<%BANNER%>

Design and Characterization of an Intensity Modulated Optical MEMS Microphone


PAGE 1

DESIGN AND CHARACTERIZATION OF AN INTENSITY MODULATED OPTICAL MEMS MICROPHONE By LEE HUNT A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ENGINEERING UNIVERSITY OF FLORIDA 2003

PAGE 2

Copywright 2003 By Lee Hunt

PAGE 3

iii ACKNOWLEDGEMENTS I wish to give my love and gratitude to my friends and family for all the love and support I have received during the many years of my college career. I would never have succeeded to the extent that I have without them. I especially appreciate the love and support from my fiance, Ms. Lisa Sewell, and for her ability to see through my “preoccupied graduate student” exterior and find that which lies beneath. I would also like to express my gratitude to my advisor, Dr. Toshikazu Nishida, for recruiting me into IMG and giving me the educational tools and support needed to complete a difficult project. Thanks also go to Dr. Mark Sheplak, Dr. Lou Cattafesta, and Dr. Peter Zory for the excellent work they have done in teaching the classes that provided the foundation for my research, and also for their guidance during the process. I also wish to thank my friends and colleagues in the Interdisciplinary Microsystems Group who have provided support. Special thanks go to Karthik Kadirval for his support in the beginning of my work on the optical microphone project, and to Stephen Horowitz and Robert Taylor for timely assistance when I needed it. Financial support for this project is provided by DARPA (Grant #DAAD19-00-10002) through the Center for Materials in Sensors and Actuators (MINSA) and is monitored by Dr. Paul Holloway.

PAGE 4

iv TABLE OF CONTENTS ACKNOWLEDGEMENTS...............................................................................................iii ABSTRACT....................................................................................................................... ..x CHAPTER 1 INTRODUCTION..........................................................................................................1 1.1. Optical Microphone Tran sduction Schemes............................................................1 1.1.1. Intensity Modulation.........................................................................................2 1.1.2. Polarization Modulation....................................................................................5 1.1.3. Phase Modulation..............................................................................................6 1.1.4. Suitability of Transduction Tec hniques for MEMS Implementation...............9 1.2. Microphone Structure............................................................................................12 1.2.1. Overview.........................................................................................................12 1.2.2. MEMS Chip....................................................................................................14 1.2.3. Optical Fibers..................................................................................................15 1.2.4. Light Source....................................................................................................18 1.2.5. Detection Electronics......................................................................................18 2 MICROPHONE SYSTEM PARTITI ONING AND PERFORMANCE METRICS...21 2.1. System Partitioning................................................................................................21 2.1.1. Acousto-Mechanical Stage.............................................................................21 2.1.2. Mechano-Optical Stage...................................................................................21 2.1.3. Opto-Electrical Stage......................................................................................22 2.2. System Performance Metrics.................................................................................22 2.2.1. System Sensitivity...........................................................................................23 2.2.2. System Linearity.............................................................................................39 2.2.3. System Frequency Response...........................................................................44 2.2.4. System Electronic Noise.................................................................................46 2.2.5. System Minimum Detectable Signal..............................................................51 2.2.6. Optical Reference Path Losses and System Performance Metrics.................55 2.2.7. Summary of Predicted System Performance..................................................57 3 DESIGN OF THE OPTICS FO R THE MEMS OPTICAL MICROPHONE.............59 3.1. Selection of the Optics...........................................................................................59 3.1.1. Performance....................................................................................................59 3.1.2. System Connectivity.......................................................................................61 3.1.3. Ease of Handling and Manufacturability........................................................61

PAGE 5

v 3.1.4. Cost.................................................................................................................62 3.2. Selection of the Tubing..........................................................................................63 3.3. Alignment Issues....................................................................................................64 3.3.1. MEMS Chip Cavity Alignment issues............................................................64 3.3.2. Fiber Bundle Geometry Issues........................................................................67 3.3.3. Application of Alignment Theory to Fiber Bundle Selection.........................68 4 FABRICATION OF THE OPTICAL MICROPHONE..............................................70 4.1. MEMS Exchange Process......................................................................................70 4.2. Packaging Process..................................................................................................71 5 EXPERIMENTAL SETUP AND RESULTS..............................................................79 5.1. Laser and Photodetector Characterization.............................................................80 5.1.1. Experimental Setup for Laser and Photodetector Characterization................80 5.1.2. Results of Laser and Photodetector Characterization.....................................81 5.2. Static Calibration...................................................................................................81 5.2.1. Experimental setup for static calibration........................................................81 5.2.2. Results of static calibration.............................................................................84 5.3. Dynamic Calibration..............................................................................................87 5.3.1. Experimental setup for dynamic calibration...................................................87 5.3.2. Results of the dynamic calibration..................................................................90 6 CONCLUSIONS AND FUTURE WORK..................................................................99 6.1. Conclusions............................................................................................................99 6.2. Future Work.........................................................................................................100 APPENDIX A MEMS OPTICAL MICROPHONE DIAPHRAGM PROCESS FLOW..................102 B FIBER BUNDLE PROCESS FLOW........................................................................105 C MECHANICAL DRAWINGS..................................................................................117 D PHOTODETECTOR SPECIFICATIONS................................................................118 E SPECIFICATIONS FOR POPP ER & SONS STEEL TUBING..............................119 LIST OF REFERENCES................................................................................................120 BIOGRAPHICAL SKETCH ..........................................................................................124

PAGE 6

vi LIST OF TABLES 1-1 – Summary of Intensity-Modula ted Optical Microphone Designs............................10 1-2 – Summary of Phase Modulat ed Optical Microphone Designs.................................11 2-1 – Acousto-Mechanical Lumped Element Parameters................................................45 2-2 – Summary of Configuration Settings for Theoretical Performance Metrics............57 2-3 – Summary of Theoretical System Performance Metrics...........................................58 3-1 – Error Analysis of Different Fiber Bundle Configurations.......................................69 4-1 – Wafers Used for Optical Microphone Fabrication..................................................70 5-1 – Experimental HP8168B Noise.................................................................................81 5-2 – Comparison between Theoretical a nd Experimental Static Calibration..................86 5-3 – Experimental Results of Unreferen ced Output Microphone Dynamic Calibration98 5-4 – Experimental Results of Referenced Output Microphone Dynamic Calibration....98

PAGE 7

vii LIST OF FIGURES 1-1 – Optical Microphone Cla ssification Based on Trans duction Mechanism [2].............2 1-2 – Radiated Wave Intensity -modulating Microphone Types.........................................3 1-3 – Evanescent Wave Intensity -modulating Microphone Types.....................................4 1-4 – Polarization Modula ting Microphone Types.............................................................6 1-5 – Grating-Type Phase M odulating Microphone Types................................................7 1-6 – Interferometric Phase Modulating Microphone Types..............................................8 1-7 – Block Diagram of the Optical Microphone.............................................................13 1-8 – Cross Section of the Fiber Bundle in the MEMS Chip...........................................14 1-9 – Cross Section of the MEMS Chip...........................................................................15 1-10 – End View of the Optical Fiber Bundle..................................................................16 1-11 – Optical Fibers in Steel Tubing...............................................................................16 1-12 – Optical Fiber Bundle Drawing..............................................................................17 2-1 – Side View of Deflec ting Plate or Membrane...........................................................24 2-2 – Method of Images (View fr om Side of Fiber Bundle)............................................27 2-3 – Ring Approximation Diagram.................................................................................30 2-4 – Theoretical Power Coupled w ith Ideal Fiber Configuration...................................32 2-5 – Theoretical Sensitivity with Ideal Fiber Configuration...........................................32 2-6 – Block Diagram of the Unrefe renced Output Configuration....................................34 2-7 – Equivalent Circuit for the PDA400 Photodetector..................................................35

PAGE 8

viii 2-8 – Block Diagram of the Refere nced Output Configuration........................................36 2-9 – Comparison of Unreferenced and Referenced Output Sensitivities........................39 2-10 – Linearity of Mechano-Optical Stage.....................................................................41 2-11 – Plot of Acousto-Mechanical Sensitivity as a Function of Radial Position............42 2-12 – Noise Contributions for the Photodetector Output................................................46 2-13 – Noise Contributions fo r the Microphone Output...................................................48 2-14 – Illustration of the Physics Behind the MO MDS...................................................53 3-1 – Bundle Position Error Illustration...........................................................................65 3-2 – Angular Misalignment Error Illustration.................................................................66 3-3 – Radial Position Error Illustration.............................................................................67 4-1 – Abeysinghe et al. Packaging Technique..................................................................71 4-2 – Beggans et al. Packaging Technique.......................................................................73 4-3 – Kadirval Pack aging Technique................................................................................74 4-4 – Proposed Package for the Optical Microphone.......................................................76 4-5 – Proposed Optical Mi crophone Array Package........................................................77 5-1 – Experimental Setup fo r Laser Characterization.......................................................80 5-2 – Block Diagram of Static Calibration.......................................................................84 5-3 – Experimental Power Coupled vs. Equilibrium Gap................................................85 5-4 – Experimental Maximum Power Coupled Regression Line Slope...........................86 5-5 – Unreferenced Output Optic al Microphone Configuration.......................................88 5-6 – Referenced Output Opti cal Microphone Configuration..........................................89 5-7 – Linearity, Unreferenced Output Configuration.......................................................91 5-8 – Linearity, Referenced Microphone Configuration..................................................92

PAGE 9

ix 5-9 – Magnitude Response, Unrefere nced Microphone Configuration............................93 5-10 – Magnitude Response, Referen ced Microphone Configuration.............................94 5-11 – Phase Response, Unreferenc ed Microphone Configuration..................................94 5-12 – Phase Response, Referen ced Microphone Configuration.....................................95 5-13 – Electrical Noise Floor, Bo th Microphone Configurations.....................................97

PAGE 10

x Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Engineering DESIGN AND CHARACTERIZATION OF AN INTENSITY MODULATED OPTICAL MEMS MICROPHONE By Lee Hunt August 2003 Chairman: Toshikazu Nishida Major Department: Electrical and Computer Engineering This thesis presents the design and characterization of an intensity-modulated optical lever microphone. Microphone noise models from previous works are expanded to include the light source and all electronics. Physical phenomena responsible for limiting the microphone minimum detectable signal (MDS) are identified, and an accurate model developed for use with an LED or laser light source. The sensitivity, minimum detectable signal, and electronics noise are characterized by a scaling analysis in which coupled equations for dependence on optical power, membrane radius, photodetector gain, and optical losses in the reference path are presented. It was discovered that, in this optical microphone geometry, the laser is the limiting factor in the microphone MDS and electronics noise, and optical losses in the reference path can improve microphone sensitivity, MDS, and noise floor for a referenced optical microphone.

PAGE 11

xi An unreferenced electronic configuration and a referenced electronic configuration were experimentally characterized using a laser as a light source. The unreferenced optical microphone achieved a sensitivity of 0.032 mV / Pa, MDS of 65 dB (re. 20 Pa), and dynamic range from 65 – 122 dB (re. 20 Pa). The referenced optical microphone achieved a sensitivity of 1.77 mV / Pa, MDS of 47 dB (re. 20 Pa), and dynamic range from 47 – 122 dB (re. 20 Pa). Both unreferenced and referenced measurements were made at 1600 Hz with a bin width of 2 Hz.

PAGE 12

1 CHAPTER 1 INTRODUCTION Optical microphones vary widely in their construction, but all possess innate resistance to electro-magnetic interference (EMI) and other harsh environments to which other types of microphones are sensitive. This innate resistance is derived from the separation of the optical sensing element from the electronics via optical fibers and assumes the electronics are remotely located with respect to the test environment. In the case where the electronics are not remotely located, the microphone package must isolate the microphone electronics from the test environment. MEMS technology provides a promising new implementation for optical microphones. MEMS devices have the capability to be smaller than conventional microphones, and MEMS microphone chips could be processed by the thousand on wafers if the market can support this volume. Despite these advantages, Professor Steve Senturia [1] notes that MEMS devices have a coupling between the package and the device, thus requiring them to be designed concurrently, which makes a MEMS microphone design inherently more complicated than a conventional (non-MEMS) microphone. 1.1. Optical Microphone Transduction Schemes In 1996, Nykolai Bilaniuk first introduced a classification scheme for optical microphones that relied on the transduction mechanism as the primary sorting criterion [2]. He also explained the methods of operation of multiple types of devices in each

PAGE 13

2 category, with emphasis on the most promising technologies, and he also discussed microphone system performance metrics. Bilaniuk defined three properties of light that could be modulated: the intensity (or irradiance), the phase, and the polarizati on [2]. Since electro-optical detectors respond to light intensity, all modulation schemes must be reduced to an intensity modulation at the electronics end of the sy stem. The figure below adapted from [2] shows a detailed classification scheme for optical microphones. Figure 1-1 – Optical Microphone Classificati on Based on Transduction Mechanism [2]. 1.1.1. Intensity Modulation Bilaniuk [2] describes an intensity-modulated microphone as one which selectively removes energy from the optical path. As shown in Figure 1-1, an intensityOptical Microphones Intensity Modulating Polarization Modulating Phase Modulating Radiated Wave Evanescent Wave Moving Grating Lever Cantilever Macrobend Microbend Coupled Waveguide Radiated Wave Evanescent Wave Grating Interferometric Michelson Interferometer Mach-Zehnder Interferometer Fabry-Perot Two-Mode Fiber Dynamic Photorefractive Grating Input Coupling Grating

PAGE 14

3 modulating optical microphone can be subdivide d into two broad categories: radiated wave and evanescent wave. All of the energy in radiated wave optical microphone leaves a controlled optical path and partially recaptured or backscattered [2]. Figure 1-2 recreates Bilaniuk’s [2] illustration of the radiated wave transduction strategies. Figure 1-2 – Radiated Wave Intensity-modulating Microphone Types. The moving grating approach relies on the motion of a “light gate” to modulate the light coupled between an input waveguide and an output waveguide. These types of devices do not make use of diffraction or any structures on the order of the wavelength of the light. An intensity-modulated lever microphone utilizes one or more waveguides to deliver light to a vibrating plate or membrane. Reflected light is co llected by one or more Moving Grating Cantilever Lever Macrobend

PAGE 15

4 waveguides and delivered to a photodetector. Lever microphones may also have focusing optics to improve light collection. In a cantilever microphone, the waveguide is discontinuous, and part is free to vibrate in an acoustic field. This varies the alignment between the fixed segment and the free segment of the waveguide, causing a modulation of the power coupled. Macrobend-type intensity-modulating schemes use acoustic waves to deform a fiber configuration, such as a coil. Optical fibers are chosen that do not completely confine the light. The deformation modulates the losses in the length of fiber, subsequently modulating the output power. Alternatively, the evanescent-wave coupling methods “rely on …mode coupling or on absorption from the evanescent field” [2]. Bilaniuk defines two classes of evanescent wave intensity-modulating microphones: microbend and coupled waveguide. Figure 1-3 recreates Bilaniuk’s [2] illustration of the evanescent wave intensity modulation techniques. Figure 1-3 – Evanescent Wave Intensity-modulating Microphone Types. Microbend Cou p led Wave g uide

PAGE 16

5 The microbend technique uses a microstructure to apply periodic deformations to a waveguide. The acoustic field modulates the pressure exerted on the waveguide by these deformations, which in turn causes leakage of power out of the waveguides. The coupled waveguide technique can work in one of two different ways. In the first way, the waveguides are fabricated on a membrane structure with a fixed separation between the two. The membrane deflects in the presence of an acoustic field, and this deflection changes the index of refraction in the two waveguides. The change in refractive index modulates the power coupled between the waveguides. Alternately, the waveguides are fabricated so that one is attached to a structure, while the other is free to vibrate. An acoustic field will modulate the separation between the waveguides, which modulates the power coupled between the two. 1.1.2. Polarization Modulation The second major category of optical microphones as defined by Bilaniuk [2] is polarization modulation. Polarization modulati on type devices alter the polarization of the light when in the presence of an acoustic field. Bilaniuk [2] subdivides polarization modulation devices into two subcategories, but he notes that alternate schemes are possible. Figure 1-4 adapted from [2] depicts the two subcategories. In the first category, a layer of liquid crystals is subjected to acoustic field induced shear stresses, which modulate the polarization of the light passing through. A polarizer is located at the output of the device to isolate the desired polarization axis. In the second category, “a moveable dielectric plate interacts with the evanescent field of a waveguide excited with both TE a nd TM modes, causing a different change in

PAGE 17

6 the refractive index of the two modes, according to Bilaniuk [2]”. A polarizer at the output isolates the desired polarization axis. Figure 1-4 – Polarization Modulating Microphone Types. 1.1.3. Phase Modulation Phase modulated optical microphones are described by Bilaniuk [2] as a mechanism that “changes either the physical length or the refractive index of an optical test path and recombining the result with the signal from a reference path.” The reference path is unaffected by the acoustic field, while the test path undergoes some form of mechanical deformation. The two defined subgroups for this category of optical microphones are grating type devices and interferometric devices. A grating type device is one with a structure machined onto a waveguide with features on the order of the wavelength of th e light. The two different subcategories of grating devices defined by Bilaniuk [2] are input coupling gratings and dynamic refractive gratings. They are shown in the following figure, adapted from Bilaniuk [2]. N ematic Li q uid Cr y stal Differential Index Shifte r

PAGE 18

7 Figure 1-5 – Grating-Type Phase Modulating Microphone Types. The input coupling grating device has a grating fabricated on the waveguide. Incident light at the proper angle, wavelength and with the proper grating spacing will be coupled into the waveguide. The acoustic fi eld modulates a nearby dielectric structure, varying the index of refraction of the system and modulating the output. The dynamic photorefractive grating uses a prism to split light onto two mirrors, one of which is free to vibrate in an acoustic field. The light reflects off the mirrors to pass through a grating, and the light from each mirror is captured by a photodetector. The light from the stationary mirror is used as a reference signal, while the light from the vibrating mirror is used as the modulated signal. The second major category of phase modulating optical microphones is interferometric-type phase-modulating microphones. They typically use one of the three most familiar types of interferometers: Fabry-Perot, Michelson, or Mach-Zehnder. Alternately, a two-mode fiber can be used to make a phase modulated microphone. The figure below (adapted from [2]) depicts the four interferometric optical microphone categories. Input Coupling Grating D y namic Photorefractive Gratin g

PAGE 19

8 Figure 1-6 – Interferometric Phase Modulating Microphone Types. The Fabry-Perot optical microphone uses an optical cavity formed between two parallel surfaces. One of the surfaces is free to vibrate in an acoustic field, while the other is fixed. Typically, the vibrating surface is a plate or membrane, and the fixed reflecting surface is the face of the fiber, but additional optics may be used instead. A Michelson optical microphone splits a free-space beam into two paths. The reference path is reflected of a stationary re flector. The test path is reflected off of a reflector that vibrates in an acoustic field. The beams recombine and interfere, and the recombined signal is received by a photodetector. In a Mach-Zehnder optical microphone, the light enters via a waveguide, which is split into two paths. The referen ce path is held constant, but the test path is free to vibrate Fabr y -Pero t Michelson Mach-Zehnde r Two-Mode Fibe r

PAGE 20

9 in an acoustic field. The light in the two paths is recombined and sent to a photodetector. Interference effects will modulate the power seen by the detector. The fourth type of interference optical microphone is a two-mode fiber microphone. In this design, a section of two-mode optical fiber is spliced at the end of a single mode fiber. The two-mode fiber is fr ee to vibrate in an acoustic field. Acoustic vibrations will modulate the index of refr action of each mode differently, and an interference pattern will be generated at the junction between the two fibers. 1.1.4. Suitability of Transduction Techniques for MEMS Implementation In general, the simplest type of microphone to analyze and build is an intensitymodulated device. The simplest intensity-modulated device can be constructed with an LED, multimode or single mode fibers, a membrane or other vibrating reflective surface, and a photodetector. Table 1-1 (see [3] – [8]) summarizes recent intensity-modulated optical microphone designs. The results indicate a large variability in performance with the implementation of the intensity-modulated microphone. While this observation may seem obvious, it reinforces the importance of optimizing the system as a whole when designing the microphone and not just an individual stage. In general, for the intensity-modulated optical microphone, increasing the diaphragm radius increases the sensitivity and decreases the minimum detectable signal (MDS). Therefore, intensity-modulated microphone performance is decreased when the diaphragm is constrained to have a diameter of less than a few hundred microns.

PAGE 21

10 Table 1-1 – Summary of Intensity-Modulated Optical Microphone Designs Author / Year Design Type Source and o Sens Noise Freq Response Linearity Range MDS V. P. Klimashin 1979 [3] Lever, -wSupport Optics, non MEMS Incandescent Lamp 7.5 mV / Pa 20-22dB (re 20 Pa) 0 20kHz -w5dB fluctuatio ns Hu and Cuomo 1992 [4] Lever, -wMylar Membrane, Non MEMS LED, 2.4 mW 36.5 mV / Pa 0-31.5 kHz De Raula and Vinha 1992 [5] Multiple Light Source non-MEMS scheme 150 W Xenon Arc Lamp 5.6 nW / Hz0.5 Lukosz and Pliska 1992 [6] Evanescent wave, microbend, 6x6 mm2 membrane Laser, = 632.8 nm 0.31 Pa-1 49 dB (re 20 Pa) Up to 10kHz 49 dB 95 dB Suhadolnik, et al. 1995 [7] Lever, fiber bundle and deflecting diaphragm, nonMEMS High due to speckle pattern of laser light MO Stage 1500 m Kadirvel 2002 [8] Lever, fiber bundle and deflecting diaphragm Laser, = 1550 nm 152 V / Pa 110 dB (re 20 Pa) 1 kHz – 6.4 kHz 110 dB – 135 dB 110 dB (re 20 Pa) The choice of light source and photodetector also plays a large role in the performance of an intensity-modulated OM. Both affect the device sensitivity and noise floor. Sensitivity increases as coupled optical power increases, so high intensity light sources provide higher sensitivities, provided that the photodetector does not saturate. A disadvantage of all intensity-modulated OMs is the large DC component of the received signal. The DC component does not contribute to the device sensitivity, but it does contribute to photodetector saturation. This limits the product of the optical received power and the detector trans-impedance gain. The maximum intensity of the light source is limited by the linearity range and gain of the detector. Table 1-2 (see [9] – [15]) summarizes recent phase modulated optical microphone (PM) designs. Since no standard method of reporting the sensitivity of an optical microphone has been agreed upon, it is difficult to compare the overall performance of

PAGE 22

11 different PM designs. Theoretically, a PM device would be able to provide higher performance than an intensity-modulated microphone in a MEMS implementation, especially for membranes constrained to be smaller than a few hundred microns in diameter. PM devices have a smaller DC component allowing for more flexibility in selecting photodetector gain settings. Table 1-2 – Summary of Phase Modulated Optical Microphone Designs Author / Year Design Type Source and o Sens SNR Freq Response Linearity Range Resolution Rao et al. 1997 [9] Bragg Grating –wFizeau Cavity 20mW LED @ o=1550nm 12 pm / 50dB > 1kHz < 5000 Furstenau et al. 1998 [10] Fabry-Perot Cavity 0.5mW (after pigtail) LED @ o=1300nm Varies by 80dB over freq range w.r.t. B&K 4134 Tested over 100Hz to 15kHz Du et al. 1999 [11] Fiber Bragg Grating LED @ o=1550nm 1.5 pm / NA < 1200 +/29 Graywall 1999 [12] Surface-machined Fabry-Perot Cavity, theoretical analysis LED @ o=650nm 8.9 mV / Pa > 100 100Hz to 2kHz Rao et al. 2000 [13] Fiber Bragg Grating and Fizeau Cavity 20mW LED @ o=1550nm ~ 540o / m Abeysinghe et al. 2001 [14] Fabry-Perot Cavity machined on surface of optical fiber LED @ o=850nm 0.11 mV / psi (16 mV / MPa) NA 0 – 80 psi (0 – 552 kPa) Wang et al. 2001 [15] Non-MEMS FabryPerot Cavity LED @ o=850nm 4 nm / psi (0.58 nm / kPa) NA 0 – 6000 psi (0 – 41.4MPa) 0.02 psi (1379 Pa) Despite these advantages, PM microphones present some significant challenges. The dimensions involved are on the order of tens of optical wavelengths, making static characterization and packaging very difficult. PM microphones are much more sensitive to misalignments and phase noise sources than an intensity-modulated microphone. Because of this, PM implementations require more complicated electronics for signal demodulation, and they have stricter require ments for the light source. Finally, a PM

PAGE 23

12 microphone has a periodic power coupled curve, constraining the microphone to either a very small membrane deflection or to a “peak-counting” scheme during demodulation. Due to the additional complexity involved in implementing a PM microphone and the mixed results achieved by previous implementations (Table 1-2), an intensitymodulated lever-type transduction scheme was chosen for this work. 1.2. Microphone Structure 1.2.1. Overview The intensity-modulated optical microphone that is the topic of this thesis can be divided into four major physical parts. They are the MEMS chip, the optical fibers, the light source, and the detection electronics. The following figure shows the block diagram for the optical microphone. In the steady-state case, light from the light source is coupled into the transmit (Tx) fiber. The Tx fiber delivers the light to the MEMS chip, where it is reflected and partially coupled into the receive (Rx) fiber. The Rx fiber then delivers the light to a photodetector, where it is converted into an electrical signal and processed by detection electronics. When an acoustic field is present at the MEMS chip, the coupled optical power is modulated. This allows the transducer to convert acoustic energy into electrical energy, which is the definition of a microphone.

PAGE 24

13 Figure 1-7 – Block Diagram of the Optical Microphone. There are four energy domains present in this system that carry information. The first domain is the acoustic domain, where the desired measurement lies. The MEMS diaphragm converts the acoustic energy into mechanical energy through its displacement. The mechanical displacement of the membrane varies the power coupled into the Rx optical fiber, converting the signal into the optical domain. At the photodetector, the signal is converted into the electrical domain for analysis. For our design, we have chosen a reflective-type intensity-modulated optical lever microphone, with the mechano-optical transduction mechanism shown in Figure 1-8. The dominant reason for this selection is that this type of intensity-modulated optical microphone is much simpler to design and package than other intensity-modulated microphones. Details of each component are described in later sections of this chapter. Acoustic Waves MEMS Chip Rx Fiber Tx Fiber Light Source Detector and Electronics

PAGE 25

14 Figure 1-8 – Cross Section of the Fiber Bundle in the MEMS Chip. 1.2.2. MEMS Chip The MEMS chip is a 2.5 mm x 2.5 mm silicon chip with a micromachined 1 mm diameter silicon nitride diaphragm. The process flow for the MEMS chip is discussed in Section 4.1. A cross section of the MEMS chip is shown in Figure 1-9. The dominant membrane material is a 1 m thick layer of silicon nitride. A 70 nm thick layer of aluminum is deposited on the membrane surface to enhance reflectivity. The cavity formed by the bulk silicon and silicon nitride membrane is fitted over the end of a steel hypodermic tube containing the optical fibers. Rx Rx Tx MEMS Chip Protective Steel Tubing Epoxy Light Cone

PAGE 26

15 Figure 1-9 – Cross Section of the MEMS Chip. 1.2.3. Optical Fibers The optical fibers selected for the optical microphone are the Thorlabs AFS105/125Y multimode optical fibers. They are used for both transmit (Tx) and receive (Rx) fibers. The end of the optical fiber that terminates at the MEMS chip is designated the device end, and the end connected to the light source / photodetector is designated the Tx / Rx end. One fiber acts as a Tx fiber, and six fibers are Rx fibers. Figure 1-10 shows the desired shape of the fiber optic bundle as seen from the nitride membrane into the steel tubing. In this figure, the cores of each fiber are color-coded, and surrounded by a white ring representing the cladding. The dashed line is a possible location for the border of the light cone reflected by the membrane. The receive fiber area inside the dotted ring is responsible for collection of the reflected light. Aluminum (70 nm) N itride (1 m) Oxide (0.7 m) Bulk Silicon (~500 m) 1 m m

PAGE 27

16 Figure 1-10 – End View of the Optical Fiber Bundle. Figure 1-11 – Optical Fibers in Steel Tubing. Figure 1-11 shows the fiber bundle inside the protective steel tubing. The end view shown is the ideal position, where the transmit fiber is located in the exact center of the tube. This allows the light to reflect off of the center of the membrane, which has the maximum acousto-mechanical sensitivity (defined here as change in membrane deflection per change in acoustic pressure). Side View End View Steel Tubing Epoxy Receive Fiber Transmit Fiber Tx Rx Rx Rx Rx Rx Rx

PAGE 28

17 Figure 1-12 is a diagram of the fiber bundle in its protective tubing. Connections to other parts of the system are noted. The dashed arrows indicate the path of light through the system. Paths three and four contain modulated data. Figure 1-12 – Optical Fiber Bundle Drawing. Based on the work of He and Cuomo [4], the mechano-optical (MO) stage sensitivity, defined as change in coupled optical power per change in membrane displacement, is maximized when a single transmit fiber is surrounded by a tightly packed ring of receive fibers (see Figure 1-10). The smaller the radius of the receive ring, the greater the sensitivity of the MO stage, and the smaller the equilibrium gap, which is defined as the equilibrium distance between the fiber bundle face and the membrane. It may be possible to increase the sensitivity of the MO stage by adding extra Tx fibers [16]. The current fiber bundle is designed to sample the displacement of the membrane at the center (at the maximum displacement). For membranes which are much larger than the accompanying fiber bundles, all bundles could illuminate areas near the center of the membrane, where Sam is high and the stage is linear. If the diameter of the Device Transmit Receive From laser To Detector 1 2 3 4 Steel Hypodermic Needle Emitted Light (no reflecting membrane present)

PAGE 29

18 bundle structures is on the order of the membrane, then sensitivity gains will be lessened and linearity of the stage may become an issue. Sensitivity and linearity are discussed later in this thesis. Since the MO stage sensitivity is a resu lt of the displacement of the illuminated region of the membrane, extra MO sensitivity can be obtained by illuminating additional portions of the membrane by additional fiber bundle structures. With the assumptions that adding additional identical bundle struct ures does not remove any light collection ability, the same electro-optic sensitivity is available to each bundle structure, and the region of the illuminated membrane is locally flat, then the MO stage will remain linear and the system sensitivity would be given by the following equation. i i am mo oeS S S S_ Equation 1-1 1.2.4. Light Source The light source used by this optical microphone is the HP8168B Tuneable Laser Source. The maximum output power of the laser at 1550 nm is 0.515 mW. An alternate laser source or an LED source could be used in place of the HP8168B. 1.2.5. Detection Electronics There are three schemes considered by this thesis for use as detection electronics. The first scheme uses a single photodetector and takes the unreferenced output of the photodetector as the microphone output. This technique is called the unreferenced output technique. This scheme can be used with an amplifier at the output of the microphone to increase the gain. The primary advantage of this scheme is simplicity. Fewer optical and electronic components are required here than for any other configuration (see Section

PAGE 30

19 5.2.1 for details). This greatly reduces the cost of the optical microphone when compared with the other opto-electronic configurations. The largest disadvantage of this configuration is the dependence of the unreferenced OE sensitivity on the optical power (shown in Section 2.2.1.3). This dependence makes the unreferenced optical microphone much less stable in the presence of laser instability and drift. Also, the overall sensitivity of the unreferenced OE microphone is less than the referenced microphone configuration. The second scheme, which was used by Kadirval [8], is the referenced output technique. It uses an optical splitter to separate the light source output into two paths. One path is connected to the Tx fiber for transmission to the MEMS chip. The Rx output is the modulated data signal, and is taken to a photodetector. The second path is connected directly to a photodetector for use as a reference signal. An analog divide circuit is used to divide the modulated signal by the reference signal, and the divided signal is taken as the microphone output. The largest advantage of the referenced output configuration is the independence of the sensitivity on light source power. This minimizes the negative effects of low frequenc y fluctuations in the light source output power, such as fluctuations due to temperature changes. Another advantage of the referenced output configuration is the ability to significantly improve microphone performance by adding optical losses to the reference signal path (shown in Chapter 2). Additionally, an amplifier may be used at the output to further increase sensitivity. Despite the advantages, the referenced optical microphone requires more optics and electronics than the unreferenced microphone (see Section 5.2.1 for details). This increases the cost compared to the unreferenced microphone. Another disadvantage is

PAGE 31

20 the increased electronics noise in the output due to the extra electrical components (Section 2.2.4 for details). The third scheme, heterodyne modulation, is designed to take advantage of the flatness of the overall noise floor of th e unreferenced optical microphone at high frequencies. In this scheme, the laser output is modulated by an external sinusoidal signal to frequencies much higher than the high frequency cutoff of the microphone. The received optical microphone signal at the photodetector will be contained as a bandpass signal centered at fo, where fo is much larger than the microphone bandwidth. After passing through the photodetector, the signal is passed through a lock in amplifier and demodulated back to the original baseband signal. This will make the noise floor of the microphone dependant on the high frequency noise floor of the laser, and not the low frequencies where 1/f noise (and other noise sources) is present. As with the previous two OE configurations, an amplifier can be used at the output to increase sensitivity. The major disadvantage of this electronic configuration is the increased electronic complexity when compared to the other electronic configurations. Additionally, the lock-in amplifier must be capable of passing frequencies at least 10 times the microphone high frequency cutoff. Also, even small laser transients will cause the lock-in amp to fail to reproduce the signal. This scheme was not implemented in this thesis, although it is likely that an optical microphone system using a laser as the light source would require heterodyne detection for satisfactory performance.

PAGE 32

21 CHAPTER 2 MICROPHONE SYSTEM PARTITIONING AND PERFORMANCE METRICS 2.1. System Partitioning The intensity-modulated optical microphone is partitioned into three stages where transduction between energy domains occurs. The three stages of an intensity-modulated optical microphone were identified by Bilaniuk [2]. They are the acousto-mechanical stage, the mechano-optical stage, and the optoelectrical stage. Kadirval [8] and Bilaniuk [2] discuss these stages in detail, and a summary is included below. 2.1.1. Acousto-Mechanical Stage The acousto-mechanical stage is where the energy in the acoustic signal is converted into the mechanical domain. This is accomplished when the pressure and volume velocity of the acoustic signal induce a displacement and restoring force in the membrane. The unit of sensitivity for this stage is a displacement per unit pressure, typically given in m / Pa. 2.1.2. Mechano-Optical Stage In the mechano-optical stage, input optical power is reflected by the displacing membrane and coupled into output (Rx) fibers. Transduction occurs when the mechanical displacement of the membrane varies the percent of the input power that is coupled into the output fibers. The unit of sensitivity of this stage is normalized power per unit displacement, typically given in m-1.

PAGE 33

22 2.1.3. Opto-Electrical Stage The third and final transduction stage in an intensity-modulated optical microphone is the opto-electrical stage. This stage uses one or more photodetectors to convert the coupled optical power into an electrical signal. The sensitivity units for this stage are normally written as in volts (V). Occasionally an author will write the OE stage sensitivity in V/(W/W). Most authors (including Bilaniuk) lump the optical power dependence of the overall sensitivity of some microphone configurations into the OE stage. 2.2. System Performance Metrics Kadirval [8] used the following performance metrics to classify the optical microphone. They are sensitivity, linearity, frequency response, noise floor and minimum detectable signal (MDS). These metrics can also be used to describe the performance of the individual stages. The th eoretically determined performance metrics for the system and each stage are summarized later in this chapter. In this thesis, a theoretical sensitivity model for the referenced output configuration is derived for the case where optical reference path losses and a low-noise amplifier at the output are present. A theoretical model of the electronic noise is derived for both unreferenced and referenced configurations. This model extends the noise model derived by He & Cuomo [16] and used by Kadirval [8] to include the intensity noise of the light source and all electronics. The physics behind the minimum detectable signal equation presented by Kadirval [8] are explained, and the equation is used with the improved noise model to predict the minimum detectable signal.

PAGE 34

23 2.2.1. System Sensitivity The sensitivity of a system is defined as the differential change of the output quantity divided by the differential change of the input quantity. For a microphone, the system output is a voltage, and the input is a pressure. The optical microphone is a multienergy domain system with three transduction stages, as previously described. The maximum ideal sensitivity is a product of the sensitivities of the individual stages. Equation 2-1 is the equation for the system sensitivity in terms of the individual stages. All reported sensitivities in this thesis ar e based on a fiber bundle with identical Tx and Rx fibers having an inner core diameter of 105 m and a cladding diameter of 125 m. oe mo amS S S S Equation 2-1 Section 2.2.1.3 examines the sensitivity of the OE stage in more detail then was done by Kadirval [8]. It derives theoretical models for the OE sensitivity in the unreferenced and referenced configurations, and it examines sensitivity limits of the stage resulting from the finite linearity range of the photodetector. 2.2.1.1. Acousto-Mechanical Sensitivity The acousto-mechanical stage converts pressure to a displacement. Equation 2-2 gives the sensitivity of the stage, where wo is the deflection of the membrane at the center, and p is the acoustic pressure at the center of the membrane. p w So am Equation 2-2 To derive Sam, first begin with Equation 2-3, the transverse deflection equation for a plate derived by Sheplak et al. [17].

PAGE 35

24 2 2 2 1 3 2 4 24 2 1 12 ) ( a r a k kI k I a kr I Eh k pa r wo o Equation 2-3 Figure 2-1 – Side View of De flecting Plate or Membrane. In the case of a membrane where a << kr, Equation 2-3 simplifies to Equation 2-4. 2 2 2 3 41 78 2 a r k Eh pa r w Equation 2-4 Letting wo = w(0) and substituting Equation 2-4 into Equation 2-3 produces the equation for the sensitivity of the membrane as a function of radial distance from the center. 2 2 2 3 41 78 2 a r k Eh a r Sam Equation 2-5 If we assume that the light spot on the membrane is very small (10% or less) with respect to the membrane diameter, then the sensitivity of the membrane can be lumped at w(r)Equilibrium Membrane Position Positive Plate Deflection Negative Plate DeflectionClamped Boundary

PAGE 36

25 the radial center. Equation 2-6 gives the final equation for the acousto-mechanical sensitivity of the membrane, lumped to the radial center. 2 3 478 2 0 k Eh a S Sam am Equation 2-6 If it cannot be assumed that the light spot is small, then the membrane sensitivity cannot be lumped into the center of the membrane. Sam will become a function of radial position, r, with respect to the membrane center, and the microphone sensitivity will decrease. In this optical microphone, the light spot illuminates less than 10% of the membrane. Based on the observed fiber position error (see Section 3.3.1) of less than 50 m for the fiber bundle used in this thesis, Sam can still be approximated as a constant for this membrane. The tension parameter k is determined by Equation 2-8. The in-plane stress (o) of the nitride layer for the nitride deposition process used in the microphone fabrication was reported to range between 50 MPa and 120 MPa by the MEMS Exchange website. Special fabrication instructions were given to minimize the in-plane stress in the nitride layer, so it is expected that the stress will be equal to the minimum reported value for the MEMS Exchange deposition process, o = 50 MPa. Using E = 270 GPa (for SixNy), h = 1 m, a = 1 mm, and o = 0.27 (for SixNy), we estimate Sam = 1.249 x 10-3 m / Pa. E h a ko o 21 12 Equation 2-7 For a discussion of the effects of the observed membrane linearity on Sam, see Sections 2.2.2.1 and 5.3.2.

PAGE 37

26 2.2.1.2. Mechano-Optical Sensitivity The mechano-optical transduction stage converts a mechanical displacement to an optical power coupling factor. The sensitivity of the stage is given by Equation 2-8, where w is the deflection of the membrane at the center, and is the coupled optical power of the stage in W/W. w Sam Equation 2-8 He and Cuomo [18] derived a formula for determining the power coupled by light reflecting off of a deflecting membrane in a microphone similar to that shown in Figure 1-8. The analysis is valid for multimode optical fibers. Theoretical work by Ruan and Felson [19] can be used to derive the power coupled as a function of membrane displacement for the case of a single mode transmit fiber and a multimode receive fiber, although that configuration is not analyzed here. Ruan and Felson’s work is applicable to membranes with a finite radius of curvature, however He and Cuomo’s work is not. The analysis here based on [18] assumes no misalignment errors and no power lost due to mismatch between fiber numerical apertures (NA). This is a good approximation when the angular alignment between the fiber surface and the membrane is less than 5 degrees (for fibers with NA = 0.22 or less). If this approximation does not hold, then the method of images (explained below) is not valid. Adjusting the method of images to account for angular alignments is complicated, and as of this writing, no work exists that rigorously solves the problem. Section 3.3.1 discusses types of alignment errors, methods to avoid them, and their implications in more detail.

PAGE 38

27 In general, the power coupled into an optical fiber can be determined by integrating the optical intensity (also known as the irradiance) over the collecting surface, assuming all light present is entering the fibers at an angle less than the acceptance angle of the fiber. If this is not the case, then only the irradiance due to rays entering the fiber at less than the acceptance angle should be integrated in Equation 2-12. The analysis of He and Cuomo [18] assumes the former. The re flected intensity profile at the surface of the fiber bundle is determined in [18] by the method of images. Figure 2-2 – Method of Images (View from Side of Fiber Bundle). In the method of images, the reflecting surface is defined to be the reflecting plane, and the surface of the fiber bundle (as shown in Figure 1-10) is defined to be the receiving plane. They are separated by a gap, g. The method of images states that the reflected optical power incident onto the Rx cores is the same as the optical power incident on the Rx core images, located at a distance of 2g from the receiving plane. Image Plane Reflecting PlaneReceiving Plane Transmit Fiber Core g g Receive Fiber Core Receive Fiber Core Rx Fiber Image Rx Fiber Image

PAGE 39

28 Using the method of images, He and Cuomo derived an equation for the intensity on the image plane as a function of radial distance from the center of the fiber bundle. This thesis uses Equation 2-9 through 2-11 from He and Cuomo [18] as the first step in determining the power coupled and sensitivity of the MO stage. Without an understanding of these equations, a microphone designer will not be able to identify miscalculations due to errors that have been observed in the output of Equation 2-9. This problem will be discussed in more detail later in this section. Equation 2-9 The quantity, A, used in Equation 2-9 is defined by Equation 2-10. In Equation 210, rtx_core is the radius of the transmit fiber core, and g is the equilibrium gap. g r Acore tx2_ Equation 2-10 2 & 2 & 2 1 1 1 1 1 1 ln 8 2 & 2 & 2 1 1 1 1 1 1 ln 8 2 & 2 1 & 2 1 1 1 1 ln 8 1 tan 1 tan tan 4 1 2 2 & 2 1 & 2 1 1 1 1 ln 8 1 tan 1 tan tan 4 1 2 1 & 2 1 1 tan 1 tan 1 tan 1 tan 1 2 2 & 1 0 & 2 1 1 1 1 ln 8 1 tan 1 tan tan 4 1 2 2 & 1 0 & 2 1 1 1 1 ln 8 1 tan 1 tan tan 4 1 2 1 2 & 2 1 1 tan 1 tan 1 tan 1 tan 1 2 2 0 & 2 1 1 tan 1 tan 1 ) (2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 2 2 2 2 2 1 1 1 2 1 1 1 1 2 2 2 2 2 1 1 1 2 2 2 2 2 1 1 1 2 1 1 1 1 2 1 1 2k k k k if k A k k A k A k k k k if k A k k A k A k k k k if k A A k A k A A k A A A A k k k k if k A A k A k A A k A A A A k k k if k A A k A A k k A A k k k k if k A A k A k A A k A A A A k k k k if k A A k A k A A k A A A A k k k if k A A k k A A k A A k k k if k A A k A A I r Ic c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c o

PAGE 40

29 The quantity k used in Equation 2-9 is defined by Equation 2-11. In Equation 211, rtx_core is the radius of the transmit fiber core, and r is the radial coordinate measured from the center of the Tx fiber axis. core txr r k_ Equation 2-11 The quantities c and kc are the critical angle of the Tx fiber and the critical value of k associated with that angle. For more details on the variables, see He and Cuomo’s work [18]. Some sets of input parameters with a gap, g, on the order of the Tx fiber diameter were observed to produce non-intuitive intensity profiles. For example, using Equation 2-9 at a gap of 50 m with a Tx fiber core diameter of 105 m resulted in I(r) = 0 at all values of r. Therefore, for the theoretical power coupled and sensitivity to be accurate, an optical microphone designer must plot Equation 2-9 for equilibrium gaps of the desired value. If the plots are erroneous, then the power coupled and sensitivity analysis will be invalid. The Mathcad code used to generate the intensity curves was carefully examined for errors, and none were found. It is possible that Equation 2-9 does not accurately predict the intensity at small gap distances. The power coupled into the receive fibers is determined by using a ring approximation with a power correction factor The ring approximation used in He and Cuomo [18] approximates the face of the recei ve fibers as an annular ring. The power coupled is determined by integrating the normalized intensity (Equation 2-9) over the ring area. Figure 2-3 shows the ring approximation. The actual light collection surfaces

PAGE 41

30 are shaded gray, and the integrated area of the ring approximation is shown by the dashed ring. Note that the relative sizes of the core and claddings are not necessarily to scale. Figure 2-3 – Ring Approximation Diagram. The power coupled and sensitivity equations for the mechano-optical (MO) stage are given by He and Cuomo [18]. These were the equations used by Kadirval [8] to predict the performance of the MO stage in his optical microphone. This thesis has modified the power coupled equation from [18] (referred to as the ideal power coupled from this point) to include the effects of radi al position errors in the receive optical fibers, and also to correct for overestimation of the power coupled by the ring approximation. Radial position error, RPE, is defined and discussed in Section 3.3.2. The power coupled correction factor, cf, is calculated by taking the ratio of the actual surface area of the receive fibers to the area of the ring in the ring approximation. The ideal power coupled is then multiplied by this correction f actor (which is a function of the optical fiber geometry and the RPE) to calculate the corrected power coupled. In this thesis, ideal power coupled and sensitivity refers to the case where cf = 1, meaning that the ring approximation area exactly matches the surface area of the receive fibers. Since this can

PAGE 42

31 never happen in practice, the ideal situation will occur only when the area mismatch is neglected, as is done by [18] and [8]. The corrected power coupled and sensitivity equations are given in Equations 2-12 and 2-13. kdk I g k I RPE c P P RPE gcore tx RPE b core tx RPE m o f i o _ 1, 2 Equation 2-12 RPE g dz d P P dz d RPE g Si o mo, Equation 2-13 Two theoretical corrected power coupled curves, based on using Equation 2-12 with a fiber bundle constructed from AFS105/125Y multimode fibers as both Tx and Rx, are shown in Figure 2-4. One of the curves corresponds to an RPE of 0 m, and the other corresponds to an RPE of 10 m (the observed RPE of the custom fiber bundle). The corresponding corrected sensitivity curves are shown in Figure 2-5. The horizontal axis on each plot is the equilibrium gap, g, between the receiving plane and the reflecting plane in the method of images. The maximum corrected theoretical MO sensitivity with RPE = 0 m is Smo = 1.094E-3 m-1 and occurs at g = 230 m. When the power coupled correction factor is taken into account, the maximum corrected theoretical power coupled is Smo = 0.784E-3 m-1 and occurs at g = 265 m.

PAGE 43

32 0% 5% 10% 15% 20% 25% 30% 01002003004005006007008009001000Gap (um)Power Coupled (W/W) RPE = 0 um RPE = 10.0 um Figure 2-4 – Theoretical Power Coupled with Ideal Fiber Configuration. -4.00E-04 -2.00E-04 0.00E+00 2.00E-04 4.00E-04 6.00E-04 8.00E-04 1.00E-03 1.20E-03 01002003004005006007008009001000Gap (um)Sensitivity (1 / m) RPE = 0 um RPE = 10 um Figure 2-5 – Theoretical Sensitivity with Ideal Fiber Configuration.

PAGE 44

33 The discontinuities observed in the sensitivity equations are due to transitions in the power coupled integral. Specifically, each discontinuity corresponds to the edge of the light cone crossing the boundary of the receive fiber ring. The discontinuity is present in the power coupled equations, but since it manifests in these plots as an integral of the discontinuity seen in the sensitivity cu rves, it is difficult to see on the viewing scale of the power coupled plot. 2.2.1.3. Opto-Electrical Sensitivity The opto-electrical stage converts an optical power coupled to an electrical signal. This is accomplished with the use of a Thorlabs PDA400 photodetector, which consists of a photodiode and a trans-impedance amplifier with five gain settings. In this thesis, the photodiode and trans-impedance amplifier are collectively referred to as a photodetector, and they are treated as one unit. The sensitivity of the stage is given by Equation 2-14, where is the optical power coupled, and V is the output voltage of the sensor. d dV Soe Equation 2-14 The output voltage of the opto-electrical stage is a function of the detection electronics and the detection method used. For the Unreferenced Output detection technique, shown in Figure 2-6, the output vo ltage is a function of the photodetector responsivity and gain, the output amplifier gain, and the laser power.

PAGE 45

34 Figure 2-6 – Block Diagram of the Un referenced Output Configuration. The output voltage of the unreferenced output configuration, Equation 2-15, is derived by applying Kirchoff’s and Ohm’s Laws to the equivalent circuit of the detector, shown in Figure 2-7. Since a voltage amplifier is connected in series with the detector, the output amplifier gain, Ga, is multiplied by Vdet_out to get the microphone output voltage, Vout. Pout is the received optical power from the fiber bundle Rx fibers. By equating Pout with times Pin, Equation 2-15 neglects losses in the fiber bundle other than those from the power coupling effect. For a real bundle, other losses are present at the connectors and in the fibers themselves. These losses have not been measured and are neglected here. in a out a outP RGG P RGG V Equation 2-15 Photodetector Lowpass Filter / Output Amplifier Photodiode Trans-impedance Amplifier Pout Vout R G Ga

PAGE 46

35 Figure 2-7 – Equivalent Circuit for the PDA400 Photodetector. Substituting Equation 2-16 into Equation 2-15 gives the equation for the electrooptical sensitivity of the Unreferenced Output detection technique. in a oeP RGG S Equation 2-16 Equation 2-16 shows a linear relation between the received optical power and the OE stage sensitivity. This linear relationship only holds when the photodetector is operated in a linear region. A Thorlabs PDA-400 photodetector, with specs given in Appendix D, has a peak response of 0.95 A/W at 1550 nm. The minimum transimpedance gain setting, G, for the PDA400 is 15,000 V/A. Using the detector response R = 0.95 A/W and the gain G = 15000 V/A, gives Soe = 14250(V/W)*Pin. If Pin = 0.7 mW, then Soe = 9.975 V. Since the photodetector saturates at 10 V, the maximum unreferenced OE stage sensitivity is 9.975 V Ga. It is very important to observe that the overall sensitivity of the unreferenced optical microphone is limited by the maximum DC optical power received by the photodetector due to detector saturation. Id eally, the photodetector would consist of only + Pout R Pout Gdet + Vdet_out Photodiode Trans-impedance Amplifier

PAGE 47

36 a photodiode, and a highpass filter would be placed at the photodiode output. Without the trans-impedance amplifier, the DC optical power can be removed before amplification, eliminating the limit of the OE sensitivity due to the DC optical power. Any sensitivity lost from removing the trans-impedance amplifier can be recovered by increasing the gain of the output amplifier. The Referenced Output OE Stage is more useful in a microphone system due to the invariance of the sensitivity with optical power, which will be proven here. Figure 2-8 is a block diagram for the referenced OE microphone configuration. By using the equivalent circuit of the photodetectors and by inspecting the block diagram, Equation 217 can be derived for the microphone output voltage, Vout. ref ref ad a ref ref ad a outP P G G G G P RG P RG G G Vmod mod mod mod Equation 2-17 Figure 2-8 – Block Diagram of the Referenced Output Configuration. Highpass Filter / Output Amplifier Vout Modulated Photodetector (MOD) Photodiode Trans-impedance Amplifier Pmod R Gm od Ga Photodiode Trans-impedance Amplifier Pref R GrefAnalo g Divide Circuit Gad Vmod Vref Reference Photodetector (REF) Vm od Vr e f

PAGE 48

37 In Equation 2-17, Pmod is identical to Pout in the unreferenced block diagram, when the same assumptions are made. If is the optical losses present in the reference signal path, then Pref can be represented as (1-) Pin. By observing that the fiber bundle power coupled, is Pout / Pin, Equation 2-17 can be rewritten in terms of the component gains and the optical power coupled into the fiber bundle. ref ad a outG G G G Vmod1 Equation 2-18 Substituting Equation 2-18 into Equation 2-15 gives the equation for the electrooptical sensitivity of the Referenced Output detection technique, where Gratio is the ratio of Gmod to Gdet. ratio ad a ref ad a oeG G G G G G G S 1 1mod Equation 2-19 Equation 2-19 varies directly with the ratio of the modulated detector gain to the reference detector gain (Gmod / Gref), with the built-in gain of the analog divide circuit (Gad), and with the gain of the output amplifier (Ga). Also, increasing optical losses in the reference path, increases the sensitivity of the OE stage and decreases the optical power incident on the photodetector. Later in this chapter, it will be shown that optical reference path losses will improve the electronics noise and microphone minimum detectable signal under some conditions. Using two PDA-400 photodetectors, an analog divide circuit hardwired for a gain of 10 V, and the minimum and maximum values of Gratio (based on available PDA400 gains settings), the sensitivity of the OE stage can range between Soe_min = (0.10 V)*Ga /

PAGE 49

38 (1) and Soe_max = (1000 V)* Ga / (1). If we take the ratio of the sensitivity of the referenced to the unreferenced OE stage, we can see how the stage sensitivities compare at varying input power levels. This is done in Figure 2-9 for Gad = 10 V and Ga = 1 V / V. From Figure 2-9, it can be concluded that the unreferenced microphone will have a lower sensitivity than the referenced microphone unless the laser is operated at the maximum power for which the photodetector remains linear, no reference path losses are present, and the photodetector gain ratio is one. When this occurs, the two stages will have identical sensitivities. Increasing Gratio and will increase the sensitivity of the referenced OE stage. These values are physically limited by the photodetector range of linearity for Gratio and and also by the analog divide circuit for Specifically, the input to the analog divide circuit must never drop below a certain threshold, and the output of the analog divide circuit can never saturate. This limits to less than 0.9 for the AD circuit used in this thesis. In practice, the PDA400 detectors are not useful when the gain is set higher than 47,000 V/A. An additional constraint is the fixed gainbandwidth product limiting the maximum gain for a minimum bandwidth.

PAGE 50

39 0 1 2 3 4 5 6 7 8 9 10 0100200300400500600700Laser Power ( W)Soe_ref / Soe_unref Gratio = 1, Gdet = 15000 V/A, Alpha = 0 Gratio = 3.1, Gdet = 15000 V/A, Alpha = 0 Gratio = 1, Gdet = 15000 V/A, Alpha = 0.5 Figure 2-9 – Comparison of Unreferenced and Referenced Output Sensitivities. 2.2.2. System Linearity The linearity of the optical microphone is determined by the linearity of the individual stages. The acousto-mechanical stage linearity is governed by the nitride membrane. The linearity of the mechano-optical stage is dominated by the assumptions that the membrane curvature is negligible, and by the local flatness of the sensitivity vs. equilibrium gap curve. The linearity of the opto-electrical stage is governed by the linearity range of the photodetector and additional electronics. The following sections establish the conditions for linearity of each stage, and therefore the whole device.

PAGE 51

40 2.2.2.1. Acousto-Mechanical Linearity The theory for the range of linearity of the membrane was investigated by Sheplak and Dugundji [20]. Using this theory Sahni et al. [21] determined that the diaphragm is linear over the region from 0 – 2000 Pa (160 dB re. 20 Pa). Sheplak et al. [20] present Equation 2-20, which relates the membrane aspect ratio to the in-plane stress and the maximum linear pressure (3% linearity). E p h ao max 2 3 max Equation 2-20 By substituting the maximum linear pressure and the microphone membrane dimensions into Equation 2-20, the in-plane stress, o, of the membrane can be estimated. The experimental linearity range of the microphone (see Chapter 5) is reached at 122 dB (re. 20 Pa). This results in the AM sensitivity increasing by a factor of 15. This effect is considered when predicting the microphone performance in Section 2.2.7. 2.2.2.2. Mechano-Optical Linearity The linearity of the mechano-optical stage is based on two factors: the linearity of the power coupled curve (flatness of the se nsitivity curve), and the assumption of the membrane curvature being negligible. The point of interest for the linearity analysis is about the point of maximum sensitivity. Figure 2-10 shows a plot of the sensitivity from an equilibrium gap of 240 m to 290 m using Equation 2-13. The vertical axis of the curve is highly magnified, and the thin horizontal lines denote the region that is within 3% of the maximum sensitivity. It is evident from Figure 2-10 that the MO stage sensitivity is linear within

PAGE 52

41 +/-10 m of the optimal gap. Since the largest maximum membrane deflection, at Pmax = 2 kPa, allowed by the variability in the nitride stress of the MEMS chip process is +/-2.49 m, the sensitivity of the MO stage will be constant within 3%. The large window for linearity holds at equilibrium gaps out to 400 m. This means that if the equilibrium gap is set at a value larger than the best case equilibrium gap, then the sensitivity will still be constant to within 3% over the range of the diaphragm motion. 5.00E-04 6.00E-04 7.00E-04 8.00E-04 9.00E-04 1.00E-03 240245250255260265270275280285290Gap ( m)Corrected Theoretical Sensitivity (1 / m) Figure 2-10 – Linearity of Mechano-Optical Stage. The second criterion for establishing the linearity of the MO stage is the flatness of the membrane over the illuminated region. An alternate way of viewing this requirement is to look at the acousto-mechanical sensitivity of the membrane over the illuminated region as a function of radial distance from the center. For linearity, the MO sensitivity of the membrane should vary by no more than 3%. It is important to note that the method of images used in Section 2.2.1.2 requires a flat membrane in the illuminated

PAGE 53

42 area. This means that the membrane must be within 3% of planar over the illuminated region, and must be parallel to the fiber bundle face. Figure 2-11 shows the normalized acousto-mechanical sensitivity, using Equation 2-6, of the membrane as a function of the radial distance from the membrane center. This plot shows that the sensitivity is within 3% of the maximum when the illuminated area is within 93 m of the membrane center. For optical fibers with an NA = 0.22, the spot radius is less than 93 m when the gap is less than 430 m. 0.76 0.84 0.92 1.00 1.08 1.16 1.24 0102030405060708090100 Radial Distance from Membrane Center ( m)Normalized Acousto-Mechanical Sensitivity Figure 2-11 – Plot of Acousto-Mechanical Sensitivity as a Function of Radial Position. It has been determined that the MO stage sensitivity varies by no more than 3% over the equilibrium gap range of 200 m to 400 m. It has also been determined that the acousto-mechanical sensitivity at every illuminated point on the membrane is within 3% of its maximum value when the equilibrium gap is less than 430 m. Therefore, the MO stage of the optical microphone is linear when the equilibrium gap is between 200

PAGE 54

43 m and 400 m. Since the ideal equilibrium gap is 230 m, the OM stage of the optical microphone is linear at the equilibrium gap of 230 m with a maximum deflection of +/2.49 m at Pmax = 2 kPa. This analysis does not include the effect of misalignments on the linearity. They are discussed in Chapter 3. 2.2.2.3. Opto-Electrical Linearity The linearity of the opto-electrical stage is effectively limited by the linearity of each electronics component in the system. The PDA-400 photodetectors are linear up to an output voltage of 10 V. The AD734 analog divide chip is linear over the input range of +/12.5 V. Therefore, the photodetectors are the limiting factor in determining the OE stage linearity. As long as the detectors are operated below saturation, the opto-electrical stage is linear. The maximum voltage that can be output in the linear range of operation for the electronics serves to limit the maximum sensitivity of the unreferenced output configuration when the output amplifier is held at a fixed gain. To illustrate, consider Equation 2-21, the equation for the photodetector output in terms of the optical power for the unreferenced OE configuration. laser outRGP V Equation 2-21 The maximum Vout that the PDA400 photodetector can output is 10 V. The responsivity is 0.95 A/W. Substituting these c onstants into the above equation shows that the product of the gain and the optical power cannot be larger than the saturation voltage of the detector divided by the detector responsivity, or 10.5 W*V / A for the PDA400.

PAGE 55

44 This product effectively limits the maximum sensitivity that the unreferenced OE stage can provide, since increasing Plaser requires G to be reduced if Plaser*G > 10 V. When the range of linearity of the measurement equipment is considered, then the gain of the output amplifier is also limited. To ensure the microphone can operate in its intended environment, the linearity range of the measurement equipment must also be considered. 2.2.3. System Frequency Response The frequency response of the optical microphone was discussed in depth by Kadirval [8]. He developed an equivalent circuit for each stage of the optical microphone. The frequency response of the microphone is the product of the frequency response of the individual stages. 2.2.3.1. Acousto-mechanical frequency response The lumped element model parameters used in the acousto-mechanical stage are shown in Table 2-1. These parameters were given by Sahni [21]. The frequency response analysis for this optical microphone membrane was presented by Kadirval [8], and it remains valid. Using the lumped element parameters in Table 2-1 and the equations for the frequency response of the AM stage, the 3 dB frequency for the upper end of the frequency response was 76.25 kHz. The lumped element approximation ia limited to below 50 kHz according to Sahni [21], so the upper limit of the frequency response will not be accurately predicted by this model. 2 2 _1 3 1 s C M a C s Heff eff eff unorm am Equation 2-22

PAGE 56

45 Hz j H s H s Hnorm am norm am norm am0 2_ _ Equation 2-23 Table 2-1 – Acousto-Mechanical Lumped Element Parameters Parameter Formula Value Units Mmea 23 a hn 1316 kg / m4 Cmea h ao 24 1.963E-15 m3 / Pa Mrad 23 aa 5.199E+5 kg / m5 Ca 2 2c h aa cav 1.246E-15 m3 / Pa fres n oa 39 0 98.6 kHz The lower end of the frequency response is governed by the vent channel of the microphone. Although the optical microphone was not designed with a vent channel, the steel tube is not an exact fit with the membrane cavity, and allows equalization of the pressure on both sides of the membrane due to low frequency pressure fluctuations. 2.2.3.2. Mechano-optical frequency response The mechano-optical frequency response is a constant. This was discussed by Kadirval [8] in detail, and will not be reproduced here. 2.2.3.3. Opto-electrical frequency response The opto-electrical frequency response is determined by the frequency response of the electrical components in the system. The photodetector has a worst case bandwidth of 50 kHz at the maximum gain and a bandwidth of 10 MHz at the lowest gain. The analog divide circuit has a bandwidth of 10 MHz. Therefore, the bandwidth of

PAGE 57

46 the photodetectors is the limiting electrical component for the upper range of the frequency response for practical gain settings. The lower range of the frequency response is limited by the cut-on frequency of the highpass filter. In the highpass filter used by this thesis, the cut-on frequency was set at 30 Hz. 2.2.4. System Electronic Noise The electronic noise is defined to be the noise, in volts, seen at the output of the microphone circuit. Previous works (He & Cuomo, 1991 [16], [22]) assume that “the light source does not contribute significantly to the (noise) background.” He and Cuomo note that these conditions do not always hold, and that voltage fluctuations due to light source noise may affect the signal. This thesis includes the effects of the detector, the light source intensity noise, and the electronics to determine the electronic noise at the microphone output. The noise analysis traces the path of RMS noise signals, in V / Hz, through both the referenced and unreferenced electronics. The goal of the first part of the electronics noise analysis is to identify and derive equations for the individual noise sources that contribute to the electronics noise of the photodetector. Figure 2-12 – Noise Contributions for the Photodetector Output. Photodetector Pdet_NEP, Plight_noise R*G Vdet, Vlight Photodiode NEP Terms Photodetector Electronics Noise Terms

PAGE 58

47 Figure 2-12 illustrates the noise sources th at produce electrical noise at the output of the photodetector. Pdet_NEP is the noise equivalent power at the input of the photodiode due to the inherent noise of the photodiode (dominated by thermal and shot noise …equations presented are referenced to Wilson and Hawkes [23]). The units for Pdet_NEP are W / Hz. From this point on, the photodetector output due solely to Pdet_NEP is referred to as the detector (or photodetector) electronics noise, Vdet, and has the units V / Hz. Thermal noise and shot noise are the dominant noise sources in a photodetector. The shot noise, Vshot, in V / Hz, is given by Equation 2-24, where G is the transimpedance gain, R is the diode responsivity, e is the charge of an electron, and Plight is the incident optical power received by the detector. 2 s hotlightVGeRP Equation 2-24 Thermal noise, Vtherm, is also present in a photodetector, and can be given by the following equation, where k is Boltzman’s constant, T is the mean temperature in Kelvin, and G is the trans-impedance gain. 4thermVkTG Equation 2-25 The total detector noise, Vdet, can be written as the RMS sum of the thermal and shot noise, shown in the following equation. 22 det s hotthermVVV Equation 2-26 The other noise source, Plight_noise, represents the noise power fluctuation incident on the photodetector due to fluctuations in the intensity of the light source integrated over the diode collection area. The units for Plight_noise are W / Hz. The voltage fluctuations

PAGE 59

48 at the output of the detector produced by Plight_noise are referred to in this thesis as the light source (or laser) electronics noise, Vlight. The units are the units V / Hz. Where necessary, subscripts will be added to distinguish between referenced and unreferenced quantities. Since Plight_noise is highly dependant on the type of light source and supporting electronics, a purely theoretical model will not be used (see the end of this section). Instead, the electronics noise at the output of the photodetector will be measured. The measured quantity will be equal to the RMS sum of Vlight_noise and Vdet. By using Equations 2-24 – 2-26 and the experimentally measured noise quantity, Vlight_noise can be determined for the light source – detector system. When Vlight_noise is known, the photodetector equivalent circuit can be used to calculate Plight_noise. Dividing Plight_noise by the photodiode active area produces an estimate of the light source intensity noise. Figure 2-13 – Noise Contributions for the Microphone Output. The second part of the electronics noise analysis propagates Vdet through the unreferenced and referenced configurations to determine the electronics noise at the output of the microphone. These noise terms are referred to as the “total electronics noise” or the “microphone electronics noise” terms. The total electronics noise is determined by the amplification of Vdet by signal amplification components and the addition of any additional electronics noise sources (see Figure 2-6 and Figure 2-8). For Microphone OE Stage Pdet_NEP, Plaser_noise Unref / Refvnoise_ref, vnoise_unrefPhotodiode NEP Terms Microphone Electronics Noise Terms

PAGE 60

49 this work, the analog divide IC used by the referenced configuration was the Analog Devices AD734 [24]. The electronics noise analysis of the unreferenced microphone configuration in Figure 2-6 is not difficult since only one sign al path exists. By inspecting Figure 2-6, Equation 2-27 and 2-28 can be derived for the total noise output at the photodetector, Vdet_total and the unreferenced noise, Vunref. In Equation 2-28, Va is the input noise of the amplifier, in V / Hz. 2 2 det_detdet_ totallightnoiseVVRGP Equation 2-27 22 det_ unrefatotalaVGVV Equation 2-28 The analysis of the referenced microphone noise is more complicated. There are two signal paths into the system, and a time-domain division. To complicate matters, the optical noise Vlight_noise in the modulated and reference signal paths (see Figure 2-8) may or may not be correlated. In a real microphone system, it is expected that the optical noise in the modulated and reference signal paths will be correlated to some degree. Theoretical models are presented here for both uncorrelated and correlated optical noise in the referenced configuration. Vlight_noise_ref and Vlight_noise_mod are given by Equations 2-29 and 2-30, which are derived from the power to voltage conversion equations of a photodetector. ___1 2lightnoiserefreflightnoiseVRGP Equation 2-29 __modmod_2lightnoiselightnoiseVRGP Equation 2-30

PAGE 61

50 When Vlight_noise_ref and Vlight_noise_mod are uncorrelated, they will not divide in the analog divide circuit. Instead, they add as shown in Equation 2-31 for the uncorrelated analog divide circuit output, Vad_uncorr. In Equation 2-31, Vad is the input noise of the analog divide circuit. 2222 _det____mod22aduncorradlightnoisereflightnoiseadVGVVVV Equation 2-31 The total electronics noise at the output of the referenced microphone for the uncorrelated case can be determined by propagating Vad_uncorr through the output amplifier, with input noise Va, as shown in Equation 2-30. 22 __ refuncorraaduncorraVGVV Equation 2-32 When the optical noise is correlated, then it will be divided by the analog divide circuit. In this case, Vad_corr is given by Equation 2-32. 2 __mod 22 _det __22lightnoise adcorradad lightnoiserefV VGVV V Equation 2-33 The total electronics noise at the output of the referenced optical microphone when the optical noise is correlated can be determined by replacing Vad_uncorr with Vad_corr in Equation 2-32. Equation 2-34 gives this result. 22 __ refuncorraadcorraVGVV Equation 2-34 Theoretical models for the intensity noise of a laser are inaccurate in predicting the performance of a commercial laser, due to the uncertainty in the laser quantum efficiencies, and the variation in the performance of the electronics controlling and

PAGE 62

51 cooling the laser. Coldren and Corzine [25] present Equation 2-41 for the relative intensity noise, RIN, of a laser source at the optical resonance frequency (corresponding to the laser output wavelength). The dampening factor, is defined as Kf2 R + and is not a parameter provided by a laser manufacturer. Experiments can be performed to quantify it, but these are not practical to do when the direct measurement of the optical noise PSD will provide the needed value. STf RIN 16 Equation 2-35 In order to make an accurate estimate of the light source noise, this thesis does not rely on Equation 2-41. The results of the experimental measurements for Plight_noise and Vdet are given in Chapter 5. 2.2.5. System Minimum Detectable Signal The minimum detectable signal is the smallest signal that can be resolved by the microphone. The minimum detectable signal is a function of physical interactions between noise sources and desired signals in each stage as well as the sensitivities of each stage. Previous work (He & Cuomo, 1991 [16]) only considered the OE Stage MDS, and also ignored the laser noise effect on the total electronics noise. This work takes additional noise sources into consideration (see Section 2.2.4). Equation 2-42 gives the System MDS for the optical microphone. Three physical MDS reducing mechanisms are considered: membrane noise due to the Brownian motion of the membrane atoms, variations in the coupled laser power due to laser intensity noise in the MO stage, and electronic noise in the OE stage. Each of these terms will be discussed in more detail later in this section.

PAGE 63

52 2 2 2 lightnoise light oe r ammoP P V MDSpf SSS Equation 2-36 The first term under the radical in Equation 2-42 is the AM stage MDS due to the Brownian noise of the membrane, given in Pa2. The second term is the MO stage MDS, which is due to the variations in the coupled laser power due to laser intensity noise, in the presence of variations due to desired membrane deflection (from an acoustic signal), given in Pa2. If the laser intensity noise is close in magnitude to the change in coupled power in the Rx fiber, then the signal can never be distinguishable from the optical noise, even with an ideal (no-noise) detector. The third term is the MDS of the OE stage, also given in Pa2. The OE stage MDS is determined by the total electronics noise and the measurement bin width. In previous works, the MO MDS was neglected [16, 22]. No previous work takes the light source electronics noise into account when calculating the OE MDS; they only consider the photodiode shot noise. 2.2.5.1. Acousto-Mechanical Minimum Detectable Signal The dominant noise source of the AM stage is the Brownian motion of the membrane. The silicon nitride membrane atoms, like all atoms above absolute zero, exhibit Brownian motion. This Brownian mo tion causes a deflection of the diaphragm in the same manner that an acoustic signal does. Due to this deflection, pressures which cause a deflection smaller than the deflection due to Brownian motion are not detectable. The equation for the mean equivalent pressure fluctuations due to the Brownian motion is given by Equation 2-43 from Chau and Wise [26].

PAGE 64

53 1122 2 232rmpmp kT p f a Equation 2-37 Using = 1, T = 300 K, k = 1.38 E-23 J / K, m1 = m2 = 4.78 E-26 kg, p1 = p2 = 101.4 kPa, and a = 500 m, the mean equivalent pressure fluctuations due to Brownian motion is = 3.642 E-11 Pa2 f. This corresponds to a MDS of -11 dB (re. 20 Pa). Based on the MDS of the other stages (presented later), it can be conc luded that this noise source is completely negligible in an intensity-modulated optical microphone. 2.2.5.2. Mechano-Optical Minimum Detectable Signal The dominant noise source in the MO stage is the intensity noise of the light source. The physical effect is illustrated in Figure 2-14. Power is coupled by light reflecting into the receive fibers from the membrane. This power can be written as the sum of three components: the optical DC component (Plight), the acoustically-modulated optical power signal (Pmodulated(t)), and the optical noise power (Plaser_noise(t)). Figure 2-14 – Illustration of the Physics Behind the MO MDS. When the membrane deflects due to an acoustic signal, Pmodulated(t) is produced. Fluctuations in the light power output superimposes Plight_noise(t) onto the desired modulated signal. If the fluctuations in power coupled due to light output noise are larger than the fluctuations in power coupled due to the acoustic signal (Plight_noise(t) >= Time Variance of Coupled Power Components Plaser + Plight_noise(t) + Pmodulated(t) Rx Fibers (more than one)

PAGE 65

54 Pmodulated(t)), then the microphone will not be able to detect the acoustic signal without the optical noise removed by the electronics. In the event that the optical noise can be removed by the electronics (see Section 2.2.4), the MO MDS effect will not be present in the microphone. The equation for the contribution (when present) of the MO stage to the MDS is: lasernoise laser MO ammoP P MDS SS Equation 2-38 The MO MDS will dominate the system MDS if the following equation is satisfied and if the optical noise cannot be removed from the system by the electronics (as is done by the referenced optical microphone when the optical noise is correlated): noiseoe laseroePV PS Equation 2-39 2.2.5.3. Opto-Electrical Minimum Detectable Signal Noise mechanisms in the OE stage are due to the electronics, the shot noise of the photodetector, and the detected noise power of the laser. Lasers operated in constant current mode are noisy, and will usually dominate the electronics noise. The equation for the OE stage contribution to the MDS is the following, where Voe_noise is the electronics noise and f is the measurement bin width of the electronics noise. oenoise OE ammooeVf MDS SSS Equation 2-40 If the inequality in Equation 2-45 does not hold, then the OE stage MDS will dominate.

PAGE 66

55 2.2.6. Optical Reference Path Losses and System Performance Metrics In this section, the effect of optical losses in the reference signal path in the referenced OE configuration will be analyzed. The effect of optical reference path losses on sensitivity, electronics noise, and MDS will be discussed. 2.2.6.1. Sensitivity and reference path losses In Equation 2-50, the referenced microphone sensitivity is presented, with Soe expanded to show the dependance on ref ad amp mo am refG G G G S S Smod1 Equation 2-41 The losses in the reference path in the referenced opto-electronic stage reduce the voltage seen by the denominator input of the analog divide circuit. Decreasing the denominator increases the circuit output, so adding optical losses to the reference path will increase the sensitivity of the microphone. For the unreferenced microphone, there is no reference path, so is undefined. 2.2.6.2. Electronics noise and reference path losses In Equations 2-29, 2-31, and 2-32, it was shown that the electronics noise for the referenced microphone is dependant on (1 – ) when the optical noise is uncorrelated. Therefore, increasing with uncorrelated optical noise will decrease the electronics noise. If the photodetector noise dominates, then the effect of increasing will be negligible, but if the light source noise dominates, then increasing can significantly improve the referenced electronics noise. Therefore, when the optical noise in the

PAGE 67

56 reference and modulated signal paths is perfectly uncorrelated, optical losses in the reference path are desirable. Equations 2-29, 2-33, and 2-34 show that the referenced electronics noise is dependent on (1 – )-1 when the optical noise is perfectly correlated. In this case, increasing the losses in the optical path will actually increase the electronics noise (in the same manner sensitivity is increased). Therefore, when the optical noise in the reference and modulated signal paths is perfectly correlated, optical losses in the reference path are undesirable. A real microphone system is likely to have some correlation between the optical noise in the reference and modulated signal paths, but the extent of the correlation in general is unknown. Therefore, it is possible for losses in the reference path to increase or decrease the referenced electronics noise. It is expected that a referenced microphone with a small correlation between optical noise signals will receive some benefit from reference path losses, while a referenced microphone with a large correlation between optical noise signals will have its noise floor slightly worsened. 2.2.6.3. Minimum detectable signal and reference path losses The referenced optical microphone MDS is strongly dependent on whether the optical noise is correlated or not. If the noise is correlated, then the MO MDS does not factor into the total MDS, since optical noise fluctuations are completely removed. If increases and the optical noise is correlated, both overall sensitivity and electronics noise will increase at approximately the same rate. Therefore, optical path losses will not affect the MDS when the REF and MOD optical noise is perfectly correlated.

PAGE 68

57 If the noise is perfectly uncorrelated, then overall sensitivity will increase and electronics noise will decrease. Therefore, optical path losses will improve the MDS when the REF and MOD optical noise is perfectly uncorrelated. Note that since the MO MDS effect is not present for the correlated case, the total MDS can be much lower for the correlated case than for the uncorrelated case. 2.2.7. Summary of Predicted System Performance Due to the quantity of theoretical data presented in previous sections of this chapter, the best realistic device performance metrics will be summarized here. No design has a problem with the frequency response since the most limiting component is the photodetector with the maximum gain, which has a bandwidth of 50 kHz. Since the lumped element model approximates the membrane for f < 50 kHz, this will be used as the upper limit of the frequency response, even though electronics may be capable of higher frequencies. The lower limit of the frequency response will be 30 Hz, which is the cut-on frequency of the highpass filter used in the OE stage. The reference path optical losses in Table 2-2 were experimentally measured. The performance specifications in Table 2-3 are those that the optical microphone is expected to have when it is experimentally characterized. Table 2-2 – Summary of Configuration Settings for Theoretical Performance Metrics Configuration Mod Detector Gain (V / A) Ref Detector Gain (V / A) Laser Output Power (W) Amplifier Gain (V / V) Ref Optical Path Losses (W / W) Unreferenced Amplified Output 15,000 N/A 350 1 N/A Referenced Amplified Output 15,000 15,000 350 1 76%

PAGE 69

58 Table 2-3 – Summary of Theoretical System Performance Metrics Configuration Sensitivity (mV / Pa) MDS (dB re. 20 Pa) Electronic Noise (V / Hz) Unreferenced Amplified Output 0.073 73.8 0.19 Referenced Amplified Output 0.612 64.4 14.3

PAGE 70

59 CHAPTER 3 DESIGN OF THE OPTICS FOR THE MEMS OPTICAL MICROPHONE This chapter examines the selection process for the optics design for the MEMS optical microphone. The design of the optics must be done in parallel with the microphone package. Some optics required by the optical microphone are optical fibers, a light source, an optical splitter, photodetector s, and opto-isolators. Other miscellaneous components that are needed are connectors, protective tubing for the optical fibers, and packaging the MEMS optical microphone. This selection process is vital to the bundle performance and to the feasibility and robustness of the microphone package. 3.1. Selection of the Optics There are many factors which were considered in the selection of the optics. The most important factors are device performance, system connectivity, ease of handling, and cost. 3.1.1. Performance Device performance considerations are the most important factor for selecting the optics used in the optical microphone. Performance specifications for the optical microphone were listed and a theoretical model was derived in Chapter 2 for the system sensitivity, minimum detectable signal, frequency response, and dynamic range of linearity. Assumptions and simplifications were part of the theoretical model of the device performance. Therefore, the microphone optics must have is the ability to adhere

PAGE 71

60 to the theoretical assumptions under the widest possible range of considerations. Ideally, the selection of the optical components will guarantee the validity of the assumptions. Practically, the optical components must be selected to minimize deviations from any inherent assumptions (discussed in Chapters 1 and 2) and the non-idealities included in model formation. As shown in Chapter 2, MDS (in Pa or dB) is dependent on the individual noise contributions of the light source, the membrane, and the electronics, as well as the overall sensitivity and the product of Sam and Smo. Therefore the components should be selected such that their combined contribution to MDS is equal to or below the desired minimum detectable signal. To determine this, their effects on the sensitivity and noise floor must be known in advance! Since sensitivity is dependent on received optical power in some microphone configurations, light sources with high power are usually more desirable than sources with low power. However the MDS may not be improved, since using a light source with more power can cause the laser intensity noise to rise to unacceptable levels. Increasing the numerical aperture of the optical fibers used for the fiber bundle increases the MO stage sensitivity. Small core fibers have higher sensitivity than large core fibers. Focusing optics can also be used to provide large MO stage sensitivity increases. Misalignments and reflection losses can decrease the sensitivity and also potentially lead to laser instability. Large core fibers should not be coupled (via standard connectors) to small core fibers, since large power (and sensitivity) losses will result. Factors that influence sensitivity and MDS also can affect the device linearity. Increasing the output of the light source without bound will eventually saturate the detection electronics. By varying the optical fiber numerical aperature, NA, in Equations

PAGE 72

61 2-9 through 2-13, it was found that using opti cal fibers with high numerical apertures compresses the power coupled and sensitivity curves towards zero, reducing the range of linearity of the MO stage when compared with the linearity using a low NA fiber. 3.1.2. System Connectivity The issue of system connectivity must be considered when selecting optical fibers and optical equipment. Specifically, it must be possible for two fibers which are to be connected to each other to be connectorized with compatible connectors. Free space coupling mechanisms are possible as a last resort, but they are undesirable since they allow ambient light to be coupled into the system, are difficult to align, and are sensitive to vibrations. Finally, a design that requires multiple fibers in one connector must take into account the available connector sizes when choosing the size and types of fibers that will be used. 3.1.3. Ease of Handling and Manufacturability Manufacturability of a device is as important to overall success with the microphone as the performance. No matter how good the predicted performance of a device, if it cannot be built efficiently and eff ectively, then it will not be useful. During fabrication of the fiber bundle, fibers must be stripped, inserted into a steel tube, and generally exposed to rough handling. In general, the smaller the diameter of the fibers, the more difficult any handling with them becomes. The yield of the process for producing fiber bundles is lower for smaller core fibers, therefore, an improved process for producing small core fiber bundles is required.

PAGE 73

62 3.1.4. Cost Finally, optical fibers and other optical components are expensive. For an optical microphone to be competitive with a capacitive microphone, the cost of each component must be minimized. Components should be chosen such that the least expensive component that satisfies the specifications is used. Although this is an intuitive statement, its application is not always easy. Packaged lasers with built in control electronics and ultra low noise floors are expensive, but they may improve the device performance significantly. However, if th e photodetector noise and MDS cannot match the laser, then the cost of the optical microphone will be needlessly high. Alternately, it is possible to buy fiber pigtailed lasers at communications wavelengths (1550 nm) that are (relatively) inexpensive. If care is not taken to protect these lasers from static electricity and thermal effects, then the lasers will have a high rate of failure, and the cost of operation of the optical microphone will again become needlessly high. Based on the sensitivity, noise and MDS analysis from chapter 2, it is best to select lasers with the highest signal-to-noise ratio. More output power is not always better, considering that the MO MDS is based on the laser SNR, while the OE MDS is based on the laser noise power, and can be worsened by increasing the laser power even if the laser SNR remains constant. Also, OE configurations are available which eliminate optical power from the sensitivity equation. Photodetectors with a high gain–bandwidth product and low noise are desirable. High photodetector built-in gain is undesirable, since the DC component of the optical signal will result in detector saturation at low power levels, limiting the sensitivity. Also,

PAGE 74

63 an intensity-modulated optical microphone will have a large dc component, so high gain detectors will saturate. A low-noise amplifier and a high-pass filter at the output of the system will recover any gain lost by low gain photodetectors or low power lasers. 3.2. Selection of the Tubing The protective steel tubing is used to protect and align the optical fibers in the MEMS chip cavity. It provides mechanical support to the fibers and the mounted MEMS chip and it isolates the fibers inside the tube from the acoustic field under test. The protective steel tubing must protect the fibers, mount the microphone chip, and properly align the fiber bundle. The fiber bundle is assumed to be tightly packed and aligned with the center of the membrane. Id eally, a tube with an inner diameter (ID) equal to the fiber bundle diameter and an outer diameter (OD) equal to the MEMS cavity diameter is used. This topic is discussed in more detail later in this chapter. The material selected for the tubing was steel, because acoustic impedance of the protective covering for the fiber bundles needs to be much larger than the acoustic impedance of the test medium (air, in this case). If the acoustic impedance of the tube was not much higher than air, then sound could penetrate the tube and cause a displacement of the fibers in the fiber bundle. This is undesirable, since the theoretical characterization of the optical microphone requires the fiber bundle face to be fixed. In an environment where the acoustic impedances of the tube and the medium are more closely matched, a more sophisticated model should be used to account for possible movement of the fibers in the tube.

PAGE 75

64 Another reason for the selection of steel as the tubing material is its ability to protect fibers from damage. Steel is much harder than other reasonably priced materials and will be able to provide adequate protection to the fiber bundle. 3.3. Alignment Issues Thus far, only fiber bundles with the optimal bundle geometry and ideal arrangement between the fiber bundle and the membrane have been analyzed. Assumptions about the geometry of the bundle have been made, but if the steel tubing and fibers are not properly chosen, these assumptions may not hold. This section attempts to develop criteria for minimizing the errors when geometry assumptions do not hold. 3.3.1. MEMS Chip Cavity Alignment Issues Ideally, the MEMS chip and fiber bundle are aligned as shown in Figure 3-1. Four geometric parameters of the device are identified: the membrane diameter (DIA), the fiber bundle diameter (FiberDIA), the inner diameter of the steel tube (ID), and the outer diameter of the steel tube (OD). Two types of position errors are also identified: the distance between the outer edge of the steel tube (Type I Error, ERR1), and the distance between the fiber bundle and the inner edge of the steel tube (Type II Error, ERR2). Type II Error also represents the worst case radial position error (see Section 3.3.2). The worst case bundle position error (BPE), defined as the radial distance between the actual location of the Tx fiber core and the desired location of the Tx fiber core, is given by the following equations. 2 1 ERR ERR B P E Equation 3-1

PAGE 76

65 OD DIA ERR 2 1 1 Equation 3-2 FiberDIA ID ERR 2 1 2 Equation 3-3 Figure 3-1 – Bundle Position Error Illustration. From the equations for bundle position error, it is evident that the bundle position error is minimized when ID = FiberDIA a nd OD = MembaneDIA. If a device with zero error could be fabricated, then the fiber bundle geometry would be perfect and the bundle would always be perfectly aligned. In prac tice, packaging issues and the unavailability of optimally sized steel tubing prevent BPE from being eliminated completely. The steel tubing (and optical fibers) was chosen to minimize the worst case BPE (and RPE; see Section 3.3.2). The BPE was analyzed for fiber designs using multiple different fiber types, from small 50 m diameter core, 55 m diameter cladding fibers to 200 m diameter core, 220 m diameter cladding fibers. In the ideal configuration, smaller fibers improve sensitivity and reduce the equilibrium gap. In practice, it is very Membrane Diameter, DIA Bundle Diameter, FiberDIA Steel Tube Inner Diameter, ID Steel Tube Outer Diameter, OD Type I Error, ERR1 Type II Error, ERR2

PAGE 77

66 difficult to build an ideal fiber bundle with 50/55 fibers, since commercially available tube gauges allows large BPE (and RPE; see Section 3.3.2) with these fibers. The 105/125 fibers used by this thesis were chosen because they provided the best mix between ideal device performance, worst case error device performance, and manufacturability. The tube size that minimizes the RPE with the 105/125 fibers is the 21HW gauge from Popper & Sons. Specs for this tube and other useable tube sizes are shown in Appendix E. Another type of alignment problem occurs when the fiber bundle face is tilted at an angle with respect to the ideal position. This angular misalignment error is illustrated in Figure 3-2. Typical misalignments enc ountered are only a few degrees. Figure 3-2 exaggerates the effect for purposes of illustration. Figure 3-2 – Angular Misalignment Error Illustration. Although the alignment problem shown in the above picture is greatly exaggerated, it illustrates how size mismatches between the OD of the steel tube and the

PAGE 78

67 MEMS chip cavity can invalidate the para llel surfaces assumption. This kind of alignment problem is usually dealt with by th e packaging. Horowitz [27] studied this type of alignment error for a solitary single mode fiber optical microphone, but it is significantly more complicated to theoretically characterize the effects of angular misalignments for a multiple multimode fiber structure. A detailed analysis of the effects of this type of alignment error will not be dealt with here. Minimization of Type I Error can eliminate this problem. For a package that virtually eliminates this type of error, see Chapter Four. 3.3.2. Fiber Bundle Geometry Issues Thus far, the fiber bundle has been assumed to be tightly packed (all fibers in contact). In practice, this is difficult to do. A fiber bundle geometry with fiber position errors is shown in Figure 3-3. Figure 3-3 – Radial Position Error Illustration. Radial Position Error (RPE)

PAGE 79

68 The fiber radial position error (RPE) shown above is modeled for small error values in Chapter 2 by shifting the ring outward by the amount of the radial position error. In general, this type of error is caused by Type II errors in the fiber bundle structure, when the Rx fibers are shifted to wards the steel tube wall by a different amount than the Tx fibers. This reduces sensitivity and increases the optimal equilibrium gap (see Chapter 2). The observed RPE with the custom fiber bundle is 10 m. Using RPE = 10 m for using the 105/125 bundle, the MO sensitivity is reduced by 28% compared to the case where RPE = 0 m (see Chapter 2 for details). For fiber position errors larger than 30 m, the actual ring area diverges significantly from the assumed ring area, and other effects such as multiple reflections can become significant. Since the corrected power coupled and sensitivity equations do not consider multiple reflections, a more sophisticated model should be used when RPE > 30 m. Finally, if all the fibers in the bundle do not have the same RPE, then the corrected ring approximation theory will be invalid, since the bundle structure assumption will no longer hold (i.e. Rx fibers randomly placed with respect to the Tx fiber). 3.3.3. Application of Alignment Theory to Fiber Bundle Selection In this section, the alignment issues are quantified for specific fiber bundle geometries, and the optical fibers and steel tubing are selected. Equations 3-1 through 33 were used to determine the gauge and wall thickness of the steel tube that minimizes the worst case alignment errors in the optical microphone. Table 3-1 shows the worst case bundle position errors when combining various steel tube gauges from Popper & Sons [43] with different Tx and Rx optical fi bers. The design selected for this thesis was

PAGE 80

69 the first entry in Table 3-1. This design provided the best balance between minimizing errors and availability of standard optical connectors. Table 3-1 – Error Analysis of Diffe rent Fiber Bundle Configurations Design Tubing ID ( m) OD ( m) FiberDIA ( m) ERR1 ( m) ERR2 ( m) BPE ( m) BPE Norm Tx: 105/125 Rx: 105/125 21HW 457.2 812.8 375 93.6 41.1 134.7 0.2694 Tx: 105/125 Rx: 105/125 20RW 622.3 901.7 375 49.15 123.65 172.8 0.3456 Tx: 105/125 Rx: 200/225 21TW 609.6 812.8 575 93.6 17.3 110.9 0.2218 Tx: 105/125 Rx: 200/225 20RW 622.3 901.7 575 49.15 23.65 72.8 0.1456 Tx: 200/225 Rx: 105/125 21RW 533.4 812.8 475 93.6 29.2 122.8 0.2456 Tx: 200/225 Rx: 105/125 20RW 622.3 901.7 475 49.15 73.65 122.8 0.2456 Tx: 200/225 Rx: 200/225 21XXTW 711.2 812.8 675 93.6 18.1 111.7 0.2234 Tx: 200/225 Rx: 200/225 20XTW 723.9 901.7 675 49.15 24.45 73.6 0.1472 Tx: 50/55 Rx: 50/55 21HW 457.2 812.8 165 93.6 146.1 239.7 0.4794 Tx: 50/55 Rx: 50/55 20RW 622.3 901.7 165 49.15 228.65 277.8 0.5556 Tx: 105/125 Rx: 50/55 21HW 457.2 812.8 235 93.6 111.1 204.7 0.4094 Tx: 105/125 Rx: 50/55 20RW 622.3 901.7 235 49.15 193.65 242.8 0.4856 Tx: 50/55 Rx: 105/125 21HW 457.2 812.8 305 93.6 76.1 169.7 0.3394 The design used by Kadirval’s optical microphone was a Tx: 50/55, Rx: 50/55 design fabricated by Romack. It is evident from Table 3-1 that this design could have significant alignment issues if special packag ing techniques are not used. Romack used a compound tube structure with a smaller tube containing the fibers and a larger tube housing the small tube.

PAGE 81

70 CHAPTER 4 FABRICATION OF THE OPTICAL MICROPHONE The fabrication of the optical microphone consists of two parts: (1) the MEMS optical diaphragm chip, and (2) the fiber bundle. The MEMS chip containing the silicon nitride diaphragm was fabricated by MEMS Exchange, an umbrella organization that combines the processing capabilities of many foundries across the country [28]. The fiber bundle was fabricated at the University of Florida. 4.1. MEMS Exchange Process The process flow used in the MEMS Exchange process was a modified version of the process flow designed by Kadirval [8]. Both mask and wafers were purchased through the MEMS Exchange. The mask was designed at the University of Florida. Table 4-1 summarizes the wafers used for fabrication of the optical microphone. Table 4-1 – Wafers Used for Op tical Microphone Fabrication [28] Number of Wafers 5 Material silicon Diameter 100 mm Surface Finish double side polished Thickness 500 – 525 m Orientation <100> Doping Type n-type Quality prime Resistivity 1 – 10 / cm Initial State virgin Source MX Price per Wafer $ 19.95

PAGE 82

71 The process grows the membrane layer out of silicon nitride, and uses a DRIE to etch the bulk silicon and leave the nitride membrane. Silicon dioxide was used as an etch stop for the DRIE. The detailed process flow is shown in Appendix A. Not shown in Appendix A are steps to mount and demount the wafer onto a handle wafer to provide mechanical support during the DRIE. 4.2. Packaging Process Different novel packaging strategies for an optical microphone have been proposed [29,30]. Abeysinghe [29] notes that “adhesives limit the range of operation of the sensors.” To minimize the amount of adhesives used, the cavity is etched at the end of the fiber and an anodic bonding process is employed to bond an ultra-thin silicon wafer to the end of the fiber to serve as a membrane [29]. A diagram of the packaging technique described in [29] is shown Figure 4-1. Figure 4-1 – Abeysinghe et al. Packaging Technique. This packaging technique provides a compact device with a sensor head that is the diameter of the optical fiber. It would func tion as a pressure sensor in the configuration Ultra thinmembrane Machined cavity FiberCore Fiber Cladding

PAGE 83

72 shown above (no vent channel), but modifications could be made to add a vent channel. With this design, it is difficult to control the mechanical properties of the membrane. Another possible drawback of the package is the size of the cavity. Depending on the dimensions of the fiber, the optimal equilibrium membrane position may have both intensity and interference modulation effect pres ent. If both these effects are present, the theoretical analysis of the device performance will be very difficult. A similar packaging technique is proposed by Beggans et al. [30]. Beggans’ technique is more useful for an intensity-modulated device. First, a cavity is machined in a glass wafer. Then, a hole with a diameter which can accompany an optical fiber is drilled at the bottom of the cavity through the glass substrate. An ultra-thin silicon wafer is anodically bonded to the glass substrate, creating a cavity. An optical fiber is inserted into the cavity through the hole drilled in the substrate and fixed with a bonding agent. The resulting device is shown in Figure 4-2. This package allows more flexibility in membrane diameter and equilibrium gap position, but it does not provide as much control over the equilibrium gap position. As the fiber is being positioned, an active measurement technique is required to verify membrane position. Both techniques presented thus far do not protect the optical fibers. The ideal package would provide some mechanical support for the optical fibers to reduce the chances of device breakage. The packaging technique proposed by Kadirval [8] provides protection for the fibers by using a fiber bundle like the type described in this thesis. The optical fibers are protected by a steel tube and furcation tubing. Multiple MEMS chips with through-wafer holes (identical to those used in this thesis) are bonded together in the

PAGE 84

73 package. All but one of these chips has the diaphragm removed, and they are used to make a handle wafer stack. The steel tube containing the fiber bundle is inserted through the handle wafer stack, and the membrane wafer is placed on the top of the wafer stack. An illustration of the proposed technique is shown in Figure 4-3. Figure 4-2 – Beggans et al. Packaging Technique. This technique provides both a compact package and support for the fibers. A drawback is the potential angular misalignment (other than normal light incidence) between the fiber and the membrane. The worst-case angular misalignment depends on the diameter of the steel tube and the thickness of the handle wafer stack. Increasing the thickness of the handle wafer stack and decreasing the difference between the diameter of the fiber and the wafer hole will reduce the worst-case angular misalignment. Ultra thinmembrane Machined cavity Fibe rCore Fiber Cladding Glass Substrate

PAGE 85

74 Figure 4-3 – Kadirval Packaging Technique. The technique for packaging the optical microphone design proposed in this thesis is similar to [8]. In this proposed package, a second steel tube is used in place of a handle wafer stack. This steel tube (package tube) would have an inner diameter equal to the outer diameter of the steel tube (bundle tube) used in the fiber bundle construction (approx 830 m). An illustration of this package is shown in Figure 4-4. Standard available tubing gauges (Popper & Sons [43]) can provide a flush fit to within tens of microns for this package configuration. The package tube can be an inch or more long. Using simple geometry, the worst-case angular misalignment can be calculated. Assuming a worst case gap of 50 m between the bundle steel tube and package steel tube (a reasonable assumption based on available tube gauges), and assuming the minimum length of 1 in. (25,400 m) for the package steel tube, the maximum angular misalignment of the bundle tube in the package tube is 0.113o. If a 2 in. package tube were used, then the maximum angular misalignment becomes 0.056 o. Therefore, it can MEMS Diaphragm Chip FiberCore Fiber Cladding Handle Wafer Stack Epoxy

PAGE 86

75 be concluded that this packaging strategy is capable of virtually eliminating angular misalignments. The primary advantages of this package are that it simultaneously provides a robust package and minimizes the worst case angular misalignment error between the fiber bundle and the membrane, as demonstrated in the previous paragraph. Also, the dimensions of the package steel tube can be chosen to either minimize package diameter or to fit a commercially available calibrator such as the Bruel & Kjaer 4231 Microphone Calibrator [31]. In addition, the outside of the package steel tube could be threaded to allow a protective screen to be attached over the membrane, protecting it from damage. For an example of this type of protective screen, see those used by Bruel & Kjaer 1” [32], ” [33], and 1/8” [34] microphones. Another advantage of this package is the ability to take multiple packaged optical microphones and easily assemble them into a microphone array bundle.

PAGE 87

76 Figure 4-4 – Proposed Package for the Optical Microphone. In the top view in Figure 4-5, seven packaged optical microphones are shown in a bundle surrounded by an array package steel tube whose dimensions would be selected to make the fit as tight as possible. The cylindrical structure of the individual microphone packages allows the microphone array geometry to be a scaled version of the fibers in the individual microphone package. This allows an arbitrary number of “rings” of packaged microphones to be used in the array, without the individual packaged microphones interfering (mechanically) with each other. In addition, the cylindrical structure of the proposed array package would be easy to construct, using only a custom MEMS Diaphragm Chip Bundle Steel Tube Package Steel Tube Epoxy Epoxy Device Furcation Tubing > 1”

PAGE 88

77 steel tube and epoxy to hold it together. Due to the steel tubing, this proposed array package would be very robust. Figure 4-5 – Proposed Optical Microphone Array Package. Top View Side View 1 mm 2.5 mm 7.5 mm 9.5 mm

PAGE 89

78 In Figure 4-5, the membranes have been removed from the square chips to show the individual microphone cavity positions. This proposed microphone array could sample the acoustic field on the center 50 m of the 1 mm diameter diaphragm. Each diaphragm is 2.5 mm from each of its neighbors. Finally, the performance of the microphones in the array is independent of the number of microphones in the array (assuming the array package does not affect the sound field and availability of sufficient light sources, detectors, etc).

PAGE 90

79 CHAPTER 5 EXPERIMENTAL SETUP AND RESULTS The experimental characterization of the optical microphone is divided into three sections. First is the experimental character ization of the laser and photodetector. This characterization measures the value of Plaser_noise(Plaser) (compare to Plight_noise in Chapter 2). The experimentally measured value Plaser_noise(Plaser) is used as an input to the theoretical model of the microphone presented in Chapter 2. Second is the static calibration of the custom fiber bundle. This static calibration attempts to verify the theoretical power coupled vs. equilibrium gap and sensitivity vs. equilibrium gap plots. The static calibration curve measured in this experiment is used to identify the location of the fiber bundle with respect to the membrane in the dynamic calibration experiments. The final experimental step is the dynamic calibration of the optical microphone. In the dynamic calibration, a plane wave tube (pwt), speaker, and calibrated microphone are used to determine the optical microphone sensitivity, linearity range, frequency response, noise floor, dynamic range, and minimum detectible signal. The experimental results are then compared with the theoretical predictions. The dynamic calibration is performed for both the unreferenced and re ferenced output microphone configurations.

PAGE 91

80 5.1. Laser and Photodetector Characterization 5.1.1. Experimental Setup for Lase r and Photodetector Characterization In the characterization of the laser (or other light source), the optical spectrum of the photodetector was measured for laser outputs of 100 W through 500 W, in steps of 50 W. The noise power spectral density of the detector output, in V / Hz, was recorded by the Pulse system from 0 – 6.4 kHz with a bin width of 1 Hz and using 500 samples. This value is the RMS sum of the detector noise and detected light intensity noise, both in V / Hz. The optical intensity noise component of the detector noise is determined by removing the theoretical Vdet from the measured value, leaving Vlight (see Section 2.2.4 for variable definitions). By dividing Vlight by the photodetector responsivity and gain, the optical intensity noise of the laser, in V / Hz1/2, can be determined. Figure 5-1 – Experimental Set up for Laser Characterization. The DC optical power was obtained by obs erving the 0 Hz frequency bin. The worst-case light source noise recorded at 660 Hz, and the best-case noise (where the laser Acoustical Signal System Component KEY Electrical Signal System Output System Input Optical Signal PDA-400 Photodetector Pulse System Computer Ethernet Port HP8168B Laser ISS-1550 Opto-isolator

PAGE 92

81 noise became constant with respect to frequency) was measured at 1600 Hz. The experimental setup for the laser noise ch aracterization is shown in Figure 5-1. 5.1.2. Results of Laser and Photodetector Characterization The experimentally measured value for Plaser_noise is given by the last column in Table 5-1. The HP8168B was observed to ha ve a flat noise floor above 1550 kHz. Below 1.55 kHz, the noise floor was not flat. Since this system would ideally be used with a heterodyne detection scheme to operate the microphone in the flat noise range of the light source, the laser linearity and MDS will be experimentally measured above 1550 Hz. For all laser characterization experiments, Gdet = 15,000 V / A. Table 5-1 – Experimental HP8168B Noise Laser Power (W) Measured Noise @ 1600 Hz (uV / Hz1/2) Calculated Laser Noise @ 1600 Hz (pV / Hz1/2) Laser SNR @ 1600 Hz (dB) 100 0.741 52.0 53.0 150 1.29 90.2 52.4 200 1.53 108 52.9 250 1.06 74.4 55.6 300 1.98 139 53.6 350 0.774 54.3 58.4 400 1.68 118 55.7 450 3.59 252 52.9 500 3.26 229 53.9 The laser was observed to have a maximum SNR at Plaser = 350 W. Therefore, this value will be used to characterize the laser. 5.2. Static Calibration 5.2.1. Experimental Setup for Static Calibration The static calibration of the optical microphone has two goals. The first goal is to verify the corrected power coupled model presented in Chapter 2. The second goal is to

PAGE 93

82 obtain a power coupled curve for the fiber bundle under consideration, so that it may be used to measure the equilibrium gap of the assembled microphone package. The procedure for this is described further in the dynamic calibration section. In the static calibration experiment used by Kadirval [8], the fiber bundle is placed flush against a metal mirror. To align the fibers normal to the mirror, Kadirval assumed that the power coupled into the receive fibers was maximum when the fibers were aligned normal to the mirror. While there may be a range of equilibrium gap distances that this assumption is valid for, it is not valid in general, and will result in the proper alignment of the fibers with the mirror. Kadirval used a micropositioner to move the fiber bundle with respect to the mirror. Due to the way in which the fiber bundle was mounted, the fibers were able to slip in the mechanical mount as the micropositioner was moved. Due to these two problems with the previous static calibration experiment in [8], the gap measurements reported by the micropositioner were significantly larger than their actual values, resulting in an incorrect calibration curve. In this thesis, these two issues have been fixed. The mechanical mounts for the fiber and the mirror were aligned normal to each other by cubic blocks, which have parallel surfaces. The mirror was mounted on the micropositioner and was moved with respect to the fiber bundle, whose position was fixed. The zero position of the mirror was set by slowly moving the mirror with the mi cropositioner until mechanical contact with the mirror was observed. After mechanical contact has been made, the fiber holder is closed, fixing the fiber bundle in place. Note: the mirror used in these experiments was slightly rusted and had scratches on its surface. The method of zeroing used here can

PAGE 94

83 potentially scratch the mirror, so it should not be used with a mirror in good condition. An electrical contact method could be used instead. The static calibration setup used for the optical microphone is similar to the setup used by Kadirval [8] except as noted in the pr evious paragraph. The experimental setup for the static calibration is illustrated in Figure 5-2. In Figure 5-2, the custom fiber bundle is oriented normal to a reflective mirror, which is mounted on a computercontrolled micropositioner. The computer-controlled micropositioner is automated by a Labview application to sample the power output of a reference channel and the output of the fiber bundle. After taking 30 samples, the micropositioner adjusts the mirror position relative to the fiber bundle and repeats the measurement. This step-and-sample process continues until cancelled by the user. The light source used is the LPS-SMF28-1550-FC laser diode operated in constant current mode by a Keithley 2400 constant current source. Laser light is passed through a Newport ISS-1550 optical isolator a nd is split by a 50/50 optical power splitter. One output of the splitter delivers light into the Tx fiber of the custom fiber bundle. This light reflects off of the mirror and is collected by the receive fibers, which transport the light to a PDA-400 detector (TEST). The s econd splitter output delivers light directly to another PDA-400 (REF). A Keithley 2000 multimeter controlled by a Labview application sampled the TEST and REF detector outputs, and sends the data to a computer where it is recorded in a file along with the micropositioner position. The gain of the PDA-400 detectors in the static calibration is 15,000 V / A. After the experiment is complete, the recorded detector outputs are corrected for the experimentally measured bias errors of each photodetector. Dividing the bias-

PAGE 95

84 corrected TEST data by the bias-corrected REF data and plotting it with respect to the measured gaps produces the experimental power coupled curve. The results of the static calibration are presented in the next section. Figure 5-2 – Block Diagram of Static Calibration. 5.2.2. Results of Static Calibration The output of the static calibration is an experimental plot of optical power coupled vs. equilibrium gap distance. This pl ot is shown in Figure 5-3 shows the static LPS-SMF28-1550-FC Laser Diode ISS-1550 Opto-isolator Optical Splitter 1550 nm PDA-400 Photodetector (REF) Transmit Fiber, Custom Bundle Keithley 2400 Sourcemeter Aluminum Mirro r Micropositioner Micropositioner Controller Computer Serial Port Receive Fiber, Custom Bundle PDA-400 Photodetector (TEST) Computer GPIB Port KEY Electrical Signal Optical Signal Mechanical Signal System Input System Output System Component Keithley 2000 Multimeter

PAGE 96

85 calibration curve for the optical microphone using the custom fiber bundle. The plot also shows the theoretically predicted curves for an RPE of 0 m and an RPE of 10 m. The maximum power coupled is slightly higher than theoretically predicted, and the location of the peak power coupled is at a larger gap than predicted. 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 01002003004005006007008009001000Gap ( m)Power Coupled (W/W) RPE = 0 um RPE = 10.0 um Experimental Figure 5-3 – Experimental Power Coupled vs. Equilibrium Gap. The location and magnitude of the max slope of the experimental curve was determined by fitting a regression line to points in a 30 m window on the experimental power coupled curve. The location of the maximum slope of the regression line was assumed to be the gap at the maximum slope, and the maximum slope of the regression line was assumed to be the peak slope of the experimental power coupled curve. The regression line at the peak slope is shown in Figure 5-4.

PAGE 97

86 Table 5-2 – Comparison between Theoretical and Experimental Static Calibration Category Theoretical (RPE = 0 m) Theoretical (RPE = 10 m) Experimental Peak Power Coupled 21.6 % 17.2 % 22.1 % Gap @ Peak Power ( m ) 500 530 600 Power Coupled @ Max Slope 5.2 % 4.2 % 10.9 % Gap @ Max Slope ( m) 230 260 360 Maximum Slope ( m-1) 1.094 x 10-3 0.784 x 10-3 0.800 x 10-3 y = 0.0008x 0.162 R2 = 0.9948 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 200240280320360400440480520560600Gap (um)Power Coupled (W / W) Figure 5-4 – Experimental Maximum Po wer Coupled Regression Line Slope. The R2 value of the regression line in Figure 5-4 indicates the MO stage is linear over the gap region from 350 m to 380 m. Since the membrane deflection is much less than 30 m, the MO stage is linear when operated at the maximum sensitivity. In both Figure 5-3 and Figure 5-4, the expe rimental power coupled curve is much less smooth than the theoretical curve. At some gap values the power coupled seems to be discontinuous with the surrounding power coupled points. This is believed to be caused by two phenomena. The first cause is ta ble vibrations. The optical table on which

PAGE 98

87 the experiments were performed was not isolated from the ground by a cushion of compressed air. This allows acoustic vibrati ons traveling through the floor of the room to vibrate the mirror during the measurement. The second cause is the roughness of the mirror surface. The mirror used in this experiment was observed to have scratches and rust spots on the mirror surface. Light interacting with the scratches and rust spots can cause discontinuities in the measured power coupled. 5.3. Dynamic Calibration 5.3.1. Experimental Setup for Dynamic Calibration The goal of the dynamic calibration is to characterize the sensitivity, minimum detectable signal, electronics noise, linearity and frequency response of the unreferenced and referenced optical microphone configurations, and to compare the experimental performance of the microphone with the theoretical performance. In the both unreferenced and referenced output calibration experiments, the cut-on frequency of the SRS 560 was 30 Hz, the gain of the SRS 560 was 1 V / V, and the transimpedance gain of the PDA-400 detector(s) was 15,000 V / A. Data is taken with the Pulse system from 0 Hz – 6.4 kHz with a 2 Hz bin width. All single tone measurements were made with a sinusoidal input at 1600 Hz A uniform window was used to measure single tone signals, and a Hanning window wa s used for measuring broadband or noise signals. In all experimental setups (F igure 5-5 and Figure 5-6), a B&K 4138 1/8” microphone [34], pre-calibrated with a B&K 4228 pistonphone [35], is used as a reference microphone. This microphone is mounted next to the optical microphone at the end of the plane wave tube, and it connected to an input channel of the Pulse system (not

PAGE 99

88 shown in Figure 5-5 and Figure 5-6). The e xperimental setups for the optical microphone dynamic calibration are similar to the setup used by Kadirval [8]. Figure 5-5 – Unreferenced Output Optical Microphone Configuration. A “power coupled alignment” technique is introduced for measuring the equilibrium gap. This technique makes a power coupled measurement as the fiber bundle and membrane chip are mounted on the custom plane wave tube (PWT) plug and relates the power coupled measurement to an equilibrium gap by the static calibration curve. Acoustical Signal System Component KEY Electrical Signal System Output System Input Transmit Fiber, Custom Bundle HP8168B Laser Sou r ce ISS-1550 Opto-isolator Computer Ethernet Port Amplifier Pulse System Pulse SystemOptical Signal Computer Ethernet Port SRS 560 Filter / Amp MEMS Chi p PDA-400 Photodetector (DET) Receive Fiber, Custom Bundle JBL Speaker and Plane Wave Tube

PAGE 100

89 The first experimental setup is the unref erenced output configuration, shown in Figure 5-5. The optical splitter is not required for operation of the microphone in this configuration. In the unreferenced output conf iguration, the system output is taken at the output of the SRS 560 Filter / Amplifier. Figure 5-6 – Referenced Output Optical Microphone Configuration. MEMS Chi p JBL Speaker and Plane Wave Tube Computer Ethernet Port Pulse System ISS-1550 Opto-isolator Optical Splitter 1550 nm Transmit Fiber, Custom Bundle HP8168B Laser Source PDA-400 Photodetector (REF) Pulse S y ste m Computer Ethernet Port 10*MOD --------REF ISS-1550 Opto-isolator Analo g Divide Circui t PDA-400 Photodetector (REF) Receive Fiber, Custom Bundle Amplifier KEY Optical Signal Acoustical Signal Electrical Signal System Output System Input System Component SRS 560 Filter / Amp (Optional)

PAGE 101

90 The next setup is the referenced output c onfiguration, shown in Figure 5-6. The referenced output is also used for the equilibrium gap measurement, since the steady state output of the analog divide is 10x the power coupled. An estimate of the power coupled can be made with a voltmeter, but a more accurate value is obtained with the pulse system. In order to add optical losses to the reference path (when desired), the reference isolator was intentionally misaligned with the optical splitter, inducing losses. The measured optical path losses induced by this intentional misalignment was = 0.76. In order to determine the effect of on the MDS, the losses were removed after the microphone characterization was completed, and the MDS was measured with = 0. 5.3.2. Results of the Dynamic Calibration The microphone output was observed to be slightly unstable for the unreferenced and referenced microphone, with the referenced microphone the more stable of the two configurations. The unreferenced configuration instabilities are expected, since the unreferenced microphone is sensitive to drift in the laser output. These instabilities were not severe, although it could affect a measurement made by the optical microphone on rare occasions. Since lasers are inherently unstable, care must be taken to ensure that the laser is operated at a stable output power, with a minimal amount of power switching between lasing modes. The most stable output of the laser does not necessarily provide the maximum intensity SNR, so operating the laser at the optimal SNR point will cause a stability tradeoff. In general, to minimize the microphone MDS, operate the laser at the peak intensity SNR. To minimize the effect of instabilities, observe the laser power stability with an optical spectrometer and opera te the microphone at the most stable point.

PAGE 102

91 An LED is a more stable light source, and could potentially provide improved MDS and stability in an optical microphone than a laser source, without requiring optical isolators. This advantage is countered by the considerably lower coupled power that is obtained by fiber-coupled LED sources compared to fiber-coupled lasers (as much as 1000 times more power for laser sources). An LED source must provide sufficient output power for the received optical intensity modulated signal to be detectable by the photodiode (it is not sufficient for just the st eady state received power to be detectable). Both microphone configurations were observed to be linear up to 122 dB (re. 20 Pa). This linearity is much less than theoretically predicted (160 dB re. 20 Pa). It is caused by two factors: (1) a significantly lower in-plane stress than the 50 MPa design goal, and (2) the thickness variation of the membrane measured by the University of Michigan after the deposition of the nitride (+/1 m). These effects cause the AM sensitivity to be as much as 15-20 times the design value (see Sec. 2.2.2.1). y = 0.3165x + 0.2808 R2 = 0.92670.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 08152330384553606875Sound Pressure (Pa)Unreferenced Microphone Output (mV) Figure 5-7 – Linearity, Unrefere nced Output Configuration.

PAGE 103

92 The unreferenced microphone has a problem with linearity. Specifically, the slope of the linearity curve varies by an unacceptable amount in the microphone linear region. The unreferenced microphone sensitivity is dependant on the output power of the light source, which is not stable for a laser. This is because the laser output is known to drift over time and has small instabilities present at its operating point. y = 1.7666x + 0.2152 R2 = 0.99160.0 4.0 8.0 12.0 16.0 20.0 24.0 28.0 32.0 36.0 40.0 08152330384553606875Sound Pressure (Pa)Unreferenced Microphone Output (mV) Figure 5-8 – Linearity, Referenced Microphone Configuration. From Figure 5-7 and Figure 5-8 the unreferenced and referenced sensitivities can be observed. The unreferenced output optical microphone has a sensitivity of 0.3165 mV / Pa. The referenced output optical microphone has a sensitivity of 1.7666 mV / Pa. The magnitude response for the unreferenced and referenced microphones was made with a periodic random noise signal with amplitude equal to 90 dB. For the unreferenced case, the amplitude of the magnitude response was not equal to the slope of the linearity regression curve. This is caused by drift and instabilities of the laser output,

PAGE 104

93 to which the unreferenced microphone is highly sensitive. The referenced configuration magnitude response was close to the slope of the referenced microphone linearity curve regression line. 0 40 80 120 160 200 240 280 320 360 400 08001600240032004000480056006400Freq (Hz)Unreferenced Magnitude Response ( V/Pa) Figure 5-9 – Magnitude Response, Unre ferenced Microphone Configuration.

PAGE 105

94 0 2 4 6 8 10 12 14 16 18 20 08001600240032004000480056006400Freq (Hz)Referenced Magnitude Response (mV/Pa) Figure 5-10 – Magnitude Response, Re ferenced Microphone Configuration. -40 -36 -32 -28 -24 -20 -16 -12 -8 -4 0 4 8 12 16 20 24 28 32 36 40 08001600240032004000480056006400Freq (Hz)Unreferenced Phase Response (degrees) Figure 5-11 – Phase Response, Unrefe renced Microphone Configuration.

PAGE 106

95 -40 -36 -32 -28 -24 -20 -16 -12 -8 -4 0 4 8 12 16 20 24 28 32 36 40 08001600240032004000480056006400Freq (Hz)Referenced Phase Response (degrees) Figure 5-12 – Phase Response, Refe renced Microphone Configuration. The phase response for both optical micr ophone configurations is shown in Figure 5-11 and Figure 5-12. The unreferenced microphone exhibits a flatter phase response than the referenced optical microphone. The referenced optical microphone was observed to have partial coherence between the optical noise in the modulated and reference signal paths. This partial coherence, which is not the same at every frequency bin, is believed to be responsible for the random nature of the referenced phase response. The measured electrical noise floor of the unreferenced microphone at 1600 Hz is 0.77 V / Hz. For the referenced microphone with = 0.76, the measured electrical noise floor at 1600 Hz was 5.63 V. When = 0, the measured electrical noise floor at 1600 Hz was 5.63 V. Since increasing alpha was experimentally observed to increase sensitivity (not shown in plots), the MDS of the referenced optical microphone was

PAGE 107

96 improved by adding optical reference path losses. If the optical intensity noise received by the REF and MOD detectors was highly correlated (more than was observed), then the optical path losses are expected to increase both the electronic noise and the same factor, resulting in no improvements to the MDS. From Figure 5-13 it is observable that the unreferenced noise floor, which is dominated by laser intensity noise, is not flat The referenced noise floors are flat in areas where the unreferenced microphone has noise peaks (the analog divide noise is negligible). Since both configurations are dominated by the optical intensity noise of the laser source, the optical intensity noise in the referenced microphone modulated and reference signal paths must be partially correlated (since only a correlated signal can be removed by a time-domain division). Since th ere are frequency ranges where the optical intensity noise dominates and the referenced noise floor is not flat, the optical noise in the referenced microphone signal paths is not perfectly coherent at all frequencies.

PAGE 108

97 0.01 0.1 1 10 100 1000 10000 08001600240032004000480056006400Freq (Hz)Noise Floor ( V / Hz1/2) Unref Noise Floor Ref Noise Floor, Alpha = 0 Ref Noise Floor, Alpha = 0.76 Figure 5-13 – Electrical Noise Floor Both Microphone Configurations. Table 5-3 summarizes the performance metrics of the optical microphone in the unreferenced output configuration, which used the experimental setup in Figure 5-5. Table 5-4 summarizes the performance metrics of the optical microphone in the referenced output configuration, which used the experimental setup in Figure 5-6. The reported MDS in both tables is taken with a 2 Hz bin width at a center frequency of 1600 Hz.

PAGE 109

98 Table 5-3 – Experimental Results of Unreferenced Output Microphone Dynamic Calibration Laser Power (W) PDA 400 Gain (V / A) SRS 560 Gain (V / V) Sensitivity (mV / Pa) Frequency Response (Hz) MDS (dB re. 20 Pa) Dynamic Range (dB) 350 15,000 1 0.320 300 – 6400* 65 65 – 122 The low end of frequency response is set by the lack of a well-defined vent channel. The upper end is limited by the range of the plane wave tube over which only the first propagation mode is supported. Table 5-4 – Experimental Results of Referenced Output Microphone Dynamic Calibration Laser Power (W) PDA 400 Gain (V / A) SRS 560 Gain (V / V) Sensitivity (mV / Pa) Frequency Response (Hz) MDS (dB re. 20 Pa) Dynamic Range (dB) 350 15,000 (both detectors) 1 1.77 300 – 6400* 47 47 – 122 The low end of frequency response is set by the lack of a well-defined vent channel. The upper end is limited by the range of the plane wave tube ove r which only the first propagation mode supported.

PAGE 110

99 CHAPTER 6 CONCLUSIONS AND FUTURE WORK 6.1. Conclusions In conclusion, the MEMS-based intensity-modulated optical microphone is an excellent choice for applications with harsh environmental or size constraints. Optical MEMS microphones are currently marketed as a surveillance technology, as an EMI and RFI immune technology, and as a suitable technology for use in automobile voice recognition systems (see [36 and 37]). Intensity-modulated MEMS optical microphones can be configured to provide a MDS, sensitivity, and electronic noise that is competitive with other microphones of comparable size. It is also possible to design the optical microphone with a significantly higher sensitivity and lower MDS by sacrificing frequency response and reducing the upper limit of the microphone’s dynamic range. It is difficult to make an intensitymodulated MEMS optical microphone with a high sensitivity, low noise floor, low MDS, high frequency response, and high upper limit to the dynamic range. This is due to inherent tradeoffs of the AM stage linearity and frequency response with the AM stage sensitivity. Alternate, more sensitive, fiber geometries are required to make an intensitymodulated optical microphone suitable for aero-acoustic measurements.

PAGE 111

100 6.2. Future Work There are many different areas for future research and development of optical microphones. The coupled equations for sensitivity, electronics noise, and minimum detectable signal presented in Chapter 2 can be used in a microphone design optimization problem (see Papila et al. [38]). This would allow a designer to optimize the microphone performance, given a set of design constraints. A laser can provide as much as 1000 times more power than an LED source when used as a light source in an intensity-modulated lever microphone. Laser sources require opto-isolators to provide output power stab ility for continuous-wave (CW) operation, and are generally less stable than LED sources. A future generation version of the optical microphone could be implemented with a single, large-core, high-NA fiber (instead of a fiber bundle) using an LED as a light source to improve stability and frequency response. Since the sensitivity of an unreferenced microphone with LED source is very low compared to a laser source, novel techniques to couple greater than 50 – 100 W of power into the transmit fiber must be used. If the optical power coupled into the transmit fiber(s) is not sufficiently high, then the microphone MDS will be limited by the MDS of the photodetectors and other electronics. When la sers are used as light sources, this is not an issue. Another area of future work is to implement a robust package for the optical microphone. A package such as the one presen ted in Figure 4-4 and Figure 4-5 would be a good choice for implementation. Since the performance of a MEMS device is application specific, multiple packages that are tuned to the specific requirements of each

PAGE 112

101 optical microphone application should be devel oped. In addition, an array packaging technique should be developed to take advantage of the small size of the MEMS device. The inherent immunity of the MEMS op tical microphone to adverse environments should be further explored. The utility of optical microphones in a high-EMI environment has been demonstrated by Phone-Or, but the separation of the electronics and the sensing element allows an optical microphone to potentially be useful in other harsh environments, such as underwater or in high temperatures. The transition to these environments brings new issues that must be addressed, such as the high acoustic impedance of water compared to the microphone package, or the effects of thermal expansion due to convection in high temper ature environments. A non-MEMS optical microphone for high temperature environments (1000 oF) has been developed by Cuomo et al. [39 and 40]. Finally, alternate configurations of op tical microphone should be explored. The bundle geometry used by this thesis is not optimal. Additional research needs to be done to determine the optimal bundle geometry by adding focusing optics, and additional or new fiber structures. Single fiber designs, such as the one used in [39], have been studied for non-MEMS optical microphones.

PAGE 113

102 APPENDIX A MEMS OPTICAL MICROPHONE DIAPHRAGM PROCESS FLOW Table A-1 – MEMS Exchange Process Flow for Optical Microphone Major Step Sub-steps Picture RCA Clean HF Dip Dry / Wet /Dry Oxidization Oxide Growth 4:1 Sulfuric / Peroxide Bath HCl Dip 50:1 HF Dip LPCVD Silicon Nitride Deposition Nitride deposition Dehydration Bake HDMS Prime Photoresist Coat Photoresist Coat with Hardbake Photoresist Hardbake ~ 500 m Silicon ~ 0.7 m SiO2 ~ 500 m Silicon 0.7 m SiO2 1 m SixNy ~ 500 m Silicon 0.7 m SiO2 1 m SixNy 2.7 m Photoresist

PAGE 114

103 103 Table A-1. Continued Major Step Sub-steps Picture Silicon Nitride RIE Reactive Ion Etch Dehydration Bake Photoresist Coat Alignment and Exposure Contact Photolithography Develop and Hardbake Buffered Oxide Etch Oxide Etch with Buffered HF ~ 500 m Silicon 0.7 m SiO2 1 m SixNy 2.7 m Photoresist ~ 500 m Silicon 0.7 m SiO2 1 m SixNy 2.7 m Photoresist 10 m Photoresist ~ 500 m Silicon 0.7 m SiO2 1 m SixNy 2.7 m Photoresist 10 m Photoresist

PAGE 115

104 104 Table A-1. Continued Major Step Sub-steps Picture Silicon DRIE Silicon Deep Reactive Ion Etch Photoresist Ashing Resist ash Buffered Oxide Etch Oxide Etch with Buffered HF Aluminum E-Beam Evaporation Aluminum EBeam ~ 500 m Silicon 0.7 m SiO2 1 m SixNy 0.08 m Al ~ 500 m Silicon 0.7 m SiO2 1 m SixNy ~ 500 m Silicon 0.7 m SiO2 1 m SixNy ~ 500 m Silicon 0.7 m SiO2 1 m SixNy 2.7 m Photoresist

PAGE 116

105 APPENDIX B FIBER BUNDLE PROCESS FLOW The fiber bundle used for the optical mi crophone was designed and fabricated at the University of Florida. The Guide to Connectorizing and Polishing Optical Fibers published by Thorlabs [41] provided a starting poi nt for the development of the process. The fabrication of the fiber bundle is done in two steps: bundle assembly and connector polishing. Table B-1 – Materials List for Bundle Assembly Thorlabs Part Number Component Quantity AFS105/125Y Multimode Optical Fiber 105/125/250 T06F13 Mechanical Buffer Stripper 1 MS403-10 Epoxy Syringe 1 F120 Pack of 24 Hr Room Temperature Curing Epoxy 1 FT030 Reinforced 3.0 mm Diameter Furcation Tubing 4’ 6” Shears or Razor Blade 1 Duct Tape 1 21HW 1 “ Length of Steel Tubing 1 NA (compare to CL-200) 100x Fiber Microscope 1 Silly Putty (or modeling clay) Small piece Tape Measure 1 30126D1 FC Connector 1 10440A or 10770A SMA Connector 1 Paper Towels 3-5 Light Source (flashlight, laser pointer) 1 Needle-nosed pliers or tweezers 1

PAGE 117

106 The detailed process for fiber bundle assembly is described next. The required components are listed in Table B-1. A clear workspace of at least 4 feet long and 2 feet wide, with the edge of a table present is required to assemble the bundles. The first step in assembling the custom fiber bundles is to cut the optical fiber and furcation tubing to length. The design goal is a bundle that is approximately 3 feet long. It is desired that each section have the same length, so three 1’ 6” long sections of furcation tubing are cut. Next, designate one section of furcation tubing as the device section, one section as Rx, and one section as Tx. Now cut seven 3’ 4” lengths of optical fiber. Use a sharp razorblade or sharp pair of scissors to avoid crushing the fiber as it is cut, which can cause cracking. The second step is to strip the buffer off of the ends of the optical fibers. It is the most labor intensive part of the custom fiber bundle construction, and requires care not to damage the unprotected fibers. The design goal is for 3.5” or more of buffer to be removed from the end of the fibers that will be threaded through the steel tubing, and for 1-1.5” to be removed from opposite end of the fibers, which will be connectorized. To remove the buffer, place the length of fiber on the table, with the end to be stripped extending approximately 2 cm off of the table. Fix the fiber to the table with tape or by applying pressure to it with your hand. It is critical that the fiber be perpendicular to the table edge and not be able to move during the stripping action, since this will cause the fiber to break. Insert the fiber into the mechanical buffer stripper (T06F13) as shown in the Thorlabs guide [41]. The fiber should extend past the cutting blades of the mechanical stripper by no more than 1-2 mm. Squeeze the cutting blades closed, and pull the mechanical stripper away from the table to strip the piece of buffer off of the fiber.

PAGE 118

107 The pulling motion must be firm and with cons tant pressure to avoid breaking the optical fiber. If resistance is encountered as the buffer section is removed, increase the pulling pressure. If there is a pause after the stripping motion begins, the fiber will break. Continue to strip the buffer in 1-2 mm long segments, repositioning the fiber as necessary, until the desired length of buffer has been stripped from the end of the fiber. If the fiber breaks during stripping, a few options are available. If the break shortens the fiber length by no more than a 0.5”, it should be possible to continue stripping the fiber. The end of the fiber to be inserted in the steel tube must be stripped to the prescribed length. Any length fluctuations in the fiber due to breakage must not affect the device end of the fibers. If the fiber break shortens the fiber length by greater than 0.5”, then the fiber will have to be replaced, or all fibers trimmed to the new length. If the break is more than 1”, re-trimming is not recommended. After the seven fibers are stripped to the desired length, place them next to each other, aligned to the buffered/unbuffered (B/U) junction at the device end of the fibers. Verify all fibers have 3.5” of buffer stripped from the device end of the fibers. Verify the fibers are all equal in length to within +/0.25”. Verify the Tx/Rx ends are all stripped to roughly 1”. Finally, verify that no cracks are vi sible in any part of the fibers, since cracks will prevent the fiber from transmitting light. After the buffer stripping is completed, the fibers must be threaded through the device section of the furcation tubing. The easiest method to accomplish this is to remove the pull-string from the inside of the fu rcation tubing, tape the tube to the table, and push the seven fibers through the tube at the same time. The fibers should extend from the end of the furcation tube by approximately 6”.

PAGE 119

108 Figure B-1 – Fiber Placement in the Steel Tube. The next step is the most difficult step in the entire process. The seven fibers must be inserted into the steel tubing. They must all be inserted together, and they must be approximately 1” of stripped fibers exte nding from the end of the steel tube. No epoxy is needed yet. For an illustration of this, see Figure B-1. Care must be taken to gently insert the fibers into the steel tube. Verify all seven fibers have passed through the tube, and that none have broken. Trim the ends of the stripped fibers so that all are the same length, as shown in Figure B-1. The steel tube was selected to be a tight fit for the stripped fibers. Note: the buffered fibers will not fit. Never push the steel tube closer than 0.25” to the Buffered / Unbuffered (B/U) boundary, since this will cause the fibers to experience a shear stress along the B/U boundary. This shear stress will cause some of the fibers to break, and this breakage may not be noticeable until after the bundle has been completed. This type of breakage can occur during the drying of the epoxy, so care must be taken to keep the steel tube away from the B/U boundary at all times. Device Section Furcation Tubing (FT030) Buffered / Unbuffered Fiber Boundary Steel Tube Cross-Section 1.5” > 0.25” > 1” Stripped ends trimmed to equal length

PAGE 120

109 The next step is identification of the cen ter fiber in the bundle. To identify the center (Tx) fiber, move the steel tube such that the ends of the fibers are flush with the end of the steel tube. Ideally, the end of the tube should be near the edge of the table to allow easy inspection of the fiber ends. If the fibers were not properly trimmed, then it will be difficult to see the bundle by looking into the end of the tube. Cover the remote ends of the fibers with a paper towel so that no light can be coupled into them from the ceiling lights. Isolate one fiber from the c overed fiber ends, and use a light source (for example, a laser pointer) to couple light into the remote end. Look into the steel tube at the fiber bundle face, and identify the illuminated fiber. If the illuminated fiber is in the center of the fiber bundle, set the remote end separate from the others. Continue until every fiber has been inspected. If any fiber does not transmit light, then it is broken and must be removed. If this happens, the fibers must be removed from the steel tube and the furcation tubing, and a replacement for the damaged fiber must be cut and stripped. By now, every fiber has been verified to transmit light, the Tx fiber has been identified, and the steel tube is properly on the end of the fibers (see Figure B-1). The next step is to put the remote ends of the optical fibers into their respective furcation tube sections. The Tx fiber should be pushed into one furcation tube segment, and the six Rx fibers should be pushed into the third (and last) segment. When the Tx and Rx furcation tubes are brought into contact with the devi ce furcation tube, the stripped fibers should extend at least ” from the end of the Tx and Rx sections. No buffered fiber should extend from the furcation tubes by more than a few millimeters.

PAGE 121

110 Figure B-2 – Illustration of Furcation Tube Segments after Assembly. Note that Figure B-2 shows the steel tube placed partially inside the device furcation tube segment. This is accomplished by pulling the fibers at the Tx and Rx end. The steel tube should be supported during the pulling process, and guided into the furcation tube. If the steel tube is pushed into the furcation tubing, it may slide close to the B/U boundary, and cause the fibers to break. At this point, the fiber bundle is confirmed to fit together properly. If necessary, the fibers were trimmed and re-stripped from the Tx/Rx end and/or the Tx/Rx tubes were resized. The next step is to epoxy the components together. First, arrange the fibers as shown in Figure B-1. Prepare the epoxy as shown in [41]. Make epoxy beads at the points shown in Figure B-3. Next, move th e steel tube along the fibers to smear the epoxy inside the tube. Care must be taken to ensure the steel tube does not come closer that 0.25” to the B/U boundary. Epoxy must cover all parts of the fiber from the end of 1’ 6” 1’ 6” > 1.1” > 0.75” Tx/Rx Section Furcation Tubing (FT030) Device Section Furcation Tubing (FT030) SteelTube Protruding Fiber Bundle(7 fibers) Rx Fiber (6 fibers) > 0.75”Tx Fiber (1 fiber)

PAGE 122

111 the steel tube to the B/U boundary. When dried, it will provide support for the stripped fibers whose positions will be fixed and are easily damaged by shear stresses. Figure B-3 – Epoxy Points for Steel Tube. Once the epoxy is inside the steel tube, arrange the fibers as shown in Figure B-3 and epoxy the steel tube into the furcation tubing. Fill the open end of the furcation tubing with epoxy, so that the steel tube is not free to move when the epoxy has dried. Use the epoxy syringe to place an epoxy bead at the free end of the steel tubing. To finish the device section of the fiber bundle, the fibers must be forced into a shape closely approximating the ideal fiber bundl e structure shown in Figure 1-10. This is done by wrapping the optical fibers protrudi ng from the steel tube with a small piece of silly putty or clay, which must hold its shape during the 24 hr. drying process. The location of the silly putty wrap is shown in Figure B-4. Device Section Furcation Tubing (FT030) Epoxy Points Move steel tube to smear epoxy

PAGE 123

112 Figure B-4 – Location of Epoxy Bead and Putty Wrap. The silly putty wrap should be squeezed with a small pair of needle-nosed pliers or tweezers to tighten the ring structure of the bundle. This is critical to the bundle performance. The final step of the bundle assembly is to connectorize the Tx and Rx furcation tubes. For the Tx segment, this process is identical to the connectorization procedure outlined in [41] using the 30126D1 FC connector. The Rx segment process differs slightly since multiple fibers are present. Also, the Rx segment requires an SMA connector. There are two connector models that can be used. The best connector is the 10440A, which has a 440 m diameter hole. All six Rx fibers will fit in the 10440A if the cladding has been stripped as described above. A bundle using this connector will be aligned more accurately with the photodetector. Alternately, if the fibers at the Rx connector have not all been properly stripped (it is difficult to correct alignment problems after the steel tube is assembled), then a 10770A SMA connector must be used. This Device Section Furcation Tubing (FT030) Epoxy Beads Putty

PAGE 124

113 connector has a 770 m diameter hole, and can hold the six Rx fibers if none have been properly stripped. Using a 10770A connector makes the polishing process more difficult, and it makes it difficult to align the Rx fibers with the photodetector. To improve the alignment of the Rx fibers with the photodetector, use a putty wrap like the kind used for the steel tube fibers. The assembled connector is illustrated in Figure II-5. Figure B-5 – Assembled Tx/Rx Connector. After both connectors have been assembled, the bundle must be maintained until the epoxy begins to harden (45 minutes). Every five minutes, retighten the putty wraps and reposition the fibers so that they come out of the center of the steel tube. When the epoxy begins to harden, wrap the junction of the three furcation tube sections with duct tape, and allow the epoxy to harden for 24 hours. After the 24-hour period has passed, the bundle is ready for connector polishing. The components needed for the polishing of the connector are shown in Table B2. The same workspace used to assemble the bundle can be used to polish the connectors. Tx/Rx Section Furcation Tubing (FT030) Epoxy Bead Putty (Rx only) Fiber(s) Strain Relief Boot

PAGE 125

114 Table B-2 –Materials List for Connector Polishing Thorlabs Part Number Component Quantity LFG5P Lapping Film, 5 m Grain 1 LFG3P Lapping Film, 3 m Grain 1 LFG1P Lapping Film, 1 m Grain 1 LFG03P Lapping Film, 0.3 m Grain 1 S90W Diamond Tipped Fiber Scribe 1 N/A (compare to CL-200) 100x Fiber Microscope 1 CTG913 Glass Polishing Plate 1 NRS913 Rubber Polishing Pad 1 D50-FC FC Polishing Disk 1 D50-SMA SMA Polishing Disk 1 Custom Steel Tube Polishing Disk 1 Razor Blade 1 KW32 Kim Wipes 1 Light Source 1 The first step in polishing the fiber bundle connectors is to inspect the bundle and verify the epoxy is dry. Also, verify the duct tape properly holds the furcation tubing sections together. Finally, ensure the connectors and putty wraps are not stuck to the table. When the bundle is verified to be mechani cally stable, polishing may begin. First, the fibers that protrude from their connector s must be cleaved. The cleaving process is described in [41], but it must be modified slightly when multiple fibers are present in a connector. The most fragile connector is th e FC connector, since it contains a single Tx fiber. To cleave the fiber, gently touch the end of the diamond scribe to the Tx fiber at a location no more than a fiber diameter beyond the epoxy bead (see Figure B-6).

PAGE 126

115 Figure B-6 – Location to Cleave Fibers in the Fiber Bundle. If the fiber is cleaved more than one fiber diameter from the epoxy bead, extra care must be taken during the initial polishing steps to avoid shattering the end of the fiber. Next, the fiber bundles protruding from the steel tube and the Rx connector must be cleaved. Since the bundle of fibers does not cleave with just a touch from the scribe, a sawing motion may be used with the scribe to cut the fibers. A razorblade may be used if the fibers do not cleave with the scribe. It is important that the fibers are not broken by Tx/Rx Section Furcation Tubing (FT030) Epoxy Bead Putty (Rx only) Fiber(s) Cleaving Location Protrudes < 1 Fiber Diameter

PAGE 127

116 bending at the cleave point. This can cause cracks to form in the fibers below the surface of the connector ferule, adversely affecting performance. After the fibers have been cleaved, they must be polished. The polishing process is identical to the process in [41]. In ge neral, the multiple fiber connectors will take longer to polish, so extra time must be spent with each grain of lapping paper. The appropriate polishing disk must be used with each connector. Instructions for using the disks are in [41]. The custom disk for the steel tube is used in the same way as the SMA and FC disks. Once the polishing is completed, the bundle must be inspected. The quality of the polish and the location of the fibers in the fiber bundle should be observed with the fiber microscope. Also, the fibers should be te sted by coupling light into the FC and SMA connectors, and observing the fiber bundle face in the steel tube. Make a note of fibers that do not transmit light. Once the inspection of the bundle is complete, the bundle is ready to use.

PAGE 128

117 APPENDIX C MECHANICAL DRAWINGS Figure C-1 – Custom Plane Wave Tube Mount. Set Screw Mounting Screw Holes Steel Tube Hole MEMS Chip Recess

PAGE 129

118 APPENDIX D PHOTODETECTOR SPECIFICATIONS [42] Table D-1 – PDA-400 Saturation Power Detector Saturation Voltage [42] 10 V Photodiode Response [42] 0.95 A/W MaxOptPower*TransGain 10.5 *W Trans-impedance Gain [42] (V/A) Maximum Incident Power (W) 15,000 701.7 47,000 224.0 150,000 70.1 470,000 22.4 1,500,000 7.01 Table D-2 – PDA-400 Measured Dark Noise Trans-impedance Gain [42] (V/A) Photodiode NEP [42] (pW / Hz ) Photodiode Dark Noise (V rms @ f = 1 Hz) 15,000 8.2 1.17 x 10-7 47,000 6.0 2.68 x 10-7 150,000 3.8 5.42 x 10-7 470,000 3.4 1.52 x 10-6 1,500,000 2.9 4.13 x 10-6 Table D-3 – PDA-400 Maximum Received Power at Gain Settings Trans-impedance Gain [42] (V/A) Max Received Power (W) 15,000 702 47,000 223 150,000 70.2 470,000 22.3 1,500,000 7.02

PAGE 130

119 APPENDIX E SPECIFICATIONS FOR POPPER & SONS STEEL TUBING [43] Tube Gauge ID Min ID Max OD Min OD Max 0.0150” 0.0180” 0.0320” 0.0325” 21HW 381.0 m 457.2 m 812.8 m 825.5 m 0.0230” 0.0245" 0.0355" 0.0360" 20RW 584.2 m 622.3 m 901.7 m 914.4 m 0.0270” 0.0285" 0.0355" 0.0360" 20XTW 685.8 m 723.9 m 901.7 m 914.4 m 0.0290” 0.0315" 0.0355" 0.0360" 20XXTW 736.6 m 800.1 m 901.7 m 914.4 m

PAGE 131

120 LIST OF REFERENCES [1] S. D. Senturia, Microsystems Design New York: Kluwer Academic, 2001. [2] N. Bilaniuk, "Optical Microphone Transduction Techniques," Applied Acoustics vol. 50, pp. 35-63, 1997. [3] V. P. Klimashin, “Optical Microphone,” Pribory i Tekhnika Eksperimenta, no. 3, pp. 135-137, May 1979. [4] Hu and F. W. Cuomo, “Theoretical and Experimental Study of a Fiber Optic Microphone,” J. Acoust. Soc. Amer., vol. 91, no. 5, pp. 3049-3056, May 1992. [5] M. H. de Paula and C. A. Vinha “High-sensitivity Optical Microphone for Photoacoustics,” American Institute of Physics, Review of Sci. Instrum., vol. 63 no. 6, pp. 3487-3491. June 1992. [6] W. Lukosz and P. Pliska, “Integrate d Optical Nanomechanical Light Modulators and Microphones,” Sensors and Materials, vol. 3 no. 5, pp. 261-280, 1992. [7] A. Suhadolnik, A. Babnik, and J. Mozina, “Microphone Based on Fiber Optic Reflective Sensor,” SPIE Proceedings, vol. 2510, pp. 120-127, 1995. [8] K. Kadirval, “Design and Characteri zation of a MEMS Based Optical Microphone for Aeroacoustic Measurement,” Master’s Thesis in Electrical and Computer Engineering. Gainesville: University of Florida, 2002. [9] P. J. Henderson, Rao, Y. J., Jackson, D. A., Zhang, L., and Bennion, I. “Simultaneous Multi Parameter Monitoring Using a Serial Fiber-Fabry-Perot Array with Low Coherence and Wavelength-Domain Detection,” Meas. Sci. Technol. vol. 9. pp. 1837-1839. 1998. [10] Norbert Furstenau, Horack, H., and Schmidt, W. “Extrinsic-Fabry-Perot Interferometer Fiber-Optic Microphone,” IEEE Transactions on Instrumentation and Measurement. vol. 47. no. 1. pp. 138-142. 1998. [11] Weichong Du, Tao X., and Tam, H. Y. “Temperature Independent Strain Measurement with a Fiber Grating Tapered Cavity Sensor.” IEEE Photonics Technology Letters. vol. 11. no. 5. pp. 596-598. 1999.

PAGE 132

121 [12] D. Greywall, “Micromachined Optical Interference Microphone,” Sensors and Actuators A Physical, vol. 75, pp. 257-268, 1999. [13] Y. J. Rao, Cooper, M. R., Jackson, D. A., Pannell, C. N., and Reekie, L. “Simultaneous Measurement of Displacement and Temperature Using In-FiberBragg-Grating-Based Extrinsic Fizeau Sensor .” Electronics Letters. vol. 36. no. 19. pp. 1610-1612. Sept. 14, 2000. [14] D. C. Abeysinghe, Dasguptam, S., Boyd J.T., Jackson H.E., "A Novel MEMS Pressure Sensor Fabricated on an Optical Fiber," IEEE Photonics Technology Letters vol. 13, pp. 993-995, 2001. [15] A. Wang, Xiao, H., Wang, J., Wang, Z., Zhao, W., May, R.G. “Self-Calibrated Interferometric-Intensity-Based Optical Fiber Sensors,” J. Lightwave Tech., vol. 19 no. 10, pp. 1495-1501, Oct. 2001. [16] G. He, Cuomo, F.W, "Displacement Response, Detection Limit, and Dynamic Range of Fiber-Optic Lever Sensors," Journal of Lightwave Technology vol. 9, no. 11, pp. 1618-1625, 1991. [17] M. Sheplak, Seiner, J.M., Bruer K.S, Schmidt M.A, "A MEMS Microphone for Aeroacoustics Measurements," 37th Aerospace Science Meeting and Exhibit AIAA 99-0606 Reno, NV. Jan 11-14, 1999. [18] G. He, Cuomo, F.W, "A Light Intensity Function Suitable for Multimode FiberOptic Sensors," Journal of Lightwave Technology vol. 9, pp. 545-551, 1991. [19] Y. Z. Ruan and L. B. Felson, “Reflection and Transmission of Beams at a Curved Interface,” J. Optical Society of America A, vol. 3 no. 4, pp. 566-579, April 1986. [20] M. Sheplak and J. Dugundji, “Large Defl ections of Clamped Circular Plates Under Initial Tensions and Transitions to Membrane Behavior,” J. Appl. Mech., vol. 65, pp. 107-115, March 1998. [21] R. Sahni, “Design of a MEMS-Based Piezoresistive Microphone,” Master's Thesis in Aerospace and Engineering Mechanics. Gainesville: University of Florida, 2001. [22] G. He, Cuomo, F.W, "The Analysis of Noises in a Fiber Optic Microphone," Journal of the Acoustical Society of America vol. 92, no. 5, pp. 2521-2526, 1992. [23] J. Wilson and J. Hawkes, Optoelectronics Hemel Hempstead Hertfordshire: Pretence Hall Europe, 1998.

PAGE 133

122 [24] Analog Devices, AD734 Product Data Sheet Norwood, MA: Analog Devices. [25] L. Coldren and S. Corzine, Diode Lasers and Photonic Integrated Circuits New York: John Wiley & Sons, Inc, 1995. [26] H. L. Chau, Wise, K.D., "Noise Due to Brownian Motion in Ultrasensitive Solid State Pressure Sensors," IEEE Transacti ons on Electron Devices, vol. ED-34, pp. 859-864, 1987. [27] S. Horowitz, “A MEMS-Based Si ngle Fiber Optical Lever Microphone,” High Honors’s Thesis in Electrical and Computer Engineering. Gainesville: University of Florida, 2000. [28] MEMS Exchange Web Site, www.mems-exchange.org Reston, Va. June 17, 2003. [29] D. C. Abeysinghe, Dasguptam, S., Boyd J.T., Jackson H.E., "A Novel MEMS Pressure Sensor Fabricated on an Optical Fiber," IEEE Photonics Technology Letters vol. 13, pp. 993-995, 2001. [30] M. H. Beggans, et al. “Optical Pressure Sensor Head Fabrication Using Ultra-Thin Silicon Wafer Anodic Bonding,” SPIE Symposium on the Design, Test, and Microfabrication of MEMS and MOEMS, Paris, France, vol. 3680, pp. 773-782, March – April 1999. [31] Bruel & Kjaer, Type 4231 Microphone Calibrator Product Data Sheet Naerum, Denmark: Bruel & Kjaer. [32] Bruel & Kjaer, Type 4145 1” Microphone Product Data Sheet Naerum, Denmark: Bruel & Kjaer. [33] Bruel & Kjaer, Type 4191 1/2” Microphone Product Data Sheet Naerum, Denmark: Bruel & Kjaer. [34] Bruel & Kjaer, Type 4138 1/8” Microphone Product Data Sheet Naerum, Denmark: Bruel & Kjaer. [35] Bruel & Kjaer, Type 4231 Microphone Calibrator Product Data Sheet Naerum, Denmark: Bruel & Kjaer. [36] Phone-Or, “Optical Microphone for Speech Recognition Applications,” Yehuda, Israel: Phone-Or. [37] Phone-Or, Surveillance Optical Microphone (SOM) Product Data Sheet OrYehuda, Israel: Phone-Or.

PAGE 134

123 [38] M. Papila, R. Haftka, T. Nishida, and M. Sheplak. “Piezoresistive Microphone Design Paretto Optimization: Tradeoff between Sensitivity and Noise Floor,” 44th Structures, Structural Dynamics, and Materials Conference AAIA 03-0710 July 2003. [39] F. W. Cuomo, T. D. Nguyen, and A. J. Zuckerwar. “High-Temperature Fiber-Optic Lever Microphone Incorporating a Single Fiber,” Journal of the Acoustical Society of America vol. 93, no. 4, p. 2523, 1993. [40] A. J. Zuckerwar, F. W. Cuomo, T. D. Nguyen, S. A. Rizzi, and S. A. Clevenson. “High-Temperature Fiber-Optic Lever Microphone,” Journal of the Acoustical Society of America vol. 97, no. 6, pp. 3605 3616, 1995. [41] Thorlabs, Inc, Guide to Connectorizing and Polishing Optical Fibers Newton, New Jersey: Thorlabs, Inc. [42] Thorlabs, Inc, PDA-400 Switchable Gain InGaAs Amplified Photodetector Product Data Sheet Newton, New Jersey: Thorlabs, Inc. [43] Popper & Sons, Type 304 Stainless Steel Needle Tubing Hyde Park, New York: Popper & Sons.

PAGE 135

124 BIOGRAPHICAL SKETCH Walter Lee Hunt, Jr. was born on July 1, 1977, in Ocala, Florida. He graduated from high school at Marianna High School in Mari anna, Florida, in 1995. Afterwards, he attended Chipola Junior College in Marianna until graduating in 1997 and transferring to the University of Florida. In August of 2000, the University of Florida awarded him Bachelor of Science degrees in computer engi neering and in electrical engineering. Since then, he has worked for the Interdisciplinary Microsystems Group at the University of Florida as a research assistant while pursuing a Master of Engineering degree, specializing in electrical engineering. Mr. Hunt’s research interests include the theoretical study and experimental character ization of optical MEMS sensors, and MEMS optical microphones in particular.


Permanent Link: http://ufdc.ufl.edu/UFE0000952/00001

Material Information

Title: Design and Characterization of an Intensity Modulated Optical MEMS Microphone
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000952:00001

Permanent Link: http://ufdc.ufl.edu/UFE0000952/00001

Material Information

Title: Design and Characterization of an Intensity Modulated Optical MEMS Microphone
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000952:00001


This item has the following downloads:


Full Text









DESIGN AND CHARACTERIZATION OF AN INTENSITY
MODULATED OPTICAL MEMS MICROPHONE



















By

LEE HUNT


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ENGINEERING

UNIVERSITY OF FLORIDA


2003

































Copywright 2003

By

Lee Hunt















ACKNOWLEDGEMENTS

I wish to give my love and gratitude to my friends and family for all the love and

support I have received during the many years of my college career. I would never have

succeeded to the extent that I have without them. I especially appreciate the love and

support from my fiancee, Ms. Lisa Sewell, and for her ability to see through my

"preoccupied graduate student" exterior and find that which lies beneath.

I would also like to express my gratitude to my advisor, Dr. Toshikazu Nishida,

for recruiting me into IMG and giving me the educational tools and support needed to

complete a difficult project. Thanks also go to Dr. Mark Sheplak, Dr. Lou Cattafesta, and

Dr. Peter Zory for the excellent work they have done in teaching the classes that provided

the foundation for my research, and also for their guidance during the process. I also

wish to thank my friends and colleagues in the Interdisciplinary Microsystems Group

who have provided support. Special thanks go to Karthik Kadirval for his support in the

beginning of my work on the optical microphone project, and to Stephen Horowitz and

Robert Taylor for timely assistance when I needed it.

Financial support for this project is provided by DARPA (Grant #DAAD19-00-1-

0002) through the Center for Materials in Sensors and Actuators (MINSA) and is

monitored by Dr. Paul Holloway.














TABLE OF CONTENTS


A C K N O W L E D G E M E N T S ............................................................................................... iii

A B STRA C T ............... ....................................... .............. ................. x

CHAPTER

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

1.1. Optical M icrophone Transduction Schemes....................... ............................. 1
1.1.1. Intensity M odulation ................ ................................ .... .. .. ...... 2
1.1.2. Polarization M odulation............................................. ........................... 5
1.1.3. Phase M odulation. ............................ ..... .. ... ............. .. .. ................ 6
1.1.4. Suitability of Transduction Techniques for MEMS Implementation ............... 9
1.2. M icrophone Structure .. ...... ................................................. .............. 12
1.2.1. Overview ......................................... ............ 12
1.2.2. M EM S Chip .................. .......................................... .......... 14
1.2.3. O optical Fibers ................................................. ......................... ....... 15
1.2.4. L eight Source .......................................................................... 18
1.2 .5 D election E electronics ........................................................................... .... 18

2 MICROPHONE SYSTEM PARTITIONING AND PERFORMANCE METRICS... 21

2 .1. Sy stem P partition in g ........................... .......................................... .................. 2 1
2.1.1. A cousto-M mechanical Stage ........................................ .................. ...... 21
2 .1.2 M echano-O ptical Stage............................................................................ ... 2 1
2 .1.3 O pto-E electrical Stage ........................................................................... .... 22
2.2. System Perform ance M etrics ........................................................ ......... ..... 22
2.2.1. System Sensitivity. ............................................. .. ............. .............. .. 23
2.2.2. System Linearity ...... .............................................................. 39
2.2.3. System Frequency R esponse................................ ............... ... ................. 44
2.2.4. System Electronic N oise ........................................................... ......... .... 46
2.2.5. System M minimum Detectable Signal ....................................... ........... ... 51
2.2.6. Optical Reference Path Losses and System Performance Metrics ............... 55
2.2.7. Summary of Predicted System Performance ............................................... 57

3 DESIGN OF THE OPTICS FOR THE MEMS OPTICAL MICROPHONE ............ 59

3 .1. Selection of th e O ptics ................................................................. ................ .. 59
3.1.1. Perform ance ....................................................... ............ ... 59
3.1.2. System Connectivity ................... ................ ......... .............. 61
3.1.3. Ease of Handling and Manufacturability ................................................... 61

iv









3 .1 .4 C o st ................................................................................................. 6 2
3 .2 Selection of the T ubing ................................................................................. 63
3.3. Alignm ent Issues. .................................................... ............... 64
3.3.1. M EM S Chip Cavity A lignm ent issues.................................. .................... 64
3.3.2. Fiber Bundle G eom etry Issues.................................................... ................. 67
3.3.3. Application of Alignment Theory to Fiber Bundle Selection........................ 68

4 FABRICATION OF THE OPTICAL MICROPHONE ......................................... 70

4.1. M E M S E change P rocess............................................................ .................... 70
4.2. Packaging Process................... .......... .................... ...................... .. ................ 71

5 EXPERIMENTAL SETUP AND RESULTS................. ................. 79

5.1. Laser and Photodetector Characterization........................................................... 80
5.1.1. Experimental Setup for Laser and Photodetector Characterization.............. 80
5.1.2. Results of Laser and Photodetector Characterization.................................... 81
5.2. Static C alibration .............. ...... ..... ................... .............. 81
5.2.1. Experimental setup for static calibration ............................................. .. 81
5.2.2. R results of static calibration ........................................................... ... ......... 84
5.3. Dynam ic Calibration........ ... ..................... .................. .............. 87
5.3.1. Experimental setup for dynamic calibration ............................................. 87
5.3.2. Results of the dynamic calibration......................................... ............. 90

6 CONCLUSIONS AND FUTURE WORK ......................................................... 99

6 .1. C o n clu sio n s...................................................... ............... 9 9
6.2. Future W ork ..................................... .............................. .......... 100

APPENDIX

A MEMS OPTICAL MICROPHONE DIAPHRAGM PROCESS FLOW.................. 102

B FIBER BUNDLE PROCESS FLOW ................................................. 105

C M ECH AN ICAL D R A W IN G S........................................................ .... .. .............. 117

D PHOTODETECTOR SPECIFICATIONS.......................... .......... .... 118

E SPECIFICATIONS FOR POPPER & SONS STEEL TUBING ........................... 119

L IST O F R E F E R E N C E S ......... ............... .................................................................. 120

B IO G R A PH IC A L SK E T C H ........................................ ............................................124
















LIST OF TABLES

1-1 Summary of Intensity-Modulated Optical Microphone Designs ............................ 10

1-2 Summary of Phase Modulated Optical Microphone Designs ............................... 11

2-1 Acousto-Mechanical Lumped Element Parameters ............................................ 45

2-2 Summary of Configuration Settings for Theoretical Performance Metrics ............ 57

2-3 Summary of Theoretical System Performance Metrics.................................... 58

3-1 Error Analysis of Different Fiber Bundle Configurations..................................... 69

4-1 Wafers Used for Optical Microphone Fabrication ............................................... 70

5-1 Experim ental H P 8168B N oise....................................................... ... ................. 81

5-2 Comparison between Theoretical and Experimental Static Calibration................ 86

5-3 Experimental Results of Unreferenced Output Microphone Dynamic Calibration 98

5-4 Experimental Results of Referenced Output Microphone Dynamic Calibration.... 98
















LIST OF FIGURES

1-1 Optical Microphone Classification Based on Transduction Mechanism [2]............. 2

1-2 Radiated Wave Intensity-modulating Microphone Types.................................... 3

1-3 Evanescent Wave Intensity-modulating Microphone Types.................................. 4

1-4 Polarization M odulating M icrophone Types.................................. .................. ... 6

1-5 Grating-Type Phase Modulating Microphone Types......................................... 7

1-6 Interferometric Phase M odulating M icrophone Types.............................................. 8

1-7 Block Diagram of the Optical Microphone. ......................... ...................... 13

1-8 Cross Section of the Fiber Bundle in the MEMS Chip. .......................................... 14

1-9 Cross Section of the MEMS Chip. .............................................. ................ 15

1-10 End View of the Optical Fiber Bundle. ............... .............................. 16

1-11 Optical Fibers in Steel Tubing.................................................................. 16

1-12 Optical Fiber Bundle Drawing. ........................................ ......................... 17

2-1 Side View of Deflecting Plate or Membrane.................................. 24

2-2- Method of Images (View from Side of Fiber Bundle). ................... ................. 27

2-3 R ing A pproxim ation D iagram ................................................................................ 30

2-4 Theoretical Power Coupled with Ideal Fiber Configuration. ................................. 32

2-5 Theoretical Sensitivity with Ideal Fiber Configuration .............. .............. 32

2-6 Block Diagram of the Unreferenced Output Configuration. ............. ................ 34

2-7 Equivalent Circuit for the PDA400 Photodetector ............................................... 35









2-8 Block Diagram of the Referenced Output Configuration................. .......... 36

2-9 Comparison of Unreferenced and Referenced Output Sensitivities..................... 39

2-10 Linearity of Mechano-Optical Stage. ............................... ....................... 41

2-11 -Plot of Acousto-Mechanical Sensitivity as a Function of Radial Position............ 42

2-12 Noise Contributions for the Photodetector Output......................................... 46

2-13 Noise Contributions for the Microphone Output ............................................ 48

2-14 Illustration of the Physics Behind the MO MDS............................................. 53

3-1 Bundle Position Error Illustration. ........................................ ...................... 65

3-2 Angular Misalignment Error Illustration........ ............ ...................... 66

3-3 Radial Position Error Illustration ............................................... 67

4-1 Abeysinghe et al. Packaging Technique............ ............................. .............. 71

4-2 Beggans et al. Packaging Technique. ...................................................... 73

4-3 Kadirval Packaging Technique................ ........... ................. ...... ..... ......... 74

4-4 Proposed Package for the Optical Microphone. .............................................. 76

4-5 Proposed Optical Microphone Array Package. ..................................... ........ .. 77

5-1 Experimental Setup for Laser Characterization...................... .... .............. 80

5-2 Block Diagram of Static Calibration. ...................................................... 84

5-3 Experimental Power Coupled vs. Equilibrium Gap. ............................................. 85

5-4 Experimental Maximum Power Coupled Regression Line Slope .......................... 86

5-5 Unreferenced Output Optical Microphone Configuration...................................... 88

5-6 Referenced Output Optical Microphone Configuration .................................... 89

5-7 Linearity, Unreferenced Output Configuration. .............................................. 91

5-8 Linearity, Referenced Microphone Configuration. .............................................. 92









5-9 Magnitude Response, Unreferenced Microphone Configuration............................ 93

5-10 Magnitude Response, Referenced Microphone Configuration. ......................... 94

5-11 Phase Response, Unreferenced Microphone Configuration............................. 94

5-12 Phase Response, Referenced Microphone Configuration. ................................... 95

5-13 Electrical Noise Floor, Both Microphone Configurations.............. ........... 97















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Engineering

DESIGN AND CHARACTERIZATION OF AN INTENSITY
MODULATED OPTICAL MEMS MICROPHONE

By

Lee Hunt

August 2003


Chairman: Toshikazu Nishida
Major Department: Electrical and Computer Engineering


This thesis presents the design and characterization of an intensity-modulated

optical lever microphone. Microphone noise models from previous works are expanded

to include the light source and all electronics. Physical phenomena responsible for

limiting the microphone minimum detectable signal (MDS) are identified, and an

accurate model developed for use with an LED or laser light source. The sensitivity,

minimum detectable signal, and electronics noise are characterized by a scaling analysis

in which coupled equations for dependence on optical power, membrane radius,

photodetector gain, and optical losses in the reference path are presented. It was

discovered that, in this optical microphone geometry, the laser is the limiting factor in the

microphone MDS and electronics noise, and optical losses in the reference path can

improve microphone sensitivity, MDS, and noise floor for a referenced optical

microphone.









An unreferenced electronic configuration and a referenced electronic

configuration were experimentally characterized using a laser as a light source. The

unreferenced optical microphone achieved a sensitivity of 0.032 mV / Pa, MDS of 65 dB

(re. 20 [tPa), and dynamic range from 65 122 dB (re. 20 tpPa). The referenced optical

microphone achieved a sensitivity of 1.77 mV / Pa, MDS of 47 dB (re. 20 [tPa), and

dynamic range from 47 122 dB (re. 20 [tPa). Both unreferenced and referenced

measurements were made at 1600 Hz with a bin width of 2 Hz.














CHAPTER 1
INTRODUCTION

Optical microphones vary widely in their construction, but all possess innate

resistance to electro-magnetic interference (EMI) and other harsh environments to which

other types of microphones are sensitive. This innate resistance is derived from the

separation of the optical sensing element from the electronics via optical fibers and

assumes the electronics are remotely located with respect to the test environment. In the

case where the electronics are not remotely located, the microphone package must isolate

the microphone electronics from the test environment.

MEMS technology provides a promising new implementation for optical

microphones. MEMS devices have the capability to be smaller than conventional

microphones, and MEMS microphone chips could be processed by the thousand on

wafers if the market can support this volume. Despite these advantages, Professor Steve

Senturia [1] notes that MEMS devices have a coupling between the package and the

device, thus requiring them to be designed concurrently, which makes a MEMS

microphone design inherently more complicated than a conventional (non-MEMS)

microphone.

1.1. Optical Microphone Transduction Schemes

In 1996, Nykolai Bilaniuk first introduced a classification scheme for optical

microphones that relied on the transduction mechanism as the primary sorting criterion

[2]. He also explained the methods of operation of multiple types of devices in each









category, with emphasis on the most promising technologies, and he also discussed

microphone system performance metrics.

Bilaniuk defined three properties of light that could be modulated: the intensity

(or irradiance), the phase, and the polarization [2]. Since electro-optical detectors

respond to light intensity, all modulation schemes must be reduced to an intensity

modulation at the electronics end of the system. The figure below adapted from [2]

shows a detailed classification scheme for optical microphones.


Figure 1-1 Optical Microphone Classification Based on Transduction Mechanism [2].

1.1.1. Intensity Modulation

Bilaniuk [2] describes an intensity-modulated microphone as one which

selectively removes energy from the optical path. As shown in Figure 1-1, an intensity-









modulating optical microphone can be subdivided into two broad categories: radiated

wave and evanescent wave. All of the energy in radiated wave optical microphone leaves

a controlled optical path and partially recaptured or backscattered [2]. Figure 1-2

recreates Bilaniuk's [2] illustration of the radiated wave transduction strategies.


Moving Grating


Lever


Macrobend


Figure 1-2 Radiated Wave Intensity-modulating Microphone Types.


The moving grating approach relies on the motion of a "light gate" to modulate

the light coupled between an input waveguide and an output waveguide. These types of

devices do not make use of diffraction or any structures on the order of the wavelength of

the light.

An intensity-modulated lever microphone utilizes one or more waveguides to

deliver light to a vibrating plate or membrane. Reflected light is collected by one or more


Cantilever









waveguides and delivered to a photodetector. Lever microphones may also have

focusing optics to improve light collection.

In a cantilever microphone, the waveguide is discontinuous, and part is free to

vibrate in an acoustic field. This varies the alignment between the fixed segment and the

free segment of the waveguide, causing a modulation of the power coupled.

Macrobend-type intensity-modulating schemes use acoustic waves to deform a

fiber configuration, such as a coil. Optical fibers are chosen that do not completely

confine the light. The deformation modulates the losses in the length of fiber,

subsequently modulating the output power.

Alternatively, the evanescent-wave coupling methods "rely on ... mode coupling

or on absorption from the evanescent field" [2]. Bilaniuk defines two classes of

evanescent wave intensity-modulating microphones: microbend and coupled waveguide.

Figure 1-3 recreates Bilaniuk's [2] illustration of the evanescent wave intensity

modulation techniques.


Figure 1-3 Evanescent Wave Intensity-modulating Microphone Types.


Coupled Waveguide


Microbend









The microbend technique uses a microstructure to apply periodic deformations to

a waveguide. The acoustic field modulates the pressure exerted on the waveguide by

these deformations, which in turn causes leakage of power out of the waveguides.

The coupled waveguide technique can work in one of two different ways. In the

first way, the waveguides are fabricated on a membrane structure with a fixed separation

between the two. The membrane deflects in the presence of an acoustic field, and this

deflection changes the index of refraction in the two waveguides. The change in

refractive index modulates the power coupled between the waveguides. Alternately, the

waveguides are fabricated so that one is attached to a structure, while the other is free to

vibrate. An acoustic field will modulate the separation between the waveguides, which

modulates the power coupled between the two.

1.1.2. Polarization Modulation

The second major category of optical microphones as defined by Bilaniuk [2] is

polarization modulation. Polarization modulation type devices alter the polarization of

the light when in the presence of an acoustic field. Bilaniuk [2] subdivides polarization

modulation devices into two subcategories, but he notes that alternate schemes are

possible. Figure 1-4 adapted from [2] depicts the two subcategories.

In the first category, a layer of liquid crystals is subjected to acoustic field

induced shear stresses, which modulate the polarization of the light passing through. A

polarizer is located at the output of the device to isolate the desired polarization axis.

In the second category, "a moveable dielectric plate interacts with the evanescent

field of a waveguide excited with both TE and TM modes, causing a different change in









the refractive index of the two modes, according to Bilaniuk [2]". A polarizer at the

output isolates the desired polarization axis.
















Nematic Liquid Crystal Differential Index Shifter

Figure 1-4 Polarization Modulating Microphone Types.

1.1.3. Phase Modulation

Phase modulated optical microphones are described by Bilaniuk [2] as a

mechanism that "changes either the physical length or the refractive index of an optical

test path and recombining the result with the signal from a reference path." The reference

path is unaffected by the acoustic field, while the test path undergoes some form of

mechanical deformation. The two defined subgroups for this category of optical

microphones are grating type devices and interferometric devices.

A grating type device is one with a structure machined onto a waveguide with

features on the order of the wavelength of the light. The two different subcategories of

grating devices defined by Bilaniuk [2] are input coupling gratings and dynamic

refractive gratings. They are shown in the following figure, adapted from Bilaniuk [2].
























Dynamic Photorefractive Grating


Figure 1-5 Grating-Type Phase Modulating Microphone Types.

The input coupling grating device has a grating fabricated on the waveguide.

Incident light at the proper angle, wavelength and with the proper grating spacing will be

coupled into the waveguide. The acoustic field modulates a nearby dielectric structure,

varying the index of refraction of the system and modulating the output.

The dynamic photorefractive grating uses a prism to split light onto two mirrors,

one of which is free to vibrate in an acoustic field. The light reflects off the mirrors to

pass through a grating, and the light from each mirror is captured by a photodetector.

The light from the stationary mirror is used as a reference signal, while the light from the

vibrating mirror is used as the modulated signal.

The second major category of phase modulating optical microphones is

interferometric-type phase-modulating microphones. They typically use one of the three

most familiar types of interferometers: Fabry-Perot, Michelson, or Mach-Zehnder.

Alternately, a two-mode fiber can be used to make a phase modulated microphone. The

figure below (adapted from [2]) depicts the four interferometric optical microphone

categories.


Input Coupling Grating






















Fabry-Perot Michelson










Mach-Zehnder Two-Mode Fiber

Figure 1-6 Interferometric Phase Modulating Microphone Types.

The Fabry-Perot optical microphone uses an optical cavity formed between two

parallel surfaces. One of the surfaces is free to vibrate in an acoustic field, while the

other is fixed. Typically, the vibrating surface is a plate or membrane, and the fixed

reflecting surface is the face of the fiber, but additional optics may be used instead.

A Michelson optical microphone splits a free-space beam into two paths. The

reference path is reflected of a stationary reflector. The test path is reflected off of a

reflector that vibrates in an acoustic field. The beams recombine and interfere, and the

recombined signal is received by a photodetector.

In a Mach-Zehnder optical microphone, the light enters via a waveguide, which is

split into two paths. The reference path is held constant, but the test path is free to vibrate









in an acoustic field. The light in the two paths is recombined and sent to a photodetector.

Interference effects will modulate the power seen by the detector.

The fourth type of interference optical microphone is a two-mode fiber

microphone. In this design, a section of two-mode optical fiber is spliced at the end of a

single mode fiber. The two-mode fiber is free to vibrate in an acoustic field. Acoustic

vibrations will modulate the index of refraction of each mode differently, and an

interference pattern will be generated at the junction between the two fibers.

1.1.4. Suitability of Transduction Techniques for MEMS Implementation

In general, the simplest type of microphone to analyze and build is an intensity-

modulated device. The simplest intensity-modulated device can be constructed with an

LED, multimode or single mode fibers, a membrane or other vibrating reflective surface,

and a photodetector.

Table 1-1 (see [3] [8]) summarizes recent intensity-modulated optical

microphone designs. The results indicate a large variability in performance with the

implementation of the intensity-modulated microphone. While this observation may

seem obvious, it reinforces the importance of optimizing the system as a whole when

designing the microphone and not just an individual stage.

In general, for the intensity-modulated optical microphone, increasing the

diaphragm radius increases the sensitivity and decreases the minimum detectable signal

(MDS). Therefore, intensity-modulated microphone performance is decreased when the

diaphragm is constrained to have a diameter of less than a few hundred microns.











Table 1-1 Summary of Intensity-Modulated Optical Microphone Designs

Author / Year Design Type Source and ,X Sens Noise Freq Linearity MDS
Response Range
20-22dB 0 20kH
V. P. Klimashin Lever, -w- Support Incandescent 7.5 mV Pa (re -w- 5dB
1979 [3] Optics, non MEMS Lamp 0 fluctuatio
20pLPa)
ns

Hu and Cuomo Lever, -w- Mylar LED, 36.5 mV 0-31.5
Membrane, Non
1992 [4] Membrane, Non 2.4mW Pa kHz
MEMS

De Raula and Multiple Light 150 W Xenon 5.6 nW
Vinha Source non-MEMS Arc La Hz05
1992 [5] scheme

Lukosz and Evanescent wave, 49 dB t -
Pliska microbend, 6x6 mm2 L6se, n= 0.31 Pa1 (re to 49 dB 95
632.8 nm 10kHz dB
1992 [6] membrane 20pPa)
High due
Suhadolnik, et Lever, fiber bundle to
al and deflecting speckle MO Stage
al.
1995 [7] diaphragm, non- pattern 1500 gm
MEMS oflaser
light

Kadirvel Lever, fiber bundle 2 t 110 dB 1 kHz 110 dB 110 dB (re
2002 [8] and deflecting 1550 nm Pa (re 6.4 kHz 135 dB 20 gtPa)
diaphragm 20tPa)




The choice of light source and photodetector also plays a large role in the


performance of an intensity-modulated OM. Both affect the device sensitivity and noise


floor. Sensitivity increases as coupled optical power increases, so high intensity light


sources provide higher sensitivities, provided that the photodetector does not saturate.


A disadvantage of all intensity-modulated OMs is the large DC component of the


received signal. The DC component does not contribute to the device sensitivity, but it


does contribute to photodetector saturation. This limits the product of the optical


received power and the detector trans-impedance gain. The maximum intensity of the


light source is limited by the linearity range and gain of the detector.


Table 1-2 (see [9] [15]) summarizes recent phase modulated optical microphone


(PM) designs. Since no standard method of reporting the sensitivity of an optical


microphone has been agreed upon, it is difficult to compare the overall performance of












different PM designs. Theoretically, a PM device would be able to provide higher


performance than an intensity-modulated microphone in a MEMS implementation,


especially for membranes constrained to be smaller than a few hundred microns in


diameter. PM devices have a smaller DC component allowing for more flexibility in


selecting photodetector gain settings.


Table 1-2 Summary of Phase Modulated Optical Microphone Designs

Source and Freq Linearity Reouto
Author / Year Design Type Source an Sens SNR Freq Linearity Resolution
o. Response Range
20mW LED
Rao et al. 1997 Bragg Grating w- 12 pm 50dB >kHz
[9] Fizeau Cavity = l5000ES
,o-1550nm

0.5mW Varies by Tested
80dB over
Furstenau et al. t C (after pigtail) r over
1998 [10] Fabry-PerotLED fwreq range B 100Hz to
,-=1300nm 41w34 15kHz
4134
Du et al. 1999 LED @
De[11] Fiber Bragg Grating L o 1.5 pm / E NA < 1200 +/- 29pE
[11] Ao-1550nm
Surface-machined
Graywall 1999 Suace-machd LED @ 100Hz to
[12] Fabry-Perot Cavity, 650n 8.9 mV / Pa > 100 2kHz
__________ theoretical analysis _________ =m
20mW LED
Rao et al. 2000 Fiber Bragg Grating 54020mW
[13] and Fizeau Cavity o=1 50m
L o 1550nm
0.11 mV/
Fabry-Perot Cavity LED @ psi 0 -80 psi
Abeysinghe et machined on surface L 80 NA (0- 552
al. 2001 [14] of optical fiber 850n (16 mV kPa)
MPa)
0 -6000
Wang et al. Non-MEMS Fabry- LED @ 4 nm psi psi 0.02 psi
2001 [15] Perot Cavity =850nm (0.58m NA (0 (1379 Pa)
kPa) _41.4MPa)



Despite these advantages, PM microphones present some significant challenges.


The dimensions involved are on the order of tens of optical wavelengths, making static


characterization and packaging very difficult. PM microphones are much more sensitive


to misalignments and phase noise sources than an intensity-modulated microphone.


Because of this, PM implementations require more complicated electronics for signal


demodulation, and they have stricter requirements for the light source. Finally, a PM









microphone has a periodic power coupled curve, constraining the microphone to either a

very small membrane deflection or to a "peak-counting" scheme during demodulation.

Due to the additional complexity involved in implementing a PM microphone and

the mixed results achieved by previous implementations (Table 1-2), an intensity-

modulated lever-type transduction scheme was chosen for this work.

1.2. Microphone Structure

1.2.1. Overview

The intensity-modulated optical microphone that is the topic of this thesis can be

divided into four major physical parts. They are the MEMS chip, the optical fibers, the

light source, and the detection electronics.

The following figure shows the block diagram for the optical microphone. In the

steady-state case, light from the light source is coupled into the transmit (Tx) fiber. The

Tx fiber delivers the light to the MEMS chip, where it is reflected and partially coupled

into the receive (Rx) fiber. The Rx fiber then delivers the light to a photodetector, where

it is converted into an electrical signal and processed by detection electronics. When an

acoustic field is present at the MEMS chip, the coupled optical power is modulated. This

allows the transducer to convert acoustic energy into electrical energy, which is the

definition of a microphone.












Acoustic Waves


MEMS Chip


Light Source









Detector and Electronics


Rx Fiber


Figure 1-7 Block Diagram of the Optical Microphone.


There are four energy domains present in this system that carry information. The

first domain is the acoustic domain, where the desired measurement lies. The MEMS

diaphragm converts the acoustic energy into mechanical energy through its displacement.

The mechanical displacement of the membrane varies the power coupled into the Rx

optical fiber, converting the signal into the optical domain. At the photodetector, the

signal is converted into the electrical domain for analysis.

For our design, we have chosen a reflective-type intensity-modulated optical lever

microphone, with the mechano-optical transduction mechanism shown in Figure 1-8.

The dominant reason for this selection is that this type of intensity-modulated optical

microphone is much simpler to design and package than other intensity-modulated

microphones. Details of each component are described in later sections of this chapter.


Tx Fiber





























Rx Rx

Figure 1-8 Cross Section of the Fiber Bundle in the MEMS Chip.


1.2.2. MEMS Chip

The MEMS chip is a 2.5 mm x 2.5 mm silicon chip with a micromachined 1 mm

diameter silicon nitride diaphragm. The process flow for the MEMS chip is discussed in

Section 4.1.

A cross section of the MEMS chip is shown in Figure 1-9. The dominant

membrane material is a 1 |jm thick layer of silicon nitride. A 70 nm thick layer of

aluminum is deposited on the membrane surface to enhance reflectivity. The cavity

formed by the bulk silicon and silicon nitride membrane is fitted over the end of a steel

hypodermic tube containing the optical fibers.


MEMS Chip







Protective Steel Tubing
Epoxy


Light Cone


























Figure 1-9 Cross Section of the MEMS Chip.


1.2.3. Optical Fibers


The optical fibers selected for the optical microphone are the Thorlabs

AFS105/125Y multimode optical fibers. They are used for both transmit (Tx) and

receive (Rx) fibers. The end of the optical fiber that terminates at the MEMS chip is

designated the device end, and the end connected to the light source / photodetector is

designated the Tx / Rx end. One fiber acts as a Tx fiber, and six fibers are Rx fibers.

Figure 1-10 shows the desired shape of the fiber optic bundle as seen from the nitride

membrane into the steel tubing. In this figure, the cores of each fiber are color-coded,

and surrounded by a white ring representing the cladding. The dashed line is a possible

location for the border of the light cone reflected by the membrane. The receive fiber

area inside the dotted ring is responsible for collection of the reflected light.


Aluminum (70 nm)
Nitride (1 !lm)
Oxide (0.7 !lm)


Bulk Silicon (-500 !lm)


m m
! !


























Figure 1-10 End View of the Optical Fiber Bundle.


End View


Side View


Steel Tubing


Epoxy


Receive Fiber

Transmit Fiber


Figure 1-11 Optical Fibers in Steel Tubing.

Figure 1-11 shows the fiber bundle inside the protective steel tubing. The end

view shown is the ideal position, where the transmit fiber is located in the exact center of

the tube. This allows the light to reflect off of the center of the membrane, which has the

maximum acousto-mechanical sensitivity (defined here as change in membrane

deflection per change in acoustic pressure).









Figure 1-12 is a diagram of the fiber bundle in its protective tubing. Connections

to other parts of the system are noted. The dashed arrows indicate the path of light

through the system. Paths three and four contain modulated data.


From laser
Steel Hypodermic Needle
/ Transmit




3 ""' To Detector

Emitted Light Device A
(no reflecting
membrane present) Receive
Receive


Figure 1-12 Optical Fiber Bundle Drawing.

Based on the work of He and Cuomo [4], the mechano-optical (MO) stage

sensitivity, defined as change in coupled optical power per change in membrane

displacement, is maximized when a single transmit fiber is surrounded by a tightly

packed ring of receive fibers (see Figure 1-10). The smaller the radius of the receive

ring, the greater the sensitivity of the MO stage, and the smaller the equilibrium gap,

which is defined as the equilibrium distance between the fiber bundle face and the

membrane.

It may be possible to increase the sensitivity of the MO stage by adding extra Tx

fibers [16]. The current fiber bundle is designed to sample the displacement of the

membrane at the center (at the maximum displacement). For membranes which are much

larger than the accompanying fiber bundles, all bundles could illuminate areas near the

center of the membrane, where Sam is high and the stage is linear. If the diameter of the









bundle structures is on the order of the membrane, then sensitivity gains will be lessened

and linearity of the stage may become an issue. Sensitivity and linearity are discussed

later in this thesis.

Since the MO stage sensitivity is a result of the displacement of the illuminated

region of the membrane, extra MO sensitivity can be obtained by illuminating additional

portions of the membrane by additional fiber bundle structures. With the assumptions

that adding additional identical bundle structures does not remove any light collection

ability, the same electro-optic sensitivity is available to each bundle structure, and the

region of the illuminated membrane is locally flat, then the MO stage will remain linear

and the system sensitivity would be given by the following equation.


S= SoeSmoZ,, Equation 1-1


1.2.4. Light Source

The light source used by this optical microphone is the HP8168B Tuneable Laser

Source. The maximum output power of the laser at 1550 nm is 0.515 mW. An alternate

laser source or an LED source could be used in place of the HP8168B.

1.2.5. Detection Electronics

There are three schemes considered by this thesis for use as detection electronics.

The first scheme uses a single photodetector and takes the unreferenced output of the

photodetector as the microphone output. This technique is called the unreferenced output

technique. This scheme can be used with an amplifier at the output of the microphone to

increase the gain. The primary advantage of this scheme is simplicity. Fewer optical and

electronic components are required here than for any other configuration (see Section









5.2.1 for details). This greatly reduces the cost of the optical microphone when compared

with the other opto-electronic configurations. The largest disadvantage of this

configuration is the dependence of the unreferenced OE sensitivity on the optical power

(shown in Section 2.2.1.3). This dependence makes the unreferenced optical microphone

much less stable in the presence of laser instability and drift. Also, the overall sensitivity

of the unreferenced OE microphone is less than the referenced microphone configuration.

The second scheme, which was used by Kadirval [8], is the referenced output

technique. It uses an optical splitter to separate the light source output into two paths.

One path is connected to the Tx fiber for transmission to the MEMS chip. The Rx output

is the modulated data signal, and is taken to a photodetector. The second path is

connected directly to a photodetector for use as a reference signal. An analog divide

circuit is used to divide the modulated signal by the reference signal, and the divided

signal is taken as the microphone output. The largest advantage of the referenced output

configuration is the independence of the sensitivity on light source power. This

minimizes the negative effects of low frequency fluctuations in the light source output

power, such as fluctuations due to temperature changes. Another advantage of the

referenced output configuration is the ability to significantly improve microphone

performance by adding optical losses to the reference signal path (shown in Chapter 2).

Additionally, an amplifier may be used at the output to further increase sensitivity.

Despite the advantages, the referenced optical microphone requires more optics and

electronics than the unreferenced microphone (see Section 5.2.1 for details). This

increases the cost compared to the unreferenced microphone. Another disadvantage is









the increased electronics noise in the output due to the extra electrical components

(Section 2.2.4 for details).

The third scheme, heterodyne modulation, is designed to take advantage of the

flatness of the overall noise floor of the unreferenced optical microphone at high

frequencies. In this scheme, the laser output is modulated by an external sinusoidal

signal to frequencies much higher than the high frequency cutoff of the microphone. The

received optical microphone signal at the photodetector will be contained as a bandpass

signal centered at fo, where fo is much larger than the microphone bandwidth. After

passing through the photodetector, the signal is passed through a lock in amplifier and

demodulated back to the original baseband signal. This will make the noise floor of the

microphone dependant on the high frequency noise floor of the laser, and not the low

frequencies where 1/f noise (and other noise sources) is present. As with the previous

two OE configurations, an amplifier can be used at the output to increase sensitivity. The

major disadvantage of this electronic configuration is the increased electronic complexity

when compared to the other electronic configurations. Additionally, the lock-in amplifier

must be capable of passing frequencies at least 10 times the microphone high frequency

cutoff. Also, even small laser transients will cause the lock-in amp to fail to reproduce

the signal. This scheme was not implemented in this thesis, although it is likely that an

optical microphone system using a laser as the light source would require heterodyne

detection for satisfactory performance.
















CHAPTER 2
MICROPHONE SYSTEM PARTITIONING AND PERFORMANCE METRICS

2.1. System Partitioning

The intensity-modulated optical microphone is partitioned into three stages where

transduction between energy domains occurs. The three stages of an intensity-modulated

optical microphone were identified by Bilaniuk [2]. They are the acousto-mechanical

stage, the mechano-optical stage, and the opto-electrical stage. Kadirval [8] and Bilaniuk

[2] discuss these stages in detail, and a summary is included below.

2.1.1. Acousto-Mechanical Stage

The acousto-mechanical stage is where the energy in the acoustic signal is

converted into the mechanical domain. This is accomplished when the pressure and

volume velocity of the acoustic signal induce a displacement and restoring force in the

membrane. The unit of sensitivity for this stage is a displacement per unit pressure,

typically given in jtm / Pa.

2.1.2. Mechano-Optical Stage

In the mechano-optical stage, input optical power is reflected by the displacing

membrane and coupled into output (Rx) fibers. Transduction occurs when the

mechanical displacement of the membrane varies the percent of the input power that is

coupled into the output fibers. The unit of sensitivity of this stage is normalized power

per unit displacement, typically given in |jm-1












2.1.3. Opto-Electrical Stage

The third and final transduction stage in an intensity-modulated optical

microphone is the opto-electrical stage. This stage uses one or more photodetectors to

convert the coupled optical power into an electrical signal. The sensitivity units for this

stage are normally written as in volts (V). Occasionally an author will write the OE stage

sensitivity in V/(W/W). Most authors (including Bilaniuk) lump the optical power

dependence of the overall sensitivity of some microphone configurations into the OE

stage.

2.2. System Performance Metrics

Kadirval [8] used the following performance metrics to classify the optical

microphone. They are sensitivity, linearity, frequency response, noise floor and

minimum detectable signal (MDS). These metrics can also be used to describe the

performance of the individual stages. The theoretically determined performance metrics

for the system and each stage are summarized later in this chapter.

In this thesis, a theoretical sensitivity model for the referenced output

configuration is derived for the case where optical reference path losses and a low-noise

amplifier at the output are present. A theoretical model of the electronic noise is derived

for both unreferenced and referenced configurations. This model extends the noise

model derived by He & Cuomo [16] and used by Kadirval [8] to include the intensity

noise of the light source and all electronics. The physics behind the minimum detectable

signal equation presented by Kadirval [8] are explained, and the equation is used with the

improved noise model to predict the minimum detectable signal.









2.2.1. System Sensitivity

The sensitivity of a system is defined as the differential change of the output

quantity divided by the differential change of the input quantity. For a microphone, the

system output is a voltage, and the input is a pressure. The optical microphone is a multi-

energy domain system with three transduction stages, as previously described. The

maximum ideal sensitivity is a product of the sensitivities of the individual stages.

Equation 2-1 is the equation for the system sensitivity in terms of the individual stages.

All reported sensitivities in this thesis are based on a fiber bundle with identical Tx and Rx

fibers having an inner core diameter of 105 |tm and a cladding diameter of 125 |tm.


S = SamSmoSoe Equation 2-1


Section 2.2.1.3 examines the sensitivity of the OE stage in more detail then was

done by Kadirval [8]. It derives theoretical models for the OE sensitivity in the

unreferenced and referenced configurations, and it examines sensitivity limits of the stage

resulting from the finite linearity range of the photodetector.


2.2.1.1. Acousto-Mechanical Sensitivity

The acousto-mechanical stage converts pressure to a displacement. Equation 2-2

gives the sensitivity of the stage, where Wo is the deflection of the membrane at the

center, and p is the acoustic pressure at the center of the membrane.


Owo
Sam = Equation 2-2
Op

To derive Sa, first begin with Equation 2-3, the transverse deflection equation for

a plate derived by Sheplak et al. [17].











12(1- v2 p 4 a2 r
S k 2Eh3 2kl,(k) 4a2


Equation 2-3


Figure 2-1 Side View of Deflecting Plate or Membrane.


In the case of a membrane where a << kr, Equation 2-3 simplifies to Equation 2-4.


)= 2.78pa4Ehk2 2
w(r)= 1-2 a 2
SEhak2 2


Equation 2-4


Letting Wo = w(0) and substituting Equation 2-4 into Equation 2-3 produces the

equation for the sensitivity of the membrane as a function of radial distance from the

center.


W 2.78a4 I r2
Eh3 k 2 L a 2


Equation 2-5


If we assume that the light spot on the membrane is very small (10% or less) with

respect to the membrane diameter, then the sensitivity of the membrane can be lumped at









the radial center. Equation 2-6 gives the final equation for the acousto-mechanical

sensitivity of the membrane, lumped to the radial center.


2.78a4
Sa = Sm (0) = -a Equation 2-6
Eh3 k2

If it cannot be assumed that the light spot is small, then the membrane sensitivity

cannot be lumped into the center of the membrane. Sam will become a function of radial

position, r, with respect to the membrane center, and the microphone sensitivity will

decrease.

In this optical microphone, the light spot illuminates less than 10% of the

membrane. Based on the observed fiber position error (see Section 3.3.1) of less than 50

|tm for the fiber bundle used in this thesis, Sam can still be approximated as a constant for

this membrane.

The tension parameter k is determined by Equation 2-8. The in-plane stress (Go)

of the nitride layer for the nitride deposition process used in the microphone fabrication

was reported to range between 50 MPa and 120 MPa by the MEMS Exchange website.

Special fabrication instructions were given to minimize the in-plane stress in the nitride

layer, so it is expected that the stress will be equal to the minimum reported value for the

MEMS Exchange deposition process, Go = 50 MPa. Using E = 270 GPa (for SixNy), h =

1 jtm, a = 1 mm, and Vo = 0.27 (for SixNy), we estimate Sam = 1.249 x 10-3 tm / Pa.



k = o Equation 2-7
h E


For a discussion of the effects of the observed membrane linearity on Sa, see

Sections 2.2.2.1 and 5.3.2.









2.2.1.2. Mechano-Optical Sensitivity

The mechano-optical transduction stage converts a mechanical displacement to an

optical power coupling factor. The sensitivity of the stage is given by Equation 2-8,

where w is the deflection of the membrane at the center, and r is the coupled optical

power of the stage in W/W.


Sam = Equation 2-8


He and Cuomo [18] derived a formula for determining the power coupled by

light reflecting off of a deflecting membrane in a microphone similar to that shown in

Figure 1-8. The analysis is valid for multimode optical fibers. Theoretical work by Ruan

and Felson [19] can be used to derive the power coupled as a function of membrane

displacement for the case of a single mode transmit fiber and a multimode receive fiber,

although that configuration is not analyzed here. Ruan and Felson's work is applicable to

membranes with a finite radius of curvature, however He and Cuomo's work is not. The

analysis here based on [18] assumes no misalignment errors and no power lost due to

mismatch between fiber numerical apertures (NA). This is a good approximation when

the angular alignment between the fiber surface and the membrane is less than 5 degrees

(for fibers with NA = 0.22 or less). If this approximation does not hold, then the method

of images (explained below) is not valid. Adjusting the method of images to account for

angular alignments is complicated, and as of this writing, no work exists that rigorously

solves the problem. Section 3.3.1 discusses types of alignment errors, methods to avoid

them, and their implications in more detail.









In general, the power coupled into an optical fiber can be determined by

integrating the optical intensity (also known as the irradiance) over the collecting surface,

assuming all light present is entering the fibers at an angle less than the acceptance angle

of the fiber. If this is not the case, then only the irradiance due to rays entering the fiber

at less than the acceptance angle should be integrated in Equation 2-12. The analysis of

He and Cuomo [18] assumes the former. The reflected intensity profile at the surface of

the fiber bundle is determined in [18] by the method of images.



Image Plane Reflecting Plane Receiving Plane


Rx Fiber
Image Receive Fiber Core







Rx Fiber
Image Receive Fiber Core


g g



Figure 2-2 Method of Images (View from Side of Fiber Bundle).


In the method of images, the reflecting surface is defined to be the reflecting

plane, and the surface of the fiber bundle (as shown in Figure 1-10) is defined to be the

receiving plane. They are separated by a gap, g. The method of images states that the

reflected optical power incident onto the Rx cores is the same as the optical power

incident on the Rx core images, located at a distance of 2g from the receiving plane.









Using the method of images, He and Cuomo derived an equation for the intensity

on the image plane as a function of radial distance from the center of the fiber bundle.

This thesis uses Equation 2-9 through 2-11 from He and Cuomo [18] as the first step in

determining the power coupled and sensitivity of the MO stage. Without an

understanding of these equations, a microphone designer will not be able to identify

miscalculations due to errors that have been observed in the output of Equation 2-9. This

problem will be discussed in more detail later in this section.

Equation 2-9

1A (ko -1)-At1an' i -111, If (1k, <2)&(0 A A tan (k -1)-Atan I(' -1))+tan (1-k)-Atan (A(1- Il, if (lk, <2)&(2-k 2(1 A 0
A 7 Atan A+tan (1-k)-Altan (A(1- II+ In fj (k >2)&(O 2(Fl- ^)[4 80, 1+ -4(k -1)
-A 7- Atan 'A+tan (1-k) Atan (A(1- +II+In f~l( 1 (,k >2)& (022)
2(1-)14 80, 1+A(k+l
1 (r) A 4 tan (k -l)-tan (k-l)+Atan (A(k -))- Atan ( li -111, f (1< k, <2)& ( 2 7 $( A Atan-'A tan -(k -1)+Atan -(A(k -1 + In if (-(k 2)&(1< k 2)&(k -k 2)

S J + Altan' -tan (k -1)+ tan -(A(k -1)) + In ] f (k >2)&(1!k< 2)&(k -k>2)
2 4 F8, 1+ (k4+1)0
A nl .,i (L_+1)2),)] if (k(>2)&(k>2)&(k2)2)
S2(1-A )4 (8k, L++Ak+l

A I1L(k+ I+A2(k I) if (k > 2)&(k > 2)(k k > 2)
nL(k 1)J 1+ -+ 1)2
80, )(k-;1)1+A4(k +1)

The quantity, A, used in Equation 2-9 is defined by Equation 2-10. In Equation 2-

10, rtx core is the radius of the transmit fiber core, and g is the equilibrium gap.

cor e
A = -- Equation 2-10
2g









The quantity k used in Equation 2-9 is defined by Equation 2-11. In Equation 2-

11, rtx core is the radius of the transmit fiber core, and r is the radial coordinate measured

from the center of the Tx fiber axis.

r
k = r Equation 2-11
rtx core


The quantities Oc and kc are the critical angle of the Tx fiber and the critical value

of k associated with that angle. For more details on the variables, see He and Cuomo's

work [18].

Some sets of input parameters with a gap, g, on the order of the Tx fiber diameter

were observed to produce non-intuitive intensity profiles. For example, using Equation

2-9 at a gap of 50 |tm with a Tx fiber core diameter of 105 |tm resulted in I(r) = 0 at all

values of r. Therefore, for the theoretical power coupled and sensitivity to be accurate, an

optical microphone designer must plot Equation 2-9 for equilibrium gaps of the desired

value. If the plots are erroneous, then the power coupled and sensitivity analysis will be

invalid. The Mathcad code used to generate the intensity curves was carefully examined

for errors, and none were found. It is possible that Equation 2-9 does not accurately

predict the intensity at small gap distances.

The power coupled into the receive fibers is determined by using a ring

approximation with a power correction factor. The ring approximation used in He and

Cuomo [18] approximates the face of the receive fibers as an annular ring. The power

coupled is determined by integrating the normalized intensity (Equation 2-9) over the

ring area. Figure 2-3 shows the ring approximation. The actual light collection surfaces









are shaded gray, and the integrated area of the ring approximation is shown by the dashed

ring. Note that the relative sizes of the core and claddings are not necessarily to scale.

















Figure 2-3 Ring Approximation Diagram.

The power coupled and sensitivity equations for the mechano-optical (MO) stage

are given by He and Cuomo [18]. These were the equations used by Kadirval [8] to

predict the performance of the MO stage in his optical microphone. This thesis has

modified the power coupled equation from [18] (referred to as the ideal power coupled

from this point) to include the effects of radial position errors in the receive optical fibers,

and also to correct for overestimation of the power coupled by the ring approximation.

Radial position error, RPE, is defined and discussed in Section 3.3.2. The power

coupled correction factor, cf, is calculated by taking the ratio of the actual surface area of

the receive fibers to the area of the ring in the ring approximation. The ideal power

coupled is then multiplied by this correction factor (which is a function of the optical

fiber geometry and the RPE) to calculate the corrected power coupled. In this thesis,

ideal power coupled and sensitivity refers to the case where cf = 1, meaning that the ring

approximation area exactly matches the surface area of the receive fibers. Since this can









never happen in practice, the ideal situation will occur only when the area mismatch is

neglected, as is done by [18] and [8]. The corrected power coupled and sensitivity

equations are given in Equations 2-12 and 2-13.

b+RPEl
P, 2c, (RPE) core (k g)
n(g, RPE=) -= -i-(k, kdk Equation 2-12
nt m-l+RPE/C .
/tx core



Sm (g, RPE) = dPo (Q(g, RPE)) Equation 2-13
dz I dz


Two theoretical corrected power coupled curves, based on using Equation 2-12

with a fiber bundle constructed from AFS105/125Y multimode fibers as both Tx and Rx,

are shown in Figure 2-4. One of the curves corresponds to an RPE of 0 itm, and the other

corresponds to an RPE of 10 |tm (the observed RPE of the custom fiber bundle). The

corresponding corrected sensitivity curves are shown in Figure 2-5. The horizontal axis

on each plot is the equilibrium gap, g, between the receiving plane and the reflecting

plane in the method of images.

The maximum corrected theoretical MO sensitivity with RPE = 0 |tm is Smo =

1.094E-3 |tm-1 and occurs at g = 230 itm. When the power coupled correction factor is

taken into account, the maximum corrected theoretical power coupled is Smo = 0.784E-3

|tm-1 and occurs at g = 265 |tm.













30%


25%


20%


S15%
O

10%


5%


0%
0 100 200 300 400 500 600 700 800 900 1000
Gap (um)

-RPE=Oum -RPE=100um


Figure 2-4 Theoretical Power Coupled with Ideal Fiber Configuration.


1.20E-03

1.00E-03 .

8.00E-04 -
E *
6.00E-04

4.00E-04 *

5 2 .0 0 E -0 4 ------ -------- .--------------
S2.00E-04
tM ,.
0.OOE+00

-2.00E-04 -

-4.00E-04
0 100 200 300 400 500 600 700 800 900 1000

Gap (um)

*RPE=Oum -RPE=10um


Figure 2-5 Theoretical Sensitivity with Ideal Fiber Configuration.









The discontinuities observed in the sensitivity equations are due to transitions in

the power coupled integral. Specifically, each discontinuity corresponds to the edge of

the light cone crossing the boundary of the receive fiber ring. The discontinuity is

present in the power coupled equations, but since it manifests in these plots as an integral

of the discontinuity seen in the sensitivity curves, it is difficult to see on the viewing scale

of the power coupled plot.


2.2.1.3. Opto-Electrical Sensitivity

The opto-electrical stage converts an optical power coupled to an electrical signal.

This is accomplished with the use of a Thorlabs PDA400 photodetector, which consists

of a photodiode and a trans-impedance amplifier with five gain settings. In this thesis,

the photodiode and trans-impedance amplifier are collectively referred to as a

photodetector, and they are treated as one unit.

The sensitivity of the stage is given by Equation 2-14, where r is the optical

power coupled, and V is the output voltage of the sensor.

dV
Soe = Equation 2-14
dr


The output voltage of the opto-electrical stage is a function of the detection

electronics and the detection method used. For the Unreferenced Output detection

technique, shown in Figure 2-6, the output voltage is a function of the photodetector

responsivity and gain, the output amplifier gain, and the laser power.












Lowpass Filter /
Photodetector Output Amplifier


Pot R G Ga -0 Vout



Photodiode Trans-impedance
Amplifier



Figure 2-6 Block Diagram of the Unreferenced Output Configuration.


The output voltage of the unreferenced output configuration, Equation 2-15, is

derived by applying Kirchoff s and Ohm's Laws to the equivalent circuit of the detector,

shown in Figure 2-7. Since a voltage amplifier is connected in series with the detector,

the output amplifier gain, Ga, is multiplied by Vdet out to get the microphone output

voltage, Vout. Pout is the received optical power from the fiber bundle Rx fibers. By

equating Pout with q times Pin, Equation 2-15 neglects losses in the fiber bundle other than

those from the power coupling effect. For a real bundle, other losses are present at the

connectors and in the fibers themselves. These losses have not been measured and are

neglected here.


Vo, = RGGP,,, = RGGEP,J7


Equation 2-15












Pout











-Photo--------------------
Photodiode


Gdet
R P,,out


----M



Vdet out



----------------------------------
Trans-impedance
Amplifier


Figure 2-7 Equivalent Circuit for the PDA400 Photodetector.


Substituting Equation 2-16 into Equation 2-15 gives the equation for the electro-

optical sensitivity of the Unreferenced Output detection technique.


So = RGGP,, Equation 2-16


Equation 2-16 shows a linear relation between the received optical power and the

OE stage sensitivity. This linear relationship only holds when the photodetector is

operated in a linear region. A Thorlabs PDA-400 photodetector, with specs given in

Appendix D, has a peak response of 0.95 A/W at 1550 nm. The minimum trans-

impedance gain setting, G, for the PDA400 is 15,000 V/A. Using the detector response R

= 0.95 A/W and the gain G = 15000 V/A, gives Soe = 14250(V/W)*Pin. If Pin = 0.7 mW,

then Soe = 9.975 V. Since the photodetector saturates at 10 V, the maximum

unreferenced OE stage sensitivity is 9.975 V Ga.

It is very important to observe that the overall sensitivity of the unreferenced

optical microphone is limited by the maximum DC optical power received by the

photodetector due to detector saturation. Ideally, the photodetector would consist of only










a photodiode, and a highpass filter would be placed at the photodiode output. Without

the trans-impedance amplifier, the DC optical power can be removed before

amplification, eliminating the limit of the OE sensitivity due to the DC optical power.

Any sensitivity lost from removing the trans-impedance amplifier can be recovered by

increasing the gain of the output amplifier.

The Referenced Output OE Stage is more useful in a microphone system due to

the invariance of the sensitivity with optical power, which will be proven here. Figure

2-8 is a block diagram for the referenced OE microphone configuration. By using the

equivalent circuit of the photodetectors and by inspecting the block diagram, Equation 2-

17 can be derived for the microphone output voltage, Vout.


S Pm Pmod
Vout = GGd R modGmod = GGad modf -od Equation 2-17
refG,,rPf ref Pref,


Figure 2-8 Block Diagram of the Referenced Output Configuration.


Modulated Photodetector


Pmod









Pref


- Ga H- Vout

Highpass Filter /
Output Amplifier


Photodiode Trans-impedance
Amplifier








In Equation 2-17, Pmod is identical to Pout in the unreferenced block diagram, when

the same assumptions are made. If a is the optical losses present in the reference signal

path, then Pref can be represented as (1- a) Pin. By observing that the fiber bundle power

coupled, r, is Pout / Pin, Equation 2-17 can be rewritten in terms of the component gains

and the optical power coupled into the fiber bundle.


Vo = Ga)Gad Gmod Equation 2-18
(1- a)G J

Substituting Equation 2-18 into Equation 2-15 gives the equation for the electro-

optical sensitivity of the Referenced Output detection technique, where Gratio is the ratio

of Gmod to Gdet.


Soe = G d Gmod= G0Gd (Grato) Equation 2-19
(I a) Gref 9 (- a)

Equation 2-19 varies directly with the ratio of the modulated detector gain to the

reference detector gain (Gmod / Gref), with the built-in gain of the analog divide circuit

(Gad), and with the gain of the output amplifier (Ga). Also, increasing optical losses in

the reference path, a, increases the sensitivity of the OE stage and decreases the optical

power incident on the photodetector. Later in this chapter, it will be shown that optical

reference path losses will improve the electronics noise and microphone minimum

detectable signal under some conditions.

Using two PDA-400 photodetectors, an analog divide circuit hardwired for a gain

of 10 V, and the minimum and maximum values of Gratio (based on available PDA400

gains settings), the sensitivity of the OE stage can range between Soe min = (0.10 V)*Ga /









(1-a) and Soe max = (1000 V)* Ga / (1-a). If we take the ratio of the sensitivity of the

referenced to the unreferenced OE stage, we can see how the stage sensitivities compare

at varying input power levels. This is done in Figure 2-9 for Gad = 10 V and Ga = 1 V /

V.

From Figure 2-9, it can be concluded that the unreferenced microphone will have

a lower sensitivity than the referenced microphone unless the laser is operated at the

maximum power for which the photodetector remains linear, no reference path losses are

present, and the photodetector gain ratio is one. When this occurs, the two stages will

have identical sensitivities. Increasing Gratio and a will increase the sensitivity of the

referenced OE stage. These values are physically limited by the photodetector range of

linearity for Gratio and c, and also by the analog divide circuit for a. Specifically, the

input to the analog divide circuit must never drop below a certain threshold, and the

output of the analog divide circuit can never saturate. This limits a to less than 0.9 for

the AD circuit used in this thesis. In practice, the PDA400 detectors are not useful when

the gain is set higher than 47,000 V/A. An additional constraint is the fixed gain-

bandwidth product limiting the maximum gain for a minimum bandwidth.











10

9

8

7

6

V) 5

4
0
3

2


0 --------------------------------
0 100 200 300 400 500 600 700
Laser Power (gW)
Gratio = 1, Gdet = 15000 V/A, Alpha = 0
Gratio = 3 1, Gdet = 15000 V/A, Alpha = 0
-Gratio = 1, Gdet = 15000 V/A, Alpha = 0 5


Figure 2-9 Comparison of Unreferenced and Referenced Output Sensitivities.


2.2.2. System Linearity


The linearity of the optical microphone is determined by the linearity of the

individual stages. The acousto-mechanical stage linearity is governed by the nitride

membrane. The linearity of the mechano-optical stage is dominated by the assumptions

that the membrane curvature is negligible, and by the local flatness of the sensitivity vs.

equilibrium gap curve. The linearity of the opto-electrical stage is governed by the

linearity range of the photodetector and additional electronics. The following sections

establish the conditions for linearity of each stage, and therefore the whole device.









2.2.2.1. Acousto-Mechanical Linearity

The theory for the range of linearity of the membrane was investigated by

Sheplak and Dugundji [20]. Using this theory, Sahni et al. [21] determined that the

diaphragm is linear over the region from 0 2000 Pa (160 dB re. 20 [tPa). Sheplak et al.

[20] present Equation 2-20, which relates the membrane aspect ratio to the in-plane stress

and the maximum linear pressure (3% linearity).

3
J02
Equation 2-20
h max Pmax E


By substituting the maximum linear pressure and the microphone membrane

dimensions into Equation 2-20, the in-plane stress, co, of the membrane can be estimated.

The experimental linearity range of the microphone (see Chapter 5) is reached at 122 dB

(re. 20 tjPa). This results in the AM sensitivity increasing by a factor of 15. This effect

is considered when predicting the microphone performance in Section 2.2.7.


2.2.2.2. Mechano-Optical Linearity

The linearity of the mechano-optical stage is based on two factors: the linearity of

the power coupled curve (flatness of the sensitivity curve), and the assumption of the

membrane curvature being negligible.

The point of interest for the linearity analysis is about the point of maximum

sensitivity. Figure 2-10 shows a plot of the sensitivity from an equilibrium gap of 240

|tm to 290 |tm using Equation 2-13. The vertical axis of the curve is highly magnified,

and the thin horizontal lines denote the region that is within 3% of the maximum

sensitivity. It is evident from Figure 2-10 that the MO stage sensitivity is linear within










+/-10 |tm of the optimal gap. Since the largest maximum membrane deflection, at Pmax =

2 kPa, allowed by the variability in the nitride stress of the MEMS chip process is +/-2.49

|tm, the sensitivity of the MO stage will be constant within 3%. The large window for


linearity holds at equilibrium gaps out to 400 |tm. This means that if the equilibrium gap

is set at a value larger than the best case equilibrium gap, then the sensitivity will still be

constant to within 3% over the range of the diaphragm motion.


1.00E-03


Z 9.00E-04



-o -



I-
2 7.00E-04 -
"0


6.00E-04

0

5.00E-04
240 245 250 255 260 265 270 275 280 285 290
Gap (jm)

Figure 2-10 Linearity of Mechano-Optical Stage.


The second criterion for establishing the linearity of the MO stage is the flatness

of the membrane over the illuminated region. An alternate way of viewing this

requirement is to look at the acousto-mechanical sensitivity of the membrane over the

illuminated region as a function of radial distance from the center. For linearity, the MO

sensitivity of the membrane should vary by no more than 3%. It is important to note that

the method of images used in Section 2.2.1.2 requires a flat membrane in the illuminated











area. This means that the membrane must be within 3% of planar over the illuminated


region, and must be parallel to the fiber bundle face.


Figure 2-11 shows the normalized acousto-mechanical sensitivity, using Equation


2-6, of the membrane as a function of the radial distance from the membrane center. This


plot shows that the sensitivity is within 3% of the maximum when the illuminated area is


within 93 ptm of the membrane center. For optical fibers with an NA = 0.22, the spot


radius is less than 93 jtm when the gap is less than 430 jtm.


'A
w 1.08
U
C








Z
E 1.00






0.84
0.76
-g 0.92



0.84



0.76


0 10 20 30 40 50 60 70 80 90 100
Radial Distance from Membrane Center (pm)

Figure 2-11 Plot of Acousto-Mechanical Sensitivity as a Function of Radial Position.


It has been determined that the MO stage sensitivity varies by no more than 3%


over the equilibrium gap range of 200 |tm to 400 |tm. It has also been determined that


the acousto-mechanical sensitivity at every illuminated point on the membrane is within


3% of its maximum value when the equilibrium gap is less than 430 jtm. Therefore, the


MO stage of the optical microphone is linear when the equilibrium gap is between 200


~









|tm and 400 |tm. Since the ideal equilibrium gap is 230 itm, the OM stage of the optical

microphone is linear at the equilibrium gap of 230 |tm with a maximum deflection of +/-

2.49 tm at Pmax = 2 kPa.

This analysis does not include the effect of misalignments on the linearity. They

are discussed in Chapter 3.


2.2.2.3. Opto-Electrical Linearity

The linearity of the opto-electrical stage is effectively limited by the linearity of

each electronics component in the system. The PDA-400 photodetectors are linear up to

an output voltage of 10 V. The AD734 analog divide chip is linear over the input range

of +/- 12.5 V. Therefore, the photodetectors are the limiting factor in determining the OE

stage linearity. As long as the detectors are operated below saturation, the opto-electrical

stage is linear.

The maximum voltage that can be output in the linear range of operation for the

electronics serves to limit the maximum sensitivity of the unreferenced output

configuration when the output amplifier is held at a fixed gain. To illustrate, consider

Equation 2-21, the equation for the photodetector output in terms of the optical power for

the unreferenced OE configuration.


Vot = RGP,,,. Equation 2-21


The maximum Vout that the PDA400 photodetector can output is 10 V. The

responsivity is 0.95 A/W. Substituting these constants into the above equation shows that

the product of the gain and the optical power cannot be larger than the saturation voltage

of the detector divided by the detector responsivity, or 10.5 W*V / A for the PDA400.









This product effectively limits the maximum sensitivity that the unreferenced OE stage

can provide, since increasing Piaser requires G to be reduced if Plaser*G > 10 V. When the

range of linearity of the measurement equipment is considered, then the gain of the

output amplifier is also limited. To ensure the microphone can operate in its intended

environment, the linearity range of the measurement equipment must also be considered.

2.2.3. System Frequency Response

The frequency response of the optical microphone was discussed in depth by

Kadirval [8]. He developed an equivalent circuit for each stage of the optical

microphone. The frequency response of the microphone is the product of the frequency

response of the individual stages.


2.2.3.1. Acousto-mechanical frequency response

The lumped element model parameters used in the acousto-mechanical stage are

shown in Table 2-1. These parameters were given by Sahni [21]. The frequency

response analysis for this optical microphone membrane was presented by Kadirval [8],

and it remains valid. Using the lumped element parameters in Table 2-1 and the

equations for the frequency response of the AM stage, the 3 dB frequency for the upper

end of the frequency response was 76.25 kHz. The lumped element approximation ia

limited to below 50 kHz according to Sahni [21], so the upper limit of the frequency

response will not be accurately predicted by this model.


H,,,, (s) (= eff Equation 2-22
1 (l+M3Cs2)
3eff










H am nor(s)
H o H,(s) ,, n.(j2"OHz)


Equation 2-23


Table 2-1 Acousto-Mechanical Lumped Element Parameters
Parameter Formula Value Units

Mmea ph 1316 kg / m4
3r a1316
4
Cmea 1.963E-15 m3 /Pa


Mrad Pa 5.199E+5 kg / 5
3z".a2
3r. a2h
Ca ca 1.246E-15 m3/Pa
Pac

fresh 0.39 98.6 kHz
a P"


The lower end of the frequency response is governed by the vent channel of the

microphone. Although the optical microphone was not designed with a vent channel, the

steel tube is not an exact fit with the membrane cavity, and allows equalization of the

pressure on both sides of the membrane due to low frequency pressure fluctuations.


2.2.3.2. Mechano-optical frequency response

The mechano-optical frequency response is a constant. This was discussed by

Kadirval [8] in detail, and will not be reproduced here.


2.2.3.3. Opto-electrical frequency response

The opto-electrical frequency response is determined by the frequency response

of the electrical components in the system. The photodetector has a worst case

bandwidth of 50 kHz at the maximum gain and a bandwidth of 10 MHz at the lowest

gain. The analog divide circuit has a bandwidth of 10 MHz. Therefore, the bandwidth of









the photodetectors is the limiting electrical component for the upper range of the

frequency response for practical gain settings.

The lower range of the frequency response is limited by the cut-on frequency of

the highpass filter. In the highpass filter used by this thesis, the cut-on frequency was set

at 30 Hz.

2.2.4. System Electronic Noise

The electronic noise is defined to be the noise, in volts, seen at the output of the

microphone circuit. Previous works (He & Cuomo, 1991 [16], [22]) assume that "the

light source does not contribute significantly to the (noise) background." He and Cuomo

note that these conditions do not always hold, and that voltage fluctuations due to light

source noise may affect the signal. This thesis includes the effects of the detector, the

light source intensity noise, and the electronics to determine the electronic noise at the

microphone output. The noise analysis traces the path of RMS noise signals, in V / Hz ,

through both the referenced and unreferenced electronics.

The goal of the first part of the electronics noise analysis is to identify and derive

equations for the individual noise sources that contribute to the electronics noise of the

photodetector.


Figure 2-12 Noise Contributions for the Photodetector Output.


Photodetector


Photodiode NEP Terms Photodetector Electronics
Noise Terms
t N, P R*Gn Vd
Pdet NEP, Plight noise Vdet, Vhlight









Figure 2-12 illustrates the noise sources that produce electrical noise at the output

of the photodetector. Pdet NEP is the noise equivalent power at the input of the photodiode

due to the inherent noise of the photodiode (dominated by thermal and shot noise

... equations presented are referenced to Wilson and Hawkes [23]). The units for Pdet NEP

are W / Hzl2. From this point on, the photodetector output due solely to Pdet NEP is

referred to as the detector (or photodetector) electronics noise, Vdet, and has the units V /

Hz 2. Thermal noise and shot noise are the dominant noise sources in a photodetector.

The shot noise, Vshot, in V / Hz 2, is given by Equation 2-24, where G is the trans-

impedance gain, R is the diode responsivity, e is the charge of an electron, and Plight is the

incident optical power received by the detector.


V hot = G2eRTPgh Equation 2-24


Thermal noise, Vtherm, is also present in a photodetector, and can be given by the

following equation, where k is Boltzman's constant, T is the mean temperature in Kelvin,

and G is the trans-impedance gain.


thern, = TG Equation 2-25

The total detector noise, Vdet, can be written as the RMS sum of the thermal and

shot noise, shown in the following equation.


Vdet hot2 +therm2 Equation 2-26

The other noise source, Plight noise, represents the noise power fluctuation incident

on the photodetector due to fluctuations in the intensity of the light source integrated over

the diode collection area. The units for Plight noise are W / Hzl2. The voltage fluctuations









at the output of the detector produced by Plight noise are referred to in this thesis as the light

source (or laser) electronics noise, Vlight. The units are the units V / Hz Where

necessary, subscripts will be added to distinguish between referenced and unreferenced

quantities. Since Plight noise is highly dependant on the type of light source and supporting

electronics, a purely theoretical model will not be used (see the end of this section).

Instead, the electronics noise at the output of the photodetector will be measured. The

measured quantity will be equal to the RMS sum of Vlight noise and Vdet. By using

Equations 2-24 2-26 and the experimentally measured noise quantity, Vlight noise can be

determined for the light source detector system. When Vlight noise is known, the

photodetector equivalent circuit can be used to calculate Plight noise. Dividing Plight noise by

the photodiode active area produces an estimate of the light source intensity noise.


Microphone OE Stage


Photodiode NEP Terms Microphone Electronics
Noise Terms
SUnref/ Ref
Pdet NEP, Plaser noise Vnoise ref, Vnoise unref



Figure 2-13 Noise Contributions for the Microphone Output.


The second part of the electronics noise analysis propagates Vdet through the

unreferenced and referenced configurations to determine the electronics noise at the

output of the microphone. These noise terms are referred to as the "total electronics

noise" or the "microphone electronics noise" terms. The total electronics noise is

determined by the amplification of Vdet by signal amplification components and the

addition of any additional electronics noise sources (see Figure 2-6 and Figure 2-8). For









this work, the analog divide IC used by the referenced configuration was the Analog

Devices AD734 [24].

The electronics noise analysis of the unreferenced microphone configuration in

Figure 2-6 is not difficult since only one signal path exists. By inspecting Figure 2-6,

Equation 2-27 and 2-28 can be derived for the total noise output at the photodetector,

Vdet total and the unreferenced noise, Vunref. In Equation 2-28, Va is the input noise of the

amplifier, in V / Hz"2


Vdet total =Vdet2 + (RGdet Phght_ noie 2 Equation 2-27


Vnref = Ga Vdet total2 + Va2 Equation 2-28


The analysis of the referenced microphone noise is more complicated. There are

two signal paths into the system, and a time-domain division. To complicate matters, the

optical noise Vlight noise in the modulated and reference signal paths (see Figure 2-8) may

or may not be correlated. In a real microphone system, it is expected that the optical

noise in the modulated and reference signal paths will be correlated to some degree.

Theoretical models are presented here for both uncorrelated and correlated optical noise

in the referenced configuration.

Vlight noise ref and Vlight noise mod are given by Equations 2-29 and 2-30, which are

derived from the power to voltage conversion equations of a photodetector.

Vght nose RGeP re ghtP Equation 2-29
hghte_ o =h fe Equation 2-3 0n


hght noise _mod = RGmod lhght noise Equation 2-30
2








When Vlight noise ref and Vlight noise mod are uncorrelated, they will not divide in the

analog divide circuit. Instead, they add as shown in Equation 2-31 for the uncorrelated

analog divide circuit output, Vad unorr. In Equation 2-31, Vad is the input noise of the

analog divide circuit.


Iad uncorr = Gad ,2Vdet- l hght l nse ref + light _nose mod2 + 2ad2 Equation 2-31

The total electronics noise at the output of the referenced microphone for the

uncorrelated case can be determined by propagating Vad uncorr through the output

amplifier, with input noise Va, as shown in Equation 2-30.


Vrefncor = Ga ad 2uncorr a2 Equation 2-32

When the optical noise is correlated, then it will be divided by the analog divide

circuit. In this case, Vad cor is given by Equation 2-32.
Jr 2
V
Vad corr = Gad 2Vdet + 2ad + nosemod Equation 2-33
Vh\ght _noise ref

The total electronics noise at the output of the referenced optical microphone

when the optical noise is correlated can be determined by replacing Vad uncorr with Vad corr

in Equation 2-32. Equation 2-34 gives this result.


Vrefcorr = G ad corr2 + V Equation 2-34

Theoretical models for the intensity noise of a laser are inaccurate in predicting

the performance of a commercial laser, due to the uncertainty in the laser quantum

efficiencies, and the variation in the performance of the electronics controlling and









cooling the laser. Coldren and Corzine [25] present Equation 2-41 for the relative

intensity noise, RIN, of a laser source at the optical resonance frequency (corresponding

to the laser output wavelength). The dampening factor, y, is defined as Kf2R + Yo, and is

not a parameter provided by a laser manufacturer. Experiments can be performed to

quantify it, but these are not practical to do when the direct measurement of the optical

noise PSD will provide the needed value.


RIN 16tr(Av)
RN 1 sr Equation 2-35
Af Y


In order to make an accurate estimate of the light source noise, this thesis does not

rely on Equation 2-41. The results of the experimental measurements for Plight noise and

Vdet are given in Chapter 5.

2.2.5. System Minimum Detectable Signal

The minimum detectable signal is the smallest signal that can be resolved by the

microphone. The minimum detectable signal is a function of physical interactions

between noise sources and desired signals in each stage as well as the sensitivities of each

stage. Previous work (He & Cuomo, 1991 [16]) only considered the OE Stage MDS, and

also ignored the laser noise effect on the total electronics noise. This work takes

additional noise sources into consideration (see Section 2.2.4).

Equation 2-42 gives the System MDS for the optical microphone. Three physical

MDS reducing mechanisms are considered: membrane noise due to the Brownian motion

of the membrane atoms, variations in the coupled laser power due to laser intensity noise

in the MO stage, and electronic noise in the OE stage. Each of these terms will be

discussed in more detail later in this section.









P 2
hlightnoise
P (V
MDS = + ght- +Af Equation 2-36
SP,,ram ,,o S



The first term under the radical in Equation 2-42 is the AM stage MDS due to the

Brownian noise of the membrane, given in Pa2. The second term is the MO stage MDS,

which is due to the variations in the coupled laser power due to laser intensity noise, in

the presence of variations due to desired membrane deflection (from an acoustic signal),

given in Pa2. If the laser intensity noise is close in magnitude to the change in coupled

power in the Rx fiber, then the signal can never be distinguishable from the optical noise,

even with an ideal (no-noise) detector. The third term is the MDS of the OE stage, also

given in Pa2. The OE stage MDS is determined by the total electronics noise and the

measurement bin width. In previous works, the MO MDS was neglected [16, 22]. No

previous work takes the light source electronics noise into account when calculating the

OE MDS; they only consider the photodiode shot noise.


2.2.5.1. Acousto-Mechanical Minimum Detectable Signal

The dominant noise source of the AM stage is the Brownian motion of the

membrane. The silicon nitride membrane atoms, like all atoms above absolute zero,

exhibit Brownian motion. This Brownian motion causes a deflection of the diaphragm in

the same manner that an acoustic signal does. Due to this deflection, pressures which

cause a deflection smaller than the deflection due to Brownian motion are not detectable.

The equation for the mean equivalent pressure fluctuations due to the Brownian motion is

given by Equation 2-43 from Chau and Wise [26].










2 32kT (x7A +Pl V _P2)
Pr = a2 Af Equation 2-37
V a


Using a = 1, T = 300 K, k = 1.38 E-23 J / K, mi = m2 = 4.78 E-26 kg, p = p2 =

101.4 kPa, and a = 500 itm, the mean equivalent pressure fluctuations due to Brownian

motion is = 3.642 E-11 Pa2 Af. This corresponds to a MDS of-11 dB (re. 20 [tPa).

Based on the MDS of the other stages (presented later), it can be concluded that this noise

source is completely negligible in an intensity-modulated optical microphone.


2.2.5.2. Mechano-Optical Minimum Detectable Signal

The dominant noise source in the MO stage is the intensity noise of the light

source. The physical effect is illustrated in Figure 2-14. Power is coupled by light

reflecting into the receive fibers from the membrane. This power can be written as the

sum of three components: the optical DC component (Plight), the acoustically-modulated

optical power signal (Pmodulated(t)), and the optical noise power (Plaser noise(t)).


Time Variance of Coupled Power Components


Plaser + Plight noise(t) + Pmodulated(t)
Rx Fibers (more than one)




Figure 2-14 Illustration of the Physics Behind the MO MDS.

When the membrane deflects due to an acoustic signal, Pmodulated(t) is produced.

Fluctuations in the light power output superimposes Plight noise(t) onto the desired

modulated signal. If the fluctuations in power coupled due to light output noise are larger

than the fluctuations in power coupled due to the acoustic signal (Plight noise(t) >=










Pmodulated(t)), then the microphone will not be able to detect the acoustic signal without the

optical noise removed by the electronics. In the event that the optical noise can be

removed by the electronics (see Section 2.2.4), the MO MDS effect will not be present in

the microphone.

The equation for the contribution (when present) of the MO stage to the MDS is:


Player noise

MDSm = -P-O e Equation 2-38
SamSmo


The MO MDS will dominate the system MDS if the following equation is

satisfied and if the optical noise cannot be removed from the system by the electronics (as

is done by the referenced optical microphone when the optical noise is correlated):


-s > VE Equation 2-39
P S
laser oe


2.2.5.3. Opto-Electrical Minimum Detectable Signal

Noise mechanisms in the OE stage are due to the electronics, the shot noise of the

photodetector, and the detected noise power of the laser. Lasers operated in constant

current mode are noisy, and will usually dominate the electronics noise. The equation for

the OE stage contribution to the MDS is the following, where Voe noise is the electronics

noise and Af is the measurement bin width of the electronics noise.


no~e .oiseA
MDSOE = V n-fs Equation 2-40



If the inequality in Equation 2-45 does not hold, then the OE stage MDS will

dominate.









2.2.6. Optical Reference Path Losses and System Performance Metrics

In this section, the effect of optical losses in the reference signal path in the

referenced OE configuration will be analyzed. The effect of optical reference path losses

on sensitivity, electronics noise, and MDS will be discussed.


2.2.6.1. Sensitivity and reference path losses

In Equation 2-50, the referenced microphone sensitivity is presented, with Soe

expanded to show the dependance on a.


Sp ( = Sm Gmp Gmod Equation 2-41
(1 a) Gre


The losses in the reference path in the referenced opto-electronic stage reduce the

voltage seen by the denominator input of the analog divide circuit. Decreasing the

denominator increases the circuit output, so adding optical losses to the reference path

will increase the sensitivity of the microphone. For the unreferenced microphone, there

is no reference path, so a is undefined.


2.2.6.2. Electronics noise and reference path losses

In Equations 2-29, 2-31, and 2-32, it was shown that the electronics noise for the

referenced microphone is dependant on (1 a) when the optical noise is uncorrelated.

Therefore, increasing a with uncorrelated optical noise will decrease the electronics

noise. If the photodetector noise dominates, then the effect of increasing a will be

negligible, but if the light source noise dominates, then increasing a can significantly

improve the referenced electronics noise. Therefore, when the optical noise in the









reference and modulated signal paths is perfectly uncorrelated, optical losses in the

reference path are desirable.

Equations 2-29, 2-33, and 2-34 show that the referenced electronics noise is

dependent on (1 a)-1 when the optical noise is perfectly correlated. In this case,

increasing the losses in the optical path will actually increase the electronics noise (in the

same manner sensitivity is increased). Therefore, when the optical noise in the reference

and modulated signal paths is perfectly correlated, optical losses in the reference path are

undesirable.

A real microphone system is likely to have some correlation between the optical

noise in the reference and modulated signal paths, but the extent of the correlation in

general is unknown. Therefore, it is possible for losses in the reference path to increase

or decrease the referenced electronics noise. It is expected that a referenced microphone

with a small correlation between optical noise signals will receive some benefit from

reference path losses, while a referenced microphone with a large correlation between

optical noise signals will have its noise floor slightly worsened.


2.2.6.3. Minimum detectable signal and reference path losses

The referenced optical microphone MDS is strongly dependent on whether the

optical noise is correlated or not. If the noise is correlated, then the MO MDS does not

factor into the total MDS, since optical noise fluctuations are completely removed. If a

increases and the optical noise is correlated, both overall sensitivity and electronics noise

will increase at approximately the same rate. Therefore, optical path losses will not affect

the MDS when the REF and MOD optical noise is perfectly correlated.









If the noise is perfectly uncorrelated, then overall sensitivity will increase and

electronics noise will decrease. Therefore, optical path losses will improve the MDS

when the REF and MOD optical noise is perfectly uncorrelated. Note that since the MO

MDS effect is not present for the correlated case, the total MDS can be much lower for

the correlated case than for the uncorrelated case.

2.2.7. Summary of Predicted System Performance

Due to the quantity of theoretical data presented in previous sections of this

chapter, the best realistic device performance metrics will be summarized here. No

design has a problem with the frequency response since the most limiting component is

the photodetector with the maximum gain, which has a bandwidth of 50 kHz. Since the

lumped element model approximates the membrane for f< 50 kHz, this will be used as

the upper limit of the frequency response, even though electronics may be capable of

higher frequencies. The lower limit of the frequency response will be 30 Hz, which is the

cut-on frequency of the highpass filter used in the OE stage. The reference path optical

losses in Table 2-2 were experimentally measured. The performance specifications in

Table 2-3 are those that the optical microphone is expected to have when it is

experimentally characterized.

Table 2-2 Summary of Configuration Settings for Theoretical Performance Metrics
Mod Ref Laser .
Mod Ref Laser TAmplifier Ref Optical
S. Detector Detector Output O i
Configuration eeor Deeor O Gain Path Losses
Gain Gain Power (V / V) (W / W)
(V / A) (V / A) (pW)
Unreferenced
Unrefer ed O15,000 N/A 350 1 N/A
Amplified Output
Referenced
Referenced 15,000 15,000 350 1 76%
Amplified Output






58


Table 2-3 Summary of Theoretical System Performance Metrics
Sensitivity MDS Electronic Noise
Configuration ,. ,
Configuratio(mV / Pa) (dB re. 20 iPa) (lV / Hz2)
Unreferenced Amplified Output 0.073 73.8 0.19
Referenced Amplified Output 0.612 64.4 14.3
















CHAPTER 3
DESIGN OF THE OPTICS FOR THE MEMS OPTICAL MICROPHONE

This chapter examines the selection process for the optics design for the MEMS

optical microphone. The design of the optics must be done in parallel with the

microphone package. Some optics required by the optical microphone are optical fibers,

a light source, an optical splitter, photodetectors, and opto-isolators. Other miscellaneous

components that are needed are connectors, protective tubing for the optical fibers, and

packaging the MEMS optical microphone. This selection process is vital to the bundle

performance and to the feasibility and robustness of the microphone package.

3.1. Selection of the Optics

There are many factors which were considered in the selection of the optics. The

most important factors are device performance, system connectivity, ease of handling,

and cost.

3.1.1. Performance

Device performance considerations are the most important factor for selecting the

optics used in the optical microphone. Performance specifications for the optical

microphone were listed and a theoretical model was derived in Chapter 2 for the system

sensitivity, minimum detectable signal, frequency response, and dynamic range of

linearity. Assumptions and simplifications were part of the theoretical model of the

device performance. Therefore, the microphone optics must have is the ability to adhere









to the theoretical assumptions under the widest possible range of considerations. Ideally,

the selection of the optical components will guarantee the validity of the assumptions.

Practically, the optical components must be selected to minimize deviations from any

inherent assumptions (discussed in Chapters 1 and 2) and the non-idealities included in

model formation.

As shown in Chapter 2, MDS (in Pa or dB) is dependent on the individual noise

contributions of the light source, the membrane, and the electronics, as well as the overall

sensitivity and the product of Sa and Smo. Therefore the components should be selected

such that their combined contribution to MDS is equal to or below the desired minimum

detectable signal. To determine this, their effects on the sensitivity and noise floor must

be known in advance! Since sensitivity is dependent on received optical power in some

microphone configurations, light sources with high power are usually more desirable than

sources with low power. However the MDS may not be improved, since using a light

source with more power can cause the laser intensity noise to rise to unacceptable levels.

Increasing the numerical aperture of the optical fibers used for the fiber bundle

increases the MO stage sensitivity. Small core fibers have higher sensitivity than large

core fibers. Focusing optics can also be used to provide large MO stage sensitivity

increases. Misalignments and reflection losses can decrease the sensitivity and also

potentially lead to laser instability. Large core fibers should not be coupled (via standard

connectors) to small core fibers, since large power (and sensitivity) losses will result.

Factors that influence sensitivity and MDS also can affect the device linearity.

Increasing the output of the light source without bound will eventually saturate the

detection electronics. By varying the optical fiber numerical aperature, NA, in Equations









2-9 through 2-13, it was found that using optical fibers with high numerical apertures

compresses the power coupled and sensitivity curves towards zero, reducing the range of

linearity of the MO stage when compared with the linearity using a low NA fiber.

3.1.2. System Connectivity

The issue of system connectivity must be considered when selecting optical fibers

and optical equipment. Specifically, it must be possible for two fibers which are to be

connected to each other to be connectorized with compatible connectors. Free space

coupling mechanisms are possible as a last resort, but they are undesirable since they

allow ambient light to be coupled into the system, are difficult to align, and are sensitive

to vibrations. Finally, a design that requires multiple fibers in one connector must take

into account the available connector sizes when choosing the size and types of fibers that

will be used.

3.1.3. Ease of Handling and Manufacturability

Manufacturability of a device is as important to overall success with the

microphone as the performance. No matter how good the predicted performance of a

device, if it cannot be built efficiently and effectively, then it will not be useful. During

fabrication of the fiber bundle, fibers must be stripped, inserted into a steel tube, and

generally exposed to rough handling. In general, the smaller the diameter of the fibers,

the more difficult any handling with them becomes. The yield of the process for

producing fiber bundles is lower for smaller core fibers, therefore, an improved process

for producing small core fiber bundles is required.









3.1.4. Cost

Finally, optical fibers and other optical components are expensive. For an optical

microphone to be competitive with a capacitive microphone, the cost of each component

must be minimized. Components should be chosen such that the least expensive

component that satisfies the specifications is used. Although this is an intuitive

statement, its application is not always easy. Packaged lasers with built in control

electronics and ultra low noise floors are expensive, but they may improve the device

performance significantly. However, if the photodetector noise and MDS cannot match

the laser, then the cost of the optical microphone will be needlessly high.

Alternately, it is possible to buy fiber pigtailed lasers at communications

wavelengths (1550 nm) that are (relatively) inexpensive. If care is not taken to protect

these lasers from static electricity and thermal effects, then the lasers will have a high rate

of failure, and the cost of operation of the optical microphone will again become

needlessly high.

Based on the sensitivity, noise and MDS analysis from chapter 2, it is best to

select lasers with the highest signal-to-noise ratio. More output power is not always

better, considering that the MO MDS is based on the laser SNR, while the OE MDS is

based on the laser noise power, and can be worsened by increasing the laser power even

if the laser SNR remains constant. Also, OE configurations are available which eliminate

optical power from the sensitivity equation.

Photodetectors with a high gain-bandwidth product and low noise are desirable.

High photodetector built-in gain is undesirable, since the DC component of the optical

signal will result in detector saturation at low power levels, limiting the sensitivity. Also,









an intensity-modulated optical microphone will have a large dc component, so high gain

detectors will saturate. A low-noise amplifier and a high-pass filter at the output of the

system will recover any gain lost by low gain photodetectors or low power lasers.

3.2. Selection of the Tubing

The protective steel tubing is used to protect and align the optical fibers in the

MEMS chip cavity. It provides mechanical support to the fibers and the mounted MEMS

chip and it isolates the fibers inside the tube from the acoustic field under test.

The protective steel tubing must protect the fibers, mount the microphone chip,

and properly align the fiber bundle. The fiber bundle is assumed to be tightly packed and

aligned with the center of the membrane. Ideally, a tube with an inner diameter (ID)

equal to the fiber bundle diameter and an outer diameter (OD) equal to the MEMS cavity

diameter is used. This topic is discussed in more detail later in this chapter.

The material selected for the tubing was steel, because acoustic impedance of the

protective covering for the fiber bundles needs to be much larger than the acoustic

impedance of the test medium (air, in this case). If the acoustic impedance of the tube

was not much higher than air, then sound could penetrate the tube and cause a

displacement of the fibers in the fiber bundle. This is undesirable, since the theoretical

characterization of the optical microphone requires the fiber bundle face to be fixed. In

an environment where the acoustic impedances of the tube and the medium are more

closely matched, a more sophisticated model should be used to account for possible

movement of the fibers in the tube.









Another reason for the selection of steel as the tubing material is its ability to

protect fibers from damage. Steel is much harder than other reasonably priced materials

and will be able to provide adequate protection to the fiber bundle.

3.3. Alignment Issues

Thus far, only fiber bundles with the optimal bundle geometry and ideal

arrangement between the fiber bundle and the membrane have been analyzed.

Assumptions about the geometry of the bundle have been made, but if the steel tubing

and fibers are not properly chosen, these assumptions may not hold. This section

attempts to develop criteria for minimizing the errors when geometry assumptions do not

hold.

3.3.1. MEMS Chip Cavity Alignment Issues

Ideally, the MEMS chip and fiber bundle are aligned as shown in Figure 3-1.

Four geometric parameters of the device are identified: the membrane diameter (DIA),

the fiber bundle diameter (FiberDIA), the inner diameter of the steel tube (ID), and the

outer diameter of the steel tube (OD). Two types of position errors are also identified:

the distance between the outer edge of the steel tube (Type I Error, ERR1), and the

distance between the fiber bundle and the inner edge of the steel tube (Type II Error,

ERR2). Type II Error also represents the worst case radial position error (see Section

3.3.2). The worst case bundle position error (BPE), defined as the radial distance

between the actual location of the Tx fiber core and the desired location of the Tx fiber

core, is given by the following equations.

BPE = ERR1+ ERR2 Equation 3-1











ERR1 (DIA OD)
RR2 = Fi IA)


ERR2 I (ID FiberDIA)
2


Equation 3-2


Equation 3-3


- I I-


Membrane Diameter, DIA


Type I Error, ERR1


Bundle Diameter, FiberDIA

Steel Tube Inner Diameter, ID

Steel Tube Outer Diameter, OD
1------------


Figure 3-1 Bundle Position Error Illustration.


From the equations for bundle position error, it is evident that the bundle position

error is minimized when ID = FiberDIA and OD = MembaneDIA. If a device with zero

error could be fabricated, then the fiber bundle geometry would be perfect and the bundle

would always be perfectly aligned. In practice, packaging issues and the unavailability of

optimally sized steel tubing prevent BPE from being eliminated completely.

The steel tubing (and optical fibers) was chosen to minimize the worst case BPE

(and RPE; see Section 3.3.2). The BPE was analyzed for fiber designs using multiple

different fiber types, from small 50 jtm diameter core, 55 jtm diameter cladding fibers to

200 |tm diameter core, 220 |tm diameter cladding fibers. In the ideal configuration,

smaller fibers improve sensitivity and reduce the equilibrium gap. In practice, it is very


Type II Error, ERR2









difficult to build an ideal fiber bundle with 50/55 fibers, since commercially available

tube gauges allows large BPE (and RPE; see Section 3.3.2) with these fibers. The

105/125 fibers used by this thesis were chosen because they provided the best mix

between ideal device performance, worst case error device performance, and

manufacturability. The tube size that minimizes the RPE with the 105/125 fibers is the

21HW gauge from Popper & Sons. Specs for this tube and other useable tube sizes are

shown in Appendix E.

Another type of alignment problem occurs when the fiber bundle face is tilted at

an angle with respect to the ideal position. This angular misalignment error is illustrated

in Figure 3-2. Typical misalignments encountered are only a few degrees. Figure 3-2

exaggerates the effect for purposes of illustration.





















Figure 3-2 Angular Misalignment Error Illustration.

Although the alignment problem shown in the above picture is greatly

exaggerated, it illustrates how size mismatches between the OD of the steel tube and the






MEMS chip cavity can invalidate the parallel surfaces assumption. This kind of
alignment problem is usually dealt with by the packaging. Horowitz [27] studied this
type of alignment error for a solitary single mode fiber optical microphone, but it is
significantly more complicated to theoretically characterize the effects of angular
misalignments for a multiple multimode fiber structure. A detailed analysis of the effects
of this type of alignment error will not be dealt with here. Minimization of Type I Error
can eliminate this problem. For a package that virtually eliminates this type of error, see
Chapter Four.
3.3.2. Fiber Bundle Geometry Issues
Thus far, the fiber bundle has been assumed to be tightly packed (all fibers in
contact). In practice, this is difficult to do. A fiber bundle geometry with fiber position
errors is shown in Figure 3-3.


Figure 3-3 Radial Position Error Illustration.


0 0
Radial Position



00@









The fiber radial position error (RPE) shown above is modeled for small error

values in Chapter 2 by shifting the ring outward by the amount of the radial position

error. In general, this type of error is caused by Type II errors in the fiber bundle

structure, when the Rx fibers are shifted towards the steel tube wall by a different amount

than the Tx fibers. This reduces sensitivity and increases the optimal equilibrium gap

(see Chapter 2). The observed RPE with the custom fiber bundle is 10 itm. Using RPE =

10 |tm for using the 105/125 bundle, the MO sensitivity is reduced by 28% compared to

the case where RPE = 0 |tm (see Chapter 2 for details). For fiber position errors larger

than 30 |tm, the actual ring area diverges significantly from the assumed ring area, and

other effects such as multiple reflections can become significant. Since the corrected

power coupled and sensitivity equations do not consider multiple reflections, a more

sophisticated model should be used when RPE > 30 |tm. Finally, if all the fibers in the

bundle do not have the same RPE, then the corrected ring approximation theory will be

invalid, since the bundle structure assumption will no longer hold (i.e. Rx fibers

randomly placed with respect to the Tx fiber).

3.3.3. Application of Alignment Theory to Fiber Bundle Selection

In this section, the alignment issues are quantified for specific fiber bundle

geometries, and the optical fibers and steel tubing are selected. Equations 3-1 through 3-

3 were used to determine the gauge and wall thickness of the steel tube that minimizes

the worst case alignment errors in the optical microphone. Table 3-1 shows the worst

case bundle position errors when combining various steel tube gauges from Popper &

Sons [43] with different Tx and Rx optical fibers. The design selected for this thesis was










the first entry in Table 3-1. This design provided the best balance between minimizing

errors and availability of standard optical connectors.


Table 3-1 Error Analysis of Different Fiber Bundle Configurations

S ID OD FiberDIA ERR1 ERR2 BPE BPE
Design Tubing Nr
g (m) (Gtm) (|tm) (|tm) (|tm) (|tm) Norm
Tx: 105/125
x: 105/125 21HW 457.2 812.8 375 93.6 41.1 134.7 0.2694
Rx: 105/125
Tx: 105/125
x: 105/125 20RW 622.3 901.7 375 49.15 123.65 172.8 0.3456
Rx: 105/125
Tx: 105/125
x: 0/25 21TW 609.6 812.8 575 93.6 17.3 110.9 0.2218
Rx: 200/225
Tx: 105/125
Tx: 105/125 20RW 622.3 901.7 575 49.15 23.65 72.8 0.1456
Rx: 200/225
Tx: 200/225
x:0/25 21RW 533.4 812.8 475 93.6 29.2 122.8 0.2456
Rx: 105/125
Tx: 200/225
x:00/25 20RW 622.3 901.7 475 49.15 73.65 122.8 0.2456
Rx: 105/125
Tx: 200/225
x:200/225 21XXTW 711.2 812.8 675 93.6 18.1 111.7 0.2234
Rx: 200/225
Tx: 200/225
Tx:200/225 20XTW 723.9 901.7 675 49.15 24.45 73.6 0.1472
Rx: 200/225
Tx: 50/55
x:5055 21HW 457.2 812.8 165 93.6 146.1 239.7 0.4794
Rx: 50/55
Tx: 50/55
x:5055 20RW 622.3 901.7 165 49.15 228.65 277.8 0.5556
Rx: 50/55
Tx: 105/125
Tx: 105/125 21HW 457.2 812.8 235 93.6 111.1 204.7 0.4094
Rx: 50/55
Tx: 105/125
x: 105/125 20RW 622.3 901.7 235 49.15 193.65 242.8 0.4856
Rx: 50/55
Tx: 50/55
x:055 21HW 457.2 812.8 305 93.6 76.1 169.7 0.3394
Rx: 105/125



The design used by Kadirval's optical microphone was a Tx: 50/55, Rx: 50/55

design fabricated by Romack. It is evident from Table 3-1 that this design could have

significant alignment issues if special packaging techniques are not used. Romack used a

compound tube structure with a smaller tube containing the fibers and a larger tube

housing the small tube.
















CHAPTER 4
FABRICATION OF THE OPTICAL MICROPHONE

The fabrication of the optical microphone consists of two parts: (1) the MEMS

optical diaphragm chip, and (2) the fiber bundle. The MEMS chip containing the silicon

nitride diaphragm was fabricated by MEMS Exchange, an umbrella organization that

combines the processing capabilities of many foundries across the country [28]. The

fiber bundle was fabricated at the University of Florida.

4.1. MEMS Exchange Process

The process flow used in the MEMS Exchange process was a modified version of

the process flow designed by Kadirval [8]. Both mask and wafers were purchased

through the MEMS Exchange. The mask was designed at the University of Florida.

Table 4-1 summarizes the wafers used for fabrication of the optical microphone.

Table 4-1 Wafers Used for Optical Microphone Fabrication [28]
Number of .
r of 5 Material silicon Diameter 100 mm
Wafers
double side 500 525
Surface Finish s Thickness Orientation <100>
polished ktm
1-10
Doping Type n-type Quality prime Resistivity / c
Q / cm
Initial State virgin Source MX Price per19.95
Wafer









The process grows the membrane layer out of silicon nitride, and uses a DRIE to

etch the bulk silicon and leave the nitride membrane. Silicon dioxide was used as an etch

stop for the DRIE. The detailed process flow is shown in Appendix A.

Not shown in Appendix A are steps to mount and demount the wafer onto a

handle wafer to provide mechanical support during the DRIE.

4.2. Packaging Process

Different novel packaging strategies for an optical microphone have been

proposed [29,30]. Abeysinghe [29] notes that "adhesives limit the range of operation of

the sensors." To minimize the amount of adhesives used, the cavity is etched at the end

of the fiber and an anodic bonding process is employed to bond an ultra-thin silicon

wafer to the end of the fiber to serve as a membrane [29]. A diagram of the packaging

technique described in [29] is shown Figure 4-1.


S IJ-Ultra-thin membrane

Machined cavity

Fiber Core

Fiber Cladding










Figure 4-1 Abeysinghe et al. Packaging Technique.

This packaging technique provides a compact device with a sensor head that is the

diameter of the optical fiber. It would function as a pressure sensor in the configuration









shown above (no vent channel), but modifications could be made to add a vent channel.

With this design, it is difficult to control the mechanical properties of the membrane.

Another possible drawback of the package is the size of the cavity. Depending on the

dimensions of the fiber, the optimal equilibrium membrane position may have both

intensity and interference modulation effect present. If both these effects are present, the

theoretical analysis of the device performance will be very difficult.

A similar packaging technique is proposed by Beggans et al. [30]. Beggans'

technique is more useful for an intensity-modulated device. First, a cavity is machined in

a glass wafer. Then, a hole with a diameter which can accompany an optical fiber is

drilled at the bottom of the cavity through the glass substrate. An ultra-thin silicon wafer

is anodically bonded to the glass substrate, creating a cavity. An optical fiber is inserted

into the cavity through the hole drilled in the substrate and fixed with a bonding agent.

The resulting device is shown in Figure 4-2.

This package allows more flexibility in membrane diameter and equilibrium gap

position, but it does not provide as much control over the equilibrium gap position. As

the fiber is being positioned, an active measurement technique is required to verify

membrane position.

Both techniques presented thus far do not protect the optical fibers. The ideal

package would provide some mechanical support for the optical fibers to reduce the

chances of device breakage. The packaging technique proposed by Kadirval [8] provides

protection for the fibers by using a fiber bundle like the type described in this thesis. The

optical fibers are protected by a steel tube and furcation tubing. Multiple MEMS chips

with through-wafer holes (identical to those used in this thesis) are bonded together in the









package. All but one of these chips has the diaphragm removed, and they are used to

make a handle wafer stack. The steel tube containing the fiber bundle is inserted through

the handle wafer stack, and the membrane wafer is placed on the top of the wafer stack.

An illustration of the proposed technique is shown in Figure 4-3.


Figure 4-2 Beggans et al. Packaging Technique.

This technique provides both a compact package and support for the fibers. A

drawback is the potential angular misalignment (other than normal light incidence)

between the fiber and the membrane. The worst-case angular misalignment depends on

the diameter of the steel tube and the thickness of the handle wafer stack. Increasing the

thickness of the handle wafer stack and decreasing the difference between the diameter of

the fiber and the wafer hole will reduce the worst-case angular misalignment.


4 Ultra-thin membrane

__ Machined cavity


Glass Substrate


Fiber Core

Fiber Cladding












MEMS Diaphragm Chip

Handle Wafer Stack




__ Epoxy

Fiber Core

Fiber Cladding


Figure 4-3 Kadirval Packaging Technique.

The technique for packaging the optical microphone design proposed in this thesis

is similar to [8]. In this proposed package, a second steel tube is used in place of a handle

wafer stack. This steel tube (package tube) would have an inner diameter equal to the

outer diameter of the steel tube (bundle tube) used in the fiber bundle construction

(approx 830 [tm). An illustration of this package is shown in Figure 4-4. Standard

available tubing gauges (Popper & Sons [43]) can provide a flush fit to within tens of

microns for this package configuration. The package tube can be an inch or more long.

Using simple geometry, the worst-case angular misalignment can be calculated.

Assuming a worst case gap of 50 |tm between the bundle steel tube and package steel

tube (a reasonable assumption based on available tube gauges), and assuming the

minimum length of 1 in. (25,400 [tm) for the package steel tube, the maximum angular

misalignment of the bundle tube in the package tube is 0.113. If a 2 in. package tube

were used, then the maximum angular misalignment becomes 0.056 o. Therefore, it can









be concluded that this packaging strategy is capable of virtually eliminating angular

misalignments.

The primary advantages of this package are that it simultaneously provides a

robust package and minimizes the worst case angular misalignment error between the

fiber bundle and the membrane, as demonstrated in the previous paragraph. Also, the

dimensions of the package steel tube can be chosen to either minimize package diameter

or to fit a commercially available calibrator such as the Bruel & Kjaer 4231 Microphone

Calibrator [31]. In addition, the outside of the package steel tube could be threaded to

allow a protective screen to be attached over the membrane, protecting it from damage.

For an example of this type of protective screen, see those used by Bruel & Kjaer

1" [32], 1/2" [33], and 1/8" [34] microphones. Another advantage of this package is the

ability to take multiple packaged optical microphones and easily assemble them into a

microphone array bundle.












N IE IS Diaphragm Chip

Epoxy


Package Steel Tube


Bundle Steel Tube



Epoxy









Device Furcation
Tubing


Figure 4-4 Proposed Package for the Optical Microphone.

In the top view in Figure 4-5, seven packaged optical microphones are shown in a

bundle surrounded by an array package steel tube, whose dimensions would be selected

to make the fit as tight as possible. The cylindrical structure of the individual

microphone packages allows the microphone array geometry to be a scaled version of the

fibers in the individual microphone package. This allows an arbitrary number of "rings"

of packaged microphones to be used in the array, without the individual packaged

microphones interfering (mechanically) with each other. In addition, the cylindrical

structure of the proposed array package would be easy to construct, using only a custom


> 1"





steel tube and epoxy to hold it together. Due to the steel tubing, this proposed array
package would be very robust.
Side View



E-



Top View


2.5 mm 1mm



7.5mm
9.5 mm


Figure 4-5 Proposed Optical Microphone Array Package.











In Figure 4-5, the membranes have been removed from the square chips to show

the individual microphone cavity positions. This proposed microphone array could

sample the acoustic field on the center 50 utm of the 1 mm diameter diaphragm. Each

diaphragm is 2.5 mm from each of its neighbors. Finally, the performance of the

microphones in the array is independent of the number of microphones in the array

(assuming the array package does not affect the sound field and availability of sufficient

light sources, detectors, etc).
















CHAPTER 5
EXPERIMENTAL SETUP AND RESULTS

The experimental characterization of the optical microphone is divided into three

sections. First is the experimental characterization of the laser and photodetector. This

characterization measures the value of Plaser noise(Plaser) (compare to Plight noise in Chapter

2). The experimentally measured value Piaser noise(Plaser) is used as an input to the

theoretical model of the microphone presented in Chapter 2.

Second is the static calibration of the custom fiber bundle. This static calibration

attempts to verify the theoretical power coupled vs. equilibrium gap and sensitivity vs.

equilibrium gap plots. The static calibration curve measured in this experiment is used to

identify the location of the fiber bundle with respect to the membrane in the dynamic

calibration experiments.

The final experimental step is the dynamic calibration of the optical microphone.

In the dynamic calibration, a plane wave tube (pwt), speaker, and calibrated microphone

are used to determine the optical microphone sensitivity, linearity range, frequency

response, noise floor, dynamic range, and minimum detectible signal. The experimental

results are then compared with the theoretical predictions. The dynamic calibration is

performed for both the unreferenced and referenced output microphone configurations.






80

5.1. Laser and Photodetector Characterization

5.1.1. Experimental Setup for Laser and Photodetector Characterization

In the characterization of the laser (or other light source), the optical spectrum of

the photodetector was measured for laser outputs of 100 pW through 500 [IW, in steps of

50 LiW. The noise power spectral density of the detector output, in V / Hz", was

recorded by the Pulse system from 0 6.4 kHz with a bin width of 1 Hz and using 500

samples. This value is the RMS sum of the detector noise and detected light intensity

noise, both in V / Hz"2. The optical intensity noise component of the detector noise is

determined by removing the theoretical Vdet from the measured value, leaving Vlight (see

Section 2.2.4 for variable definitions). By dividing Vlight by the photodetector

responsivity and gain, the optical intensity noise of the laser, in V / Hz1/2, can be

determined.


Pulse Computer
ISS-1550 PDA-400
ISS-1550 PDA-400 System Ethernet Port
Opto-isolator Photodetector S E

-.......... ......... -- -
HP8168B
Laser

KEY
Electrical Signal .................. System Input [F[fj

Optical Signal System Output

Acoustical Signal ---- System Component

Figure 5-1 Experimental Setup for Laser Characterization.

The DC optical power was obtained by observing the 0 Hz frequency bin. The

worst-case light source noise recorded at 660 Hz, and the best-case noise (where the laser









noise became constant with respect to frequency) was measured at 1600 Hz. The

experimental setup for the laser noise characterization is shown in Figure 5-1.

5.1.2. Results of Laser and Photodetector Characterization

The experimentally measured value for Plaser noise is given by the last column in

Table 5-1. The HP8168B was observed to have a flat noise floor above 1550 kHz.

Below 1.55 kHz, the noise floor was not flat. Since this system would ideally be used

with a heterodyne detection scheme to operate the microphone in the flat noise range of

the light source, the laser linearity and MDS will be experimentally measured above 1550

Hz. For all laser characterization experiments, Gdet = 15,000 V / A.

Table 5-1 Experimental HP8168B Noise
Measured Noise @ Calculated Laser Laser SNR
Laser Power
1600 Hz Noise @ 1600 Hz @ 1600 Hz
(lW) (uV / Hz1/2) (pV / Hz1/2) (dB)
100 0.741 52.0 53.0
150 1.29 90.2 52.4
200 1.53 108 52.9
250 1.06 74.4 55.6
300 1.98 139 53.6
350 0.774 54.3 58.4
400 1.68 118 55.7
450 3.59 252 52.9
500 3.26 229 53.9


The laser was observed to have a maximum SNR at Plaser =350 |LW. Therefore,

this value will be used to characterize the laser.

5.2. Static Calibration

5.2.1. Experimental Setup for Static Calibration

The static calibration of the optical microphone has two goals. The first goal is to

verify the corrected power coupled model presented in Chapter 2. The second goal is to









obtain a power coupled curve for the fiber bundle under consideration, so that it may be

used to measure the equilibrium gap of the assembled microphone package. The

procedure for this is described further in the dynamic calibration section.

In the static calibration experiment used by Kadirval [8], the fiber bundle is

placed flush against a metal mirror. To align the fibers normal to the mirror, Kadirval

assumed that the power coupled into the receive fibers was maximum when the fibers

were aligned normal to the mirror. While there may be a range of equilibrium gap

distances that this assumption is valid for, it is not valid in general, and will result in the

proper alignment of the fibers with the mirror. Kadirval used a micropositioner to move

the fiber bundle with respect to the mirror. Due to the way in which the fiber bundle was

mounted, the fibers were able to slip in the mechanical mount as the micropositioner was

moved. Due to these two problems with the previous static calibration experiment in [8],

the gap measurements reported by the micropositioner were significantly larger than their

actual values, resulting in an incorrect calibration curve.

In this thesis, these two issues have been fixed. The mechanical mounts for the

fiber and the mirror were aligned normal to each other by cubic blocks, which have

parallel surfaces. The mirror was mounted on the micropositioner and was moved with

respect to the fiber bundle, whose position was fixed. The zero position of the mirror was

set by slowly moving the mirror with the micropositioner until mechanical contact with

the mirror was observed. After mechanical contact has been made, the fiber holder is

closed, fixing the fiber bundle in place. Note: the mirror used in these experiments was

slightly rusted and had scratches on its surface. The method of zeroing used here can









potentially scratch the mirror, so it should not be used with a mirror in good condition.

An electrical contact method could be used instead.

The static calibration setup used for the optical microphone is similar to the setup

used by Kadirval [8] except as noted in the previous paragraph. The experimental setup

for the static calibration is illustrated in Figure 5-2. In Figure 5-2, the custom fiber

bundle is oriented normal to a reflective mirror, which is mounted on a computer-

controlled micropositioner. The computer-controlled micropositioner is automated by a

Labview application to sample the power output of a reference channel and the output of

the fiber bundle. After taking 30 samples, the micropositioner adjusts the mirror position

relative to the fiber bundle and repeats the measurement. This step-and-sample process

continues until cancelled by the user.

The light source used is the LPS-SMF28-1550-FC laser diode operated in

constant current mode by a Keithley 2400 constant current source. Laser light is passed

through a Newport ISS-1550 optical isolator and is split by a 50/50 optical power splitter.

One output of the splitter delivers light into the Tx fiber of the custom fiber bundle. This

light reflects off of the mirror and is collected by the receive fibers, which transport the

light to a PDA-400 detector (TEST). The second splitter output delivers light directly to

another PDA-400 (REF). A Keithley 2000 multimeter controlled by a Labview

application sampled the TEST and REF detector outputs, and sends the data to a

computer where it is recorded in a file along with the micropositioner position. The gain

of the PDA-400 detectors in the static calibration is 15,000 V / A.

After the experiment is complete, the recorded detector outputs are corrected for

the experimentally measured bias errors of each photodetector. Dividing the bias-







84


corrected TEST data by the bias-corrected REF data and plotting it with respect to the

measured gaps produces the experimental power coupled curve. The results of the static

calibration are presented in the next section.


Micropositioner
mController
..................................- -
Compur Micropositioner
Computer
Serial Port
Transmit Fiber,
Custom Bundle


LPS-SMF28-1550-FC
Laser Diode

-................. --


Keithley 2400 ISS-1550
Sourcemeter Opto-isolator

PDA-400
Keithley 2000 Photodetector
Multimeter (REF)
Computer ..................................
GPIB Port


Aluminum Mirror


Receive Fiber,
Custom Bundle








PDA-400
Photodetector
(TEST)


KEY

Electrical Signal

Optical Signal

Mechanical Signal


System Input

System Output

System Component


Figure 5-2 Block Diagram of Static Calibration.


5.2.2. Results of Static Calibration

The output of the static calibration is an experimental plot of optical power

coupled vs. equilibrium gap distance. This plot is shown in Figure 5-3 shows the static


DIWIIIII


1550 nm


.......-----....

---

- -










calibration curve for the optical microphone using the custom fiber bundle. The plot also

shows the theoretically predicted curves for an RPE of 0 ptm and an RPE of 10 ptm. The

maximum power coupled is slightly higher than theoretically predicted, and the location

of the peak power coupled is at a larger gap than predicted.


0 100 200 300 400 500 600 700
Gap (p.m)
-RPE =0 um -RPE =100 um r Expenmental

Figure 5-3 Experimental Power Coupled vs. Equilibrium Gap.


800 900 1000


The location and magnitude of the max slope of the experimental curve was

determined by fitting a regression line to points in a 30 |tm window on the experimental

power coupled curve. The location of the maximum slope of the regression line was

assumed to be the gap at the maximum slope, and the maximum slope of the regression

line was assumed to be the peak slope of the experimental power coupled curve. The

regression line at the peak slope is shown in Figure 5-4.







86


Table 5-2 Comparison between Theoretical and Experimental Static Calibration

Theoretical Theoretical
Category Experimental
Category (RPE = 0 jim) (RPE = 10 m) experimental
Peak Power Coupled 21.6 % 17.2 % 22.1%
Gap @ Peak Power (,m) 500 530 600
Power Coupled @ Max 52 4.2 10.9
Slope
Gap @ Max Slope (rtm) 230 260 360
Maximum Slope (tm-1) 1.094 x 10-3 0.784 x 10-3 0.800 x 10-3


25.0%


20.0%


15.0%


10.0%


5.0%


0.0%


200 240 280 320 360 400 440 480 520 560 600
Gap (um)

Figure 5-4 Experimental Maximum Power Coupled Regression Line Slope.


The R2 value of the regression line in Figure 5-4 indicates the MO stage is linear

over the gap region from 350 |tm to 380 |tm. Since the membrane deflection is much less

than 30 |tm, the MO stage is linear when operated at the maximum sensitivity.

In both Figure 5-3 and Figure 5-4, the experimental power coupled curve is much

less smooth than the theoretical curve. At some gap values the power coupled seems to

be discontinuous with the surrounding power coupled points. This is believed to be

caused by two phenomena. The first cause is table vibrations. The optical table on which


y=0.0008x- 0.162
R 2 0.9948









the experiments were performed was not isolated from the ground by a cushion of

compressed air. This allows acoustic vibrations traveling through the floor of the room to

vibrate the mirror during the measurement. The second cause is the roughness of the

mirror surface. The mirror used in this experiment was observed to have scratches and

rust spots on the mirror surface. Light interacting with the scratches and rust spots can

cause discontinuities in the measured power coupled.

5.3. Dynamic Calibration

5.3.1. Experimental Setup for Dynamic Calibration

The goal of the dynamic calibration is to characterize the sensitivity, minimum

detectable signal, electronics noise, linearity, and frequency response of the unreferenced

and referenced optical microphone configurations, and to compare the experimental

performance of the microphone with the theoretical performance.

In the both unreferenced and referenced output calibration experiments, the cut-on

frequency of the SRS 560 was 30 Hz, the gain of the SRS 560 was 1 V / V, and the trans-

impedance gain of the PDA-400 detector(s) was 15,000 V / A. Data is taken with the

Pulse system from 0 Hz 6.4 kHz with a 2 Hz bin width. All single tone measurements

were made with a sinusoidal input at 1600 Hz. A uniform window was used to measure

single tone signals, and a Hanning window was used for measuring broadband or noise

signals. In all experimental setups (Figure 5-5 and Figure 5-6), a B&K 4138 1/8"

microphone [34], pre-calibrated with a B&K 4228 pistonphone [35], is used as a

reference microphone. This microphone is mounted next to the optical microphone at the

end of the plane wave tube, and it connected to an input channel of the Pulse system (not






88


shown in Figure 5-5 and Figure 5-6). The experimental setups for the optical microphone

dynamic calibration are similar to the setup used by Kadirval [8].


Pulse System Amplifier JBL Speaker and
m Plane Wave Tube
...... ............. > .....P.a.ub.

Computer
Ethernet Port

ISS-1550
Opto-isolator


HP8168B Laser
Source


MEMS Chip


Transmit Fiber,
Custom Bundle


Receive Fiber,
Custom Bundle


PDA-400
Photodetector
(DET)


Computer
Ethernet Port Pulse System


SRS 560 Filter / Amp


KEY
Electrical Signal

Optical Signal

Acoustical Signal


System Input

System Output

System Component


Figure 5-5 Unreferenced Output Optical Microphone Configuration.

A "power coupled alignment" technique is introduced for measuring the

equilibrium gap. This technique makes a power coupled measurement as the fiber bundle

and membrane chip are mounted on the custom plane wave tube (PWT) plug and relates

the power coupled measurement to an equilibrium gap by the static calibration curve.


DIEE1


.................. 01.







89


The first experimental setup is the unreferenced output configuration, shown in

Figure 5-5. The optical splitter is not required for operation of the microphone in this

configuration. In the unreferenced output configuration, the system output is taken at the

output of the SRS 560 Filter / Amplifier.


Pulse System Amplifier



JBL Speaker and
Computer Plane Wave Tube
Ethernet Port
ISS-1550
Opto-isolator


HP8168B Laser
Source


Optical Splitter
1550 nm


PDA-400 Photodetector
(REF)


MEMS Chip


e Fiber,
i Bundle


(REF)


Pulse System
Computer 1 I0*MOD
Ethernet .. .
Port REF
SRS 560 Filter / Amp Analog Divide Circuit
(Optional)


KEY

Electrical Signal

Optical Signal

Acoustical Signal


System Input

System Output

System Component


Figure 5-6 Referenced Output Optical Microphone Configuration.


F-D


....------. .

---

- -