275

7DNLQJ WKH GHULYDWLYH RI )Of DQG DSSO\LQJ WKH UHVXOW WR )f UHVXOWV LQ >(JD@ 4 Q n )f 7KXV ULQJ RVFLOODWRUV KDYH YHU\ ORZ 4fV $V D UHVXOW WKH\ FDQ EH LQMHFWLRQ ORFNHG RYHU D YHU\ ODUJH IUHTXHQF\ VLQFH WKH ORFNLQJ IUHTXHQF\ UDQJH LV LQYHUVHO\ GHSHQGHQW RQ 4 DV VKRZQ LQ $SSHQGL[ ( $OVR GXH WR WKH ORZ 4 ULQJRVFLOODWRUV W\SLFDOO\ KDYH SRRU SKDVH QRLVH SHUIRUPDQFH UHVXOWLQJ LQ KLJKHU MLWWHU

PAGE 276

$33(1',; 5(/$7,216+,3 %(7:((1 -,77(5 $1' 3+$6( 12,6( 7KH URRWPHDQVTXDUHG SKDVH MLWWHU LV GHILQHG DV WKH VTXDUH URRW RI WKH YDULDQFH RI WKH GLIIHUHQFH EHWZHHQ WKH SKDVH RI D VLJQDO DW WKH EHJLQQLQJ DQG HQG RI D SHULRG 7 >(JD@ 7KLV YDULDQFH LV &c e^>W! UfMf@` *Lf 7KLV LV FRQYHUWHG WR MLWWHU LQ WLPH E\ VLPSO\ GLYLGLQJ WKH SKDVH MLWWHU YDULDQFH E\ WKH VTXDUH RI WKH DQJXODU IUHTXHQF\ RI WKH VLJQDO 7KH DXWRFRUUHODWLRQ IXQFWLRQ RI Mf LV (P [0Wf@ IBAIfHMQIOGI *f ZKHUH /AMM LV WKH VLQJOHVLGHEDQG SKDVH QRLVH GHQVLW\ DQG 5[f DQG e-cf DUH D IRXULHU WUDQVIRUP SDLU 7KH VLQJOHVLGHEDQG GHQVLW\ LV RQH KDOI WKH GRXEOHVLGHEDQG SKDVH SRZHU VSHFWUDO GHQVLW\ 6AIf 7KHUHIRUH WKH MLWWHU YDULDQFH LQ WLPH FDQ EH H[SUHVVHG DV >+DM@ f§ LI 0IfHMGII /IfHMQIUGI? *f LF If -f ‘ c‘ ^F ZZ ‘ -n ^-7nLrr1 6AL.-7fI OI M 0I 6f>FRVN7f@ Wff O-r r f§Af§IVMIf6P.M7fGI .II -} A

PAGE 277

/,67 2) 5()(5(1&(6 >$GO@ 5 $GOHU f$ VWXG\ RI ORFNLQJ SKHQRPHQD LQ RVFLOODWRUVf 3URF ,5( YRO SS -XQH >$@ $OOHU %HUQVWHLQ 8 *KRVKDO + 6FKHWWOHU 6 6FKXVWHU < 7DXU DQG 7RUUHLWHU f&026 FLUFXLW WHFKQRORJ\ IRU VXEDPELHQW WHPSHUDWXUH RSHUDn WLRQf ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >%DO@ & $ %DODQLV $QWHQQD 7KHRU\ $QDO\VLV DQG 'HVLJQ QG HGLWLRQ 1HZ %DQ@ 5 %DQVDO f7KH IDUILHOG KRZ IDU LV IDU HQRXJK"f $SSOLHG 0LFURZDYH DQG :LUHOHVV SS 1RY >%HO@ $ %HOODRXDU DQG 0 (OPDVU\ /RZ3RZHU 'LJLWDO 9/6, 'HVLJQ &LUFXLWV DQG 6\VWHPV %RVWRQ 0$ .OXZHU $FDGHPLF 3XEOLVKHUV SS f >%L\@ 0 %L\DQL f$ GLYLGHE\ GXDO PRGXOXV SUHVFDOHU XVLQJ D ILP &026 SURFHVVf 0 6 GLVVHUWDWLRQ 8QLYHUVLW\ RI )ORULGD *DLQHVYLOOH )/ >%RD@ $ %RDUGPDQ (OHFWURPDJQHWLF 6XUIDFH 0RGHV 1HZ %RH@ : %RHUVWOHU f$ ORZMLWWHU 3// FORFN JHQHUDWRU IRU PLFURSURFHVVRUV ZLWK ORFN UDQJH RI *+]f ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS $SU >%RK@ 0 7 %RKU f,QWHUFRQQHFW VFDOLQJf§WKH UHDO OLPLWHU WR KLJK SHUIRUPDQFH 8/6,f ,('0 7HFKQLFDO 'LJHVW SS 'HF >%RZ@ & %RZLFN 5) &LUFXLW 'HVLJQ ,QGLDQDSROLV ,1 +: 6DPV 3XEOLVKLQJ >%UDY22D@ %UDYR f,QYHVWLJDWLRQ RI EDFNJURXQG QRLVH LQ LQWHJUDWHG FLUFXLWV UHODWLQJ WR WKH GHVLJQ RI DQ RQFKLS ZLUHOHVV LQWHUFRQQHFW V\VWHPf 0 6 GLVVHUWDn WLRQ 8QLYHUVLW\ RI )ORULGD *DLQHVYLOOH )/

PAGE 278

>%UDY22E@ >%XU@ >&KH@ >&KH@ >&RO@ >&UD@ >'HX@ >(JD@ >(JD@ >(JD@ >)DO2O@ %UDYR +
PAGE 279

>)ORD@ % $ )OR\G 0HKWD & *DPHUR DQG . 2 f$ 0+] _LP &026 ORZ QRLVH DPSOLILHU ZLWK G% QRLVH ILJXUHf 3URF ,((( &XVWRP ,QWHJUDWHG &LUFXLWV &RQIHUHQFH SS 0D\ >)ORE@ % $ )OR\G DQG . 2 f7KH SURMHFWHG SRZHU FRQVXPSWLRQ RI D ZLUHOHVV FORFN GLVWULEXWLRQ V\VWHP DQG FRPSDULVRQ WR FRQYHQWLRQDO GLVWULEXWLRQ V\Vn WHPVf 3URF ,QWHUQDWLRQDO ,QWHUFRQQHFW 7HFKQRORJ\ &RQIHUHQFH SS 0D\ >)OR22D@ % $ )OR\G .LP DQG . 2 f:LUHOHVV LQWHUFRQQHFWLRQ LQ D &026 ,& ZLWK LQWHJUDWHG DQWHQQDVf ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >)OR22E@ % $ )OR\G &0 +XQJ DQG . 2 f7KH HIIHFWV RI VXEVWUDWH UHVLVWLYLW\ RQ 5) FRPSRQHQW DQG FLUFXLW SHUIRUPDQFHf 3URF ,QWHUQDWLRQDO ,QWHUFRQn QHFW 7HFKQRORJ\ &RQIHUHQFH SS 0D\ >)OR2OD@ % $ )OR\G &0 +XQJ DQG . 2 f$ *+] ZLUHOHVV LQWHUFRQQHFW LPSOHPHQWHG LQ D SP &026 WHFKQRORJ\ XVLQJ LQWHJUDWHG WUDQVPLWn WHUV UHFHLYHUV DQG DQWHQQDVf 6\PS 9/6, &LUFXLWV 'LJ 7HFK 3DSHUV -XQH >)OR2OE@ % $ )OR\G / 6KL < 7DXU /DJQDGR DQG . 2 f$ *+] 62, &026 WXQHG DPSOLILHUf VXEPLWWHG IRU SXEOLFDWLRQ WR ,((( 7UDQV RQ 0LFURZDYH 7KHRU\ DQG 7HFKQLTXHV >)OR2OF@ % $ )OR\G / 6KL < 7DXU /DJQDGR DQG . 2 f62, DQG EXON &026 IUHTXHQF\ GLYLGHUV RSHUDWLQJ DERYH *+]f VXEPLWWHG IRU SXEOLn FDWLRQ WR ,(( (OHFWURQLFV /HWWHUV >)RV@ )RVVXP f62,63,&( 9HU f )'62, DQG 1)'62, 026)(7 0RGHOVff 8QLYHUVLW\ RI )ORULGD *DLQHVYLOOH )/ -DQ >*LH@ % $ *LHVHNH 5 / $OLPQ : %DLOH\ % %HQVFKQHLGHU 6 0 %ULWn WRQ &ORXVHU + 5 )DLU P $ )DUUHOO 0 *RZDQ & / +RXJKn WRQ % .HOOHU 7 + /HH / /HLEKRO] 6 & /RZHOO 0 0DWVRQ 5 0DWWKHZ 9 3HQJ 4XLQQ f$ 0+] VXSHUVFDODU 5,6& PLFURSURFHVn VRU ZLWK RXWRIRUGHU H[HFXWLRQf ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >*RQ@ *RQ]DOH] 0LFURZDYH 7UDQVLVWRU $PSOLILHUV $QDO\VLV DQG 'HVLJQ QG HGLWLRQ 8SSHU 6DGGOH 5LYHU 13UHQWLFH+DOO >*UD@ 3 5 *UD\ DQG 5 0H\HU $QDO\VLV DQG 'HVLJQ RI $QDORJ ,QWHJUDWHG &LUn FXLWV UG HGLWLRQ 1HZ
PAGE 280

>*UD22@ *UDPHJQD $ 0DJD]] & 6FODIDQL DQG 0 3DSDUR f8OWUDZLGH G\QDPLF UDQJH G% QRLVHILJXUH 0+] &026 /1$f ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >*XR2O@ ; *XR 5 /L DQG f,PSDFW RI D SURSDJDWLQJ OD\HU XQGHUQHDWK D VLOLn FRQ ZDIHU RQ WKH SHUIRUPDQFH RI LQWHJUDWHG DQWHQQDVf WR EH SXEOLVKHG >*XW22@ 9 *XWQLN DQG $ 3 &KDQGUDNDVDQ f$FWLYH *+] FORFN QHWZRUN XVLQJ GLVn WULEXWHG 3//Vf ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS 1RY >+DM@ $ -DKLPLUL 6 /LPRW\UDNLV DQG 7 + /HH f-LWWHU DQG SKDVH QRLVH LQ ULQJ RVFLOODWRUVf ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS -XQH >+DU@ +DUWPDQQ DQG 0 6WUXWW f&KDQJHV RI WKH IRXU QRLVH SDUDPHWHUV GXH WR JHQHUDO FKDQJHV RI OLQHDU WZRSRUW FLUFXLWVf ,((( 7UDQV (OHFWURQ 'HYLFHV YRO SS 2FW >+D\@ +D\DVKL + .LPXUD + 6LPRPXUD DQG $ 0DWVX]DZD f$ P: 0+] &026 /1$ ZLWK PHVK DUUD\HG 026)(7Vf 6\PS 9/6, &LUFXLWV 'LJ 7HFK 3DSHUV SS -XQH >+R@ <& +R .LP % )OR\G & :DQQ < 7DXU /DJQDGR DQG 2 f *+] DQG *+] WXQHG DPSOLILHUV LPSOHPHQWHG LQ D _LP &026 WHFKn QRORJ\ RQ 62, 626 DQG EXON VXEVWUDWHVf ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS 'HF >+R22@ <& +R f,PSOHPHQWDWLRQ DQG LPSURYHPHQW IRU 5) ORZ QRLVH DPSOLILHUV LQ FRQYHQWLRQDO &026 WHFKQRORJLHVf 3K 'LVVHUWDWLRQ 8QLYHUVLW\ RI )ORULGD *DLQHVYLOOH )/ >+XD@ 4 +XDQJ 3 2UVDWWL DQG ) 3LD]]D f%URDGEDQG _LP &026 /1$V ZLWK VXEG% 1) IRU *60 DSSOLFDWLRQVf 3URF &XVWRP ,QWHJUDWHG &LUFXLWV &RQIHUHQFH SS 0D\ >+XQ@ &0 +XQJ <& +R DQG 2 f+LJK4 FDSDFLWRUV LPSOHPHQWHG LQ D &026 SURFHVV IRU ORZSRZHU ZLUHOHVV DSSOLFDWLRQVf ,((( 7UDQV 0LFUR ZDYH 7KHRU\ DQG 7HFKQLTXHV YRO SS 0D\ >+XQ22D@ &0 +XQJ f,QYHVWLJDWLRQ RI D PXOWL*+] VLQJOHFKLS &026 3// IUHn TXHQF\ V\QWKHVL]HU IRU ZLUHOHVV DSSOLFDWLRQVf 3K 'LVVHUWDWLRQ 8QLYHUn VLW\ RI )ORULGD *DLQHVYLOOH )/

PAGE 281

>+XQ22E@ &0 +XQJ / 6KL /DJQDGR DQG . f*+] YROWDJHFRQWUROOHG RVFLOODWRU IDEULFDWHG LQ D &026 SURFHVVf 6\PS 9/6, &LUFXLWV 'LJ 7HFK 3DSHUV SS -XQH >+XQ22F@ &0 +XQJ DQG .HQQHWK 2 f3DFNDJHG *+] &026 9&2 ZLWK SKDVH QRLVH RI G%F+] DW D N+] RIIVHWf ,((( 6ROLG6WDWH &LUn FXLWV YRO QR SS -DQ >+XQ2O@ &0 +XQJ % )OR\G DQG f)XOO\ LQWHJUDWHG &026 9&2V DQG SUHV FDOHUVf ,((( 7UDQV RQ 0LFURZDYH 7KHRU\ DQG 7HFKQLTXHV YRO QR SS -DQ >-LQ@ 5 3 -LQGDO f+LJK IUHTXHQF\ QRLVH LQ ILQH OLQH 1026 ILHOG HIIHFW WUDQVLVn WRUVf ,('0 7HFKQLFDO 'LJHVW SS 'HF >.DH@ 9 5 YRQ .DHQHO f$ KLJKVSHHG ORZSRZHU FORFN JHQHUDWRU IRU D PLFURn SURFHVVRU DSSOLFDWLRQf ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS 1RY >.LP22D@ .LP f'HVLJQ DQG FKDUDFWHUL]DWLRQ RI 5) FRPSRQHQWV IRU LQWHU DQG LQWUDFKLS ZLUHOHVV FRPPXQLFDWLRQVf 3K' 'LVVHUWDWLRQ 8QLYHUVLW\ RI )ORULGD *DLQHVYLOOH )/ >.LP22E@ .LP + .LV@ 6 9 .LVKRUH &KDQJ $VPDQLV & +XOO DQG ) 6WXEEH f6XEVWUDWH LQGXFHG KLJKIUHTXHQF\ QRLVH LQ GHHS VXEPLFURQ 026)(7V IRU 5) DSSOLn FDWLRQVf 3URF &XVWRP ,QWHJUDWHG &LUFXLWV &RQIHUHQFH SS 0D\ >.OH@ % .OHYHODQG & + 'LD] 9RRN / 0DGGHQ 7 + /HH DQG 6 6 :RQJ f0RQROLWKLF &026 GLVWULEXWHG DPSOLILHU DQG RVFLOODWRUf ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >.QD22@ + .QDSS 7 ) 0HLVWHU 0 :XU]HU =RVFKJ $XILQJHU DQG / 7UHLW LQJHU f$ *+] G\QDPLF IUHTXHQF\ GLYLGHU LQ 6L*H ELSRODU WHFKQRORJ\f ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >.XR2O@ 7 & .XR DQG % % /XVLJQDQ f$ : FODVV) 5) SRZHU DPSOLILHU LQ SP &026 WHFKQRORJ\f ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >.XU@ .XURNDZD f1RLVH LQ V\QFKURQL]HG RVFLOODWRUVf ,((( 7UDQV RQ 0LFUR ZDYH 7KHRU\ DQG 7HFKQLTXHV YRO QR SS $SU

PAGE 282

>.XU@ 0 .XULVX 0 1LVKLNDZD + $VD]DZD $ 7DQDEH 0 7RJR DQG $ )XUXNDZD f$Q *+] P: &026 IUHTXHQF\ GLYLGHUf 6\PS 9/6, &LUFXLWV 'LJ 7HFK 3DSHUV SS -XQH >/HH@ % /HHVRQ f$ VLPSOH PRGHO RI IHHGEDFN RVFLOODWRU QRLVH VSHFWUXPf 3URF ,((( YRO SS )HE >/HH@ 7 + /HH 7KH 'HVLJQ RI &026 5DGLR )UHTXHQF\ ,QWHJUDWHG &LUFXLWV &DPEULGJH 8. &DPEULGJH 8QLYHUVLW\ 3UHVV >/OR@ 2 /ORSLV 0 5HJLV 6 'HVJUHD DQG *UDIIHXLO f3KDVH QRLVH SHUIRUn PDQFH RI PLFURZDYH DQDORJ IUHTXHQF\ GLYLGHUV DSSOLFDWLRQ WR WKH FKDUDFn WHUL]DWLRQ RI RVFLOODWRUV XS WR WKH PPZDYH UDQJHf ,((( ,QW )UHTXHQF\ &RQWURO 6\PS SS 0D\ >/RQ@ 5 /RQJ DQG 0 $ &RSHODQG f7KH PRGHOLQJ FKDUDFWHUL]DWLRQ DQG GHVLJQ RI PRQROLWKLF LQGXFWRUV IRU VLOLFRQ 5) ,&fVf ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS 0DU >0DL@ & $ 0DLHU $ 0DUNHYLWFK & 6 %UDVKHDUV 7 6LSSHO ( 7 &RKHQ + %ORPJUHQ %DOODUG 3DWWLQ 9 0ROGHKQDXHU $ 7KRPDV DQG 7D\ORU f$ 0+] %L&026 VXSHUVFDODU 5,6& PLFURSURFHVVRUf ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS 1RY >0DQ@ 7 0DQNX f0LFURZDYH &026f§GHYLFHV DQG FLUFXLWVf 3URF &XVWRP ,QWHn JUDWHG &LUFXLWV &RQIHUHQFH SS 0D\ >0HD@ & 0HDG DQG / &RQZD\ ,QWURGXFWLRQ WR 9/6, 6\VWHPV 5HDGLQJ 0$ $GGLVRQ:HVOH\ SS >0HK@ 0HKWD f$Q LQYHVWLJDWLRQ RI EDFNJURXQG QRLVH LQ ,&V DQG LWV LPSDFW RQ ZLUHOHVV FORFN GLVWULEXWLRQf 0DVWHUfV 7KHVLV 8QLYHUVLW\ RI )ORULGD *DLQHVYLOOH >0LO@ $ % 0LOOHU f2SWLFDO LQWHUFRQQHFW WHFKQRORJLHV IRU 6L 8/6,f ,('0 7HFKQLFDO 'LJHVW SS 'HF >0XU@ 0XUDWD 7 2WVXML ( 6DQR 0 2KKDWD 0 7RJDVKL DQG 0 6X]XNL f$ QRYHO KLJKVSHHG ODWFKLQJ RSHUDWLRQ IOLSIORS +/2))f FLUFXLW DQG LWV DSSOLFDWLRQ WR D *EV GHFLVLRQ FLUFXLW XVLQJ D SP *D$V 0(6)(7f ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS 2FW >@ . 2 .LP % )OR\G DQG 0HKWD f,QWHU DQG LQWUDFKLS FORFN VLJQDO GLVWULEXWLRQ XVLQJ PLFURZDYHVf ,((( 6ROLG 6WDWH &LUFXLWV DQG 7HFKn QRORJ\ &RPPLWWHH :RUNVKRS RQ &ORFN 'LVWULEXWLRQ 2FW $WODQWD *$

PAGE 283

>@ . f(VWLPDWLRQ PHWKRGV IRU TXDOLW\ IDFWRUV RI LQGXFWRUV IDEULFDWHG LQ VLOLFRQ LQWHJUDWHG FLUFXLW SURFHVV WHFKQRORJLHVf ,((( 6ROLG6WDWH &LUn FXLWV YRO QR SS $XJ >@ . 2 .LP % $ )OR\G 0HKWD DQG + 3DF@ / 3DFLRUHN f,QMHFWLRQ ORFNLQJ RI RVFLOODWRUVf 3URF ,((( YRO QR SS 1RY >3HO@ 0 0 3HOHOOD )RVVXP 'RQJZRRN 6XK 6 .ULVKQDQ $ -HQNLQV f/RZYROWDJH WUDQVLHQW ELSRODU HIIHFW LQGXFHG E\ G\QDPLF IORDWLQJERG\ FKDUJLQJ LQ 3'62, 026)(76f 3URF ,((( ,QW 62, &RQI SS 2FW >3UD@ $ 3UDWW DQG 1JX\HQ f'LVWULEXWHG V\QFKURQRXV FORFNLQJf ,((( 7UDQV 3DUDOOHO DQG 'LVWULEXWHG 6\VWHPV YRO QR SS 0DU >5DW@ + 5 5DWHJK DQG 7 + /HH f6XSHUKDUPRQLF LQMHFWLRQORFNHG IUHTXHQF\ GLYLGHUVf ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS -XQH >5D]@ % 5D]DYL ) /HH DQG 5 + 5D]@ % 5D]DYL ) /HH DQG 5 + 5HV@ 3 5HVWOH DQG $ 'HXWVFK f'HVLJQLQJ WKH %HVW &ORFN 'LVWULEXWLRQ 1HWn ZRUNf 6\PS 9/6, &LUFXLWV 'LJ 7HFK 3DSHUV SS -XQH >6FK@ 6FKPLGHJ f+DUPRQLF V\QFKURQL]DWLRQ RI QRQOLQHDU RVFLOODWRUVf 3URF ,((( SS $XJ >6HG@ $ 6 6HGUD DQG & 6PLWK 0LFURHOHFWURQLF &LUFXLWV UG HGLWLRQ )RUW :RUWK 7; 6DXQGHUV &ROOHJH 3XEOLVKLQJ SS >6KD@ 6KDHIIHU DQG 7 /HH f$ 9 *+] &026 ORZ QRLVH DPSOLILHUf ,((( 6ROLG6WDWH &LUFXLWV YRO SS 0D\

PAGE 284

>6KL2O@ $ 6KLUYDQL 6X % $ :RROH\ f$ &026 5) SRZHU DPSOLILHU ZLWK SDUDOOHO DPSOLILFDWLRQ IRU HIILFLHQW SRZHU FRQWUROf ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >6,$@ 6HPLFRQGXFWRU ,QGXVWU\ $VVRFLDWLRQ 7KH ,QWHUQDWLRQDO 7HFKQRORJ\ 5RDGn PDS IRU 6HPLFRQGXFWRUV 6DQ -RVH &$ 6,$ >6PL@ 5 6PLWK 0RGHP &RPPXQLFDWLRQ &LUFXLWV QG HGLWLRQ %RVWRQ 0$ 0F*UDZ+LOO SS >6WX@ ) 6WXEEH 6 9 .LVKRUH & +XOO DQG 9 7RUUH f$ &026 5) IURQWHQG IRU *+] DSSOLFDWLRQVf 6\PS RQ 9/6, &LUFXLWV 'LJ 7HFK 3DSHUV SS -XQH >6XK@ 6XK DQG )RVVXP f'\QDPLF IORDWLQJERG\ LQVWDELOLWLHV LQ SDUWLDOO\ GHSOHWHG 62, &026 FLUFXLWVf ,('0 7HFKQLFDO 'LJHVW 'HF >6XK@ 6XK DQG )RVVXP f$ SK\VLFDO FKDUJHEDVHG PRGHO IRU QRQIXOO\ GHSOHWHG 62, 026)(7fV DQG LWV XVH LQ DVVHVVLQJ IORDWLQJERG\ HIIHFWV LQ 62, &026 FLUFXLWVf ,((( 7UDQV (OHFWURQ 'HYLFHV YRO QR SS $SULO >6XW@ + 6XWRK 7DP22@ 6 7DP 6 5XVX 8 1 'HVDL 5 .LP =KDQJ DQG 7DX@ < 7DXU DQG 7 + 1LQJ )XQGDPHQWDOV RI 0RGHP 9/6, 'HYLFHV VW (GLn WLRQ SS 1HZ 7VL@ < 7VLYLGLV 2SHUDWLRQ DQG 0RGHOLQJ RI WKH 026 7UDQVLVWRU QG HGLWLRQ %RVWRQ 0$ 0F*UDZ +LOO SS >8]X@ 9 8]XQRJOX DQG 0 + :KLWH f7KH V\QFKURQRXV RVFLOODWRU D V\QFKURQL]Dn WLRQ DQG WUDFNLQJ QHWZRUNf ,((( 6ROLG6WDWH &LUFXLWV YRO VF QR SS 'HF >:DQ@ 5 :DQJVQHVV (OHFWURPDJQHWLF )LHOGV QG HGLWLRQ 1HZ :DQ22@ + 0 :DQJ f$ 9 P: *+] IUHTXHQF\ GLYLGHU LQ SP &026f ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE

PAGE 285

>:DU@ :DUH +6 /HH DQG & 6RGLQL f$ 0+] &026 SKDVHORFNHG ORRS ZLWK GXDO SKDVH GHWHFWRUVf ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS 'HF >:DV D@ :DVKLR ( 2KXH 2GD 5 +D\DPL 0 7DQDEH + 6KLPDPRWR 7 +DUDGD 0 .RQGR f *+] G\QDPLF IUHTXHQF\ GLYLGHU LQ SV (&/ 6L*( +%7Vf ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >:DV22E@ :DVKLR 5 +D\DPL ( 2KXH 2GD 0 7DQDEH + 6KLPDPRWR DQG 0 .RQGR f*+] VWDWLF IUHTXHQF\ GLYLGHU XVLQJ _LP VHOIDOLJQHG 6L*H +%7Vf ,((( 5DGLR )UHTXHQF\ ,QWHJUDWHG &LUFXLWV 6\PS SS -XQH >:HE@ & ) :HEE & $QGHUVRQ / 6LJDO / 6KHSDUG 6 /LSWD\ :DU QRFN % &XUUDQ % : .UXPP 0 0D\R 5 &DPSRUHVH ( 0 6FKZDU] 0 6 )DUUHOO 5 5HVWOH 5 0 $YHULOO ,,, 7 6LHJHO : 9 +RXWW < + &KDQ % :LOH 1JX\H f$ 0+] 6 PLFURSURFHVVRUf ,((( 6ROLG6WDWH &LUFXLWV YRO QR SS 1RY >:HV@ 1 :HVWH DQG (VKUDJKLDQ 3ULQFLSOHV RI &026 9/6, 'HVLJQ $ 6\VWHPV 3HUVSHFWLYH QG (GLWLRQ 5HDGLQJ 0$ $GGLVRQ :HVOH\ >:HW22@ 0 :HW]HO / 6KL -HQNLQV 5 GHO OD +RXVVD\H < 7DXU 3 $VEHFN DQG /DJQDGR f$ *+] VLOLFRQ 026)(7 G\QDPLF IUHTXHQF\ GLYLGHUf ,((( 0LFURZDYH DQG *XLGHG :DYH /HWWHUV YRO QR SS 2FW >:XU22@ 0 :XU]HU 7 0HLVWHU + .QDSS $XILQJHU 5 6FKUHLWHU 6 %RJXWK / 7UHLWLQJHU f *+] VWDWLF IUHTXHQF\ GLYLGHU LQ D 6L6L*H ELSRODU WHFKQROn RJ\f ,66&& 'LJHVW 7HFKQLFDO 3DSHUV SS )HE >'3@f ODWFK DQG LWV DSSOLFDWLRQ LQ D GXDOPRGXn OXV SUHVFDOHUf ,((( 6ROLG 6WDWH &LUFXLWV YRO QR SS 2FW >
PAGE 286

>=KD@ ; =KDQJ ; =KRX % $OLHQHU DQG $ 6 'DU\RXVK f$ VWXG\ RI VXEKDUn PRQLF LQMHFWLRQ ORFNLQJ IRU ORFDO RVFLOODWRUVf ,((( 0LFURZDYH DQG *XLGHG :DYH /HWWHUV YRO QR SS 2FW >=LH@ $ YDQ GHU =LHO 1RLVH LQ 6ROLG 6WDWH 'HYLFHV DQG &LUFXLWV 1HZ
PAGE 287

%,2*5$3+,&$/ 6.(7&+ %ULDQ $ )OR\G ZDV ERP LQ :HVW 3DOP %HDFK )ORULGD RQ $XJXVW +H UHFHLYHG WKH %6 ZLWK KLJKHVW KRQRUV DQG WKH 0(QJ GHJUHHV LQ HOHFWULFDO DQG FRPSXWHU HQJLQHHULQJ IURP WKH 8QLYHUVLW\ RI )ORULGD *DLQHVYLOOH LQ DQG UHVSHFWLYHO\ 'XULQJ WKH VXPPHUV RI DQG KH ZRUNHG IRU 0RWRUROD ,QF %R\QWRQ %HDFK )ORULGD LQ WKH DUHDV RI 5) SURGXFW GHYHORSPHQW DQG ,& GHVLJQ ,Q KH MRLQHG WKH 6LOLFRQ 0LFURZDYH ,QWHJUDWHG &LUFXLWV DQG 6\VWHPV 6L0,&6f 5HVHDUFK *URXS DW WKH 8QLYHUVLW\ RI )ORULGD 'HSDUWPHQW RI (OHFWULFDO DQG &RPSXWHU (QJLQHHULQJ )URP KH KHOG WKH 5REHUW & 3LWWPDQ *UDGXDWH )HOORZVKLS ZKLOH IURP KH KHOG WKH ,QWHUVLO DQG 6HPLFRQGXFWRU 5HVHDUFK &RUSRUDWLRQ *UDGXDWH )HOORZVKLS +LV UHVHDUFK LQWHUHVWV LQFOXGH 5) LQWHJUDWHG FLUFXLW GHVLJQ LQ ERWK &026 DQG %L&026 WHFKn QRORJLHV ZLUHOHVV FRPPXQLFDWLRQV DQG KLJKVSHHG FORFN DQG GDWD GLVWULEXWLRQ

PAGE 288

, FHUWLI\ WKDW KDYH UHDG WKLV VWXG\ DQG WKDW LQ P\ RSLQLRQ LW FRQIRUPV WR DFFHSWDEOH VWDQGDUGV RI VFKRODUO\ SUHVHQWDWLRQ DQG LV IXOO\ DGHTXDWH LQ VFRSH DQG TXDOLW\ DV D GLVVHUWDWLRQ IRU WKH GHJUHH RI 'RFWRU RI3KLORVSSKf .HQQHWK .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


158
= SNR(gt) ,
(5.45)
[a,(/0/3)A.n^
where SNR(aT) is given in (5.35). Nonlinearity will generate a third-order harmonic of the
interferer, which will corrupt the SNR. The DP3 is specified for the case when the SNR is
degraded by 3 dB due to this third-order harmonic. Thus, the output power of the
fo
third-order harmonic at fa should be equal to the output power of the interferer at -j-. This
results in
Wt>J [Ht'K,]-
The ratio of j to oc3 can be related to the IIP3 at that frequency [Raz95]. Since ocj is
not a readily-known quantity the following assumption is made:
(5.46)
/oV
1 (fo)
a3(/0)
3.2
~ 4A//P3
(5.47)
This assumption states that the third-order nonlinearity has the same frequency response
fo
as the gain, or equivalently, that IIP3(/"0) is equal to IIP3(-^).
Using (5.45) and (5.47) in (5.46) results in the following expression for AIIP3 (the
input amplitude in Volts corresponding to IIP3 in Watts):
AIIP3 ~ -
4 1 (/,)
3
JSNR(aT)
(5.48)
This can be expressed in dBm as follows:
7/P3I = P\
'dBm sens\dBm
+ GLNA(fo)\dB-GLNA{j)
dB
SNR(ct) I 4.8 (5.49)


177
(a) (b)
time (ns)
Figure 6-9 (a) Measurement setup used to characterize clock receiver, (b) Input and output
waveforms for the clock receiver (fin=7.4 GHz and fout=7-4 GHz/8 = 925 MHz)
for a interconnection distance of 3.3 mm.
transmitted signal for lower input power levels, the minimum detectable signal of the
receiver is -43 dBm. The receiver consumes 62.5 mW from a 2.5-V supply.
This result is the first demonstration of an on-chip wireless interconnect. A
7.4-GHz global clock signal is successfully received over a 3.3-mm distance with interfer
ence structures located in between the antennas, and a 925-MHz local clock signal is gen
erated. Also, this result is the first to integrate antennas and CMOS circuitry on a single
chip. Therefore, plausibility of a wireless system has been demonstrated using a standard
0.25-|i.m CMOS technology at ~8 GHz. Based on this result, superior performance is
expected by using a more advanced CMOS technology. This allows the operating fre
quency to increase, improving the efficiency of the antennas, which in turn allows longer
interconnection distances.


189
was 32.9, while the measured Q was ~10. This reduction in Q was due to the thinning of
the top-layer metals, once again discussed in section 3.5.2. The varactor is implemented
using an accumulation-mode high-Q MOS capacitor [Hun98]. Measurements show that
the varactor Q is approximately 47 at 15 GHz.
A VCO test structure consisting of the VCO core plus a single-stage 50-0 output
buffer was connected to a spectrum analyzer, yielding the output single-sideband
phase-noise plot and spectrum shown in Figure 6-20. With Vcti = 1.2 V, the measured cen
ter frequency is 14.92 GHz, with an output power level of -21.5 dBm. The VCO core con
sumes 7.2 mW from a 1.5-V supply. The phase noises at 1-MHz and 3-MHz offsets are
-105 and -113 dBc/Hz, respectively. Finally, an oscillation frequency versus control volt
age plot is shown in Figure 6-21. The tuning range is 690 MHz, for Vctl between 0 and 1.8
V. The linear portion of this curve reveals a VCO gain of approximately 600 MHz/V.


32
(a) (b)
(c)
(d)
ASSUMPTIONS
y=1.2 lcl = 0.395
>T = 4c0o RSub=10
8 = 2y Qgs = 3.0
9mb = 0.25gm
Cgb = 0.3Cgs
Cgd = 0.2CgS
Cdb = Cgs
Figure 2-7 Noise parameters (a) F^,,, (b) Gn, (c) Ropt, and (d) X^, for the following
normalized parameters: Qgs with GIN on and off, Rsub, gmb, Cdb, Cgb, and Cgd.
For each parameter studied, all other parameters are held constant using the
values listed in the assumption table, and GIN is on (except, obviously, for Qgs,
GIN off case).
increases. Therefore, Fm¡n decreases accordingly. For very low values of Qgs, Fmjn
decreases due to the transistor and its parasitic capacitances being so large. Specifically,
Csb begins to shunt the substrate noise to ground, decreasing its contribution to Fmin. The
value of Qgs for which Fmin is maximum depends on the value of R^.


3
constant for low-K (conventional) dielectrics. This corresponds to a downward shift in the
global delay plotted in Figure 1-1. Unfortunately, since the delay scales with chip area,
despite the addition of copper and low-K dielectrics, global interconnect delay will con
tinue to increase with succeeding generations of microprocessors. Copper and low-K
dielectrics only extend the lifetime of conventional (i.e., traditional conductor) intercon
nect systems a few technology generations. In particular, the global delay is still much
larger than the gate-delay for the 0.10-pm technology node and beyond.
The global interconnect delay is especially detrimental to global clock signal dis
tribution. Global clock signals need to be distributed across the microprocessor with skews
of less than ten percent of the global clock period. With each succeeding generation of
microprocessor, the clock frequency increases, decreasing the clock period and thus the
skew requirement in absolute time. This is in contrast to chip area and propagation delay,
which are both increasing. Hence, techniques are required to equalize the increasingly
large delays of each distributed clock signal to even greater accuracies or lower absolute
clock skews. Another serious issue with global clock distribution is the dispersion associ
ated with interconnect resistance. A non-zero interconnect resistance causes the harmonics
of the clock signal to travel at different velocities through the interconnect, resulting in an
increase (i.e., slowing) of the rise and fall times of the signal. As the interconnect length is
increased, the rise and fall times increase, and they can ultimately limit the maximum fre
quency of the signal [Deu98].
Both of the previously stated problems have been somewhat circumvented in the
ITRS for 0.13-p.m generations and beyond by distributing a lower frequency global clock
and allowing functional units to operate off of a higher frequency local clock. However,


150
To verify the dependence of jitter on SNR for a frequency divider, a 0.25-fim fre
quency divider was simulated at 4 different SNRs. To generate transient noise, an array of
30,000 data points randomly distributed according to a gaussian probability density func
tion was generated using Matlab3. The mean was set to zero and the standard deviation
was set to 1. This array was imported into SPICE at 1-ps time intervals, generating a noise
transient voltage 30 ns long. The 1-ps time interval sets the maximum frequency of the
noise, while the 30-ns total length of time sets the number of data points which will be
obtained from the simulation. In other words, these two numbers basically set the limits of
integration for (5.32), with the lower limit set by -1/(30 ns) and the upper limit set by -1/
(1 ps). By amplifying this noise and adding it in series with the input signal, a given SNR
can be obtained. The SNR is simply equal to the ratio of the signal to the standard devia-
standard deviation is equal to the amplification factor
applied to the noise data. The periods of the 2:1 and 4:1 output signals were then obtained
and the standard deviation was calculated. The result is shown in Figure 5-8, which plots
the simulated jitter versus SNR, compared with the jitter predicted by (5.35). As can be
seen, the simulated jitter for the 2:1 divider agrees very well with the predicted jitter.
Clearly, the logarithm of the jitter decreases linearly with SNR (in dB). Also, Figure 5-8
shows that the jitter is decreasing as N becomes larger. Both of these results verify (5.35).
However, the simulated jitter for the 4:1 divider is slightly less than the predicted jitter.
This is thought to be due to only having half as many data points for the 4:1 periods as
compared to the 2:1 periods. Since jitter increases as more data are included
3. This simulation procedure was developed by Jim Caserta, of the University of Florida.


APPENDIX F
QUALITY FACTOR OF RING OSCILLATOR
It is useful to model the quality factor (Q) of a ring oscillator. This Q can then be
applied to the results presented in Appendix E on injection-locked oscillators. Using the
model provided in [Ega99], a ring oscillator can be modeled as an ideal inverter with zero
delay (providing a 180 phase shift) in a loop with an element which has a delay of T, as
shown in Figure F-l. The delay element causes a phase-shift in the loop equal to
0 = co7\
(F-l)
The loop will oscillate at the frequency where 0 is equal to n, since the inverter provides
the other 180 phase shift. Thus, the oscillation frequency is equal to
The quality factor (Q) of an oscillator is related to the oscillation frequency and phase-
shift through the oscillator, as follows [Raz94]:
(F.3)
Ideal Inverter (no delay)
Delay
T
Figure F-l Model of ring oscillator as an ideal inverter followed by a delay element, used to
find the Q of a ring oscillator.
265


= 0.57 {CLoady
(7.2)
c
buffer
= I (J
n = 1
A second component to CT is Cw, which is the capacitance of the interconnecting wires
for delivering the clock from the output of the sector buffers to the local distribution sys
tem. Finally, CL is the load capacitance representing the input capacitance of the local
clock generators.
The capacitances for the wireless system are based on measured results. The
capacitances for the grid and H-tree systems are based on published 0.25-pm results
[Res98, Gie97, Web97], which are then scaled to correspond to a 0.1-pm generation. The
initial data for the total global capacitance for grid and H-tree systems are shown in Table
7-1. The scaling methodology used assumes that the number of sectors or spine locations
scales with area. According to projections in [SIA99], the projected microprocessor areas
are 300 and 622 mm2 for 0.25- and 0.1-pm generations, respectively. Thus, CG, Cw, and
CL each scale by a factor of 2.07. However, this scaling methodology does not account for
all of the >10X increase in transistor density. To accommodate this increased load, either
the drive capability in the local clock generators has to be increased or the grid and H-tree
Table 7-1 Total global capacitance for 0.25-pm technology
Capacitance (pF)
Initial
0.25-pm, Al, k=4
2 V, 750 MHz
Grid
H-tree
cG
1343
597
cw
1400
95
CL
600
600
CGlobal = Cq+Cw+Cl
3343
1292
% Total Power
14.3
5.5


240
impedance and transmission gain affect the power delivered to the input of the receiver.
Second, the area-efficiency of antenna structures should be improved, since, currently, the
antennas occupy about as much area as the receiver and transmitter circuits. Third, the
substrate antennas should be further examined, simulated, and optimized. Fourth, a
test-chip containing more realistic interference structures should be implemented. The test
structures should include metal-fill patterns as well as multiple layers of metal. Design
rules for the antennas should then be explicitly defined. Finally, a test setup should be
developed to analyze how the package affects system performance. The package setup
should include solder balls used for flip-chip bonding, ground planes embedded in the
package and the board, and the heatsink.
System-Level Implementation and Demonstration. First, the initialization and
start-up methodology for the programmable receivers should be developed beyond that
presented in Chapter 4. In particular, a method to synchronously release each initialized
receiver is required. Using this methodology, a multiple-receiver system should be imple
mented (which could include phase detectors to measure the static phase error and ana-
log-to-digital converters to translate this phase-error into a specific count for the receiver).
Second, the accumulation of phase-errors in the static phase-correction system should be
quantified. Third, the jitter due to noisy supplies and realistic switching noise has to be
quantified. Finally, an intra-chip wireless clock distribution system consisting of a trans
mitter and multiple receivers should be used to provide a clock signal to an actual micro
processor core. Using this platform, the clock skew and jitter of the wireless clock
distribution can be obtained, and realistic system operation can be demonstrated.


18
Vdd
Figure 2-2 Schematic of the source-degenerated CMOS LNA
amplifier, which has a reduced Miller effect. Also, a cascode exhibits high reverse isola
tion, which simplifies the design procedure by decoupling the input and output matching
conditions. While this topology is similar to other reported CMOS LNAs, fundamental
differences include the omission of a 50-0 output buffer, the use of shielding structures,
and the use of a capacitive transformer at the output.
2.3 Input Matching for Source-Degenerated LNA
2.3.1 Input Impedance
It can readily be shown that the input impedance of the source-degenerated LNA,
neglecting gate-to-drain capacitance (Cgdi), is as follows:
z<. = iCA+t.) + 7s- + rlt.-
JLgs\ Lgs\
(2.5)


234
technology generation, which has a 2-GHz global clock, and a 622-mm2 area [SIA99]for
an evenly-spaced grid of 16 receivers, the longest distance from the center to the comer
receiver is 1.3 cm.). The latency through the clock receiver was simulated with SPICE,
having the following components: LNA and source-follower delay = 30 ps, first 2:1
divider delay = 26 ps, second 2:1 divider delay = 55 ps, and third 2:1 divider delay = 77 ps.
Thus, the total latency through the 0.18-|im receiver is 188 ps, and the wireless intercon
nect has a total latency from the transmitter to the (unbuffered) output of the receiver of
approximately 338 ps. This corresponds to 63% of a 1.875-GHz output waveform. The
latency through the output buffer was not included since this would depend on the load
that the receiver is driving. Also, the delay through this buffer would be the same for any
interconnect system, since the local clock loads are equal. A rough estimate of the clock
skew could be approximately 10% of the latency, which is 34 ps or 6.3% of the output
clock period. This is less than the estimated worst-case skew of 11.2%, and is within the
allowable skew range of 5-10%.
The latency through the interconnect is still fairly large. What makes wireless
interconnects attractive, though, is that the latency through the receiver will decrease with
subsequent technology generations; whereas the latency through conventional intercon
nect continues to increase (refer to Figure 1-1). For example, scaling from the 100-nm to
the 70-nm generation, the gate delay decreases by -30%, whereas the global interconnect
delay increases dramatically [SIA99, Boh95]. The maximum interconnect distance
increases according to the square-root of the area increase, or from 1.3 cm to 1.4 cm (7%).
The TOF delay therefore increases by 7%, while the delay through the receiver decreases
by -30%. Thus, the wireless interconnect latency will decrease from 338 to 292 psa 14%


54
imposes severe restrictions on the achievable noise figure. For standard wireless applica
tions at this frequency range, Lg is approximately 2 nH, which is suitable for bondwire
implementation. This would increase the Q dramatically and help the noise figure. How
ever, for the wireless interconnect application, since the antenna is integrated, Lg has to be
integrated as well.
3.4.2 Inductor Characteristics
To demonstrate the effect of substrate resistance on inductor quality factor, 1.8-nH
spiral inductors have been implemented in the 5-level-metal, 0.25-pm CMOS process on
both p' and p+ (with epitaxial layer) substrates. These inductors were used for Lg in the
LNA. The inductor is implemented using shunted metals 3, 4, and 5 layers with 2.5 turns
and an area of 100x100 fim2. The quality factor is estimated using both the half-bandwidth
method [098] and the conventional method (Q11=-imag(yu)/real(yi1)). Figure 3-9 shows
Figure 3-9 Quality factors and inductance of Lg versus frequency for different substrate
resistivities (p+ ~ 0.01 i2-cm, p' with epi ~ 8 il-cm). The inductor is implemented
in a 5-metal-layer 0.25-|im CMOS process.


67
causing the total resistance at the output node of the LNA to be negative. By increasing
Vgc to 0.9 V, the gain of the LNA/buffer circuit could be increased to 25 dB; however, S22
becomes greater than 1, indicating unstable operation. Therefore, Vgc is chosen such that
stability is ensured (i.e., S22 < 0 dB).
3.6 A 23.8-GHz. 0.1-|im SOI CMOS Tuned Amplifier
3.6.1 Circuit Implementation
Since the wireless clock distribution system requires RF circuits operating above
15 GHz, it is necessary to evaluate the potential of deep submicron CMOS technologies
for this frequency range. As CMOS technologies approach the sub-tenth-micron regime,
the cutoff frequencies will exceed 100 GHz. Also, multiple interconnect layers in these
processes will facilitate on-chip passive components with acceptable quality factors.
These trends point to CMOS being a viable option for RF applications at 18 GHz and
above. Recent results which attest to this fact include a 0.18-|im distributed amplifier used
as a 16.6-GHz oscillator [Kle99], a 0.18-|lm, 23-GHz amplifier using a simple cascode
and on-chip transmission lines for matching [Yan99b], and a 0.1 -|im, 25.9-GHz volt
age-controlled oscillator [HunOOb]. From these, it can be seen that CMOS transistors can
be used to implement circuits operating above 20 GHz.
To further investigate the RF potential of deep submicron CMOS technology
including noise and linearity performance, and to evaluate the utility of lumped passive
components in 20-GHz amplifiers, a tuned amplifier operating at 23.8 GHz has been
implemented [FloOlb] in a partially-scaled 0.1-|im silicon-on-insulator (SOI) CMOS
technology from IBM. The process uses a 0.35-p.m design rule set for all dimensions
except for the 0.1-p.m effective gate length (0.15-jxm drawn gate length) and 3-nm gate


211
In a 0.18-|im copper technology, a 15-GHz wireless clock distribution system composed
of integrated transmitters, receivers, and antennas has been implemented. A wireless trans
mitter with an integrated antenna generated and transmitted a 15-GHz global clock signal
across a 5.6-mm test-chip, and this signal was detected using receiving antennas. The
transmitter consisted of a VCO, PA, and antenna. Also, a wireless clock receiver with an
integrated antenna detected a 15-GHz global clock signal supplied externally to an on-chip
transmitting antenna located 5.6 mm away from the receiver, and generated a 1.875-GHz
local clock signal. These results demonstrate wireless interconnection at 15 GHz for a
5.6-mm distance. This is the first known demonstration of an on-chip clock transmitter
with an antenna and the second demonstration of a clock receiver. This result also virtually
obsoleted the 0.25-fim result. Thus, by advancing the technology one generation, both the
frequency and interconnection distance have been approximately doubled. This is quite
significant, because it shows that the frequency of operation and the interconnection dis
tance scale roughly with Moores Law (using two data points).
Finally, a wireless interconnect composed of two clock receivers locked to the
same global clock signal has been demonstrated at 15 GHz. The distances between the
transmitting antenna and the two clock receiver are 5.95 mm each in one case, and 5.6 mm
and 6.8 mm for the other case. The 6.8-mm interconnection distance is the longest to date.
The results show that the two receivers are synchronized, having an output clock skew of
25 ps at 1.875 GHz (or 4.7%). The measured peak-to-peak clock jitter with quiet supplies
is 6.6 ps at 1.875 GHz (or 1.2%). These results suggest that a wireless distribution system
can meet the skew and jitter requirements for clock distribution systems. All of these
results demonstrate the plausibility of an intra-chip wireless interconnect system contain
ing integrated antennas and CMOS transmitter and receiver circuitry.


using 0.25-, 0.18-, and 0.10-|im technologies, respectively. A design methodology based
on injection locking is developed for CMOS frequency dividers, and a programmable
divider which limits clock skew is presented. Dividers operating up to 10, 15.8, and 18.8
GHz are demonstrated, implemented in 0.25-, 0.18-, and 0.10-pm technologies,
respectively.
Results for the overall wireless interconnect system are then presented. System
requirements (gain, matching, noise, linearity) for wireless clock distribution are derived,
including specifications for signal-to-noise ratio versus clock jitter, and amplitude mis
match versus clock skew. Wireless interconnect systems are demonstrated for the first time
using on-chip antenna pairs, clock receivers, and clock transmitters. The interconnects
operate across 3.3 mm at 7.4 GHz, using a 0.25-pm technology, and across 6.8 mm at 15
GHz, using a 0.18-|im technology. Using the 6.8-mm, 15-GHz interconnect, a 25-ps clock
skew and 6.6-ps peak jitter have been measured at 1.875 GHz for two receivers separated
by ~3 mm. Finally, the wireless interconnect system is analyzed in terms of power dissipa
tion, synchronization, process variation, latency, and area. These results indicate the
feasibility of an intra-chip wireless interconnect system using integrated antennas.
IX


227
Table 7-3 Summary of simulated process variation for 0.25-pm receiver circuits.
Parameter
Absolute Variation
Percentage
LNA Gain
0.7 dB
5.7%
Input Matching (Sjj)
1 dB
7.4%
Noise Figure
0.2 dB
6.7%
Divider Output Signals,
When Locked
42 ps
4.5%
fout= 1.0625GHz
Divider fiso
1.28 GHz
12% (input)
Receiver Gain
(to divider input)
9 dB, fin = 8.5 GHz
6.3 dB, fin = 8.39 GHz
28%, fin = 8.5 GHz
16%, fin = 8.39 GHz
Phase-Shift Due To
Amplitude Mismatch
69, fin = 8.5 GHz
61, fin = 8.39 GHz
2.4%, fout= 1.0625 GHz
2.1%, fout= 1.048 GHz
Receiver Phase-Shift
(to divider input)
27, fin = 8.5 GHz
109, fin= 8.39 GHz
1%, fout = 1.0625 GHz
3.8%, fout= 1.048 GHz
Receiver Output Signals
36 ps, Pin=-16 dBm
3.8%, fout= 1.0625 GHz
(Divider Fails at Pin = -31 dBm)
at the main components which contribute to skew and jitter for a wireless clock distribu
tion system, estimate the (current) worst-case skew and jitter, and then these estimates
compare with the measured results from Chapter 6.
7.4.1 Clock Skew
For a wireless clock distribution system, the main elements which will contribute
to clock skew are as follows: process variation affecting receiver gain and phase-shift;
time-of-flight delay mismatches due to the receivers being different distances away from
the transmitter; amplitude mismatches at the input of the divider causing AM-to-PM con
version (section 5.4); interference structures located between the transmitting and receiv
ing antennas affecting antenna transmission gain, impedance, and phase-shift; and finally,
mismatches between the loads at the clock receiver outputs. The worst-case skew can be
estimated as the sum of the mean skew and 3 standard deviations of the random skew,
where the variances (a2) of all of random skews will add to yield the total skew variance.


210
with the SNR of 21.8 dB estimated using SPICE, which was presented in section 5.7 of
Chapter 5. Since the system requirement for the SNR to obtain 0.5% RMS jitter is 12 dB,
a 7.7-dB margin is present. However, this margin could of course be consumed (and per
haps exceeded) by the interference due to digital switching noise.
6.8 Summary
In this chapter, the implementation of on-chip wireless interconnects with inte
grated antennas has been presented. Two different test chips have been implementedone
in a 0.25-|im CMOS process and one in a 0.18-|im CMOS process. Measured antenna
characteristics for both chips show that antenna gain increases with frequency where the
antenna becomes electrically longer. Also, the presence of interference structures between
the antennas can alter the gain by 5-10 dB. Loop antennas and antennas in direct contact
with the substrate have been characterized, and show promising results for higher frequen
cies. Additionally, the phase versus frequency response for the antennas is linear indicat
ing wave propagation, while the gain between two sets of pads is at least 15 dB less than
that for the antennas. These two facts together show that the signal coupling is due to wave
propagation, and that these waves are launched much more efficiently from the antennas
than from just the pads. Finally, the impedance mismatch between the antenna and
receiver degrades the minimum detectable signal of a clock receiver and hence, this
impedance has to be accurately predicted a priori.
In the 0.25-|im test chip a clock receiver with integrated antenna was imple
mented. A 7.4-GHz global clock signal was successfully received over a 3.3-mm distance
with interference structures located in between the antennas, and a 925-MHz local clock
signal was generated. This result is the first demonstration of an on-chip wireless intercon
nect. Also, this result is the first to integrate antennas and CMOS circuitry on a single chip.


25
2.5 Gain of Source-Degenerated LNA
For an amplifier driven by a source Vs, with source impedance Rs, and loaded at
the output by RL via transmission lines with characteristic impedances of Rs and RL,
respectively, the forward transducer power gain is given by [Gon97]
Gt ~ p
Load
A VS
= |s2if = 4
V
Out
^ ^S i 12 RS
(2.19)
where is the power delivered to the load, and PAVS is the power available from the
source. For RS=RL=50 £2, the transducer power gain is then
Gt lS'211 |2 2lv| .
(2.20)
2.5.1 Gain Driving Resistive Load
It can be shown that the voltage gain from the source to the load, A^ assuming that
the input is at series resonance and again neglecting Cgdl, is as follows:
Ia
V| (0 = (0
= (G,'")(|*2 + j<*>0CJl
(2.21)
joG2^L
1 +7C0o(C1 + C2)Rl
where Cdl is the total capacitance at the node connecting the drain of Mj to the source of
M2, and Qin is the quality factor of the input series resonant circuit, as given by equation
(2.8). Here, Z^q is the total equivalent impedance at the drain node of M2, as shown in
Figure 2-2. Since the output matching network essentially matches the equivalent parallel
resistance of inductor Ld to the load resistance, RL, then is approximately equal to
the following:
1 2 1
^Leq^o) = 2 QLdrLd = ^Ld^o^d
(2.22)


270
[Flo99a] B. A. Floyd, J. Mehta, C. Gamero, and K. K. O, A 900-MHz, 0.8-|im
CMOS low noise amplifier with 1.2-dB noise figure, Proc. IEEE Custom
Integrated Circuits Conference, pp.661-664, May 1999.
[Flo99b] B. A. Floyd and K. K. O, The projected power consumption of a wireless
clock distribution system and comparison to conventional distribution sys
tems, Proc. International Interconnect Technology Conference, pp. 248-
250, May 1999.
[FloOOa] B. A. Floyd, K. Kim, and K. K. O, Wireless interconnection in a CMOS
IC with integrated antennas, ISSCC Digest Technical Papers, pp. 328-329,
Feb. 2000.
[FloOOb] B. A. Floyd, C.-M. Hung, and K. K. O, The effects of substrate resistivity
on RF component and circuit performance, Proc. International Intercon
nect Technology Conference, pp. 164-166, May 2000.
[FloOla] B. A. Floyd, C.-M. Hung, and K. K. O, A 15-GHz wireless interconnect
implemented in a 0.18-p.m CMOS technology using integrated transmit
ters, receivers, and antennas, Symp. VLSI Circuits Dig. Tech. Papers, June
2001.
[FloOlb] B. A. Floyd, L. Shi, Y. Taur, I. Lagnado, and K. K. O, A 23.8-GHz SOI
CMOS tuned amplifier, submitted for publication to IEEE Trans, on
Microwave Theory and Techniques, 2001.
[FloOlc] B. A. Floyd, L. Shi, Y. Taur, I. Lagnado, and K. K. O, SOI and bulk
CMOS frequency dividers operating above 15 GHz, submitted for publi
cation to IEE Electronics Letters, 2001.
[Fos97] J. G. Fossum, SOISPICE-4 (Ver. 4.41) (FD/SOI and NFD/SOI MOSFET
Models), University of Florida, Gainesville, FL, Jan. 1997.
[Gie97] B. A. Gieseke, R. L. Alimn, D. W. Bailey, B. J. Benschneider, S. M. Brit
ton, J. D. Clouser, H. R. Fair m, J. A. Farrell, M. K. Gowan, C. L. Hough
ton, J. B. Keller, T. H. Lee, D. L. Leibholz, S. C. Lowell, M. D. Matson, R.
J. Matthew, V. Peng, Quinn, A 600-MHz superscalar RISC microproces
sor with out-of-order execution, ISSCC Digest Technical Papers, pp. 176-
177, Feb. 1997.
[Gon97] G. Gonzalez, Microwave Transistor Amplifiers: Analysis and Design, 2nd
edition, Upper Saddle River, NJ: Prentice-Hall, 1997.
[Gra93] P. R. Gray and R. G. Meyer, Analysis and Design of Analog Integrated Cir
cuits, 3rd edition, New York: John Wiley & Sons, Inc., 1993.


53
L=2.2nH
Figure 3-8 Schematic of 8-GHz fully differential LNA showing parasitics.
This LNA is a fully differential version of that given in Figure 2-2. As mentioned
in the first chapter, a differential topology is required for the clock receiver application.
Transistor M5 provides tail current to the differential amplifier, where Vbias *s externally
generated. An important difference between this LNA and the 900-MHz LNA, is that
inductors Lgl, 2 and Lsl 2 are integrated. This LNA was originally designed to provide 12.6
dB of gain while driving 50 £2, and the anticipated NF was approximately 3 dB. The
higher anticipated NF, compared to the 900-MHz result, is due to short-channel effects
increasing y [Jin85], as well as the integration of Lg. According to simulations, the integra
tion of Lg increases the NF by approximately 0.6 dB. First, the increase is due to thermal
noise generated by the series resistance of Lg (rLgi), whose Q factor is 10 (5) for p' (p+)
substrates. Second, the parasitic capacitances at the input and output ports of Lg (CLg¡)
move the input match away from the optimal noise match. Therefore, the integration of Lg


167
Table 6-1 Wavelength in millimeters versus frequency for different mediums
Frequency
1 GHz
7.4 GHz
15 GHz
20 GHz
24 GHz
^free-space
300
40.5
20
15
12.5
^oxide
151
20.5
10.1
7.6
6.3
Tl
^silicon
87.2
11.8
5.8
4.4
3.6
(a)
(b)
Figure 6-1 Die photograph of (a) linear and (b) zigzag dipole antennas.
silicon dioxide. Since the antenna is not operating at resonance (i.e., the length is not X/2)
then the antenna impedance has a reactive component. Therefore, this reactance has to be
conjugately matched with the impedances of the transmitter and receiver.
Both linear and zigzag dipole antennas have been implemented. Die photographs
of example antennas with pads are shown in Figure 6-1 [KimOOa]. The axial length of the
antennas is 2 mm and the line trace width is 10 |im. For the zigzag antenna shown in Fig
ure 6-l(b), the arm element length is 75 |im and the bend angle is 120, while the zigzag
antennas implemented with clock receivers and transmitters have an arm element length of
80 |im and a bend angle of 30. Although the lateral field components cancel each other in
a zigzag structure, the longitudinal components reinforce each other. Since the physical
length of a zigzag antenna is longer than the linear dipole antenna, a zigzag dipole antenna


59
Vdd
Figure 3-12 Schematic of differential 14-GHz LNA with source-followers.
(a)
(b)
Figure 3-13 Die photographs of (a) 0.18-|xm, 14-GHz differential CMOS LNA and
(b) 0.1-|im, 23.8-GHz SOI CMOS tuned amplifier.


APPENDIX C
DERIVATION OF NOISE PARAMETERS
C. 1 Equivalent Input Noise Generators
Any noisy two-port network can be represented as a noiseless two-port net
work with equivalent voltage (vn) and current (in) noise generators at the input of the net
work [Gra93], as shown in Figure C-l. The two noise generators allow the noise
properties of the network to be correctly represented for any input impedance (admit
tance). To calculate these noise generators, the output short-circuit (isc) current can be
equated for both networks with the input first short-circuited and then open-circuited,
yielding vn and in, respectively.
Traditionally, the noise factor of the network is characterized in terms of four noise
parameters, as follows:
(a)
Noisy
Two-Port
O
O
Noiseless
Two-Port
O
C
Figure C-l Representation of a noisy two-port network as a noiseless two-port with
equivalent input noise generators.
247


224
shifting. This 9-dB gain variation is severe, and can result in clock skew due to amplitude
mismatch. Referring to section 5.4, Figure 5-5, a 9-dB amplitude variation can result in as
much as 23 ps of skew at 8.5 GHz divided by 8, or 2.4% output clock skew.
Additionally, the variation in the resonant frequency will cause different
phase-shifts through the amplifier at a given input frequency. Figure 7-6(b) shows the
phase-shift to the input of the divider versus frequency. At 8.5 GHz, the variation in
phase-shift is 27.6, yielding 9 ps of output skew or ~1% of the local clock period. The
phase variation increases around the resonant frequency, though. At 8.39 GHz, corre
sponding to the worst-case variation, the phase variation is -109, which is -30% of the
input period and 3.8% of the output period (i.e., 30% divided by 8).
Even more severe than the skew due to amplitude mismatch and resonant fre
quency variations is the fact that this gain variation can cause the signal at the input of the
divider to be too low for the divider to lock. In that case, the divider would self-oscillate,
and the system would not be synchronized. The gain variation in Figure 7-6(a) is com
pounded by the variation in the input sensitivity of the divider. As was just mentioned, the
input-referred self-oscillation frequency (fiS0) can vary by -1.28 GHz. Thus, the conver
sion gain of the divider, and hence, the gain of the receiver will vary considerably.
Figure 7-7 shows the output clock waveforms from the Monte Carlo simulations
(30 iterations) for three different input power levels to the receiver. Reading top to bottom,
the input power levels are -16 dBm, -24 dBm, and -31 dBm. At a -16-dBm input
powerFigure 7-7(a)the output clock waveforms are all at the same frequency, which is
8.5 GHz divided by 8 = 1.0625 GHz. The skew at the rising and falling edges are 30 and
36 ps, respectively, which corresponds to a 3.8% skew. At a -24-dBm input powerFigure


87
used to adjust the oscillation frequency during design, while the voltages can be used to
adjust the oscillation frequency during circuit testing.
4.3.3 Simulation of Injection-Locked Divider
This section will present simulation results for an ILFD, showing the fiSO and
demonstrating the dependence of the input swing on the locking range. To extract fiSO the
8:1 divider is simulated using a transient analysis, with both CLK and CLKb held at
approximately mid-supply. As with other oscillator simulations, initial conditions are
required, where nodes Q¡ and Q are initialized to the supply and Qbi and Qb are initialized
to ground. Finally, since this is an analog simulation, the SPICE tolerances should be
decreased as compared to a standard digital simulation.
Figure 4-3 shows the simulated oscillation of the 8:1 ELFD, for a 0.25-|im CMOS
process. Since Vdd = 2.5 V, CLK and CLKb are set to ~1.3 V. The output of the first 2:1
divider is oscillating at 6.1 GHz, while the following 2:1 dividers are locked at 3.05 and
Figure 4-3 Simulation of self-oscillation of 8:1 source-coupled divider. The clock inputs
are held at 1.3 V, causing the first 2:1 divider to oscillate at 6.1 GHz for
Vdd=2.5V(fiSO = 12.2 GHz).


138
trigger level as well as the difference between the low-level and high-level input signals,
referred to as amplitude mismatch.
The clock skew as a function of trigger level and input signal levels can readily be
obtained. The phase angles, 0¡ and 02, at which the maximum and minimum input signals
are equal to the trigger level, Vref, are as follows:
The clock skew in terms of phase of the input signal is then
(5.12)
(5.13)
A0f = sin \ y
v min' > mm J max'
V V ,
min max'
(5.14)
Solving for the amplitude mismatch ratio Vmin/Vmax yields the allowable signal level dif
ference for a given input angular clock skew, A0¡n, as follows:
V
min
V
max
= sin (sin \S)-AQin)],
(5.15)
where
5 =
ref
V
min
(5.16)
The maximum clock skew will occur when 5=1. For this case, (5.15) can be reduced to
V
min
V
max
= cos(A0n) = cosWtc- ^-"1,
k 1 out'
(5.17)
where the input angular clock skew has been converted to a clock skew, At, with respect to
the output local clock period, using the division ratio of the frequency divider, N.


27
factor is maximized2. Second, to maximize the gain, coj should be maximized. This can be
achieved in two ways, as follows: (1) bias the transistor for maximum CDj, which can
increase the power consumption, or (2) use a more advanced (i.e., costly) technology
(shorter channel length) for a given power consumption. Thus, cost and power consump
tion should be balanced to achieve the correct trade-off for o>p Third, looking in the
denominator of (2.26), it can be seen that the gain and the input matching criteria trade-off
via the term 0>pLs. Since series feedback is used in the source of M1? the gain is reduced
while the input series resistance (Rs) is increased by a factor of one plus the loop gain
(l+Oj-LjVRs). Returning to (2.9), the input matching condition can still be met by choos
ing cOj-Ls = 35 Q, as opposed to 50 2 Thus, by allowing a small amount of mismatch at the
input (i.e., Sj] = -15 dB), the gain is increased by approximately 18%, or 1.4 dB. Finally,
the term 18/nl should be maximized, where 5 can be maximized by reducing Cdl through
judicious layout, and n can be minimized by decreasing or even eliminating Cj.
(a) (b)
Figure 2-5 S2i at resonance versus (a) Ld for constant and (b) for constant Ld. The
output is always matched to 50 £1
2. Note that if becomes too large, the LNA can become intolerant to process variations.


226
(a)
w
o
>
(b)
w
o
>
(C)
w
4*
o
>
Pin = -16 dBm
Time (ns)
Figure 7-7 Output clock signals from frequency divider with process variation:
Input power levels are (a) -16 dBm, (b) -24 dBm, (c) -31 dBm.
7.4 Synchronization
Probably the single-most important criterion for a global clock distribution system
is that all of the output clock signals should be synchronized, having low clock skew and
low clock jitter. The clock skew should be less than 5-10% of the clock period, while the
peak-to-peak clock jitter should be less than 3% of the clock period. This section will look


141
Once again, a clock skew can be defined in terms of the input phase difference in
radians. Taking the difference between the steady-state phase difference given in (5.18) for
the maximum and minimum input swing locking signals (VLmax, VLmin), assuming that
the output swings are the same for each case, and making the following substitution
o
Aco 0A
ILO
VLmin A v2
(5.19)
Jl + (Aco aAY
results in (5.15). Thus, the relationship between clock skew and amplitude mismatch at the
divider input is the same for dividers operating in the latched or injection-locked mode. In
fact, an equivalent trigger level, V^?, can be defined for the ILO divider as follows:
VIL? = V
v ref vO
Aco0A
^l+fAco 0A):
(5.20)
As the locking frequency approaches the natural frequency of the ILO, the equivalent trig
ger level approaches zero.
The amplitude mismatch for the worst case clock skew can again be quantified. For
an injection-locked frequency divider (ILFD), the minimum input signal level will be that
for just locking the ILO. This locking condition is given in appendix E in (E.7), where the
minimum VL is
VLmin = V0 A(oA (5.21)
Plugging this into (5.19), and assuming that Aco0A 1 (which is valid in this case since
VLmin V0), results in 5 = 1. With these substitutions, (5.15) becomes
^ = 1 [sin(sin_1(l)-A0(/J)] = sin^-AoJ = cos(A6), (5.22)
as in (5.17), and Figure 5-5 remains valid. To obtain less than 0.5% of clock skew from
amplitude mismatch, the inputs to all of the frequency dividers should be matched to


255
(C.18) for Fmin. Now that explicit equations are in hand for the noise parameters for the
complete model of a MOSFET, a numerical simulator (e.g., Matlab) can be utilized to
probe the effect of each model parameter on the noise parameters. Graphical results for
this exercise are shown in Figure 2-7. Further intuition can be gained by looking at case
studies considering the noise parameters for various second-order effects, while neglecting
all other second-order effects, as discussed in the next section.
D.3 Case Studies: Noise Parameters for Second-Order Effects
D.3.1 Case 1 Effect of
The simplest case for evaluating the noise parameters is to consider the first-order
effect, which is thermal noise at the drain (id), as well as the effect of Cgd. Substrate resis
tance and all other capacitances except Cgs are neglected. Under these assumptions, y2i =
gm jooCgd, CGT = Cgs + Cgd, and the input noise generators become
_ jd_ = ld
Vnl ?2l 8m~jCgd
(D.18)


100
25 mm
(a)
E
E
in
C\J
25 mm
(b)
Figure 4-15 Illustrations (to scale) of grids upon which clock receivers should be placed, (a)
Grid spacing of one wavelength from transmitter, with solid for wavelength in
silicon dioxide and dash-dot for wavelength in silicon, (b) Grid spacing of one
eighth of a wavelength in oxide, used when the dividers are programmed to initial
states of 0 to 7 (for divide-by-8).
all of the receivers to be a multiple of a wavelength away from the transmitter. This con
cept is illustrated (to scale) in Figure 4-15(a), which shows circular grids spaced multiple
wavelengths from the transmitter. The solid line represents wavelength in silicon dioxide
and the dash-dotted line represents wavelength in silicon substrate. This begs the question,
though, of in which medium does the electromagnetic wave propagate. Also, multipath
effects will be anisotropic; therefore, the wavelength in one direction will not necessarily
be the same as the wavelength in another direction. These issues will be addressed further
in chapter 6. Assuming, however, that the wavelengths are known a priori, then the receiv
ers can be placed on such a grid.
From examining Figure 4-15(a), it can be seen that the grid is not dense enough to
accommodate clock receivers in a feasible arrangement. This is where the concept of pro
gramming or initializing the frequency divider is useful. Frequency dividers are basically


66
decreasing channel length due to increased hot-carrier-induced noise) and the body resis
tance can greatly influence the noise figure. According to SPICE simulations, increasing y
to 2, and Rsub from 10 to 20 Q, increases the noise figure by 2 dB. Finally, BSIM3v3s
modeling of the transistors noise parameters is not very accurate; therefore, BSIM3v3s
optimal noise matching condition versus the actual optimum can be very different. This
can result in operating at a suboptimal input noise-matching condition. All of these effects
taken together can yield the 4.7-dB increase in noise figure.
Figure 3-17 shows the measured S-parameters for the differential LNA with
source-follower buffers (driving a 100-2 load). For Vgc=0.5 V, the gain of the LNA/buffer
circuit increases to 21 dB, while consuming a total of 28.2 mW from a 1.5-V supply. Table
3-5 summarizes the measured results. The resonant frequency has increased to 14.4 GHz.
The gain is increased compared to the single-stage LNA result due to removal of the
capacitive transformer and the presence of negative conductance at the output nodes of the
LNA. Too much negative conductance, however, can result in unstable operation by


124
Figure 4-33 Output wave-form for bulk 128:1 frequency divider for 15.36-GHz input signal
and Vdd=1.5 V. Output frequency is 120 MHz.
shows the output spectrum of the divider for an 18.75-GHz input signal and a 4.6875 GHz
output signal. The spectrum also contains higher order harmonics of the output signal. The
power consumption of the SOI 4:1 divider is 13.5 mW. From the maximum operating fre
quency, the extracted SOI [DP]2 level-sensitive latch delay is 13.3 ps at 1.5 V.
A bulk 128:1 divider was also measured on-wafer. This divider consists of the 4:1
[DP] divider followed by an asynchronous 32:1 divider implemented with TSPC logic.
At 1.5 V, the bulk divider can operate between 3 and 15.4 GHz. As expected, the lower
frequency cutoff of the bulk divider is less than that for the SOI divider, due to SOIs tran
sient bipolar leakage. At 15.4 GHz, a 120-MHz output signal is generated, shown in Fig-
ure 4-33. The bulk divider power consumption is 9.8 mW. The extracted bulk [DP]
level-sensitive latch delay is 16.2 ps at 1.5 V. As expected, the bulk latching delay is
longer than the SOI latching delay.


CHAPTER 2
CMOS LOW NOISE AMPLIFIERS
2.1 Overview
2.1.1 Scope of LNA Research
A key building block for the clock receiver, as well as for the front-end of super
heterodyne and direct conversion receivers, is the low noise amplifier (LNA). Due to its
moderate to high gain as well as its low noise figure, the LNA approximately sets the over
all signal-to-noise ratio of the receiver by reducing the impact of noise from subsequent
stages. This is equivalent to saying that when the LNA is properly designed, the total
receiver noise figure is roughly that of the LNA. For a cascaded system, the total noise fac
tor (F) and noise figure (NF) are given by the following formulas [Gon97]:
(2.1)
(2.2)
NF = 101og(F),
where (S/N) is a signal-to-noise ratio, and F¡ and Gj are the noise factors and available
power gains, respectively, of individual stages in the receiver. As can be seen, the noise
contributions from latter stages are divided by the total gain preceding thema process
known as input-referring. Thus, to minimize the total noise figure, the first stage in the
receiver should amplify the input signal while adding minimal noise.
For the wireless interconnect application, a CMOS technology is required to be
consistent with the ITRS. However, RF CMOS circuitry has only recently been under
12


204
O)
CO
4
£
-4-*
3
Q_
C
fin=15 GHz^ !/\
-: -
ID !.: I
i
as the transmitting antenna, with a wood as the back dielectric. Input frequency is
15 GHz; output frequencies are 1.875 GHz; distances are 5.95 mm, each.
receiver, and the impedance of the input balun, the mismatch loss in the system is calcu
lated to be 0.6 dB. With a 20.7-dBm available input power, the power delivered to each
receiver is -39.9 dBm, corresponding to the minimum detectable signal for the receiver.
The second double-receiver interconnect was demonstrated using Zgc as the trans
mitting antenna. This case is of particular interest since Z6c is located 6.8 mm from Rx2
and 5.6 mm from Rx4. At 15 GHz, the wavelength of the transmitted signal is 10.1 mm in
silicon-dioxide and 5.8 mm in silicon. Therefore, a time-of-flight delay mismatch between
8-12 ps is present in the system, where the actual value depends on the effective dielectric
constant for this multi-path medium. Additionally, the transducer gains for each antenna
pair are not equal; thus, there will be a skew due to amplitude mismatch at the input of the
divider. These skews will be tuned out using the programmable divider.


188
lower mobility. This implies that the total gate area is larger if PMOS is used for the
oscillator. Since 1/f noise is approximately inversely proportional to the gate area, the
close-in phase noise can be reduced. Furthermore, PMOS transistors have much lower
hot-carrier noise, according to [Che90]; thus, the far-out phase noise would also be
improved. Hence, PMOS transistors are exclusively used in this VCO design [HunOOa].
The inductance of both Lj and L2 was designed to be 0.33 nH, while measure
ments reveal an inductance of 0.41 nH. This increase in inductance as well as increased
capacitance from the transistors, as presented in section 3.5.3, decreased the VCO operat
ing frequency from ~21 GHz to 15 GHz. The inductors were implemented with metal 5
and 6 layers shunted together above a patterned ground shield. As mentioned in Chapter 3,
this improves the inductor Q by providing a low resistance path to ground for capacitively
r\
coupled current. The area of the inductor is 70.6x70.6 pm and the metal spacing, width,
and number of turns are 2.2 |im, 6.6 pm, and 1.75, respectively. The expected inductor Q


218
Table 7-2 Total global capacitance for 0. l-(im technology
Capacitance (pF)
Case 1
0.10-pm, Al, k=4
1.2 V, 2 GHz
Case 2
0.10-pm, Cu, k=2
1.2 V, 2 GHz
Grid
H-tree
Wireless
Grid
H-tree
Wireless
cG
2523
1219
1129
1281
870
1064
cw
2450
166
166
772
52
52
CL
1242
1242
1242
1242
1242
1242
CGlobal = Cg+C\v+Cl
6215
2627
2537
3295
2164
2358
% Total Power
11.2
4.7
4.6
5.9
3.9
4.2
receiver consumes 32 mW. The equivalent capacitance can then be used to make a com
parison to the grid and H-tree systems.
7.2.3 Results and Conclusions for Power Consumption
Table 7-2 shows a breakdown of the global capacitive loading for the three distri
bution systems for cases 1 and 2. All capacitance units are in picofarads. The final row
gives the power consumed by the global clock distribution systems as a percentage of total
microprocessor power. This percentage represents the relative amount of power dissipated
in the global system to drive a load of CL. To obtain these numbers, the total power con
sumption projected in the ITRS is converted into an equivalent capacitance using (7.3) and
then compared with Cglobai. For clarification, this percentage is not the relative amount of
power dissipated in the entire clock system, which is estimated to be between 30 and 40%
[Gie97]. This is because the percentage does not include the dissipation in the local clock
distribution system, which will be much greater than the dissipation in the global system.
The results show that for both cases, the wireless system is comparable in perfor
mance to the H-tree system and better in performance than the grid-based system, in terms


239
Therefore, by modulating this carrier, data communications can be obtained. Such data
communications could be used for on-chip interconnection, multi-chip interconnection,
and general wireless communications. Obviously, the circuitry will be more complicated;
however, this work has shown that CMOS can support RF circuitry operating up to 20
GHz. Also, this work has demonstrated transmitter and receiver circuits with on-chip
antennas. Potentially, this could be the largest application for this work, in that a true sin
gle-chip radio (including antenna) could be implemented for low-cost applications. There
fore, in summary, this work is applicable to both wireless interconnect systems and
general wireless communication systems.
7.7.4 Suggested Future Work
Circuit Design and Implementation. First, the state-dependent initialization failure
in the programmable divider has to be corrected using the new latches presented in Figure
4-24. Second, automatic gain (and frequency) control circuits should be implemented in
the receiver to tune the gain in the amplifier and the self-oscillation frequency of the injec
tion-locked divider. Potentially, the first 2:1 divider could be changed to an LC injec
tion-locked oscillator. Third, the systems robustness against process variation should be
improved by stabilizing or potentially eliminating the source-follower buffer and using
more sophisticated biasing techniques for the transmitter and receiver. Finally, a
phase-locked transmitter operating above 15 GHz has to be implemented. Particularly, the
CMOS PA requires optimization, such that it delivers more power to the antenna.
Antenna Design and Characterization. First, the modeling and predictions of
antenna performance should be improved, using three-dimensional electromagnetic simu
lation tools. This is very important to the overall system operation, since both the antenna


22
Table 2-1 Required Value of Qgs for Component Variations with S j t < -10 dB
Band
GSM
935-960
MHz
GSM
935-960
MHz
GSM
935-960
MHz
ISM
2.4-2.5
GHz
ISM
2.4-2.5
GHz
ISM
2.4-2.5
GHz
Qb
37.9
37.9
37.9
24.5
24.5
24.5
ITlI, ITCI, ITtI
5%
5%
10%
5%
5%
10%
tLs
35
50
50
35
50
50
Qgs
2.9
4.8
2.4
2.6
4.3
2.3
Table 2-1 shows the required value of Qgs for the GSM and ISM (Industrial, Scien
tific, and Medical) bands, for varying component tolerances and values for 0>pLs. It can be
seen that as QB (the quality factor implied by the operating frequency band) decreases, so
too does Qgs, as expected. Also, as the component tolerance becomes larger, the required
value of Qgs decreases, indicating that high-Q networks are more sensitive to component
variations, as originally asserted. Finally, as (OjLs becomes closer to 50 Q, Qgs increases,
indicating that the network can withstand larger component variations. Note that choosing
a smaller value of Qgs will result in additional margin for input matching variations. The
results presented in Table 2-1 show that choosing Qgs between 2.3 and 4.8 will result in
Su < -10 dB over the entire band, while withstanding between 5 to 10% of component
variations. Once Qgs is specified, Cgsl and hence can be determined, allowing the
inductor values to be chosen as given by (2.9) and (2.10).
2.4 Output Matching for Source-Degenerated LNA
Driving a 50-2 load while providing sufficient power gain requires either an out
put matching network or a 50-2 buffer. Since a buffer typically would dissipate additional
power while also degrading the circuit linearity, a single-stage design with an output


238
addition, the upper frequency limit is set by how fast the RF circuits can operate, rather
than how fast the interconnects can operate. This is a hidden benefit of wireless intercon
nects, in that the performance improves as the frequency increases (the improvements are
largely due to the antennas and inductors becoming more efficient). Thus, wireless inter
connects shift the microprocessor performance from being back-end (i.e., interconnect)
limited to once again being front-end (i.e., transistor) limited.
Therefore, this work has demonstrated both the plausibility and feasibility of a
wireless clock distribution system. First, the design and implementation of LNA and fre
quency divider circuits operating between 1 and 24 GHz was presented. The results dem
onstrate the competitiveness of CMOS for both interconnect and general wireless
communication systems. Second, the system requirements of a wireless clock distribution
system were derived, translating the clock requirements of skew and jitter into standard
radio requirements. Third, the plausibility of a wireless interconnect system was demon
strated by implementing on-chip antenna pairs, clock receivers, and clock transmitters.
These are the first demonstrations of wireless interconnects and the first time antennas and
circuitry have been incorporated on the same silicon chip. Finally, the feasibility of a wire
less interconnect system was evaluated in terms of power dissipation, synchronization,
process variation, latency, area, and design verification. All of these results indicate the
potential of an on-chip wireless clock distribution system.
7.7.3 Broader Applicability
The results from a wireless clock distribution system can be applied to' general
wireless interconnect systems as well as RF systems with on-chip antennas. Basically, a
wireless clock distribution system consists of communication of just a carrier signal.


179
o
UMC test-chip, whose area is 7 x 6 mm The locations of relevant zigzag antennas (Zx)
and the loop antenna (LPj) have been noted. Clock transmitters (Txl, Tx2) and receivers
(Rxl-Rx4) have been labeled as well. Located in between the antennas are test structures
which interfere with the clock transmission and reception. These test structures contain
multiple metal interconnects, vias, substrate connections, passivation openings, and
metal-fill patterns (not shown) for metal layers 1 through 6. Therefore, the density of
structures between the antennas is high.
6.5.2 Antenna Descriptions
The majority of the antennas implemented are 2-mm long zigzag dipole antennas,
labeled Zx in Figure 6-10. The zigzag antennas, illustrated in Figure 6-11, have a 10-|im
trace-width, an 80-|im arm element length, and a 30 bend angle. These values were based
on the best results currently available from antenna design experiments [KimOOb]. Anten
nas Z6a to Z6f are implemented in metal 6 at various locations throughout the chip, with
spacings of d={6.7, 6.7, 6.7, 5.7, 4.3, 3.2} mm, for antenna pairs Z{6a 6b bc
respectively. The distance from metal 6 to the substrate is ~7.2 |im. Also, different metal
layers were used to fabricate the antennas to examine their dependence on antenna gain.
These antennas are labeled Z¡, and Z3, and are separated by 3.6 mm.
Two antennas are evaluated for use in the transmittera zigzag dipole antenna and
a loop antenna (LPj) which has an omnidirectional in-plane radiation pattern [Bal97]. The


137
transmitter and receiver. Assuming that the receivers do not correct for this difference in
signal strength, then the voltage swing used to trigger or lock the dividers will be much
lower in receivers farther from the transmitter. This can potentially lead to a clock skew,
due to a conversion from amplitude modulation to phase modulation (AM to FM). The
conversion mechanism depends on whether the divider is operating in a latched mode or
an injection-locked mode, and each case will be examined.
5.4.1 Latched Mode
For a divider operating in the latched mode, AM-to-FM conversion will result if
the divider switches at a point other than the zero-crossing of the input signal. This is illus
trated in Figure 5-4. On the left is the case for a divider which triggers at the zero crossing.
The output signals for both the high input level and low input level transition at the same
time; hence, the divider rejects AM and there is no added clock skew. On the right is the
case for a divider which triggers at a point other than the zero crossing, resulting in
AM-to-PM conversion or clock skew. The amount of clock skew induced depends on the
Figure 5-4
Illustrations of AM-to-PM conversion for a divider operating in the latched mode
with trigger levels either at the zero crossing (left) or not (right).


219
of power dissipation. The results also show that technology developments such as Cu or
low-K will have the greatest positive impact on systems whose total equivalent capacitance
is dominated by Cw, such as the grid. Finally, the results show that the power dissipated in
the clock receivers, given by Cq, should be a small fraction (1.9%) of the total power dis
sipated in the microprocessor. These results show that power dissipation does not impose
limitations for wireless clock distribution systems.
7.3 Process Variation
The wireless clock distribution system consists of multiple clock receivers distrib
uted throughout the microprocessor. These receivers will have to be matched across the
chip in terms of gain and phase so that the output local clock signals are synchronized.
However, process variation and temperature gradients make this matching difficult; thus,
there will be a nonzero clock skew due to process variation. The effects of process varia
tion on a low noise amplifier, frequency divider, and clock receiver will be simulated in
this section using Monte Carlo analyses, and the resultant local clock skew will be shown.
7.3.1 Simulation Methodology
Two ways process variation can be simulated are by using either comer models or
Monte Carlo analyses. The comer models provide a worst-case variation that can be
expected lot-to-lot, representing approximately a +/- 3a variation. These models can basi
cally be used to confirm circuit operation across all wafers. While this is important for the
wireless clock distribution system, evaluating the variation within a single die is important
as well. The within-die variation is simulated using Monte Carlo analyses in SPICE. With
these analyses, select model parameters are defined by a probability density function


230
4.6.2). Additionally, this skew correction technique is static; hence, the receivers could
accumulate phase errors if, for example, a single clock pulse is missed due to interference.
More work is required to analyze this phase-error accumulation.
In Chapter 6, the measured skew between two receivers with programmable divid
ers was 4.7%. This skew included process variation, TOF delay mismatches, amplitude
mismatch, and different interference structures. As expected, the measured skew is less
than 6.25%. These results show that the wireless clock distribution system should be able
to meet the skew requirements of global clock distribution systems. However, additional
work is needed to perfect the start-up technique for the programmable dividers (and cor
rect for the state-dependent initialization failure of the dividers described in 4.6.4). This
start-up methodology appears to be the ultimate key for demonstrating feasibility in terms
of clock skew for a wireless clock distribution system.
7.4.2 Clock Jitter
The main elements contributing to clock jitter for a wireless clock distribution sys
tem are as follows: thermal noise received by the antenna and generated in the receiver and
antenna, switching noise or digital interference received by the antenna, power-supply
noise, and phase noise from the phase-locked loop (PLL) which synthesizes the global
clock signal in the transmitter. Since these sources of jitter are all independent, their vari
ances add. As was shown in Chapter 5, thermal noise and interference degrade the sig-
nal-to-noise ratio (SNR) at the input of the frequency divider, resulting in jitter at the
output of the divider (given by (5.35)). Noise on the power supplies will translate into
noise on the signals themselves, depending on the power-supply-rejection-ratio of the


60
spacing = 1.25-|im. Patterned-ground shields were again used to decrease the substrate
resistance. Since this technology has a high-resistivity substrate, the increase in series
resistance, and hence degradation in Q, caused by eddy currents in the substrate is small,
as was shown in section 3.4.2. Ld was implemented using metals 5 and 6 shunted together,
decreasing the series resistance and maximizing Q. Simulations show that for Lg, mini
mizing the parasitic capacitance has a greater effect on NF than minimizing the series
resistance (provided the Q is > 20). This is due to the capacitance moving the input match
ing condition away from the optimum noise match. Therefore, due to the availability of
copper, only metal 6 was used for Lg.
A key benefit of copper metallization is its reduction of interconnect series resis
tance. For inductors, the Q will be almost doubled as compared with the same-thickness
aluminum (equal to the ratio of the A1 and Cu resistivities). Both metals 5 and 6 were to
have a nominal thickness of 1 pm. However, during fabrication these two metal layers had
to be thinned from 1 pm to approximately 0.5 pm. Therefore, rs was increased and Q was
decreased. Figure 3-14 shows the measured inductance and quality factor (Qbw [098])


178
7 mm
Figure 6-10 Layout of UMC test-chip showing locations of relevant zigzag antennas (Zx), loop
antenna (LPj), clock transmitters (Tx), and clock receivers (Rx).
6.5 Antenna Characteristics in 0.18-flm CMOS
6.5.1 Chip Implementation
Based on the results achieved in the 0.25-|im CMOS test chip, a test chip was
implemented in a 0.18-pm CMOS technology with copper interconnects. This technology
was obtained from UMC as part of the SRC Copper Design Challenge, sponsored by the
Semiconductor Research Corporation (SRC), Novellus, SpeedFam-IPEC, and UMC. The
substrate resistivity is 15-25 Q-cm. Six layers of copper interconnect were provided.
Multiple antenna test structures, LNAs, frequency dividers, clock receivers, and
clock transmitters have been included in this test chip. Figure 6-10 shows the layout of the


70
which has a magnitude less than one, a pole at co=(gm3+gT)/(CgS3+CLT) < 0>p, and a zero at
0)=%. Negative conductance indicates that energy is being generated. When a source-fol
lower is used after an amplifier containing an inductive load, this energy actually replen
ishes some of the energy lost in the inductor, causing the quality factor to increase. Hence,
the negative conductance of the source-follower increases the gain of the LNA through
increasing the load quality factor.
Figure 3-19 shows the measured input conductance versus control voltage (Vgc),
versus frequency for a source-follower implemented in the 0.25-pm CMOS test chip. This
buffer is driving an open pad structure having a capacitance of ~ 40 fF. The conductance
becomes more negative with increasing frequency in agreement with (3.2). Also, the con
ductance becomes more negative as Vgc increases to 1.2-1.4 V, and then begins to become
less negative for further increases in Vgc. This is due to gm3 increasing for Vgc between 0
to 1.2 V, causing Gsf to decrease. However, the transistors in the source-follower eventu
ally enter the linear regime, causing gT to increase and gm3 to decrease; thus, Gsf becomes
less negative.
3.6.3 Gain of 24-GHz LNA
The voltage gain of the three-stage LNA can be readily derived using equations
(A.3), (3.4), and (2.23). Assuming Gsf < 0, the total voltage gain is as follows:
v|co = a,
l^v, eg ^v, sf ^v, cj|
(3.5)
n((0o)
8m3+j<*Cgs3
8m3 + 8t + j((Cgs3 + Clt)
2smfiLd2(£>Ld2
8(co0)
(GO


6.3 Antenna Characteristics in 0.25-|im CMOS 170
6.3.1 Measured Characteristics for Antennas 171
6.3.2 Integrated Antenna with LNA 174
6.4 Wireless Interconnect in 0.25-|im CMOS 175
6.5 Antenna Characteristics in 0.18-|im CMOS 178
6.5.1 Chip Implementation 178
6.5.2 Antenna Descriptions 179
6.5.3 Measured Antenna Characteristics 181
6.6 Wireless Interconnects in 0.18-|im CMOS 186
6.6.1 Clock Transmitter 186
6.6.2 Voltage Controlled Oscillator 187
6.6.3 Power Amplifier 191
6.6.4 Clock Transmitter with Integrated Antenna 194
6.6.5 Single Clock Receiver with Integrated Antenna 197
6.6.6 Simultaneous Transmitter and Receiver Operation 199
6.7 Double-Receiver Wireless Interconnect 200
6.7.1 Measurement Setup 201
6.7.2 Demonstration of Double-Receiver Interconnect 203
6.7.3 Measured Skew Between Two Clock Receivers 206
6.7.4 Measured Jitter of Clock Receivers 208
6.8 Summary 210
7 FEASIBILITY OF WIRELESS CLOCK DISTRIBUTION SYSTEM 212
7.1 Overview 212
7.2 Power Consumption Analysis 212
7.2.1 Power Consumption Comparison Methodology 213
7.2.2 Clock Distribution Systems 216
7.2.3 Results and Conclusions for Power Consumption 218
7.3 Process Variation 219
7.3.1 Simulation Methodology 219
7.3.2 LN A and Frequency Divider Variation 221
7.3.3 Clock Receiver Variation 222
7.3.4 Conclusions for Process Variation 225
7.4 Synchronization 226
7.4.1 Clock Skew 227
7.4.2 Clock Jitter 230
7.4.3 Conclusions for Synchronization 233
7.5 Latency of 0.18-p.m Wireless Interconnect 233
7.6 Intangibles 235
7.7 Conclusions and Future Work 236
7.7.1 Feasibility Summary 236
7.7.2 Conclusions for Wireless Clock Distribution Systems 237
7.7.3 Broader Applicability 238
7.7.4 Suggested Future Work 239
vi


148
Now all the tools are available to derive the relationship between signal-to-noise
ratio (SNR) and clock jitter due to input additive noise. The output jitter due to an injec
tion-locked divider will be less than that for a latched divider; hence, (5.30) is used for the
output phase noise of the divider, representing the worst case. Plugging (5.30) into (5.33)
the total output jitter of the clock receiver due to input additive noise, is as follows:
(5.34)
Substituting (5.27), results in
(5.35)
The timing jitter can be converted to a phase jitter C7cycie (in cycles or, equivalently, as a
percentage of the period) by simply multiplying (5.35) by the frequency squared, yielding
(5.36)
Figure 5-7 shows a plot of cycle-to-cycle RMS clock jitter versus SNR (in dB) for
a division ratio of N=8. As can be seen, as the SNR improves, the clock jitter improves as
well. The slope of the line is linear when plotting the jitter on a logarithmic scale, with a
slope of 6-dB (SNR) per octave decrease in jitter. To obtain an RMS jitter of 0.5% of the
local clock period, a 12-dB SNR is required at the input of the divider. When the SNR is 6
and 0 dB, the local clock jitter is 1% and 2%, respectively. The fact that 0 dB SNR yields
only 2% RMS jitter is due to the 8:1 divider reducing the jitter (in percentage) by the divi
sion ratio. Thus, the RMS jitter at the input of the divider is 16%. The peak-to-peak jitter
at the input of the divider for a 0-dB SNR is then 6ain = 96%.


157
Desired
Interferer
A¡ nt
Desired
GCLK

^mds
f Signals Into Receiver
Interferer
a>(xK<
1 ifo)Amds
Harmonic
1 (fo\A3
43I 3 IAint

fo
Signals Out of LNA
Into Divider
Figure 5-10 Illustration of scenario used to specify LNA IIP3.
where x and y are the input and output of the LNA, OC] is the gain of the LNA, and a2 and
a3 are the second- and third-order nonlinearity voltage coefficients, respectively. For a
fully-differential circuit, the even-order nonlinearity coefficients are close to zero [Raz95].
/
Therefore, the output from the LNA for the interfering signal at is as follows:
y(0 = [a1(y)Al, + |a3(j)Ajcos(24<) + ia3(^)A^coS(2It4, (5.44)
where the frequency dependence of the a parameters has been included. The third-har
monic term in this expression will interfere with the desired GCLK signal, degrading the
SNR and eventually overtaking the desired signal.
The scenario used to specify JIP3 is now described. Integral to any IIP3 specifica
tion is the constraint on the power of the interferer. In the case of the clock receiver, the
power of the interferer is constrained by setting the ratio between the output power of the
fo
desired signal at/0, when it is at its sensitivity, to the output power of the interferer at
to the required SNR to meet a given jitter specification. In equation form, this is written as


113
vb
(a)
V,
dd
DFF for high-speed 2:1
(b)
Figure 4-24 (a) Schematic of fixed programmable DFF (for lower frequency divide-by-4 and
divide-by-8). (b) Schematic of improved high-frequency (divide-by-2)
programmable DFF.
modifications. The evaluate and hold portions are turned off during INI=1 by including
two transistors, driven by ENI, in series with the CLK and CLKb transistors (performing a
logical AND function). To write both high and low values to the drain nodes, the PMOS
transistors are replaced with inverters. The size of the NMOS transistor in this inverter is
kept small. Simulations show that using this new DFF in the lower frequency 4:1 and 8:1
stages eliminates the state-dependent initialization failure. However, using this DFF in the
high-frequency 2:1 stage decreases the operating frequency, due to the increased capaci
tive load from the inverter. Thus, the modified DFF shown in Figure 4-24(b) is used for the
first 2:1 stage. Simulations show that this stage will initialize properly due to the sizing
ratio between the PMOS and NMOS. However, in future implementations of this divider,


107
S1 SO
one of the PMOS transistors pulls its drain up to Vdd, the positive feedback in the regener
ative pair causes the other drain to be pulled down to ground. Since these PMOS
transistors originally had their gates grounded to act as pseudo-NMOS loads, the initial
ization data are multiplexed with a ground signal, where the INI signal controls the multi
plexer. Thus, for INI = 1, Vx is defined as above, and for INI = 0, Vx = 0.
The logic blocks are implemented using complimentary pass-transistor logic
(CPL) [Bel95], in a logic array. Inputs to the array, shown in Figure 4-19, are SI, SO, INI,
and their complements. Inverters are required at the output of each logic function to


APPENDIX E
INJECTION LOCKING OF OSCILLATORS
E.l Overview
Any oscillator can be locked or synchronized to an external signal whose fre
quency (or harmonic) is close to the natural frequency of the oscillator. This type of circuit
is known as an injection-locked oscillator (ILO), where the oscillator is locked to either
the fundamental, subharmonic, or superharmonic of the natural oscillating frequency. The
first presentation of ILOs was by Adler in 1946 [Adl46], where he developed a differential
equation relating the phase difference between the locking (input) and oscillator (output)
signals for small-level input signals near the natural frequency of the oscillator. From this,
the locking frequency range and voltage levels were related, and then the transient pull-in
process was investigated. This theory was extended by [Pac65] to accommodate large
input-locking levels. The noise properties of ILOs have also been investigated, where the
ILO behaves as a first-order phase-locked loop with regards to noise on the input locking
signal [Kur68, Llo98]. Multiple other investigations and applications of ILOs exist
[Sch71, Uzu85, Zha92, Rat99], and ILOs remain a topic of active research.
The function of this appendix is to review the basic operation of an ILO and to
emphasize the results that are important for clock receivers which utilize a superharmonic
ILO in the frequency divider. Therefore, the basic differential equation for the phase offset
between the input and output signals will be reviewed, followed by a presentation of the
locking range, its dependence on signal level, and the steady-state phase error between the
260


17
the short-channel regime and when taking into account other sources of noise (e.g., sub
strate resistance, inductor parasitic resistance, and gate-induced noise [Zie86, Sha97,
Tsi99]), with typical experimental values exceeding 3 dB.
The high noise figure for the common-gate topology is a major disadvantage. Fun
damentally, the high noise figure is due to gm being constrained by the input matching
condition, which thereby constrains the noise figure. This is analogous to saying that the
input matching conditions for optimal power match and noise match are not coincident.
Therefore, a topology which decouples gm from the input matching condition would add
an additional degree of freedom to the design, and hopefully superimpose the power and
noise matching conditions. A topology which provides this decoupling is shown in Figure
2-1(b), which consists of a common-source amplifier with inductive degeneration. To
obtain the extra degree of freedom in the design, an inductor (Lg) is added in series with
the gate. As will be shown in the next section, this topology allows the power and noise
matching conditions to be met simultaneously, while exhibiting sufficient linearity and
allowing for low power consumption.
2.2.2 Source-Degenerated CMOS LNA
Figure 2-2 shows a simplified schematic of a CMOS LNA with inductive source
degeneration. This LNA is matched to 50 Q at both the input and the output. A sin
gle-stage topology is used to minimize the power dissipation and to improve 1-dB com
pression point (PldB) and IP3 performance. The circuit gain is provided by a cascode
1. If gm = 1/50 2'1, then the noise figure would be approximately 2.2 dB for the
long-channel limit.


19
A salient feature of this topology is that source degeneration provides a real term to the
input impedance, which can then be used to match to 50 1 This real term is not a resis
tance per se (hence it will not generate thermal noise), but rather when a current is applied
to the input node, the voltage that develops at that node has components both in phase
(real term) and 90 degrees out of phase (imaginary terms due to Lg, Ls, and Cgsl). The
in-phase component results from the series feedback in the source of Mj.
As can be seen from (2.5), this network takes the form of a series resonant circuit,
with resonant frequency
_i
= [(S + L,)CSJ,] 2. (2.6)
Thus, at series resonance, the input impedance becomes
ry Sm 1 j
Z¡n ~ r Ls~ (0tLs ,
(2.7)
which is a function of the bias condition, the channel length of Mj, and Ls. The quality
factor of this network, including the source resistance, Rs, is
in 0Cgil (R5 + corLJ)
Thus, to design the input matching network for a given Cgsi to achieve an
and series resonance, the following relations are used:
(2.8)
< -10 dB
26 96
< L <
(0r j T
Ls =
T Ls-
*oC 1
2.3.2 Input Match to Withstand Component Variations
(2.9)
(2.10)
A methodology is then needed to choose Cgsl (i.e., the width of Mj) and yield Ls
and Lg. One way is to choose Cgsi to meet the input matching condition over an entire


73
Examining these results, it can be concluded that the gain performance of this
23.8-GHz amplifier is limited primarily by the inductor quality factor. Since this design
was completed before the effects of substrate resistance on inductor Q [FloOOb] were fully
understood, patterned ground shields were not used even though the substrate resistivity of
1-2 Q-cm is in the range where ground shields would have enhanced the inductor Q. The
substrate resistance for these inductors is on the order of 300 Q while the parasitic
capacitance to the substrate is 25 fF (refer to inductor model in Figure 3-1(a)). According
to equation (3.1), the inductor Q is greatly reduced at 24 GHz, due to the substrate resis
tance. In fact, had patterned ground shields been used, then the substrate resistance would
have been reduced to -10-15 Q, while the parasitic capacitance would have increased.
This would have approximately doubled the inductor Q, potentially increasing the gain by
201og{2} =12 dB (for the case when the source-follower is adjusted for close to zero input
conductance). Further improvements could be obtained by improving the back-end pro
cess by increasing the top-level metal thickness, increasing the distance between the
top-level metal and the substrate, and using copper rather than aluminum for the metal (as
will be the case for fully-scaled 0.1 -|im technology). Such improvements should easily
lead to inductor Q > 30 [Bur98] at 20 GHz, as has been shown in this chapter. This would
dramatically increase the amplifier gain and decrease the power consumption.
Using simulated values for all quantities except in (3.7), and then inserting the
extracted value of 2, results in S2i = 9.2 dB, which is close to the measured S2i- The
discrepancy is attributed to differences between the transistor model and the actual device.
The estimated input conductance to the source-follower is -2 mS, resulting in a Qeff of 2.8.
Therefore, the negative conductance of the source-follower improves S21 by a factor of


228
The non-random or systematic skew component will be due to time-of-flight
(TOF) delay mismatches. Since the receiver locations cannot be precisely known (the
evenly-spaced grid of 16 receivers is more conceptual than practical) there can be a situa
tion where the TOF mismatch is half of the input clock period. For example, the speed of
light in a silicon-substrate propagation medium2 is 0.87xl08 m/s. If the distance between
the transmitter and two receivers differs by 2.9 mm plus any integer wavelength (7.s¡ = 5.8
mm at 15 GHz), then the TOF delay mismatch is 33 ps, which is half of the period of a
15-GHz input signal. Therefore, the worst-case TOF mismatch is 50% with respect to the
input signal, resulting in 6.3% of skew with respect to the output signal.
Random skew components will be due to process variation, amplitude mismatches,
and load mismatches. The effects of process variation on skew were presented in the pre
vious section. The results show that for the 0.25-|lm receiver, the simulated skew due to
process variation is 3.8%. This was for the case when the input signal was large enough to
lock the receiver, which is a necessary assumption for this analysis. Although the process
variation analysis contained skew due to amplitude mismatch generated in the receiver,
there will also be amplitude mismatch due to different antenna gains. The longer the
path-length, the lower the transmission gain. Also, interference structures will modify the
antenna gain further. Referring to section 5.2.2, the specified antenna gains for 0.5- and
1.5-cm distances are -40 and -56 dB, respectively. Metal interference structures have been
shown experimentally to degrade the gain by ~7 dB. Thus, there is a maximum of 23 dB of
amplitude mismatch at the input of the divider. Applying (5.17), this results in 3% of skew
2. Since multiple paths exist from the transmitting to the receiving antenna, the effective
dielectric constant and hence, speed of light, will be between that of silicon and SO2.


63
-Measured y22
Simulated y22
PMOS
rVgs=0V, Vds=-0.2V
=19.2 fF .
2 2.5 3
Frequency (GHz)
Figure 3-15 Measured versus simulated yu (with transistor in linear region) and y22 (with
transistor off) for NMOS and PMOS transistors. The slope of these lines over 2k
is equal to the capacitances Cgg and Cdd for yu and y22, respectively.
tor biased in the linear region, yn yields the total gate capacitance (Cgg = Cgs + Cgd + Cgb
= WLCox + 2WC0V). With the transistor turned off, the channel conductance is zero, and
y22 yields the total drain capacitance (Cdd = Cdb + Cgd), while y12 yields Cgd. Although
there are more accurate ways to obtain these capacitances, this measurement provides a
quick and easy way to observe any capacitance trends. Figure 3-15 shows the measured
versus simulated yu and y22 for the NMOS and PMOS transistors. The slope of each of
these lines is proportional to capacitance, therefore the steeper the line, the higher the
capacitance. As can be seen, the measured capacitances are all higher than the simulated
capacitances (notebecause the open pad structure used for de-embedding does not
include the interconnects to the transistor, the measurements overestimate the capacitances


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E419W17BA_GVG68C INGEST_TIME 2015-02-05T21:07:30Z PACKAGE AA00024537_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


95
Table 4-2 Summary of measured characteristics for SCL frequency dividers
Frequency
Divider
0.25-|lm Divider
0.18-|im Divider
Division Ratio
128:1
64:1
f
Amax
6.6 GHz @ 1.5 V
8.8 GHz @ 2.0 V
9.98 GHz @ 2.5 V
15.8 GHz @ 1.5 V
20.4 GHz @ 2.1 V
fiSO
6.1 GHz @ 2.0 V
7.3 GHz @ 2.5 V
9.15 GHz @ 1.5 V
14.2 GHz @ 2.1 V
Power
3.3 mW @ 1.5 V
9.4 mW @ 2.0 V
18.4 mW @ 2.5 V
4.5 mW @ 1.5 V
12.2 mW @ 2.1 V
4.5 A 15.8-GHz. 0.18-^im CMOS SCL Divider
4.5.1 Circuit Implementation
Building on the results achieved in the 0.25-(im CMOS technology, a 15.8-GHz
64:1 source-coupled frequency divider was implemented using a 0.18-|lm CMOS technol
ogy with six layers of copper interconnects [FloOla], This technology was obtained from
UMC as part of the SRC Copper Design Challenge, sponsored by the Semiconductor
Research Corporation (SRC), Novellus, SpeedFam-IPEC, and UMC. The block diagram
shown in Figure 4-11 is similar to that in Figure 4-5, except the division ratio is now 64:1
and there are two TSPC low-frequency 8:1 dividers in opposite phase. This is done to pro
vide two output signals for testing and triggering, as well as to minimize the supply-line
bounce. All of the devices for the SCL and TSPC dividers shown in Figures 4-1 and 4-6
are implemented with channel lengths of 0.18 (im and channel widths given in Table 4-3.
Figure 4-12 shows a die photograph of the 64:1 divider, with a die size of 370x590
(im2. On-chip 20-pF bypass capacitors are included between Vdd and ground for the 8:1


222
7.5
8
Time (ns)
8.5
Figure 7-5
Process variation of frequency divider, obtained with Monte Carlo analysis (30
iterations), for an 8.5-GHz, 500-mV amplitude input signal, Vdd = 2 V.
self-oscillating. A 640-MHz variation in the output frequency of the 2:1 divider was
observed, corresponding to a 1,28-GHz variation in the input-referred self-oscillation fre
quency, which is significant.
7.3.3 Clock Receiver Variation
The variability of the entire clock receiver was analyzed using both small-signal
(AC) analysis and transient analysis. Referring to Figure l-3(b), the receiver consists of a
fully-differential LNA, a pair of source-follower buffers, an 8:1 frequency divider, and
output buffers. Figure 7-6(a) shows the simulated gain variation from a Monte Carlo anal
ysis (30 iterations), equal to the voltage gain versus frequency from the input of the
receiver to the input of the frequency divider. Due to both the LNA driving a capacitive
load and the negative input conductance of the source-followers, the gain is much larger
than that of the LNA driving a resistive load. Furthermore, since the resonant frequency of
the LNA in the receiver is set by the small input capacitance of the source-follower, the


269
[BravOOb]
[Bur98]
[Che90]
[Che98]
[Col98]
[Cra95]
[Deu98]
[Ega90]
[Ega91]
[Ega99]
[FalOl]
D. Bravo, H. Yoon, K. Kim. B. Floyd, and K. K. O, Estimation of the sig-
nal-to-noise ratio for on-chip wireless clock signal, IEEE International
Interconnect Technology Conf., pp. 9-11, June 2000.
J. N. Burghartz, D. C. Edelstein, M. Soyuer, H. A. Ainspan, K. A. Jenkins,
RF circuit design aspects of spiral inductors on silicon, IEEE J. Solid-
State Circuits, vol. 33, no. 12, pp. 2028-2034, Dec. 1998.
J. Y. Chen, CMOS devices and technology for VLSI, Englewood Cliffs, NJ:
Prentice Hall, 1990, pp. 211-227.
T. Chen, K. Kim, and K. K. O, Application of a new circuit design ori
ented Q extraction technique to inductors in silicon ICs, IEDM Technical
Digest, pp. 527-530, Dec. 1998.
J. T. Colvin, S. Bhatia, and K. K. O, A bond-pad structure for reducing
effects of substrate resistance on LNA performance in a silicon bipolar
technology, Proc. 1998 Bipolar Circuits, and Technology Meeting, pp.
109-112, Sep. 1998.
J. Craninckx and M. Steyaert, A 1.8-GHz CMOS low-phase-noise volt
age-controlled oscillator with prescaler, ISSCC J. Solid-State Circuits, vol.
30, no. 12, pp. 1474-1482, Dec. 1995.
A. Deutsch, H. Harrer, C. W. Surovic, G. Hellner, D. C. Edelstein, R. D.
Goldblatt, G. A. Biery, N. A. Greco, D. M. Foster, E. Crabbe, L. T. Su, P.
W. Coteus, Functional high-speed characterization and modeling of a six-
layer copper wiring structure and performance comparison with A1 on-chip
interconnections, IEDM Technical Digest, pp. 295-298, Dec. 1998.
W. F. Egan, Modeling phase noise in frequency dividers, IEEE Trans, on
Ultrasonics, Ferroelectronics, and Frequency Control, vol. 37, no. 4, pp.
307-315, July 1990.
W. F. Egan, Phase noise modelling in frequency dividers, IEEE Int. Fre
quency Control Symp., pp. 629-635, May 1991.
W. F. Egan, Frequency Synthesis by Phase Lock, 2nd edition, New York: J.
Wiley & Sons, 1999.
C. Fallesen and P. Ashbeck, A 1W 0.35|im CMOS power amplifier for
GSM-1800 with 45% PAE, ISSCC Digest Technical Papers, pp. 158-159,
Feb. 2001.


256
: J = 0(CGr) =
^21 ^ J(C^j\(>T)
(D.19)
where the cutoff frequency of the MOSFET is equal to
Sn
COnn = ^
r cr + c
(D.20)
'gs gd
The alternative and impedance-based noise parameters are easily obtained and are shown
in the following table, D-2. To obtain the noise parameters without Cgd, simply set Cgd=0.
Table D-2 Noise parameters considering Cgd
Alternative Noise Parameters
Impedance-Based Noise Parameters
*'= 2 r, (a21)
aH,+U JJ
Fminl = 1 (D.22)
G> = *(£) r, + (D23)
L + v 8m J J
Pni ~ (D.24)
z0' UcGT (D25)
Looking at these noise parameters, it can first be seen that the optimal source
impedance is an inductor which series resonates with Cgs. Second, the optimal source
resistance is a short, meaning that when driving the network with a resistive load, a noise
mismatch occurs. Furthermore, Fmin = 1, meaning that with perfect noise matching (i.e.,
Rs=0), the MOSFET does not add noise. This is due to there being only one noise source,
which is fully correlated to itself; hence, Ru = 0, while Rc = 0 as well. However, Fm¡n = 1
is physically unreasonable, meaning that second-order effects will determine Fmjn.


156
clock jitter. Though not the main concern, the harmonics and intermodulation products
will still degrade the SNR.
One of the main reasons the global clock is distributed at a higher frequency than
the local clock is to try to operate in a band which is not as noisy. The system will be gen
erating a significant amount of noise at and below the local clock frequency, and at fre
quencies due to the edge-rates of the signals. One potential problematic frequency is
around one-third of the global clock frequency4 (GCLK). Due to nonlinearity, the LNA
will generate a harmonic near GCLK. When this signal is added to the desired global
clock, the SNR will decrease and jitter will increase. This scenario can then be used to
specify the IIP3 of the LNA. Note that a scenario exists for the case of two interferers gen
erating an intermodulation product. However, since these interferers must be far enough
away from GCLK to prevent divider locking, the HP3 would be very similar to the case
discussed here.
Figure 5-10 shows an illustration of the case being considered. Two sinusoidal
input signals are present at the input of the receiver. One is at the global clock frequency
(f0) having a voltage amplitude, Asens, corresponding to the sensitivity. The interfering
fo
signal is at a frequency, with an amplitude, Aint. The response of a nonlinear LNA can
be modeled with a power series as
y(f) = tXj x(t) + a2 x2(t) + a3 x3(t) + ..., (5.43)
4. For the wireless clock distribution system, the local clock frequency will be at 1/8 of
GCLK. Assuming that this local clock signal is square, its third harmonic will be quite
strong, occurring at a frequency of 3/8 (or 0.375) of the GCLK. This subsequently can
generate a third-order harmonic at 9/8 (or 1.125) of the GCLK due to LNA nonlinearities.


220
(PDF), and for each Monte Carlo iteration, SPICE randomly selects values for the param
eters according to the PDFs (i.e., the values are more likely to be closer to the mean than to
the mean +/- 3a). This then provides statistical data for the simulated circuit, from which
the single-die variation can be obtained.
The Monte Carlo model file was generated for the 0.25-|im CMOS circuits from
TSMC. Every parameter that was varied in the comer models was replaced by a PDF,
where the typical model parameters provided the mean value and the fast and slow comers
provided the 6-a variation. For example, in the comer models, the oxide thickness varies
from a 5.5-nm thickness at the fast comer to a 6.1-nm thickness at the slow comer, having
a typical value of 5.8 nm. Thus, the PDF of the oxide thickness is Gaussian1 with a mean
of 5.8 nm and a 3-a variation of 0.3 nm. For the TSMC process, the following model
parameters for both NMOS and PMOS were replaced with PDFs: oxide thickness, length
and width variation, the threshold voltage at zero body-to-source voltage and low drain
bias, the junction and sidewall capacitances, and the gate-overlap capacitances. In addition
to these transistor parameters, the polysilicon sheet resistance was varied according to the
process data (4.5 +/- 4 Q/square). Finally, the resistance associated with the metal spi
ral-inductors was varied assuming a 3-a variation of 10%.
1. In HSPICE, the parameter statement is either variable = AGAUSS(mean, variation, sig),
or GAUSS(mean, variation, sig). The mean and variation are specified, with sig being the
number of standard deviations the variation represents. Here, AGAUSS denotes an abso
lute gaussian where the variation is an absolute quantity, and GAUSS denotes a relative
gaussian where the variation is with respect to the mean. For the oxide-thickness PDF
example above, equivalent statements would be tox = AGAUSS(5.8n, 0.3n, 3) or tox =
GAUSS(5.8n, 0.052, 3), where the 3-a variation of 0.3n is 5.2% of the mean-value 5.8n.


248
(C.l)
where Fmin is the minimum obtainable noise factor, Rn is the noise conductance, Yopt =
Gopt + jBopt is the optimum source admittance resulting in F = Fmin, and Ys = Gs + jBs is
the actual source admittance. These noise parameters are easily derived from the input
noise generators, and they assume that the input is in an admittance form. However, for
cases when the input is in an impedance form (e.g., for a source-degenerated LNA), it is
advantageous to represent the noise parameters in an impedance form as well. Such a rep
resentation yields design insight, since the optimal noise factor will occur when the volt
age-source impedance is equal to the optimum noise impedance. Unfortunately, most
electronics textbooks only represent the noise parameters in an admittance form; there
fore, the impedance form of these parameters is derived here.
C.2 Noise Parameters in Impedance Form
Assume that the two-port network is represented using equivalent input noise gen
erators, as shown in Figure C-l(b). A source with an impedance Zs=Rs+jXs is connected
to the input and generates an input thermal noise of
v2s = 4kTRsAf,
(C.2)
where k is Boltzmanns constant (1.38 x 10'23 J/K) and T is the absolute temperature in
Kelvin. The noise factor of the network is then equal to the total input-referred noise
divided by the noise generated at the input by the source, as follows:
(C.3)
2


91
Table 4-1 Transistor sizing for 0.25-pm 128:1 frequency divider
SCL 8:1 Divider
Transistors
First 2:1
Second 2:1
Third 2:1
WpW2.Wn.Wj2
12 pm
3 pm
3 pm
W3,w4,w13, w14
6 pm
6 pm
6 pm
w5, w6, w15, w16
2.1 pm
2.1 pm
4.2 pm
w7, w8, w9, w10
5.1 pm
2.55 pm
2.55 pm
TSPC 16:1 Divider
Sizes Axe Same for each 2:1 TSPC Divider
w2, w3, w4, W5, W6, W7, Wg, w10
6 pm
Wj,Wn
3 pm
w9
12 pm
Figure 4-7 Die photograph of 0.25-pm, 128:1, 10-GHz source-coupled divider.
Figure 4-7 shows a die photograph of the 128:1 divider [FloOOa]. The die size is
730x510 pm2. On-chip 20-pF bypass capacitors are included between Vdd and ground for
the 8:1 SCL divider, 16:1 TSPC divider, and output buffer. Separate supplies are used for
each circuit to prevent any low-frequency supply noise from the latter stages of the divider
from corrupting the high-frequency operation of the first 8:1 divider. In particular, when


50
points 1 and 3, respectively. To match the input to obtain Fmin, Lg should be increased and
an off-chip capacitor should be added. This transforms the 50-2 source impedance to Zopt
(equivalent to reflection coefficient ropt). However, Topt for the LNA is not equal to Sj j*,
therefore the noise matching condition and the power matching condition do not coincide.
Table 3-1 also shows the NF corresponding to perfect input power matching (i.e., Sj j =
-dB). The noise parameters can therefore be used to derive the optimal input matching
network, such that minimum noise figure is obtained while S] j is constrained to be -10 dB.
As shown in Table 3-1, these noise figures are 1.12 and 1.56 dB for operating points 1 and
3, which are very close to their respective Fmjn. To obtain these NF at Su=-10 dB, Lg
should be increased by 4 nH and a 4-pF shunt capacitor should be added to the LNA input.
The reverse isolation for each operating point is very good with more than 42 dB
of isolation for each bias condition. As mentioned earlier, this is due to the use of a pack
age with low ground-inductance, ground-shielded pad structures, a ground-shielded induc
tor, and a cascode amplifier. The measured PldB and IP3 data for each operating point are
shown in Figure 3-7. As can be seen, the power of the intermodulation product (P2fi-o)
unchanged by the current level in the LNA. However, the output power of the fundamental
(Pfl) depends on the operating power gain at the specific current level. Therefore, the IIP3
relatively scales in accordance with the gain. The output power level at which the LNA
compresses is a function of the supply voltage, which is 2.7 V for operating points 2-4 and
3 V for operating point 1. Therefore, the PldB (output) is approximately the same for each
operating point, while PldB (input) scales with the gain. The measured HP3s at 30 mW
and 6.3 mW are -1 dBm and -3.8 dBm. Due to the use of a single-stage topology, the
amplifier is very linear, and is suitable for GSM applications.


28
Looking at these four options, QLd and <% have the most significant impact on
gain. Intuitively, high-quality factor inductors result in LNAs which consume lower
power. This is the case for [Hay98], where external matching networks are used to obtain
high gain and low noise at a very small power. However, for a fully integrated LNA, exter
nal matching networks are taboo, since the extra components increase the cost and since
their high Qs are not very repeatable. Thus, to obtain the desired gain using on-chip
inductors with limited quality factors, the power consumption has to be increased. The
quality factor of Ld is limited by the number of metal layers present, the oxide thickness
between the top-level metal and the substrate, the substrate resistance, and the type of
material used for the metal interconnect (aluminum or copper).
2.5.3 Gain Driving Capacitive Load
When the LNA does not drive 50 2, and instead drives a capacitive load (i.e.,
mixer or frequency divider), the gain is significantly increased. Thus, when the capacitive
transformer (C¡ and C2) is removed, consists only of the drain inductor with its para-
sitics in parallel with an equivalent load capacitance composed of C^, CM2, and the
capacitance of the subsequent stage. Thus, ZLeq(co0) becomes twice that given by equation
(2.22). Also, with the output matching network removed, n(on) is no longer required. The
voltage gain then becomes the following:
(2.27)
With the elimination of rc(co) and the additional factor of two, the gain can be significantly
larger. This means that the desired gain of ~15 dB can be achieved at a reduced power con
sumption (i.e., Op). Also, since Q and C2 are removed, the total capacitance resonating
with Ld is small. Thus, Ld can be increased, further increasing the gain.


84
Figure 4-2 Representation of source-coupled latch as a fully-differential amplifier and the 2:1
divider as an oscillator.
representing each source-coupled latch as a fully-differential amplifier. The small-signal
loop gain (T) of the 2:1 divider can be shown to be as follows:
T = -I
Sm3
Sp-gcc + J^L
(4.3)
For oscillations to begin, the phase of T should be 360 and ITI > 1. Choosing 1TI > 1
ensures start-up and causes the amplitude of oscillation to increase until non-linearities
within the amplifier limit the amplitude. This amplitude condition requires gm3 > 0)oCL.
For the phase of T to be 360, the phase-shift through each differential amplifier should be
90. This corresponds to gP = gcc. This Barkhausen criterion for oscillators applies to
small-signal sinusoidal oscillations. However, with a loop gain greater than one, the oscil
lator will become nonlinear and its amplitude and frequency will be determined by the
large-signal characteristics of the circuit. The divider then behaves as a relaxation oscilla
tor (also known as an astable multivibrator) [Sed91].
Relaxation oscillators are hysteretic comparators which charge and discharge
capacitances located in a feedback loop. The oscillation frequency of a relaxation oscilla
tor depends on the capacitance being charged, as well as the current drive of the


65
Table 3-5 Summary of measured results for 14-GHz, 0.18-|im LNAs with copper.
Single-Stage Differential LNA
Differential LNA with Source-Followers
Center Frequency
13.3 GHz
Center Frequency
14.4 GHz
Gain
8 dB
Gain at Vgc=0.5V
21 dB
Sn
-8 dB
Gain at Vgc=0.9V
- 25 dB
S22
-15 dB
Sn
-5 dB
Reverse Isolation
27 dB
Reverse Isolation
32 dB
Noise Figure
7.7 dB
Noise Figure
8 dB
vdd
1.5 V
vdd
1.5 V
Power Consump
tion
12 mW
Power Consump
tion (Vgc=0.5V)
28.2 mW
Simulations predicted a 14-dB gain, a 3-dB NF, and a 21-GHz resonant frequency.
The reduced resonant frequency has already been accounted for, due to the increased
inductance and the larger transistor capacitances. Referring to equation (2.26), a 50%
reduction in Q results in a 6-dB reduction in lS2il2. This accounts for the drop in S2i from
the expected value of 14 dB to the measured value of 8 dB. The understanding of the
higher noise figure is not as straight-forward, though. First, these noise measurements
were obtained through on-wafer probing using external baluns, which yields higher noise
figures (on the order of 1-2 dB) as compared to packaged chips, due to variations in the
contact resistance of the probe. Second, generally speaking, as the gain is decreased, the
noise figure increases, due to higher input-referred noise. Third, decreased inductor Q
results in more thermal noise generated by the resistive loss of the inductors. This is par
ticularly severe for Lg and Ls, which are located at the input of the LNA. Fourth, the mod
eling of the transistors thermal noise coefficient (y, which was assumed to be 1.17 in
simulations based on 0.25-fxm measurement results, is expected to increase with


2.6.4 Optimum Width of M2 36
2.7 Design Methodologies for Source-Degenerated LNA 38
2.7.1 Derivation-Based Methodology 38
2.7.2 Alternative Constant Available Gain Methodology 39
2.8 Summary 41
3 CMOS LNA IMPLEMENTATION AND MEASUREMENTS 42
3.1 Overview 42
3.2 Passive Components 42
3.2.1 Inductor Design Techniques 42
3.2.2 Capacitor implementation 44
3.3 A 900-MHz, 0.8-|im CMOS LNA 45
3.3.1 Circuit Implementation 45
3.3.2 Measured Results 47
3.3.3 Summary for 900-MHz LNA 51
3.4 An 8-GHz, 0.25-|im CMOS LNA 52
3.4.1 Circuit Implementation 52
3.4.2 Inductor Characteristics 54
3.4.3 Measured Results 55
3.5 A 14-GHz, 0.18-p.m CMOS LNA 57
3.5.1 Circuit Implementation 57
3.5.2 Inductor Characteristics 58
3.5.3 Transistor Characterization 62
3.5.4 Measured Results 64
3.6 A 23.8-GHz, 0.1-fim SOI CMOS Tuned Amplifier 67
3.6.1 Circuit Implementation 67
3.6.2 Source-Follower 69
3.6.3 Gain of 24-GHz LNA 70
3.6.4 Measured Results 72
3.7 Summary 75
4 CMOS FREQUENCY DIVIDERS 77
4.1 Overview 77
4.2 Frequency Divider Using Source-Coupled Logic 78
4.2.1 Circuit Description 78
4.2.2 Latched Operating Mode 80
4.2.3 Quasi-Dynamic Operation 81
4.3 Injection Locking of SCL Frequency Divider 82
4.3.1 Oscillation of SCL Divider 83
4.3.2 Description of Injection Locking 85
4.3.3 Simulation of Injection-Locked Divider 87
4.3.4 Implications of Injection Locking for Clock Distribution 89
4.4 A 10-GHz, 0.25-p.m CMOS SCL Divider 89
4.4.1 Circuit implementation 89
4.4.2 Measured Results 92
iv


2
Table 1 -1 Technology Trends of Semiconductor Industry for Microprocessors'1-
Year
Technology Node
1999
180nm
2002
130nm
2005
lOOnm
2008
70nm
2011
50nm
Microprocessor Gate
Length (nm)
140
85-90
65
45
30-32
Microprocessor Chip
Size (mm2)
450
-508
622
713
817
Linear Dimension of
Chip (mm)
21.2
22.5
24.9
26.7
28.6
Local CLK (MHz)
1,250
2,100
3,500
6,000
10,000
Global CLK (MHz)
1,200
1,600
2,000
2,500
3,000
Metal Layers
7
8
9
9
10
Power Dissipation (W)
90
130
160
170
174
10% Global Skew
Requirement (ps)
83
62
50
40
33
+Source: 1999 International Technology Roadmap for Semiconductors [SIA99]
Figure 1-1 Global interconnect delay and gate-delay versus technology generation for
aluminum and copper metallization and conventional and low-K dielectrics.
To offset this problem, copper (Cu) interconnects and low-K dielectrics have been
introduced, decreasing the global propagation delay by a ratio of approximately PcuKlov/
pAjKconv where p is the resistivity of copper (aluminum) and K is the relative dielectric


159
The LNA gain at iJ2> can be found from simulations. These simulations show that the
attenuation at J3 is approximately 10 dB less than the gain at f0. If half of the gain in the
receiver is the gain in the LNA, then for a -54-dBm sensitivity, a 63-dB receiver gain, and
a 6-dB SNR (for 1% RMS jitter), the specified HP3 is -54 + 31.5 (-21.5) 6 4.8 = -11.8
dBm. Although this 1% RMS jitter is more than the 0.5% target, this results in a more
aggressive IIP3 specification. Thus, the system specification for IIP3 is -12 dBm.
5.9 Summary
In this chapter, system specifications have been developed for a wireless clock dis
tribution system, converting standard clock metrics into RF metrics. To maximize the
power transfer from the clock source to the local clock system, matching, gain, and
antenna requirements have been specified. In particular, the receiver gain should approxi
mately equal the loss through the antennas. Clock skew and jitter have been defined and
are used to set the remaining performance metrics of the system. Amplitude mismatch at
the input of the frequency divider results in clock skew, thus the signal-levels at the input
of the divider should be equal. Thermal noise at the input of the divider will result in clock
jitter, due to its conversion to phase noise. As the signal-to-noise ratio (SNR) improves,
the jitter decreases, and a minimum SNR at the input of the frequency divider is specified.
A 6-dB increase in the SNR halves the clock jitter. Using the SNR, a receiver sensitivity is
specified, the total output noise from the LNA with source-follower buffers is determined,
a noise figure is specified, and an IIP3 for the LNA and source-follower buffers is derived.
Table 5-1 summarizes the system specifications. These specifications are listed for
a 15-GHz global clock frequency, where the following are dependent on this frequency:
antenna-to-antenna gain, receiver gain, receiver sensitivity, noise figure, and IIP3. If a


85
comparator. It is interesting to note that any ring oscillator whose propagation delay is set
by the charging of RC circuits can be considered to be a relaxation oscillator [Ega99];
thus, the two models used to describe the dividers oscillation are equivalent.
4.3.2 Description of Injection Locking
Since the 2:1 divider can self-oscillate, it is possible to injection-lock the divider to
an input signal whose frequency is close to either the fundamental, subharmonic, or super
harmonic of the oscillation frequency. Thus, for a 2:1 injection-locked frequency divider
(ELFD), the input signal is close to the second harmonic of the oscillation frequency,
herein referred to as fiSO- The locking mechanism of the oscillator, for an input signal at
twice the natural frequency (a superharmonic), is created by nonlinear properties in the
oscillator which generate intermodulation products close to the natural frequency [Rat99].
Referring to Figure 4-l(c), the drain node of Mj oscillates at twice the natural frequency;
hence, this node is naturally suited to injecting a signal at this double frequency. An alter
native and equivalent understanding of the circuit is that injection locking is being used to
synchronize this double-frequency oscillator (drain of Mj) at a frequency near fiSo This
understanding is more straightforward and avoids the complications surrounding super
harmonic injection locking.
The range of input frequencies over which the ILFD can be locked or pulled is a
function of the input voltage swing and the frequency-dependent energy-loss in the oscil
lator. Appendix E shows that the voltage swing of the locking signal (VL) is related to the
locking frequency range (Aco0) and the oscillator output voltage swing (V0), as follows:
^
v o
(4.4)


203
place of four original probes. Therefore, a 23-pin probe was required (obtained from GGB
Industries), having a ground-signal-signal-ground ac probe for each receiver, as well as 15
dc pins to bias each receiver and measure the phase detector. Bypass capacitors were
placed on each dc supply pin to filter low frequency noise. Due to the pitch and number of
probes, these surface-mount capacitors were placed by GGB during probe manufacture.
6.7.2 Demonstration of Double-Receiver Interconnect
First, the double-receiver interconnect was measured using Z6b as the transmitting
antenna. This antenna is located 5.95 mm from each clock receiver. For this measurement,
the UMC testchip was mounted on a glass slide and placed on a 1-cm thick wood slab
(refer to Figure 6-4), which rested on the metal chuck of the probe station. Figure 6-28
shows the measured results for the double-receiver wireless interconnect. The input fre
quency is 15 GHz and the output frequencies for both clock receivers are 1.875 GHz. The
input power available to the transmitting antenna is set to 20.7 dBm.
As can be seen, this result demonstrates two receivers operating simultaneously.
From the figure, the two clock signals appear to have very low skew. However, this can be
misleading, because these waveforms were obtained through the external cables in the
measurement system. Thus, there is skew induced by mismatches between each cable and
its connectors. The extracted on-chip skew for this case obtained with the phase detector
will be presented in the next section. To obtain the power delivered to each receiver, the
antenna transducer gain was measured for each antenna pair with a wood dielectric layer
beneath the chip. The measured antenna transducer gains at 15 GHz are -60 dB for both
antenna pairs Z6b_6a and Z6b_6c. From the antenna impedance, the input impedance of the


86
Here, VQ is the voltage swing of the double-frequency oscillator (drain of Mj), and A is
equal to the derivative of phase-shift around the oscillator with respect to frequency. For a
tuned oscillator, A is equal to 2Q/(0o. Appendix F shows that the equivalent quality factor
of a ring-oscillator is approximately equal to 7t/2. Thus, an injection-locked ring oscillator
has the following input-swing to locking-range relationship (also known as the input sen
sitivity of the divider):
(4.5)
A voltage conversion gain can be defined for the frequency divider, equal to the
ratio of the output swing to the input swing. Since the gain occurs at two different frequen
cies, it is referred to as conversion gain. This gain can be related to (4.5) by relating the
output swing of the divider to the swing of the double-frequency oscillator. Thus, the cir
cuit has infinite gain at fiSO Also, (4.5) shows that the further away the input frequency
is from fiSO the larger the input voltage-swing has to be.
To control the oscillation frequency and conversion gain of the divider, the transis
tor sizes, input common-mode voltage, and supply voltage can all be adjusted. Referring
to Figure 4-1(c), the speed of switching node Q¡ bi HIGH depends on the size and IVgsl of
M7 8 (note Vdd dependence), while the speed of switching node Q¡ bi LOW depends on the
sizes of M3 4 and Mi 2 as well as the common-mode voltage on the clock inputs. For both
cases, the switching speed is decreased by the presence of regenerative pair M56, since the
positive feedback has to be overcome by the charging or discharging mechanism. There
fore, all of the transistor sizes, the supply voltage, and the clock common-mode voltage
determine the self-oscillation frequency for the 2:1 ELFD. Note that the transistor sizes are


121
Figure 4-30 Die photograph of 4:1 frequency divider.
The actual maximum operating frequency for the SOI divider is difficult to simu
late however, due to the hysteresis of the body voltages. Since the frequency divider will
operate continuously, it is important to simulate a steady-state condition for all of the body
nodes. The physical mechanism involved in charging Vbs at low drain bias is generation
current from the drain-body junction [Suh94, Suh95]. This current has to charge what can
be modeled as an RC network, with the R defined by recombination current at the
source-body junction and the C defined by the capacitance associated with the body, defin
ing a RC time-constant. The combination of a short time-constant at the input switching
frequency with a long RC time-constant for the body results in very long computation
times to reach a steady state Vbs. Therefore, the maximum operating speed of an SOI
divider is difficult to simulate due to the amount of time that it takes to reach a steady-state
condition for Vbs and the number of simulations it takes to scan for the maximum operat
ing frequency.
A final effect of SOI on the [DP]2 divider performance is an increased lower fre
quency limit due to the presence of a parasitic bipolar device in the transistor. Figure


132
input sensitivity is evident. To maximize the receiver gain, the LNA2 peak gain should
occur at a frequency close to the input self-resonant frequency of the divider. This mini
mizes the minimum detectable signal of the receiver by allowing for a small-level signal to
injection-lock the divider. If needed for testing purposes, the supply voltage of the divider
can be adjusted to tune the divider self-resonant frequency. As can be seen, if a divider
without conversion gain is used, then either the LNA gain would have to be increased by
20-30 dB, or the transmitter would have to broadcast 20-30 dB more power. Increasing the
LNA gain to ~40 dB could lead to circuit instability, while increasing the PA power to >30
dBm is unreasonable. Therefore, dividers with conversion gain are preferable for wireless
clock receivers.
5.2.4 Matching Between Antennas and Circuits
To maximize the power transfer from the transmitting antenna to the receiving
antenna, the antenna impedance and the impedance of its surrounding circuitry should be
conjugately matched. Straying from this conjugate match results in mismatch loss, defined
by the following equation [Gon97]:
Pdel = PAVS' Mant'
where Pdel is the power delivered from the antenna to the circuit, PAVS is the available
power from the antenna, and Mant is the mismatch factor. The mismatch factor is a power
transmission coefficient (transmittance) between two mediums. The antenna mismatch
factor is given by [Gon97]
2. For the rest of this chapter, when referring to the LNA, both the LNA and source-follow
ers are assumed.


83
locking range, its dependence on signal level, the steady-state phase error between input
and output signals, and the phase noise of ILOs.
4.3.1 Oscillation of SCL Divider
To understand the phenomenon of injection locking, the oscillation of the 2:1
divider with no input present must first be understood. When the input clock signal swings
are very small, then CLK and CLKb are approximately equal at their common-mode
value. Thus, both the master and slave latches are semi-transparent, allowing signals to
propagate through both latches. Referring to Figure 4-1(b), if the delay from the gate to
the drain of M3 is equal to xpd, then the total delay around the loop is equal to 4xpd. Thus,
the divider oscillates at a frequency equal to and the signal at the drain of M3 will
lag the signal at the gate of M3 by 90. The divider circuit can thus be likened to a
two-stage ring oscillator with an inversion in the loop1.
Additional understanding of the oscillator can be gained by solving for the
small-signal loop gain of the divider. First, assume the two latches are identical and that
each latch is symmetric (i.e., W3=W4, W5=W6, and W7=W8). Consider the case when
CLK and CLKj, are held at a constant and equal value. For this case, M] and M2 become
current sources. Cross-coupled transistors M5 6 can be represented as negative resistances
with values of -l/gcc, where gcc is the transconductance of M5. The PMOS load can be
modeled as a conductance, gp Finally, the total capacitance at the output nodes of the latch
can be represented by CL. Figure 4-1(c) can then be simplified to Figure 4-2, by
1. An actual expression for the propagation delay through each latch was not derived, since
this delay can readily and more accurately be obtained with SPICE.


36
(2.46)
Thus, the input should be designed such that the series resonance occurs at a frequency
higher than the operating frequency. Note that for Sn = 0, QL = Qgs, as expected.
Substituting (2.46) into (2.44), taking the partial derivative with respect to Qgs, and setting
the result to zero yields the optimum Qgs for minimum F-Fmin as follows:
(2.47)
For the assumptions listed in Figure 2-7 with Su = -10 dB, and setting t=l, then
Qgps equals 2.62 and QL is 1.95. For = -15 dB, Qpt is 2.32 and QL is 1.96, while for
Sji = -o dB, corresponding to a perfect input match, QgPs Ql = 2.18. At these optimal
Qgs values, F-Fmin is 0.02, 0.07, and 0.17 for Sn = -10, -15, and - dB, respectively3,
when considering GIN only (i.e., (2.44)). When considering all of the second-order effects
(i.e., as in Figure 2-7), F-Fmin is 0.24, 0.38, and 0.65, respectively4. Therefore, matching
the input for perfect series resonance results in a higher NF, as is the case for [Sha97].
2.6.4 Optimum Width of Mo
The optimum sizing of M2 can be obtained by minimizing £(W2). Since gm2 and
gdo2 are directly proportional to W2, while cdl has components directly proportional to
Wj and to W2, then t,(W2) has components proportional to and inversely proportional to
3. The impact in decibels depends on the particular value of Fmin. For Fm¡n = 1 dB, the cor
responding noise figures are 1.1,1.2, and 1.6 dB, respectively.
4. For Fjjjj,, = 1 dB, the corresponding noise figures are 1.8, 2.2, and 3.8 dB, respectively.


190
Table 6-3 shows a comparison between the expected and measured VCO results.
The overall Q of the LC tank is the major limiting factor of the phase noise performance,
where the tank consists of Lj, L2, Cvl, Cv2, the gate and drain capacitances of Mj and M2,
and the gate capacitances of M3 and M4. Since the measured Q of the inductor is ~5 times
smaller than that of the varactor, the Q of the overall LC tank is limited by the inductor.
The phase noise of a VCO in 1/f2 region (refer to Figure 6-20) can be approximated by
Leesons formula [Lee66], as follows:
L(Aco) = 10 log
2 FkT(Qlrs)
[Ks&J]
(6.6)
where F is the noise factor of the amplifier, k is Boltzmanns constant, T is the tempera
ture, V is the voltage across the tank, QL is the Q of L12, rs is the series resistance of the
inductor, and Aco is the frequency offset from the carrier. Assuming that the last term in the
equation is much greater than 1, and all other parameters except rs in the equation are con
stant, a 50% Q degradation increases the phase noise by ~3 dB (10 log(2)), which is


173
Figure 6-6 Antenna impedance versus frequency for linear dipole antenna.
matched differentially to 100 £2, the mismatch for the antennas in the measurement system
is virtually the same as when the antenna is integrated with the LNA. The measured
antenna impedance, shown in Figure 6-6, at 7.4 GHz is approximately 20-j300 £1 Com
paring this impedance to 100 Q, the total mismatch loss for the antenna pair can be calcu
lated to be 15 dB, using equations (5.3) and (5.4).
The reason for such a low antenna impedance is still not known for certain. Previ
ous antenna implementations indicate that this same antenna should have an impedance
closer to 100 ft [KimOOa], A potential reason for this low impedance is the effect of inter
ference structures, including metal dummy patterns for chemical-mechanical polishing, on
antenna impedance. What is known for certain, is that a 15-dB mismatch loss is unaccept
able. Clearly, better techniques are required to predict the antenna impedance, so that
proper matching circuitry can be designed. Alternatively, adjustable matching networks
can be designed to allow for the receiver input match to be tuned to account for antenna
impedance variations.


7
from the VCO is then amplified using a power amplifier (PA), and fed to the transmitting
antenna. The VCO is phase-locked to an external reference using a phase-locked loop
(PLL), providing frequency stability. The PLL consists of a phase-frequency detector
(PFD), a loop filter, the VCO, and a frequency divider.
Figure l-3(b) shows a block diagram for an integrated clock receiver. The global
clock signal is detected with a receiving antenna, amplified using a low noise amplifier
(LNA), and divided down to the local clock frequency. These local clock signals are then
buffered and distributed to adjacent circuits. The amplifier is tuned to the clock transmis
sion frequency to reduce interference and noise. Since the microprocessor is extremely
noisy at the local clock frequency and its harmonics, transmitting the global clock at a fre
quency higher than the local clock frequency provides a level of noise immunity for the
system [Meh98]. Also, operating at a higher frequency decreases the required antenna
size. The receiver is implemented in a fully differential architecture, which rejects com
mon-mode noise (e.g., substrate noise) [Meh98, BravOOa], obviates the need for a bal-
anced-to-unbalanced conversion between the antenna and the LNA, and provides
dual-phase clock signals to the frequency divider.
1.2.4 Potential Benefits
The wireless clock distribution system would address the interconnect needs of the
semiconductor industry in providing high-frequency clock signals with short propagation
delays. These needs would be met while providing multiple benefits. First, signal propaga
tion occurs at the speed of light, shortening the global interconnect delay without requir
ing integrated optical components. Second, the global interconnect wires used in
conventional clock distribution systems are eliminated, freeing up these metal layers for


68
oxide thickness. Two metal layers (0.7 |lm each) are supported and the substrate resistivity
is 1-2 Q-cm. Partially-depleted, floating-body SOI transistors are used for all circuitry.
Figure 3-18 shows a schematic of the fully-integrated tuned amplifier including,
important parasitics. A three-stage topology is used to boost the circuit gain and reverse
isolation. As can be seen, this topology is very different than the single-stage
source-degenerated LNAs previously presented, providing a look at the high-frequency
performance of other circuit topologies. For ease of input matching, a common-gate input
stage with a shunt inductor is used. An overview of the common-gate topology is con
tained in section 2.2 and Appendix A. At resonance, the input impedance to the amplifier
is l/gml. Therefore, gml is designed to be 0.02 Q'1 to provide a 50-Q match. Following the
input stage is a source-follower buffer which then drives a common-source cascode stage
with a tuned load. The output of this stage is matched to 50 2 using a capacitive trans
former. This stage is a non-degenerated version of the LNAs discussed earlier. Figure


184
Figure 6-15 (a) Impedance smith chart showing reflection coefficients of zigzag and loop
antennas (Impedances are normalized to 100 Q). (b) Antenna impedance for
zigzag and loop antennas (Z6b and LPj).
-40
-50
-60
-70
-40
1111>111I11
metal 6, d=3.2mm

tn
o
p+/nwell, d=3.6mm
-
o
CO

ap
-
d=3.6mm
-70
-80
i i i i i i i i i i i
6 8 10 12 14 16 18 6 8 10 12 14 16 18
Frequency (GHz) Frequency (GHz)
(a) (b)
Figure 6-16 (a) Zigzag transmission gains at d=3.6 mm for antennas in metals 1, 2, or 3. (b)
Zigzag transmission gains at d=3.65 mm (Zpp, Znp) for antennas implemented in
direct contact with substrate, with n-plus in p-sub and p-plus in n-well. Metal-6
antenna at d=3.2 mm is shown for comparison purposes.


15
In superheterodyne and direct-conversion receivers, the LNA is preceded by the
antenna, a duplexer filter or transmit/receive switch, and an optional pre-select filter. Since
these components are not typically integrated, the input to the LNA is driven through a
50-Q transmission line. Therefore, the LNAs input should be matched to 50 Q For the
wireless clock receiver application, the input of the LNA should be conjugately matched
to the antenna impedance, since transmission lines are not required. The output matching
of the LNA depends on whether the LNA drives an off-chip component, such as an
image-reject filter for superheterodyne architectures, or an on-chip component, such as a
mixer for direct-conversion architectures or a frequency divider for the clock receiver
application. When driving an off-chip component, the LNAs output should be matched to
50 £L The input and output matching criteria for a 50-H match are specified in terms Sjj
and S22, where both should be less than -10 dB.
2.2 Possible LNA Topologies
Designing an LNA consists of meeting the gain, noise, matching, and linearity per
formance metrics while minimizing power consumption and cost (where all of the afore
mentioned tend to trade-off to a certain extent with one another). Towards this end, there
are two main circuit topologies for CMOS that will be discussedcommon-gate and com
mon-source with inductive degeneration.
2.2.1 Common-Gate CMOS LNA
The first possible topology employs a common-gate amplifier with source induc
tance, shown in Figure 2-1(a). Appendix A contains derivations for the input impedance,
gain, and noise figure for this common-gate topology. Here, the input is a parallel resonant


135
-j1
Figure 5-2 Constant Teq circles in the rc plane. This shows the allowable circuit impedances
for a given antenna reflection coefficient (Tant for zant=(89-j45)/100).
5.3 Definition of Clock Skew and Jitter
A critical requirement for clock distribution systems is ensuring that all of the dis
tributed clock signals are synchronized throughout the chip. Since these clock signals are
used to maintain system timing, any uncertainty in the edge of the clock signal reduces the
amount of time available to perform useful operations. As a result, the minimum clock
cycle time has to be increased by the uncertainty, slowing the system down. Thus, inter
connect systems which can deliver high frequency clock signals with minimal timing
uncertainty (typically < 5-10% of a clock cycle) are required.
The origin of clock timing uncertainty is mismatch or differences in the propaga
tion delays of the signals. The difference has both time-invariant (static) and time-variant
(dynamic) components. Clock skew is defined as the static difference between clock edges
throughout the chip, while clock jitter is defined as the dynamic difference between the


72
avoided. Thus, the conductance of the source-follower is adjusted such that the gain is
improved while stability is maintained.
3.6.4 Measured Results
The measured S-parameters of the SOI amplifier are shown in Figure 3-20(a). At
23.8 GHz the amplifier is perfectly matched to 50 Q. at the input, with an of -45 dB,
and well-matched at the output, with an S22 of -9.4 dB. The transducer gain (S2i) is 7.3 dB
while the reverse isolation is 27 dB, which are both excellent for a CMOS circuit at this
frequency range. Also, the gain is greater than 0 dB beyond 26 GHz. The total supply cur
rent is 53 mA for a 1.5-V supply, which is high due to the use of multiple amplifier stages.
The measured quality factor (Q) of the on-chip spiral inductors is ~2 at 24 GHz.
This value is lower than the expected value of 4.8, which resulted in the gain being ~15 dB
(201og{4.8/2}2) lower than its simulated value of 22 dB. The reason for this Q degrada
tion, however, is not fully understood.
(a) Frequency (GHz) (b) Frequency (GHz)
Figure 3-20 (a) Measured S-parameters and (b) measured operating power gain (Gp) and
50-0 noise figure (NF) vs. frequency.


35
a = . (2.43)
&do
From (2.31),(2.33), and (2.35), it can be seen that Gn is inversely proportional to Qgs,
while Ropt and Xopt are directly proportional to Qgs. This agrees very well with the noise
parameters plotted in Figure 2-7.
Source degeneration modifies the noise parameters [Har73], For the specific case
corresponding to neglecting Cgd and transistor output impedance (the model used to calcu
late the original noise parameters), all but Xopt remain unchanged (a benefit of using
impedance noise parameters). These modified noise parameters are also shown in Table
2-2. Through setting £ = 1 and = 0, the noise parameters for a single transistor, exclud
ing GIN and Rsub, can be obtained. Examining (2.28),(2.31), and (2.37), it can be seen that
as with the gain analysis, o>p should be maximized to reduce noise figure.
These noise parameters for the source-degenerated cascode are substituted into
(2.28), along with Zs = Rs + jco0Lg, yielding the following:
F-F
min
,(2.44)
where
Ql =
o(Lg + LS)
Rc
(2.45)
By inspection of (2.44), setting QL = Qgs (corresponding to series resonance at the
input) does not result in minimum noise; thus, there is a trade-off between the noise
matching and the power matching conditions. For a given value of Sj} and assuming o>x-Ls
= R$, the following required value of Ql for minimum noise can be obtained through the
aid of (2.13):


90
Figure 4-5
CLK Q
CLK
Q
Inputs
(2:1)
(2:1)
-0
CLKb Qb
>0
CLKb
Qb
CLK
Q
(2:1)
CLKb
Qb
SCL Divide-by-8
c\k O
rlk O
rlk O
Hk n
D Qb
ir
D Qb
ir
D Qb
~ir
D Qb
H>
Output
TSPC Divide-by-16
Block diagram of 128:1 dividen
Figure 4-6 Schematic of true single-phase clocked (TSPC) latch.
4-5. The first 8:1 divider is implemented with SCL, while the lower frequency 16:1 asyn
chronous divider is implemented using true-single-phase-clocked (TSPC) logic [Yua89].
Figure 4-6 contains a schematic of the TSPC latch. Simulations indicate that this latch can
operate up to ~ 2 GHz for the chosen device sizes. All of the devices for the SCL and
TSPC dividers shown in Figure 4-1 and Figure 4-6 are implemented with a channel length
of 0.25 p.m, whereas the channel widths are given in Table 4-1.


78
up to 10 and 15.8 GHz, respectively. The [DP]2 divider is implemented in a par
tially-scaled, 0.1-|im CMOS technology with bulk and SOI substrates. A design method
ology for SCL dividers based on injection locking which allows for maximizing the
operating frequency and minimizing the required voltage swing of the dividers input sig
nal is discussed. Also, using SCL, a new programmable divider is developed which allows
the start-up state of the divider to be controlled. Such programming will decrease the sys
tematic clock skew, provided that the dividers are properly initialized and synchronously
released from the start-up state. The initialization and start-up methodology for the divid
ers are also presented in this chapter.
4.2 Frequency Divider Using Source-Coupled Logic
4,2.1 Circuit Description
Figure 4-1(a) shows a block diagram of a divide-by-eight (8:1) frequency divider
employing SCL. The divider consists of three cascaded 2:1 dividers with dual-phase
inputs and outputs. Each 2:1 divider, shown in Figures 4-1 (b) and (c), consists of two SCL
D-latches in a master-slave configuration [War89, FloOOa, HunOl]. The outputs are tied to
the inputs with inverted phase to perform a toggle operation.
Source-coupled logic (also known as MOS current-mode logic (MCML)) has
some distinct advantages for the wireless interconnect application. First, a small-level
input voltage can be detected on the clock input. This is due to the differential structure of
the latch as well as the mechanism of division, which will be discussed shortly. Therefore,
the input sensitivity for the divider is high, improving the MDS of the receiver. Second,
the latch can be designed such that the outputs do not swing rail-to-rail, decreasing the


275
[ShiOl] A. Shirvani, D. K. Su, B. A. Wooley, A CMOS RF power amplifier with
parallel amplification for efficient power control, ISSCC Digest Technical
Papers, pp. 156-157, Feb. 2001.
[SIA99] Semiconductor Industry Association, The International Technology Road
map for Semiconductors, San Jose, CA: SIA, 1999.
[Smi98] J. R. Smith, Modem Communication Circuits, 2nd edition, Boston, MA:
McGraw-Hill, 1998, pp. 83-84.
[Stu98] F. Stubbe, S. V. Kishore, C. Hull, and V. D. Torre, A CMOS RF front-end
for 1 GHz applications, Symp. on VLSI Circuits Dig. Tech. Papers, pp. 80-
83, June 1998.
[Suh94] D. Suh and J. G. Fossum, Dynamic floating-body instabilities in partially
depleted SOI CMOS circuits, IEDM Technical Digest, Dec. 1994.
[Suh95] D. Suh and J. G. Fossum, A physical charge-based model for non-fully
depleted SOI MOSFETs and its use in assessing floating-body effects in
SOI CMOS circuits, IEEE Trans. Electron Devices, vol. 42, no. 4, pp.
728-737, April 1995.
[Sut95] H. Sutoh, K. Yamakoshi, and M. Ino, Circuit technique for skew-free
clock distribution, Proc. IEEE Custom Integrated Circuits Conference,
pp. 163-166, May 1995.
[TamOO] S. Tam, S. Rusu, U. N. Desai, R. Kim, J. Zhang, and I. Young, Clock gen
eration and distribution for the first IA-64 microprocessor, IEEE J. Solid-
State Circuits, vol. 35, no. 11, pp. 1545-1552, Nov. 2000
[Tau98] Y. Taur and T. H. Ning, Fundamentals of Modem VLSI Devices, 1st Edi
tion, pp. 250-256, New York: Cambridge University Press, 1998.
[Tsi99] Y. Tsividis, Operation and Modeling of the MOS Transistor, 2nd edition,
Boston, MA: McGraw Hill, 1999, pp. 492-495.
[Uzu85] V. Uzunoglu and M. H. White, The synchronous oscillator: a synchroniza
tion and tracking network, IEEE J. Solid-State Circuits, vol. sc-20, no. 6,
pp. 1214-1226, Dec. 1985.
[Wan86] R. K. Wangsness, Electromagnetic Fields, 2nd edition, New York: John
Wiley & Sons, 1986, chapter 28.
[WanOO] H. M. Wang, A 1.8 V, 3 mW, 16.8 GHz frequency divider in 0.25-pm
CMOS, ISSCC Digest Technical Papers, pp. 196-197, Feb. 2000.


191
significant. Thus, doubling the thicknesses of metals 5 and 6 should improve the phase
noise to -108 and -116 dBc/Hz at 1-MHz and 3-MHz offsets, respectively. The power con
sumption should also decrease, since the voltage across the tank would remain approxi
mately constant while the effective resistance of the tank (Q2rs) would increase. This
explains why, to overcome the lower Q, the current level was increased, explaining the
added power consumption listed in Table 6-3.
Finally, to understand the competitiveness of this VCO result with previously pub
lished results, the phase noise can be scaled to a 5-GHz regime. Thus, for a given inductor
Q, the 15-GHz VCO would correspond to a 5-GHz VCO achieving a phase noise of -114.5
dBc/Hz at a 1-MHz offset. This result is ~2.5 dB less than that achieved in [HunOOc] at
5.35-GHz. However, if thicker metal is used for the inductors, the scaled result would then
be comparable to [HunOOc], with a benefit of reduced power consumption.
6.6.3 Power Amplifier
The PA consists of 2 stages of inductively-loaded common-source amplifiers.
Unlike the transistors in the VCO core, whose output noise is directly injected into the LC
tanks, the transistor noise in the PA does not significantly degrade the VCO phase-noise
performance. Also, because the transistors are turned on only half of the time, the effective
output noise becomes even lower. Hence, NMOS transistors with larger current-per-
unit-channel-width are used.
The first class-A stage serves as a pre-amplifier for the final power amplifier stage.
The transistor widths of stage 1 are only 2/3 of that of Mj in the VCO core, to avoid signif
icantly loading the VCO output. At the tuned frequency, the single-sided output has an
amplitude approximately equal to the supply voltage, with an offset voltage of Vdd PA. The


52
LNA implemented in a 6-level-metal 0.35-|im BiCMOS process [GraOO], which is cur
rently the lowest NF for any CMOS LNA. However, considering that [GraOO] employs
technology two generations beyond 0.8-pm, with an coy approximately five times as large,
and that the back-end metal process and substrate in [GraOO] allow for inductor quality
factors approximately 3 times as large as the 0.8-|im process, the result achieved in this
work is excellent. Also, the performance goals of NF < 2 dB and power <10 mW have
been achieved, demonstrating the competitiveness of CMOS for wireless applications.
Finally, the LNA exhibits excellent linearity and is suitable for wireless applications.
3.4 An 8-GHz. 0.25-um CMOS LNA
3.4,1 Circuit Implementation
Having demonstrated the design methodology for source-degenerated CMOS
LNAs at 900 MHz, both a single-ended and differential version of the source-degenerated
LNA was implemented in a standard 0.25-|im CMOS technology, operating at ~8 GHz.
This LNA was also implemented with an entire clock receiver, as will be presented in
Chapter 6. The technology provides 5 metal layers, and both p' (~8 G-cm) and p+ with epi
taxial layer (-0.01 G-cm) substrates, allowing the effects of substrate resistivity on LNA
performance to be studied. This technology was obtained once again through MOSIS.
Figure 3-3(b) shows a die photograph of the 8-GHz differential LNA, while Figure
3-8 shows a schematic of the differential LNA, including important parasitics (pad capaci
tances are implied for each labelled input). The die size is 730x510 fim2. All three
inductors, Ld, Lg, and Ls are integrated, as well as 20-pF bypass capacitors. Symmetry was
maintained top-to-bottom for the differential layout to avoid any systematic errors from
layout mismatch. Again, ground-shielded pads are used throughout the layout.


Ill
1.6
i2 1.2:
0.8
0.4
0
Outputs
iLj
Tn
U
40
vLl
rt 1
(V
1 n 1 4n '
INI
5n
1-2:
| o.s^
0.4:
0 1 "in 1 n 1 n 1 4n 1 5n
2:1 Outputs
2n 3n
4:1 Outputs
n 1 3n 1 4n
8:1 Outputs
2n Time 3n
Simulated waveforms of two programmable 8:1 dividersprogrammed to #000
(solid lines) and #111 (dashed lines)~for repeated initializations. On the third
initialization the circuit fails, as indicated.
during the third initialization, the 8:1 output of the #111-divider stays low, changing the
programmed state from #111 to #110. This accounts for the extra 45 phase shift.
Depending on the state that each 2:1 divider is in when INI transitions high, the
circuit may not initialize to the correct state. Figure 4-23(a) shows plots of the drain volt
ages for the third 2:1 divider (i.e., 8:1 outputs) around the failure time of the third


88
Divider stops locking
Output!
2 3
Time (ns)
(b)
Figure 4-4 (a) Simulated input sensitivity of 0.25-|lm CMOS 8:1 divider (b) Exponentially
decreasing input signal at 8 GHz and output signal of 8:1 divider, illustrating
dividers dependence on input voltage swing.
1.5 GHz, respectively. Thus, fiSO is equal to 12.2 GHz. This simulation can be iterated
through various device sizes to set fiS0 to the desired operating frequency. Additionally,
this simulation can be run for the second and third 2:1 dividers to set their respective
self-oscillation frequencies to one-fourth and one-eighth of the desired operating fre
quency. Note that for an 8:1 asynchronous divider, any of the three 2:1 dividers can oscil
late provided that the input signal level to each divider is small. However, for the typical
case where the first 2:1 divider oscillates with a large output signal, the following 2:1
dividers operate in a latched mode.
The input-swing versus locking-frequency-range relationship is shown in Figure
4-4(a), which plots the simulated input peak-to-peak voltage swing versus input fre
quency. To obtain these input swings using a single simulation, a damped sinusoid was
injected into the divider, as shown in Figure 4-4(b). Eventually, the divider cannot lock to
the input signal, at which point the input voltage swing is determined. This simulation is


251
or alternatively,
(C.16)
v *i
n n
p *
n
~R~
Finally, to obtain Fm¡n, the following relationships are used:
(C.17)
Ftn = 1 +2GJR +RC) = i+2GRc + 2jGRu + G2X (C.18)
= 1 + 2 Re(Pn) + 2 JgJ^, -1Z/O,) + fte2(P)
= 1 + 2 Re(Pn) + 2 |f|2 + Re\pn)
= 1 + 2Re(P n) + 2^Gn/in /m2(Pn)
Thus, Rn, Gn, and Pn can be used to obtain any of the other noise parameters. Through the
authors experience, these alternative parameters are more convenient for derivations and
numerical techniques, particularly for derivations of Fmin.


217
Figure 7-3 Illustration of single quadrant of global clock distribution systems for (a) grid, (b)
H-tree, and (c) wireless.
Wireless System. The wireless system, shown in Figure 7-3(c), consists of a clock
transmitter broadcasting a microwave signal to a grid of distributed receivers. Each
receiver corresponds to a spine location and has a sector buffer at the output. Due to their
low capacitance, balanced H-trees are used to distribute the signal from the receivers to the
local clock generators. Thus, Cw for both the H-tree and wireless scheme is equal.
In a wireless clock distribution system, the long interconnects for delivering the
clock from its source to spine/sector locations are not present, and the associated compo
nent of CG is zero. However, since the wireless system contains components with static
power dissipation, an equivalent global capacitance [Flo99b] representing this power dis
sipation is needed. The equivalent capacitance is defined as follows:
PTX + N PRX
'Geq V2f
(7.3)
where PTX and Prx are the power consumptions of the transmitter and receiver, and N is
the total number of receivers. Based on measured results for the 0.18-pm 15-GHz wireless
interconnect, and assuming a 1.2-V supply, the transmitter consumes 38.4 mW while each


101
counters, where the division ratio is equal to the count (e.g., an 8:1 divider counts from 0
to 3 while outputting a one and from 4 to 7 while outputting a zero). If receivers close to
the transmitter are initialized to a smaller count, and receivers far from the transmitter are
initialized to a larger count, the mismatch in time-of-flight delays among the receivers will
tend to be cancelled. Graphically, this reduces the grid spacing by the division ratio.
Hence, for an 8:1 divider, the receivers can be placed on a grid with spacing A/8, which
then equalizes the time-of-flight delays and decreases the systematic skew to 0. This is
illustrated in Figure 4-15(b), which shows the grid for wavelength in silicon-dioxide.
Since random clock skew also exists, it is impossible to design a grid to equalize
the total skew. However, assuming that the total skew in the system can be estimated or
detected, then the start-up state of the divider can be modified accordingly. Therefore, the
worst-case total skew percentage then becomes ^ where M is the division ratio of the
divider. For an 8:1 programmable divider, the maximum output skew is +/-6.25%. Note,
that if the skew can be estimated, then the receivers do not have to be placed on a grid.
The initialization methodology is conceptually illustrated in Figure 4-16, which
shows two clock receivers with an initial skew of 20%. Also shown are the 8 possible
start-up states/phases for the 8:1 dividers in the clock receiver. As can be seen, by select
ing the correct states for each divider (requiring initial skew estimation or detection), the
skew can be reduced from 20% to 3% (for this particular example).
Therefore, circuits are required to initialize the SCL divider, and a methodology is
needed to start-up the clock distribution system in phase. This is complicated by the fact
that the SCL dividers will self-oscillate in the absence of an input signal. Therefore, a way
is needed to prevent these circuits from self-oscillating during initialization as well.


112
1.5
1.3
1.1
0.9
in
|0.7
0.5
0.3
0.1
-0.1
.
2.4n
(a)
Figure 4-23
i i i i
H-C
va = HIG
^7a
Q = HIGH
D 1[^ ^3a M
pull-
up
IIM8a
J- Vb = LOW
Q
CLK-I
iS
pull
down
lbD
'4a
tM1aM9||-INI ^
^ON^
CLKb
^2a
2.8n 3.2n 3.6n 4n 4.4n
Time (b)
(a) Simulated waveforms of programmable 2:1 divider during state-depen
initialization failure, (b) Schematic of DFF indicating conflict of states.
initialization. These waveforms, Q, Q, D, and D are identified in the schematic shown in
Figure 4-23(b). The desired states of Q and Q during initialization are high and low,
respectively; thus, Vb and Va are 0 and 1, respectively. For Q to be high during initializa
tion, it has to be pulled up through M8a with Vb low. However, depending on the values of
Q and D, there will also be a discharge path to ground for Q. The pull-down paths are pos
sible since Mia 2a,9a 31-6 all on- As seen in Figure 4-23(a), Q is ~ 1.1 V; thus, the pull-down
path through M6a is quite strong. The actual value of Q1 is set through a voltage divider,
with the value depending on the current state of the divider when INI switches high.
To rectify this problem, both the evaluate and hold portions of the DFF should be
turned off when INI = 1. Also, both high and low values have to be written to the drain
nodes of the divider (as opposed to the PMOS only being able to pull the node high). Fig
ure 4-24(a) shows the new programmable DFF to be used in the 2:1 divider with these


149
Figure 5-7 Local clock jitter versus signal-to-noise ratio at input to divider.
A survey of jitter results obtained in the literature for high-performance micropro
cessors [TamOO, Boe99, Kae98, Mai97, You97] reveals that the peak-to-peak
cycle-to-cycle jitter should be less than 3% of the clock period, or the RMS cycle-to-cycle
jitter should be less than 0.5%. Therefore, the system specification for the clock receiver is
that the SNR at the input of the divider should be at least 12 dB to result in less than 0.5%
RMS cycle-to-cycle jitter from input additive noise and interference.
Finally, as was mentioned earlier, the total phase noise at the output of the divider
will have components due to input additive noise and output additive noise. Equation
(5.35) calculates the jitter from only input additive noise. Since the output additive noise is
independent of input SNR, for high enough SNRs the jitter will no longer decrease.
Instead it will level off at some SNRknee, where the jitters from input and output additive
noises are equal. As the division ratio increases, SNRj^gg will decrease. Therefore, the
output additive noise sets a floor on the obtainable jitter.


64
by a few fF). For NMOS, the gate capacitance is ~1.3 times larger, while for PMOS it is
1.2 times larger. For both NMOS and PMOS, the total drain capacitance is approximately
twice that which is simulated, with the PMOS difference being more severe. This can
result in as much as a 1.4 times (Jl) reduction in resonant frequency for a tuned circuit.
3.5.4 Measured Results
Figure 3-16(a) shows the measured gain (S21) and noise figure (NF) for the sin
gle-stage LNA matched to 100 Q at the input and output, while Figure 3-16(b) shows the
measured input and output reflection coefficients (Sn and S22). Driving 100 Q
differentially, the LNA provides 8 dB of gain at 13.3 GHz, with Sn and S22 equal to -8
and -15 dB, respectively. The LNA consumes 12 mW from a 1.5-V supply. The reverse
isolation is better than 27 dB. Finally, the 100-0 noise figure is 7.7 dB at resonance. The
results are summarized in Table 3-5.
Frequency (GHz) Frequency (GHz)
(a) (b)
Figure 3-16 (a) Measured gain and noise figure for 0.18-gm LNA.
(b) Measured input and output reflection coefficients for 0.18-pm LNA.


205
Figure 6-29 Measured input and output waveforms of double-receiver interconnect using Z6c
as the transmitting antenna, with a AINi as the back dielectric. Input frequency is
15 GHz; output frequencies are 1.875 GHz; distances are 6.8 mm for Rx4 and 5.6
mm for Rx2. Time-of-flight delay mismatch and amplitude mismatch are present.
For this second measurement, the UMC testchip was mounted on a glass slide and
placed on a 0.76-mm thick slab of aluminum nitride (AIN). This thickness was chosen
based on the results presented in [GuoOl]. Recall that AIN has a high thermal conductivity
(approximately the same as silicon), while still being a good dielectric (ka1n=8.5). Thus, it
is a suitable medium for the backside dielectric layer which will be connected to the heat
sink. Figure 6-29 shows the measured waveforms for the double-receiver interconnect
with Z6c as the transmitting antenna. Once again, the input frequency is 15 GHz and the
output frequencies are 1.875 GHz, demonstrating two receivers operating simultaneously.
The clock skew for this case will be presented in the following section, however, by
inspection, the skew appears to be similar to the result shown in Figure 6-28.


79
(b)
2:1 Divider
(Toggle F/F)
Qb
Q
Master
Slave
Figure 4-1 Block diagrams of (a) 8:1 divider consisting of three cascaded 2:1 dividers, and
(b) 2:1 divider consisting of two D-latches in master/slave, (c) Schematic of 2:1
divider implemented using source-coupled logic.


213
requires 16 or more clock receivers operating at 20 GHz or higher to be implemented
on-chip, this added power dissipation by the receivers is a concern. To evaluate the impact
these receivers have on the total power consumption of microprocessors, the power
requirements for a wireless clock distribution system are estimated for the 0.1-pm genera
tion, and then compared to those for conventional global clock distribution systems.
1.2.1 Power Consumption Comparison Methodology
The entire clock distribution system can be divided into a global system and a local
system. The global distribution system is assumed to be the network which delivers the
clock from its source (off-chip or on-chip) to various locations throughout the chip. The
local distribution system is assumed to be that which takes the global clock, generates dif
ferent clock phases as needed, and distributes these signals to all of the local circuits.
The power dissipation of the global distribution system can be modeled as
dynamic, according to the following formula:
PD=CTV2ddf, (7.1)
where CT is the total (equivalent) global capacitance being switched, including the capaci
tance of buffers, Vdd is the system supply voltage, and f is the clock frequency. To com
pare the power requirements between different types of global clock distribution systems,
the system voltage and frequencies are assumed to be equal. Also, an equal capacitive
load, representing the local clock generators or distribution system, is assumed for each
type of global distribution system. Under these assumptions, the power dissipation can be
converted to capacitances and these can be used to compare the power dissipation of dif
ferent global distribution systems, similar to an approach taken in [Res98].


145
(J) = tan"
-if Vnsin¥ ^ = tan-l rV"Sin^ = VnsinV
A + vcos\j/J
A )'
(5.23)
where approximation is due to a small-angle assumption. This shows that the amount of
phase-shift is approximately equal to the perpendicular component (in volts) divided by
the signal level (in volts). Due to the small-angle approximation, the phase noise can either
refer to <|) directly or to the vn sin(\)/) component.
The mean squared value of the phase shift is:
2 = £(< = e
/ 2 2 \
vn sin \|/
= VTif2 sin2(v|/)fi|/ = V (5-24)
A2 2kJq
2 A
where E is the expected-value operator, vn and \|/ are independent, and \\f has a uniform
probability density function. The factor of 2 in (5.24) means that half of the additive noise
power is in the phase, whereas the other half is in the amplitude. The available power from
2 oo
a resistor is vn = kT A/, where k is Boltzmanns constant (1.38 x 10 J/K) and T is
absolute temperature of the system. This PSD is a double-sideband density, meaning that
it is for positive frequencies only. Therefore, the conversion from an additive noise source
to a phase PSD, S^f), is
c (n Nadd
V-'A Ay 2 P
kT
sig
Af 2 P
(5.25)
Slg
where Psig is the power of the signal (= A^2) and is the additive noise spectral den
sity. The units of (5.25) are rad2/Hz.
Phase noise is frequently defined as a single-sideband density, £(f). Since the phase
PSD, S^(f), is a double-sideband density, then Lif) is simply one half of S^(f) [Ega99], or


74
~1.4 or 3 dB. However, IAvsfl is -0.7; therefore, the net gain in S2i due to the source-fol
lower is 1.4*0.7 s 1 (0 dB). Thus, for this amplifier, while the source-follower is providing
a needed dc level shift between stages 1 and 3, the gain is unaffected. A potential better
use for the source-follower would be to use it purely as a negative conductance generator,
without passing a signal through the stage. The power consumption could be reduced by
reusing the current in the gain stage with the source-follower.
Multistage amplifiers can exhibit poor linearity, and negative conductance can
potentially affect this linearity even further. The measured PldB is -10 dBm (-16.2 dBm
referred to the input), while the DP3 is -7.8 dBm. These linearity data are considerably
better than those obtained in [Ho98], an encouraging result, indicating that acceptable lin
earity can be achieved with amplifiers utilizing source-followers.
Figure 3-20(b) shows the measured noise figure (NF) and associated gain (Gas)
versus frequency. The gain is 8.1 dB at 23.8 GHz, while the noise figure is very high at 10
dB. The high noise figure is caused by the reduced first-stage gain, due to the low inductor
Q. Also, the noise associated with the input inductor, Ls, increases NF as well. The low
gain allows noise from stages two and three to refer to the input of the circuit. While such
a noise figure certainly is too high for typical wireless communication systems, it can be
decreased through circuit and inductor optimization. A final clarification on the amplifiers
performance is that although the NF is larger than the gain, it does not mean that the
amplifier is unusable (i.e., that the output signal-to-noise ratio (SNR) will always be < 1).
Noise figure is a relative quantity, representing degradation of SNR. Therefore, a high NF
degrades the dynamic range of a circuit through raising the minimum detectable signal.


235
decrease. This decrease will be lessened when the output buffer latency is included, which
will be increasing with each technology generation. However, this example shows that the
wireless interconnect latency is decreasing with technology generation, whereas the con
ventional interconnect latency is increasing.
1.6 Intangibles
While feasibility issues such as power consumption, synchronization, and latency
have been quantified, there are other feasibility issues which are harder to quantify, but
which contribute to the overall system feasibility. These intangible feasibility issues
include the area consumed by the system, and design verification.
Since the wireless clock distribution system employs antennas and active circuitry,
there is an area penalty incurred in using wireless interconnects. For the 0.18-pm circuits,
the areas of the transmitter and receiver, excluding antenna and pad area, are 0.116 mm
y *y
and 0.215 mm respectively. Each antenna occupies 0.15 mm For a wireless clock dis
tribution system consisting of one transmitter, 16 receivers, and 17 antennas, the total area
consumed is 6.2 mm2. The ITRS projects the die size of a 0.1-pm generation microproces-
sor to be 622 mm Thus, ~1% of the microprocessor area will be consumed by the wire
less clock distribution system. Approximately half of this area is due to the 2-mm long
antennas; thus, more area-efficient antennas would be desirable.
The area consumed by the wireless clock distribution system has to be weighed
against the metal area saved. Since the global clock tree will be eliminated, there will be
less area consumed by routing global signals. This will then free up more of the top-level
metal for the functional blocks across the microprocessor. As a result, their area could


41
Figure 2-12 Output matching network design to match 50 Q to rout
2.8 Summary
Common-gate and source-degenerated LNAs have been presented. An explicit
design methodology has been developed for a source-degenerated CMOS LNA, consider
ing gain, NF, input matching, and output matching. In particular, the optimal sizing of the
transistors has been presented for robustness against input matching variations and for
minimal noise figure. The noise parameters (in impedance form) for a MOSFET have
been derived to understand the effect of various second-order effects on noise
performance. Gate-induced noise and substrate resistance both affect the noise parameters
significantly. Using these results, a design methodology for minimum noise while con
straining the input matching condition has been presented, which yields an optimal value
for the size of the primary transistor in the cascode. An optimum sizing for the cascoded
device was obtained through minimization of noise figure. Finally, an alternative design
methodology based on the constant available gain method has been presented.


116
Figure 4-26 (a) Schematic of phase detector and low-pass filter (b) Output of phase detector
versus input skew
conventional XOR or gilbert cell architectures. Another benefit is the use of single-ended
inputs; thus, only a single-ended clock signal has to be routed to the PD. Figure 4-26(b)
shows the simulated output voltage versus input clock skew for the PD and LPF. The volt-
age-skew plot is symmetric with respect to skew=0; hence, the absolute value of skew is
measured. Since the output of the PD is close to linear, it can be characterized by measur
ing the output voltage for 0 and 180 phase offset. The measured skew can then be
extrapolated based on these two data points. Note that the voltage-skew plot repeats every
period, therefore, the measured clock skew will be modulo TcUc (tskew extracted = tskew md
Tcik), where Tcik is the period of the local clock signal.
4.6.6 Measured Results
To evaluate the operation of the initialization circuitry, a phase-detector and a
low-pass filter were implemented on-chip with two identical frequency dividers. These
frequency dividers were implemented with the original DFFs (Figure 4-19) since the


129
20 r
Frequency (GHz)
37
co
;o
g
'to

27
17
7
Frequency (GHz)
-3
Divider/Buffer v
Conversion Gain
Frequency (GHz)
Avoid
Mismatch
\ Avoid
i Mismatch
Frequency
Divider
(8:1)
Required
^ Jitter/Skew
Local Clock
Outputs
Required
S/N Ratio
Output
Buffers
t
Figure 5-1
(0 dBV) (-50 dBV) (-30 dBV)
Wireless interconnect optimization.
(odX
5.2.1 Transmitter Power
The amount of power that the power amplifier (PA) in the transmitter can deliver
depends on the supply voltage, the CMOS technology (e.g., gate-oxide thickness), the
power amplifier circuit topology, the frequency, and the antenna impedance. Depending on
the PA topology (class AB, E, F, etc.), the peak-to-peak voltage swing across the output
transistor can be between 2Vdd and 3.6Vdd [Lee98]. While recent results have shown
CMOS PAs capable of delivering 20-30 dBm of power at 900-1800 MHz using
0.2-0.35-|im technologies [FalOl, KuoOl, ShiOl], to the authors knowledge there are no
CMOS PA results in the 8 to 20-GHz region other than that which will be presented in
chapter 6. Thus, a 10-dBm output power is set as a goal for the PA at 15 GHz and above.
This corresponds to a peak-to-peak voltage swing of 2 V for a 50-Q single-ended load.
Further work is required to develop CMOS PAs operating above 15 GHz, and the 10-dBm
specification should be modified accordingly.


136
Static (skew) Dynamic (jitter)
Figure 5-3 Probability density function of propagation delay, including both static and
dynamic components resulting in clock skew and jitter (after EDN, July 1995, pp.
35).
clock edges throughout the chip. These quantities are related to the variances (a2) of the
static and dynamic delay probability density functions (PDFs), respectively.
Figure 5-3 illustrates each type of component as it relates to the total delay PDF.
Here, the total static variance is the sum of individual static variances (e.g., process varia
tion, capacitive load mismatch, etc.), yielding the clock skew. Likewise, the total dynamic
variance is the sum of individual dynamic variances (e.g., thermal noise, source jitter,
etc.), yielding the clock jitter. Both static and dynamic variances add to give the total delay
variance. The variances can be quantified as either a peak-to-peak variance (assumed to be
~ 6a), or a root-mean-squared (RMS) variance (a), where typically peak-to-peak is used
to quantify clock skew and either can be used to quantify clock jitter.
5.4 Clock Skew Versus Amplitude Mismatch
In the wireless clock distribution system, the signals will arrive at the antennas
throughout the chip at different power levels, depending on the distance between the


21
The worst case matching condition occurs either at the lower end of the frequency band
when the component tolerances cause the resonant frequency to increase or at the upper
end of the frequency band when the component tolerances cause the resonant frequency to
decrease. Both conditions yield approximately the same value for Qgs. Looking at the first
condition (i.e., L=L(1-Tl), Cgsl=Cgsl(l-Tc), to = coQ--, and cot=x(1-Tt)), the required
value of Qgs can be solved by normalizing (2.12) and then applying (2.13), as follows:
Qgs
0(1-Tc)
(1 o (1 -TL)(1 -Tc))i\
|5n|V+l)2 (r-1)2
1-|S
(2.14)
li
where
r
(x)tLs
(1-Tl)(1-Tt),
(2.15)
2Qb~ 1
2<2fi
Qb
(2.16)
(2.17)


103
operating as a delay-locked loop (e.g., [Sut95]), where the phase of all of the receivers will
be dynamically adjusted to a steady-state phase lock. This system is very similar to a dis
tributed synchronous clocking approach [Pra95]. The benefit of this system is that very
small skew can be obtained, equalizing both random and systematic phase errors. Also,
phase errors will not accumulate, since the system is controlled dynamically. However,
distributed synchronous clocking eliminates the need for integrated antennas and wireless
interconnects. Instead, multiple independent oscillators can be distributed throughout the
chip, and then phase-locked to one another, as in [GutOO]. Distributed clocking has the fol
lowing disadvantages, though: the benefit of having another communication medium on
chip is eliminated; jitter will spread and accumulate across the chip, since phase noise
from each input signal is transferred through the phase-locked loops via low-pass filter
responses; and there is added complexity to the system.
A second option for programming the dividers and starting up in a synchronized
state is to train the system at start-up, detecting the phase error for each receiver. Here, the
receiver phase errors are detected during training and a CNT value is extracted. The appro
priate CNT values are held in memory and used during normal system operation. This
phase-correction is static rather than dynamic, happening only once during system opera
tion. This static phase correction has the drawback of accumulating phase errors. In other
words, once the system is programmed, there is no other way to correct for phase errors
caused by noise or system fluctuations (e.g., power line fluctuations). Therefore, the sig-
nal-to-noise ratio of the system would have to be high.
The training mode would have the following conceptual steps. First, the clock
transmitter broadcasts the global clock (GCLK) used as a training signal. Second, each


I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor ofPhilospph)
Kenneth K0, UfthiTman
Associate Professor of Electrical and Computer
Engineering
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Mry G. Fossum
Professor of Electrical and Computer Engineering
I certify that I have read this study and that in my opinion- it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Robert M. Fox
Associate Professor of Electrical and Computer
Engineering
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and qualit
as a dissertation for the degree of Doctor of Philosophy
Michael P. Fra
Assistant Professor of Computer and Information
Science and Engineering
This dissertation was submitted to the Graduate Faculty of the College of
Engineering and to the Graduate School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
May 2001
M. Jack Ohanian
Dean, College of Engineering
Winfred M. Phillips
Dean, Graduate School


122
>y
4-31(a) shows the schematic of a [DP] inverter, while Figure 4-31(b) shows the relevant
node voltages and device currents simulated using SOISPICE ver4.41 [Fos97]. When in a
hold state (CLK=low), a single [DP] inverter should retain its previous value at the output
node regardless of the input value to the inverter. However when holding a high value at
the output with a nonzero input, Mn3 is turned on and its drain is pulled down to ground.
This is the situation shown in Figure 4-31(b), where the input is held at ~0.3 V (a low level
from pseudo-NMOS) for hold and evaluate. Since the average dynamic Vbs of Mn4 is -0.8
V and Vd n3 is now at ground, a transient bipolar leakage current (Ibjt,n4) is induced in
Mn4, despite the input of Mn4 being held low [Pel95]. This drains the charge stored at the
output, causing the inverter to begin to change states when it should be holding. These
effects can be seen in Figure 4-31(b), where a BJT current is noticeable, and the output
voltage is decreasing. The time constant associated with this bipolar leakage is on the
order of nanoseconds for this technology. If the divider is being switched at a rate much
(b) time (ns)
Figure 4-31 Transient bipolar effects of [DP]2 logic, (a) Schematic of [DP]2 inverter and (b)
corresponding node voltages and currents.


115
1.6
B 1.2:
^ o.8:
0.4-
Outputs
0
w
1.2-=
o 0.8-
0.4-
0-
1.6
w 1.2
o 0.8
0.4
0
rm
ff[ fry
"sr
I
INI
3n
2:1 Outputs
8:1 Outputs
4n
ft
5n
-
O' 1 In n 1 n 1 4n 1 5n
i i i i i i i.i
)m{
\
s/\AA(
>
m
0
1n
2n
' 3n
4n
5n
Figure 4-25 Simulated waveforms of two programmable 8:1 dividers with the new
DFFs--programmed to #000 (solid lines) and #111 (dashed lines)~for repeated
initializations. The circuit functions correctly and the output phase difference
between the two dividers remains constant.
Figure 4-26(a) shows a schematic of the phase detector (PD) and low-pass filter
(LPF) [Raz95]. The PD functions as an exclusive NOR circuit (transistors M^g). A
pseudo-NMOS load is used. A key feature is that this PD is symmetric for inputs CLK1
and CLK2. This eliminates any systematic phase error that would be introduced using


46
(a)
Figure 3-3 Die photographs of (a) 0.8-pm, 900-MHz single-ended CMOS LNA and (b)
0.25-pm, 8-GHz differential CMOS LNA.
LNA was packaged, therefore gold bond-wires are used to implement Ls as well as to
improve the quality factor of Ld. The on-chip portion of Ld consisted of a ground-shielded,
4.25-tum, 8-nH spiral inductor with a quality factor of 4.5 at 900 MHz. It was imple
mented using metals 2 and 3 shunted together. The bondwire and pin added an additional
2.2 nH to Ld. For this operating frequency and technology, Ls is ~1.4 nH. An off-chip
inductor is used for Lg, while an off-chip bypassing capacitor is required for Vdd. Finally,
ground-shielded pads [Col98] are used to minimize the substrate resistance, improve the
noise figure and reverse isolation, and control the pad impedance.
The 900-MHz LNA was packaged in an SOIC-like test package with low ground
inductance, and mounted on a multilayer board. A parasitic inductance therefore exists
between the on-chip ground and the board-level ground. If this inductance is too large, the
reverse isolation is greatly reduced, due to feedback [Col98]. To minimize the ground
inductance, 11 ground pads are included on chip and down-bonds from the on-chip ground
to the paddle of the package are used, reducing the inductance to less than 0.4 nH. The
paddle is then directly soldered to the board ground.


47
3.3.2 Measured Results
The LNA was characterized at four different operating points to study the trade-off
between power consumption and NF. Figure 3-4(a) shows a plot of measured gain and NF
versus bias current. Operating points 1 to 4 correspond to minimum to maximum NF and
maximum to minimum power, respectively. At 10 mA of current (operating point 1), the
LNA achieves a NF of 1.2 dB, which is competitive with most GaAs and bipolar LNAs.
However, the power consumption is 30 mW, which is quite large. As the power consump
tion is decreased to 8.1 mW (operating point 3), the NF only increases to 1.78 dB. This
result meets both of the desired performance metrics for CMOS LNAs (NF < 2 dB, and
power <10 mW). Measurements show that to achieve a NF of 2 dB, only 6.3 mW is
required. Operating points 3 and 4 have comparable NFs to [Stu98] and [Hua98], but at
only a fraction of their power. Much of this power savings comes from the ability to drive
Operating Point
(a) (b)
Figure 3-4 (a) Measured gain (S2i) and noise figure (NF) versus bias current. Four different
operating points are shown, numbered 1 to 4, corresponding to minimum to
maximum NF and maximum to minimum current, respectively, (b) Gain and NF
versus frequency for operating points 1 & 3.
(9P) L2S


197
Figure 6-25 Die photograph of 0.18-p.m clock receiver with zigzag dipole antenna.
simulations. In summary, although operation of the wireless transmitter has been demon
strated, the power that the transmitter can provide to the receivers through the antenna pair
is very low and as will be shown in the next section, is below that which the receiver can
detect. As can be seen, the PA appears to be the weakest link in the transmitter, and
requires circuit optimization.
6.6.5 Single Clock Receiver with Integrated Antenna
Figure 6-25 shows a die photograph of the 0.18-|im clock receiver with an inte
grated zigzag dipole antenna. The size of the receiver, including the antenna, is 0.66x2
mm while the active area is 0.37x0.58 mm The receiver consists of a zigzag dipole
antenna, a differential LNA with source-follower buffers, an 8:1 frequency divider, and
output buffers. The performance of the 0.18-|im LNA and source-followers was presented
in Chapter 3, while the divider performance was presented in Chapter 4. A summary of the
results is shown in Table 6-4, for easy reference. To decrease the minimum detectable sig
nal of the receiver, the peak LNA/buffer gain should coincide with the self-resonance of
the frequency divider. To achieve this, the supply voltage of the frequency divider was


98
Vdd=1.5 V, f¡n = 15.8 GHz Vdd=2.1 V, fin = 20.4 GHz
respectively. The output frequencies are 247 and 319 MHz, respectively. Once again, the
reduced output swing is from driving a 50-0 output load.
The f¡so is plotted in Figure 4-13(a) versus supply voltage. At 1.5 V, fis0 is 9.1
GHz, while at 2.1 V, fiS0 is 14.2 GHz. It can be seen that the difference between the maxi
mum operating frequency and fiso is ~ 6 GHz. Figure 4-13(b) shows the input sensitivity
for the 64:1 divider. The input sensitivity dips at the self-resonance frequency, as expected.
From the self-oscillation frequency at Vdd=1.5 V, is extracted to be 55 ps/DFF. Table
4-2 summarizes the measured results for the 0.18-|im frequency divider.
4.6 Programmable Frequency Divider Using SCL
4.6.1 Motivation and General Concept
As will be discussed in Chapter 5, a fundamental requirement for a wireless clock
distribution system is ensuring that all of the clock receivers distributed throughout the
chip are synchronized. This means that the delay mismatch, caused by differences in


82
Likewise, during hold operation of the master D-latch, the current pulling Qbi low is mod
eled as f2(l2), where f2 depends on the sizing of M5 and M6, the voltages Q¡ and Qbi, and
the hold current, I2. Thus, Qbi has a voltage during hold equal to
LOWHold = Vdd-f2(I2)RpM1. (4.2)
If f2(l2) > fi(Ij), the output voltage swing of the latch is decreased, and the voltages Q¡ and
Qbi might not be large enough to drive the next master-slave flip-flop.
For a given fj(Ij) to f2(l2) ratio, a lower frequency limit for the 2:1 divider exists,
set by how fast the nodes can change from LOWEvai to LOWHoid. As long as the input fre
quency is high enough, then LOWEvai cannot fully charge to LOWHold, and the divider
functions properly. This lower-frequency limit was observed in simulations, where the
minimum frequency was approximately one-eighth of the maximum frequency for the
chosen transistor widths. To make the quasi-dynamic latch fully static, first, tail current
sources should be added and Wj should be set equal to W2. This then makes Ij equal to I2.
Second, the widths of M3.6 should all be equal [Mur95], forcing fj(Ii) to be equal to f2(l2)-
4.3 Injection Locking of SCL Frequency Divider
The second mode of operation for the 2:1 SCL divider is injection locking. Injec
tion locking is the process of synchronizing an oscillator with an incident signal. Oscilla
tors can be injection-locked to the fundamental [Adl46, Pac65, Uzu85], subharmonics
[Zha92], and superharmonics [Sch71, Rat99] of the input signal. This section will exam
ine the oscillation of the SCL divider and discuss how the divider can be injection-locked.
The actual theory behind injection locking is presented in Appendix E. This appendix
reviews the basic operation of injection-locked oscillators (ILOs), and presents the


APPENDIX G
RELATIONSHIP BETWEEN JITTER AND PHASE NOISE
The root-mean-squared phase jitter is defined as the square root of the variance of
the difference between the phase of a signal at the beginning and end of a period, T
[Ega99]. This variance is:
C¡1 = £{[(t>( + r)- This is converted to jitter in time by simply dividing the phase jitter variance by the square
of the angular frequency of the signal. The autocorrelation function of = Em + xMt)] = f_^(f)ej2nfldf, (G.2)
where L^(jj is the single-sideband phase noise density, and R(x) and £J¡) are a fourier-
transform pair. The single-sideband density is one half the double-sideband phase power
spectral density, S^(f). Therefore, the jitter variance in time can be expressed as [Haj99]
<4 = 2 -if Mf)ej0df-f L(f)ej2nfrdf\ (G.3)
(2ic f0) J- J
-¡J {c ww -
-J' {JT'i* S^i2KJTf lf0 j
Mf .S,(/)[1-cos(2k/T)]/
(27t/)2 lJ J
= ^2fsjf)Sm(KjT)df
(2Kff J
267


147
of B (Hz) given in (E.l 1). The output phase-noise is then (refer to appendix E)
^(h \J ) 9
2
(5.31)
where S^ is given in (5.29). Clearly, the output phase noise of the injection-locked divider
is less than that of a divider operating in the latched mode.
As was discussed in 5.5.1, the total output phase noise of the frequency divider
will be due to the input additive noise described above, the output additive noise, and the
sampling effects of input additive noise. The sampling effects will be accounted for
implicitly by integrating the phase noise density over all frequencies which will result in
the same total power as the integral of the aliased spectrum from 0 to the sampling fre
quency. Output additive phase noise will simply add to (5.30).
5.5.4 Jitter in Clock Receiver
As derived in Appendix G and given in [Ega99], the clock jitter at a node is related
to the phase noise at that node through the following formula:
£st(/)sin2(jc/rW,
(5.32)
2
where T is the observation time. The units of (5.32) are in sec2. The RMS cycle-to-cycle
jitter would be the square root of (5.32) with T equal to the period of the output clock sig
nal. The sine-squared function has nulls at / = ^, or harmonics of the output frequency.
If the phase PSD is not varying very fast versus frequency with respect to sin (7tfT), then
the average value of sin2(7tfT) can be used. This results in the following jitter variance:
(5.33)


186
Transmitting
Antenna
(15 GHz)
Figure 6-17 Block diagram of clock transmitter for 0.18-pm test chip.
these fill patterns to be blocked around the antenna, while no blocking was used in the
0.25-|im chip. Currently, it is not known whether this difference can account for the differ
ence in the impedance. More experimental and modeling work is required to understand
these discrepancies, since predicting the antenna performance is critical to a working wire
less interconnect system.
6.6 Wireless Interconnects in 0.18-fim CMOS
6.6.1 Clock Transmitter
Figure 6-17 shows a simplified block diagram for the clock transmitter. The
15-GHz signal is generated using a differential voltage-controlled oscillator (VCO). The
VCO is required to have low phase noise to decrease the local clock jitter. The signal from
the VCO is then amplified by a two-stage output power amplifier and delivered to the
transmitting antenna. In the final clock distribution system implementation, the VCO will
be phase-locked to an external reference. However, to ease the implementation require
ments for this chip, the transmitter was operated open-loop, where the frequency of the
VCO was controlled directly with its dc input. This clock transmitter was primarily
designed by Chih-Ming Hung and characterized by the author. Accordingly, the basic cir
cuit design will only be briefly reviewed, specifically for the VCO where the interested


114
the initialization has to be carefully checked for the first 2:1 divider, and the appropriate
DFF shown in Figure 4-24(a) or (b) should be selected.
The correct operation of the programmable 8:1 divider with the new DFFs was
verified in simulations. The scenario is the same as in Figure 4-22, where two dividers are
programmed to CNT = #000 and #111, with INI switching from high to low repeatedly.
Figure 4-25 shows the simulated results. As can be seen, each time INI goes high, the out
puts of the #111-divider go high and those of the #000-divider go low. There is no longer
any situation where the output nodes of the DFFs are mistakenly pulled low when they
should be high. As a result, the outputs between the 8:1 dividers always maintain a 45
phase shift. Thus, these new DFFs remove the dependence of initialization on the current
state of the divider, and should allow the start-up state of the divider to be programmed in
45 increments as originally planned.
4.6.5 Testing Methodology
A methodology is needed to characterize the clock skew between independent
dividers, and to demonstrate the initialization circuitry. Measuring the clock skew between
two clock circuits using wafer probing would require matching the delay through the
external measurement system (probes, 3.5-mm coaxial cables, RF connectors), such that
the measured clock skew is due only to the skew in the circuits themselves. However,
matching these external components to within -20% of the required skew (which is 10%
of the local clock period) is a difficult task, or in other words the external components
need to be matched to within 8 ps for a 2.5-GHz local clock. Therefore, the clock skew is
detected on chip using a phase detector, which acts as a skew-to-voltage converter.


13
investigation by researchers, with most of the research occurring at frequencies below 6
GHz. The limited work in CMOS is due to the stringent performance requirements of con
ventional wireless standards (e.g., Global System for Mobile (GSM) or Global Positioning
Systems (GPS)), which has necessitated the use of GaAs or silicon bipolar technologies.
However, CMOS technologies can potentially reduce the overall cost of the transceiver by
allowing increased levels of integration, with the single-chip radio being an ultimate goal.
For CMOS technology to compete with silicon bipolar or GaAs technologies for wireless
applications, it must at least deliver the minimum necessary performance at a reduced sys
tem-level cost. From a performance standpoint, to be competitive with bipolar or GaAs
LNAs, CMOS LNAs must equal or surpass their low power consumptions of approxi
mately 10 mW and their low noise figures of approximately 2 dB [Sha97]. From a cost
standpoint, the LNA should be implemented in a standard digital CMOS technology with
out any specialty passive components, and the LNA should require a minimal number of
external components.
While recent works have demonstrated the potential of CMOS LNAs for ~l-GHz
applications, they generally have difficulty in attaining both low noise figure and low
power consumption simultaneously [Sha97, Stu98, Hua98]. Alternately, if they do meet
the low noise and low power, it is made possible by using either modified CMOS technol
ogies or external matching networks [GraOO, Hay98]. Also, at this time, other than work
that the author has participated in or originated, CMOS LNAs operating above ~5.8 GHz
have not been reported. The following two chapters present the design and implementa
tion of CMOS LNAs operating between 0.9 and 24 GHz. An explicitly defined design
methodology is presented in this chapter which considers gain, noise figure, input


33
2.6.3 Optimum for Minimum Noise
While Fmin is strongly dependent on multiple effects (GIN, Rsub, gmb, Cgd, and
Cgb), Gn and Xopt are only strongly dependent on Qgs with GEN on or off, and Ropt is only
strongly dependent on Qgs with GIN on. Therefore, only GIN and Qgs need to be consid
ered for Gn and Zopt. With this simplification, a design procedure for minimum noise can
be developed. First, Fmin is minimized with respect to Rsub and each of the capacitances.
Second, the actual noise factor (F) is minimized with respect to Qgs, including the effects
of GIN, through choosing the noise matching condition to approximately coincide with
the power matching condition. Figure 2-8 shows Fmin and F versus Qgs with GIN on.
Here, Sjj is constrained to be -10 or -15 dB, with 0>pLs set to 50 £1 Constraining Sj j pre
vents Zs from being equal to ^opt, resulting in a higher noise figure. As can be seen from
Figure 2-8, optimizing F with respect to Qgs is much more valuable than optimizing Fmin.
Also, since Fmin is relatively flat compared to F, only the quantity F-Fmin has to be
Figure 2-8 Minimum noise figure and actual noise figure when the input matching is
constrained by Sn* The difference between F and F^n indicates that constraining
Su causes a noise mismatch.


277
[YooOO] H. Yoon, K. Kim, and K. K. O, Interference effects on integrated dipole
antennas by a metal cover for an integrated circuit package, IEEE AP-S
bit. Symp. and USNC/URSI National Radio Science Mtg., pp.782-785, July
2000.
[You97] I. A. Young, M. F. Mar, B. Bhushan, A 0.35|i.m CMOS 3-880MHz PLL N/
2 clock multiplier and distribution network with low jitter for microproces
sors, ISSCC Digest Technical Papers, pp. 330-331, Feb. 1997.
[Yua89] J. Yuan and C. Svenson, High-speed CMOS circuit technique, IEEE J.
SolidState Circuits, vol. 24, no. 1, pp. 62-71, Feb. 1989.
[Yue98] C. P. Yue and S. S. Wong, On-chip spiral inductors with patterned ground
shields for Si-based RF IC's, IEEE J. SolidState Circuits, vol. 33, no. 5,
pp. 743-752, May 1998.
[Zha92] X. Zhang, X. Zhou, B. Aliener, and A. S. Daryoush, A study of subhar
monic injection locking for local oscillators, IEEE Microwave and Guided
Wave Letters, vol. 2, no. 3, pp. 97-99, Oct. 2000.
[Zie86] A. van der Ziel, Noise in Solid State Devices and Circuits, New York:
Wiley, 1986.


39
(a)
Rs
iWV-
Input
Matching
Network
r
Output
Matching
Network
:R.
Figure 2-10 Illustration of constant available gain methods, where for the gain stage either, (a)
a single transistor is used or (b) a source-degenerated cascode with inductive load
is used.
2.7.2 Alternative Constant Available Gain Methodology
The above design methodology is sensitive to the derived noise parameters. As
SPICE intrinsic noise models continue to improve, it is easier and more accurate to use
CAD tools to obtain the noise parameters of an intrinsic model with additional extrinsic
parasitics also modeled. A methodology for choosing Lg, Q, and C2 well-suited to CAD,
is the constant available gain method. Traditionally, the constant available gain method is
used to generate ideal matching networks, when given the S-parameters and noise param
eters of a single transistor [Gon97], as illustrated in Figure 2-10(a). However, since lossy
inductors are used at both the drain and the source, a composite circuit including Ls, Ld,
and the cascode is used in place of the active device, as illustrated in Figure 2-10(b). Val
ues for Wj, W2> Ls and Ld are chosen using the previous methodology. The S-parameters
and noise parameters of the composite circuit including the inductors are extracted using


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
A CMOS WIRELESS INTERCONNECT SYSTEM
FOR MULTIGIGAHERTZ CLOCK DISTRIBUTION
By
Brian A. Floyd
May 2001
Chair: Kenneth K. O
Major Department: Electrical and Computer Engineering
As the clock frequency and chip size of high-performance microprocessors
increase, it becomes increasingly difficult to distribute signals across the chip, due to
increasing propagation delays and decreasing allowable clock skew. This dissertation pre
sents the design, implementation, and feasibility of a wireless interconnect system for
clock distribution. The system consists of transmitters and receivers with integrated anten
nas communicating via electromagnetic waves at the speed of light. A global clock signal
is generated and broadcast by the transmitting antenna. Clock receivers distributed
throughout the chip detect the signal using integrated antennas, amplify and divide it down
to a local clock frequency, and buffer and distribute these signals to adjacent circuitry.
First, the design and implementation of CMOS receiver circuitry used for wireless
interconnects is presented. A design methodology is developed for CMOS low noise
amplifiers and demonstrated with a 0.8-|im, 900-MHz amplifier achieving a 1.2-dB noise
figure and a 14.5-dB gain. Amplifiers are also demonstrated at 7.4, 14.4, and 23.8 GHz,
Vlll


233
7.4.3 Conclusions for Synchronization
This section has estimated the (current) worst-case clock skew and jitter that can
be obtained with a wireless clock distribution system. The worst-case uncorrected skew
due to process variation, TOF mismatch, amplitude mismatch (from different antenna
gains), and mismatches in the output load is estimated to be 11.2%. Using a programma
ble 8:1 divider, though, will reduce the skew to under 6.25%. This necessitates a start-up
methodology to program and release the dividers synchronously. The worst-case
peak-to-peak jitter due to thermal noise, switching noise (digital interference), and phase
noise on the global clock signal, is estimated to be 2.4%. However, more work is required
to evaluate the jitter introduced by power-supply noise and the jitter of a 15-GHz PLL.
Measured results of a 15-GHz double-receiver wireless interconnect have shown a clock
skew of 4.7% and a peak-to-peak clock jitter of 1.24%. These results help to confirm the
estimations made in this section, showing that a wireless clock distribution system should
be able to meet the skew and jitter requirements for microprocessors. However, more work
is required to validate these estimations and to demonstrate the clock skew and jitter in a
working microprocessor (or close proximity thereof).
1.5 Latency of 0.18-p.m Wireless Interconnect
The increasing latency of conventional interconnects is a primary motivation for
investigating alternative interconnect systems. Therefore, the latency of the 0.18-(im wire
less interconnect has been calculated. The latency will include TOF delay and delay
through the clock receiver. If the silicon-substrate is assumed as the propagation medium
(which would have the worst-case latency due to higher dielectric constant), then the TOF
delay for a 1.3-cm propagation distance is 150 ps (this distance is assumed for the 100-nm


194
Table 6-3 Summary of measured characteristics for 0.18-gm CMOS VCO and PA
VCO
Expected
Measured
PA
Expected
Measured
Frequency
Range
20-21.6
GHz
14.3-15
GHz
fo
21 GHz
18 GHz
Power
Consumption
3 mW
7.2 mW
Power
Consumption
24 mW
(Vdd=0.9V)
41 mW
(Vdd=1.3V)
Phase Noise
@ 1MHz
-
-105
dBc/Hz
Output Power
to Antenna
+ 12 dBm
-13.2 dBm
Phase Noise
@ 3 MHz
-113
dBc/Hz
Operating Power
Gain (f0)
20 dB
6 dB
measurement of the full transmitter with an integrated antenna (discussed in the next sec
tion), revealing an available power of -13.2 dBm at 1.3 V to the antenna at 15 GHz.
Table 6-3 shows a comparison between the expected and measured PA results. As
can be seen, the power gain of the amplifier is less than expected and the resonant fre
quency has been reduced. Decreasing the inductor Q by half should reduce the small-sig
nal PA gain by approximately 7 dB. The first 6 dB comes from the pre-amplifier stage,
whose output is a parallel resonant circuit composed of L35 and shunt capacitances. Thus,
the load impedance at resonance is proportional to the Q of L35. Another 1-dB degrada
tion comes from the second stage, where its output is loaded by the antenna. However, 7
dB of the gain reduction is still unaccounted for. Finally, the reduction in resonant fre
quency is due to increased transistor capacitances.
6.6.4 Clock Transmitter with Integrated Antenna
Referring back to Figure 6-17, the clock transmitter consists of a VCO connected
to a PA, which then drives the transmitting antenna. Operation of the clock transmitter is
demonstrated by on-chip generation of a global clock signal and reception of this signal


8
other applications. Third, referring to Figure l-2(b), the inter-chip clock distribution sys
tem can provide global clock signals with a small skew to an area much greater than the
projected IC size. This is an additional benefit, possibly allowing synchronization of an
entire PC board or a multi-chip module (MCM). Fourth, in the wireless system, dispersive
effects are minimized since a monotone global clock signal is transmitted. Fifth, another
benefit is a more uniformly distributed power load equalizing temperature gradients across
the chip. Sixth, by adjusting the division ratio in the receiver, higher frequency local clock
signals [SIA99] can be obtained, while maintaining synchronization with a lower fre
quency system clock. Seventh, an intangible benefit of wireless interconnect systems is the
effect they could have on microprocessor or system implementations, potentially allowing
paradigm shifts such as drastically increased chip size. Finally, compared to other poten
tial breakthrough interconnect techniques, such