JOurn.31 of iLn.derr.3dua.3- Re--search
.Oluini ', issue 2 - r.o, ein er - ..' ::Deceinier -.
Design and Fabrication of a Phased Acoustic Array to Analyze Noise
Generation of Aircraft Components
This paper presents an overview of the development of a microphone array for testing within the University of
Florida Aeroacoustic Wind Tunnel Facility. Acoustic measurements are to be taken from a scaled model of a NACA
63-215 airfoil. The array design contains a total of 33 electret microphones - each of which is 6.0 mm (0.236
inches) in diameter - distributed on an 8.5-inch diameter circular plate. The microphones are distributed into
four concentric circles, eight per circle, with an additional microphone located at the center of the array. The
primary purpose of this array is to obtain directivity and spectral content information of aerodynamic noise
emanating from regions of interest on the airfoil model under different flow conditions. This is accomplished by
a process known as beamforming, in which the focus of the array is electronically steered to points in space.
The secondary objective of this research is to perform a benchmark comparison to directional array data obtained
by the NASA Langley Quiet Flow Facility (QFF).
In recent years, increasingly strict regulations on aircraft noise have imposed large economic penalties on
aircraft companies and airlines that fail to comply. As engine technology leads to quieter engines, airframe noise
- defined as the "...non-propulsive component of aircraft noise which is due to unsteady flow about the
airframe components...." 1 - has become a major contributor to the overall aircraft noise levels. The physics
behind airframe noise generation is still not fully understood and must be characterized before reduction
techniques can be implemented 2
Early techniques of airframe noise analysis involved the concept of an "acoustic mirror," which consisted of a
single microphone positioned in the acoustic far field of a large concave elliptical mirror. The origin of
acoustic mirrors can be traced back to the north and southeast coasts of England in the early 1920s, where they
were used to provide early warning of incoming enemy aircraft planning to attack coastal towns 3. These
coastal "listening ears" were eventually rendered obsolete with the development of faster aircraft and the invention
Such mirrors have been proven to accurately locate individual sound sources, but are limited in that they
require physical adjustments in order to determine the distributions of these noise sources around models.
This prevents wide-scale use of such mirrors, although they are still applicable in some larger-scale
research facilities. Other methods of airframe noise characterization have involved independent distributions
of microphones, which have proven to be useful in the analysis of airfoil self-noise generation 4. These
microphone distributions are considered independent due to the fact that the microphone outputs
remain individualized and are not combined. These microphone arrangements are considered to be precursors
to modern microphone directional arrays, based on their use of amplitude and phase relationships
between microphone clusters 1.
Phased or directional arrays capitalize on the amplitude and phase variations sensed by a spatially assorted
collection of microphones in the acoustic far-field of a noise source 5. The primary benefit of utilizing such an array
is its ability to extract a desired noise source location and information regarding relative source strengths, even
when located in a reverberant, noisy environment. Another important benefit is having the ability to
electronically "steer" its focus in a region in space, requiring no physical movement. This concept is a
vast improvement from an acoustic mirror, which relies on mechanical movement to alter its focus. This
electronic steering is done via a process known as beamforming, or the appropriate weighting and time-delays of
the individually measured signals 5.
DIRECTIONAL ARRAY THEORY
Principle of Operation
The primary objective of a microphone directional array is to have the capability of focusing on a noise source
of interest. In order to do this, the outputs of the individual microphones must be phase-shifted by an
amount corresponding to their modeled propagation delay from the point of interest and summed together.
The basic principle of such an array is presented in 5 and is summarized as follows.
Assume there is a simple acoustic monopole source located a distance ro from the origin of a circular array of
N microphones. The origin, or phase center, of the array c is defined as
where n is the location of the nth microphone. The acoustic signal can be denoted as a pressure
wave propagating spherically in all directions as:
p(,r, t) = - exp[jcot - kr)]
where C is a constant, r is the radial distance from the origin of the noise source, a) is the wave frequency, and k
is the corresponding wavenumber. Note that the wavenumber is defined as k = a/c where c is the speed of
sound. For an array of N microphones located a finite distance from the noise source, each microphone detects
a slightly different phase-shifted waveform depending on its distance. The measured pressure at the
nth microphone, pn(t), is represented as:
pf (t) = e -r/
where n is the distance from the location of the noise source to the nth microphone, c is the speed of sound,
(t -- , /C)
and n represents the delayed time from the noise source to the microphone. A positive time delay
is used for the microphones closer to the noise source, whereas a negative time delay is used for those
located farther from the source. This scenario is illustrated below in Figure 1.
Figure 1. Acoustic point source located a distance X from the origin of a microphone array.
(Adapted from 5.)
Beamforming provides a microphone array with the ability to effectively amplify sound in a region of interest
while diminishing sound from other regions. This process is useful in its ability to adapt the performance of
a directional array to the conditions associated with aeroacoustic testing of a model in a reverberant
environment, such as a wind tunnel. The two most general classifications of beamforming are time- and
frequency-domain delay-and-sum beamforming 5.
Considering the previously presented scenario of an acoustic source located a distance ro from the origin of a
circular array of N microphones, the general form of the continuous-time beamforming equation can be presented:
l(^ =- ,o-pjt-A,)
where wn and ?n are the weighting factor and time delay for the nth microphone, respectively. These terms can
be further expanded into the following:
0 0ï¿½- 0
i'-o- A - 7c
Inserting Equations (2) and (4) into Equation (3) yields the following expression for array response:
i 11, rï¿½ rï¿½-rï¿½
( N 0 0 0 1
n=1 11 C C
N 1 0 0\
.n=1 0 C
- N. f0(t)
Thus, the output of the array is reduced to the pressure signal received at the center of the array multiplied by
the total number of microphones in the array. Note that this simplification only applies to the situation when
the array's focus is directed at the source of the noise and it represents the maximum possible output of the array
for the given system configuration. Otherwise, considering an arbitrary location in space whose parameters are r,
rn, and wn, instead of ro, ron, and wOn, the response of the array is:
!n=1 r frn C
Equation (6) represents the response of a directional array using basic weighting and time-delays indicative
of conventional beamforming to focus the array on an arbitrary location in space. In practice, the output signal of
the array z(t) is monitored while its focus is scanned over a pre-defined spatial region (see Figure 2).
Figure 2. Illustration of directional array steering. (Adapted from 1.)
In signal processing, a time-delay for a signal corresponds to a phase shift in the frequency (Fourier)
domain. Applying this concept to beamforming yields the following general correlation:
p (t - to)( P (o)(mxp(- j ,)
Furthermore, this transformation can be applied to Equation (3) to yield the beamformed array response in
the frequency domain:
z(t) - Z():
Z(au)= wP,ï¿½(w)exp(- jwA,)
For this transformation to be implemented for experimentation purposes, the Fast Fourier Transform (FFT) must
be applied. For a set of data collected at regular intervals by a data acquisition system p[m], at a
sampling frequency fs, the FFT used to obtain a discrete representation of M points is given by
P = pm xp(2M ) , k = 0,1,2,...M -1
where Pk is the kth FFT coefficient. Thus, the expression for the array response in the frequency domain
from Equation (8) can be redefined as
where wk = 2nkfs/M is the radial frequency.
In addition to delay-and-sum beamforming in which each individual microphone output is multiplied by a
single complex weight factor, another method exists in which the microphone outputs are multiplied by a vector
of complex weightings6. This method, known as matrix weighting, will theoretically yield a better spatial
resolution. It will also be considered during the data analysis phase of this project.
Since beamforming to a point in space is dependent on the vectors from this point to the microphone locations,
any difference between the desired and actual microphone locations will result in errors. One of the primary
causes of such differences in microphone locations is offset errors made during the manufacturing process. In
order to ensure that the computed weight vectors for each of the microphones match the experimentally-
measured data, a calibration of the array must be performed. A common method of directional array calibration
is with the use of a small speaker used to simulate a point source7. It is advantageous to place the speaker near
the location of the model to be analyzed, with it and the array acoustically insulated from the walls and
other reflective surfaces present in the testing facility. Array data is then collected and adjusted for a range
of analysis frequencies using calibration guidelines similar to the following7.
For an array center microphone with an ideal location at the array origin
and an actual offset location denoted by mic, a quantity known as the phase factor denoting the nominal
and actual coefficients of propagation can be defined respectively as
w = exp( k.x - x )
where s represents the speaker location in space. With this information, a corrected weighting factor for a
point source location at can be defined:
wTCI expo Oï¿½I - li'c )-a -
w = I - xp(k1-
Wn., exp 6-1
Therefore, application of Equation (12) to all microphones in the directional array will yield a corrected weight
factor for each microphone. These weight factors are then summed to yield a single calibrated array response,
the formulation of which is outlined in the next section. A visual representation of the parameters involved in
the array calibration process can be seen in Figure 3.
nominal mic locations
actual mic locations
Figure 3. (a) Speaker/array apparatus for calibration of center microphone, and (b)
corrected microphone weightings for a noise source after calibration.
The theoretical response of a directional array is based on the assumption that a noise source seen by the
array exhibits monopole, or omnidirectional, behavior. The actual performance of a directional array with
the application of beamforming is almost always different from that predicted by the theoretical model due to the
fact that real-world noise sources practically never behave as independent monopole sources. For the sake
of simplicity, however, the ideal response of a directional array to a noise source is considered with the
assumption that the source exhibits monopole behavior1:
WiL j, co w expfko -r)-Y (- r
where is an arbitrary location in space and is the noise source location. Another way to express the
array response is in decibels (dB) relative to the level obtained at the noise source location:
dB(1) = 20logo1 u ,
In this form, the array response can be plotted as a contour map representing the computation of Equation (14)
over a series of steering locations located a specified distance from the array. This type of plot allows one to
examine the spatial selectivity of the array, indicated by a mainlobe and sidelobes. The mainlobe of the
array response pattern is of particular interest since its acoustic energy represents the desired response of the
array to a noise source, the width of which is defined as the array beamwidth5. In addition, the array response
may be represented in terms of normalized pressure, in which the peak response of unity represents the origin of
the noise source2. Examples of such plots are provided in the Preliminary Results section of this report.
The experiments to be conducted will be performed with the intent of examining the root causes of sound
generation on, for example, a wing configuration. They will be performed in the University of Florida
Aeroacoustic Wind Tunnel Facility. The wind tunnel is an open-jet, quiet-flow facility designed for anechoic
(echo-free) testing of airframe components. Characterization of the tunnel has shown a maximum attainable
test section speed of approximately 250 ft/s. The test stand, which is the apparatus designed to hold and orient
the test model within the test section consists of aluminum rectangular channels manufactured by 80/20 Inc.
This test stand was designed to have a large safety factor for withstanding the forces imposed on it from the air
flow through the test section. Figure 4 is a rendered model of the wind tunnel test section, including the test
stand and sample directional array placement.
Figure 4. Digital reproduction of wind tunnel test section with mounted directional array. (Courtesy of
3. Sanford and A. Hart.)
Airfoils or wing profiles are an important component of airframe noise generation. The model that will be used to
test the response of the array is a ---scaled NACA 63-215 Mod B airfoil (Figure 5). One of the primary
acoustic features of interest of the airfoil in this study is the rear, or trailing edge. Generally, the trailing edge is
the section of the airfoil that possesses the greatest amplitude of sound generation and is thus a feature
that deserves special attention. Figure 6 is a scaled profile view of the wind tunnel test section, including the
NACA airfoil model as well as a sample array placement.
NACA 63(2) z15 MOD B
0 01 02
Figure 5. Plot of the
Percent of chord length
NACA 63-215 airfoil (Courtesy of ).
Test Test Model
-.-. ^ ^
Diffuser * i-
Figure 6. Profile view of wind tunnel test section (units are in inches).
Directional Array Pattern and Structure
The most important design parameter of a directional array is the microphone layout. Selecting a microphone
layout is heavily dependent on the nature of the noise source distributions to be analyzed. The general procedure
for delay-and-sum beamforming previously mentioned is based on the assumption that any detected noise source
is a monopole type, and that a source distribution is comprised of a series of uncorrelated simple monopole
sources. Unfortunately, noise sources from airframe components practically never exhibit such behavior.
Instead, the face of the array sees fluctuations in the phase and amplitude of the noise sources, which may
cause array response errors in noise source localization. Therefore, designing the array to have the
microphones placed within close proximity of each other would place them approximately within the same
source directivity - yielding a smaller array size .
The array implemented for this project is a Small Aperture Directional Array (SADA) , designed to
provide directivity and spectra information from regions of interest on the previously mentioned NACA airfoil
model under different flow conditions. It consists of 33 Panasonic electret microphones, each one having a
diameter and height of 6.0 mm and 3.4 mm, respectively (Figure 7).
Figure 7. Panasonic 6.0-mm diameter electret microphone, Model # WM-61A.
The array pattern is comprised of four concentric circles of eight microphones each, centered around a
final microphone located at the array center. Each circle of microphones has twice the diameter of the circle
it encloses. The pattern is consistent with that chosen by Humphreys  that was implemented in the NASA
Langley QFF (Figure 8). Note that each circle in Figure 8 represents one 6.0-mm diameter electret microphone.
" -0-04- d _- 0 mnrnm I1, 2 j6 in.) 0
4 .3 2 .1 0 1 2 3 4
Figure 8. SADA microphone pattern. (Courtesy of T. Yardibi.)
Each microphone signal is to be output via a BNC cable to a National Instruments PXI Data Acquisition module.
Due to a current difference between an input channel of the PXI module and the maximum possible current that
can be handled by each microphone, additional circuitry was required to have an effective microphone circuit. It
was decided that the microphone array be based off of a printed circuit board (PCB), which would both hold all
the electrical components and provide a stable means of support for the array. In addition, it was deemed
necessary to include a cover, or array faceplate, for the purpose of eliminating the possible occurrence of
acoustic scattering by the electrical components. This faceplate was manufactured out of hard plastic using a
rapid prototyping machine. Figure 9(a) shows the completed array PCB circuitry and Figure 9(b) shows the
array with the manufactured faceplate installed.
Figure 9. (a) Printed circuit board SADA circuitry, and (b) array with faceplate installed.
As was mentioned in the Array Response section, application of beamforming to a simulated response of the SADA
to a noise source was performed for frequencies of interest. Figure 10 is a sample contour plot of the
theoretical response of the SADA for a noise source located approximately 4 feet above the array center at
a frequency of 10 kHz. As the figure shows, the sampling grid scanned by the array is a 4-foot by 4-foot square
with the peak noise source occurring at the center of the plot, represented by 0 dB. Note that this contour plot
was generated with the implementation of a weighting algorithm by Tarik Yardibi of the University of Florida
Spectral Analysis Laboratory.
-2 -15 -1 -0J5 0 05 1 1,5 2
Figure 10. Theoretical SADA response in dB at a frequency of 10 kHz. (Courtesy of T. Yardibi.)
Another way to view to the relative response of the array is in terms of normalized pressures. In other words,
the pressure distribution within the sampling area seen by the SADA varies from 0 to 1 with 1 representing the
peak pressure and origin of the noise source. Figures 11(a) and (b) show the array response in terms of
normalized pressures to a noise source with the same parameters previously mentioned for Figure 10. Note
that these figures were generated using MATLAB source code implementing beamforming methods presented
SADA Theoretical Response (f - 10 kHz)
y tin I
-20 -20 x (in.)
.20 -10" x0 " i 10 2
x [in I
Figure 11. (a) 3D surface plot and (b) contour plot of theoretical array response in terms of
normalized pressure at a frequency of 10 kHz. (Adapted from .)
In addition to these array response simulation tools, a microphone array software (MAS) package provided by
the National Instruments Corporation has also been tested. Once several modifications are made to this
software package, it is expected to be the primary method of SADA data analysis. The functions of the MAS are
to test the array with a simulated noise source of a given strength at a given location as well as to
apply beamforming algorithms to measured microphone time data. Figure 12 shows the graphical user interface
of the noise simulation portion of the software.
v i f) Amp (dB) A
Sound Pressure Distribution
-0 3 -
-0 6 -.4 -0.2 0 0.2
Figure 12. Microphone array software (MAS) user interface for noise source array response simulation.
As the figure shows, the user inputs a desired scanning range, distance of the noise source from the array,
frequency of interest, and other parameters (not displayed). The result of the noise source simulation (seen at
SAOA Theoretical Response (f w 10 HHm]
right of Figure 12) is a contour map showing the sound pressure distribution in dB within the scanned spatial
region. A comparison of Figures 10, 11, and 12 show similar SADA response simulations to a centrally-located
noise source. This is beneficial because it shows that there are multiple simulation tools with which
the experimentally measured array data can be compared.
In this paper, an overview of the design and fabrication of a directional microphone array for aeroacoustic testing
has been presented. A Small Aperture Directional Array (SADA) has been successfully constructed and is
currently undergoing calibration procedures and ideal placement within the University of Florida Aeroacoustic
Wind Tunnel Facility. To date, experimental wing acoustic data obtained with the fabricated SADA has not yet
been collected. As was mentioned, a series of array response simulations utilizing calibration and
beamforming correlations have been obtained and tested to validate the theoretical results with those obtained
by the NASA Langley Quiet Flow Facility (QFF). In addition, the University of Florida Wind Tunnel Facility is
currently undergoing acoustic insulation modifications in order to ensure ideal aeroacoustic testing
conditions. Experimentation will begin by mid-June of 2007.
I would like to express special thanks to my adviser, Dr. Lou Cattafesta, for the invaluable guidance he has
provided me during the three years I have worked for the Interdisciplinary Microsystems Group. The students of
IMG also deserve thanks, particularly Christopher Bahr, who has been an excellent source of knowledge and
advice for this project, as well as Dylan Alexander for his assistance in the array fabrication process, and,
finally, Adam Hart, Jeremy Sanford, and Drew Wetzel for their improved design work on the Aeroacoustic
Wind Tunnel Facility. I would also like to acknowledge Tarik Yardibi of the University of Florida Spectral
Analysis Laboratory for his work on the application of beamforming techniques to this study. Finally, I would like
to thank my family and friends for their support of me in all of my endeavors.
1. Humphreys Jr., W.M., Brooks, T.F., Hunter, Jr., W.W., and K.R. Meadows, "Design and Use of Microphone
Directional Arrays for Aeroacoustic Measurements", AIAA Paper 98-0471, 1998.
2. J.R. Underbrink, "Practical Considerations in Focused Array Design for Passive Broad-band Source
Mapping Applications", Master's Thesis, Pennsylvania State University, 1995.
3. A. Grantham, "Early Warning Sound Mirrors." Available online: http://www.ajg41.clara.co.uk/mirrors/index.html
4. Brooks, T.F., Pope, D.S., and M.A. Marcolini, "Airfoil Self-Noise and Prediction", NASA Reference Publication
1218, July, 1989.
5. D. P. Arnold, "A MEMS-Based Directional Acoustic Array for Aeroacoustic Measurements", Master's Thesis,
University of Florida, 2001
6. Li Y. Xie, J. Li, X. Zheng, and J. Ward, "Optimal beampattern synthesis via matrix weighting," submitted to
IEEE Transaction on Signal Processing, July 2006
7. R. P. Dougherty, "Beamforming in Acoustic Testing", Aeroacoustic Measurements, p. 62-97, Springer, 2002.
8. "UIUC Airfoil Coordinates Database". Available online: http://www.ae.uiuc.edu/m-selig/ads/coord_database.html#N
Back to the Journal of Undergraduate Research
College of Liberal Arts and Sciences I University Scholars Program I University of Florida I U UNIVERSITY of
ï¿½ University of Florida, Gainesville, FL 32611; (352) 846-2032.