1 DESIGN, FABRICATION AND CHARACTERIZATION OF A MEMS PIEZORESISTIVE MICROPHONE FOR USE IN AEROACOUSTIC MEASUREMENTS By BRIAN HOMEIJER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008
234
Embed
DESIGN, FABRICATION AND CHARACTERIZATION OF A MEMS ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
DESIGN, FABRICATION AND CHARACTERIZATION OF A MEMS PIEZORESISTIVE MICROPHONE FOR USE IN AEROACOUSTIC MEASUREMENTS
By
BRIAN HOMEIJER
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
1.2.1 Pressure, Free and Diffuse Field Microphones .................................................19 1.2.2 Linearity and Total Harmonic Distortion..........................................................21 1.2.3 Noise Floor........................................................................................................21 1.2.4 Requirements for an Aeroacoustic Microphone ...............................................22
3.1.7 Validation Using Finite Element Analysis........................................................55 3.2 Electroacoustics ............................................................................................................56
3.3 Lumped Element Modeling ..........................................................................................63 3.3.1 LEM of piezoresistive microphone...................................................................65
3.3.1.1 Diaphragm ..........................................................................................65 3.3.1.2 Cavity..................................................................................................67 3.3.1.3 Vent.....................................................................................................68 3.3.1.4 Equivalent circuit................................................................................70 3.3.1.5 Cut-on frequency and cavity stiffening ..............................................71
4.2 Optimization Results.....................................................................................................92 4.2.1 Optimization with Constant Voltage.................................................................92 4.2.2 Optimization with a Constant Current Source ..................................................94 4.2.3 Constraining Devices to a Single Wafer ...........................................................94 4.2.4 Sensitivity Analysis...........................................................................................96 4.2.5 Uncertainty Analysis.........................................................................................96
6.2 Experimental Results ..................................................................................................128 6.2.1 Electrical Characterization ..............................................................................128 6.2.2 Noise Floor......................................................................................................131 6.2.3 Linearity and Total Harmonic Distortion........................................................131
7
6.2.4 Frequency Response .......................................................................................132 6.3 Model Validation ........................................................................................................132
6.3.1 Variables and Standard Deviations .................................................................133 6.3.2 Model Validation Results................................................................................133
3-1. Overview of the microphone modeling process. ...............................................................76
3-2. Schematic of composite plate. ...........................................................................................76
3-3. Kirchoff's hypothesis showing the neutral axis and transverse normal. ............................77
3-4. Non-dimensional center deflection per unit pressure of devices with varying in-plane forces..................................................................................................................................77
3-5. Pressure that results in a 5% deviation from linearity for various inplane forces. ............78
3-6. Analytical deflection of clamped plate, at the onset of non-linearity (2000 Pa), compared to FEA results....................................................................................................78
11
3-7. Center deflection per non-dimensional pressure as a function of P* for various values of in-plane stresses.............................................................................................................79
3-8. Description of the Euler’s angles.......................................................................................79
3-9. Crystallographic dependence of the piezoresistive coefficients for p-type silicon............80
3-10. Piezoresistive factor dependence on doping concentration at room temperature..............81
3-11. Geometry of piezoresistors. ...............................................................................................81
3-12. Differential elements of the arc and taper resistor. ............................................................82
3-14. Stressed arc and taper resistors configured in a Wheatstone bridge. .................................83
3-15. Schematic of MEMS microphone and associated lumped elements. ................................83
3-16. Example of a the distributed system and the lumped equivlent.........................................84
3-17. Equivalent circuit model of the microphone......................................................................84
3-18. Accuracy of first terms of cotangent expansion. ...............................................................85
3-19. Magnitude and phase response of LEM normalized by the flat band response.................85
3-20. Equivalent circuit illustrating the effect of the cavity compliance ....................................86
4-1. Operational parameter space for a microphone. ..............................................................101
4-2. Multiobjective optimization Pareto front illustrating the trade-off between minimizing function J1 and J2. .........................................................................................102
4-3. Ideal linear output of a microphone or pressure transducer.............................................102
4-4. Features that are constrained to be larger than wline. ........................................................103
4-5. MDP vs. Bandwidth for various Pmax constraints. ...........................................................103
4-6. MDP vs. Pmax for various bandwidth constraints.............................................................104
4-7. MDP vs. Bandwidth of various Pmax constraints for a constant current source device. ..104
4-8. MDP vs. Bandwidth of various Pmax constraints for a constant current source device. ..105
4-9. Sensitivity analysis for constant current source varying 4mA by 3%. ............................105
4-10. Sensitivity analysis for constant current source varying 10mA by 10%. ........................106
12
4-11. MDP dependence on Hooge parameter for device A. .....................................................106
4-12. Dependence of MDP with respect to each variable. ........................................................107
4-13. Dependence of MDP on silicon thickness overlaid with compression coefficient..........107
4-14. Monte Carlo simulation schematic ..................................................................................108
4-15. Uncertainty of MDP of Device A. ...................................................................................108
4-16. Uncertainty of Pmax for device A. ....................................................................................109
4-17. Uncertainty of the bandwidth for device A......................................................................110
4-18. 95% probability yield limit illustration............................................................................110
5-1. Front side process steps. ..................................................................................................115
5-2. First four masks for microphone fabrication ...................................................................116
5-3. Last three masks for front side fabrication ......................................................................117
5-4. Back side process steps....................................................................................................118
5-5. Backside masks for microphone fabrication....................................................................118
5-6. Array of microphone die after dicing. Each die is 2mm x 2mm. ....................................119
5-7. Individual type A microphone die after dicing. ...............................................................119
5-8. Type B device pictured on a dime. ..................................................................................120
5-9. Backside cavity and vent of an individual type A microphone die after dicing. .............120
5-10. Packaging for acoustical characterization........................................................................121
5-11. Interface circuitry showing power supply, ac filter and amplifier...................................121
5-12. Printed circuit board for mounting the microphone and its associated components. ......122
5-13. Assembled device on PCB. Device is protected under a TO can. ..................................122
5-14. Populated PCB board inserted into PWT endplate. .........................................................123
6-1. Circuit representation of reversed biased p+ doped resistors in an n substrate. ..............138
6-2. Van der Pawl test structure schematic. ............................................................................138
6-3. Line width test structure schematic..................................................................................139
13
6-4. Kelvin test structure schematic. .......................................................................................139
6-5. Experimental setup for noise measurements....................................................................140
6-6. Experimental setup for acoustic characterization. ...........................................................140
6-7. Boron concentration in silicon device layer determined by SIMS, the accompanying curve fit and the desired model profile. ...........................................................................141
6-8. Input I-V curve of 12 BUF1-A and 12 BUF1-B devices.................................................141
6-9. Output I-V curve of 12 BUF1-A and 12 BUF1-B devices. .............................................142
6-10. Input I-V curve of a BUF1-A device with a linear curve fit............................................142
6-11. Output I-V curve of a BUF1-A device with a linear curve fit .........................................143
6-12. I-V curve of diode characteristics of a BUF1-A device. .................................................143
6-13. I-V curve of diode characteristics of a BUF1-A device focusing on the reverse region. ..............................................................................................................................144
6-14. Noise PSD of a test taper resistor. ...................................................................................144
6-15. Noise PSD from a test taper resistor minus the setup noise and the associated model curve fit. The horizontal line is the thermal noise floor for this device..........................145
6-16. Noise power spectral density of a BUF1-A device..........................................................145
6-17. Noise PSD minus the setup noise and the associated model curve fit. The horizontal line is the thermal noise floor for the device....................................................................146
6-18. Sensitivity of BUF1-A devices normalized by the bias voltage. .....................................146
6-19. Total harmonic distortion of BUF1-A microphones........................................................147
6-20. Magnitude frequency response for a BUF1-A device. Vertical dotted lines mark the piecewise FRFs that were stitched together.....................................................................147
6-21. Magnitude FRF for each device with 95% CI bounds.....................................................148
6-22. Phase response for each device tested. ............................................................................149
6-23. Phase FRF for each device with 95% CI bounds.............................................................150
6-24. Coherence function between device A-5 and the reference microphone. .......................151
6-25. Microphone photograph with and without backlighting..................................................151
14
6-26. Minimum detectable pressure probability density function.............................................152
6-27. Sensitivity probability density function. ..........................................................................152
6-28. Voltage noise probability density function. .....................................................................153
E-1. Layout of PCB package. ..................................................................................................215
15
Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
DESIGN, FABRICATION, AND CHARACTERIZATION OF A MEMS PIEZORESISTIVE MICROPHONE FOR USE IN AEROACOUSTIC MEASUREMENTS
By
Brian Homeijer
December 2008
Chair: Mark Sheplak Major: Mechanical Engineering
With air traffic expected to increase dramatically in the next decade and urban sprawl
encroaching on airports, a reduction in the sound radiated from commercial airplanes is needed.
To lower aircraft noise, manufacturers perform extensive scale model wind tunnel tests to locate
and eliminate sound sources. One of the most important pieces of equipment needed is a robust
microphone that is able to withstand large sound pressure levels on the order of 160 dB SPL,
while possessing an operating bandwidth on the order of 100 kHz and a low noise floor at or
below 26 dB SPL. This work attempts to address the needs of aircraft manufacturers with the
design, fabrication and characterization of a microelectromechanical systems piezoresistive
microphone for use in aeroacoustic measurements.
This microphone design addresses many of the problems associated with previous
piezoresistive microphones such as limited dynamic range and bandwidth. This design focuses
on improving the minimum detectable pressure over many current technologies without
sacrificing bandwidth. To accomplish this, a novel nonlinear circular composite plate mechanics
model was employed to determine the stresses in the diaphragm, which was designed to be in the
compressive quasi-buckled state. With this model, the effects of residual in-plane stresses that
result from the microelectronic fabrication process on the sensitivity of the device are predicted.
16
Ion implanted doped silicon was chosen for the piezoresistors and an integrated circuit
compatible fabrication recipe was formulated to minimize the inherent noise characteristics of
the material. The piezoresistors are arranged in a Wheatstone bridge configuration with two
resistors oriented for tangential current flow and two for radial current flow. A lumped element
model was created to predict the dynamic characteristics of the microphone diaphragm and the
integrated cavity/vent structure. The device geometry was optimized using a sequential
quadratic programming scheme. Results predict a dynamic range in excess of 120 dB for
devices possessing resonant frequencies beyond 120 kHz. Future work includes the completion
of the fabrication process and characterization of the microphones.
The characterization of the fabricated device revealed two major problems with the
piezoresistors. The diffusion of the resistors was too long and resulted with the resistor thickness
being the entire thickness of the diaphragm. The result of this error dropped the sensitivity two
orders of magnitude. In addition to the doping profile error, the inherent noise characteristic of
the resistors was also higher then expected. This increased the noise signature of the device two
orders of magnitude higher then expected. These two factors couple together and increase the
MDP of the device by 4 orders of magnitude, or 80 dB. The optimized device A had an expected
MDP of 24.5 dB . The realized device had a MDP of 108dB, or 83.5 dB higher than the desired
value. Despite the error in resistor fabrication, the models developed in this dissertation showed
that they correctly represent the realized device and therefore will be sufficient to design a
second generation microphone.
17
CHAPTER 1 INTRODUCTION
With air traffic expected to increase dramatically in the next decade and urban sprawl
encroaching on airports, the Federal Aviation Administration (FAA) has taken steps to regulate
aircraft noise. For certification, the US Code of Federal Regulations stipulates that commercial
aircraft must pass airworthiness tests, which state the maximum allowable effective perceived
noise level (EPNL) that aircraft can emit. The EPNL is the measured noise level normalized to
sound duration, atmospheric conditions and jet engine operating conditions[1]. To lower the
noise radiating from aircraft, manufacturers perform extensive wind tunnel tests to locate and
eliminate sound sources on planes. This industry is in need of a robust, and low cost alternative
to instrumentation grade condenser microphones. A microelectromechanical systems (MEMS)
microphone has the potential for a substantial cost reduction and does not have the installation
drawbacks that the current microphones possess. This work focuses on the development of a
robust microelectromechanical systems (MEMS) microphone as a low cost alternative to the
industry standard condenser microphones.
This chapter begins with an introduction to noise restrictions for commercial aircraft flying
in US airspace. Next, the differences between audio and aeroacoustic microphones are explained
and requirements for aeroacoustic microphones are given.
1.1 Noise Restrictions for Commercial Airplanes
In response to the anticipated doubling of world air traffic over the next 20 years [1],
noise restrictions continue to grow more stringent. The total number of noise restrictions,
comprised of curfews, charges and levels has increased 10 fold, (Figure 1-1) [1]. Curfews
designate quiet time around airports during the night. If aircraft are louder then the curfew limit
18
then they can only land in the daytime. Currently, aircraft that do not meet the noise
requirements are fined, with increased noise levels corresponding to a larger fine.
To meet the increasingly restrictive noise requirements, aircraft manufactures have reduced
noise signatures an order of magnitude from the original jet aircraft of the 1970’s (Figure 1-2).
To determine an aircraft’s noise signature, the FAA takes noise samples at every airport in at
least three locations each of which are illustrated in Figure 1-3. The largest improvement to date
has been in the reduction of noise in the lateral and take-off areas [2]. This improvement is
primarily the result of noise reduction in turbofans, which are loudest during takeoff. The
takeoff noise signature is also less problematic because planes gain altitude quickly, and noise
ceases to reach ground level. However, landing requires a low, slow approach. At this stage,
airframe noise is significant because the plane is in a noisy configuration with the landing gear
down and flaps fully extended (Figure 1-4).
Aircraft manufacturers are focusing on reducing the noise of commercial aircraft even
further. Next generation aircraft engines are being equipped with serrated edges called chevrons
for the back of the engine nacelle and exhaust nozzle [2]. In addition researchers are looking
into toboggan fairings to reduce landing gear noise [3]. To accomplish their goals, aircraft
manufactures need a robust aeroacoustic microphone that meets the requirements for
aeroacoustic testing.
1.2 Aeroacoustic Microphones
Transducers convert energy from one form to another; microphones are transducers that
specifically convert acoustical energy to electrical energy or modulate the electrical energy due
to the acoustic energy. This energy conversion is achieved in different ways, however, one thing
they all have in common is that they first convert acoustical energy to mechanical energy via a
diaphragm. A diaphragm is a thin structure, pictured in Figure 1-5, that vibrates when sound
19
waves strike it. A mechanical to electrical transduction scheme then determines how the
mechanical energy is converted into a readable electrical signal.
In addition to the diaphragm and the transduction mechanism, the other important elements
of a microphone are the vent and cavity. The vent is used to equilibrate the pressure acting on
the device so that the microphone only senses an ac signal, and the cavity connects the vent to
the diaphragm (Figure 1-5). To ensure that microphone output is representative of the acoustic
input, the sensitivity of the microphone must not change with frequency, and, ideally, the
microphone should have no phase shift. Figure 1-6 illustrates a normal frequency response for an
under-damped aeroacoustic microphone. The bandwidth of the microphone is defined as the
range of frequencies where the microphone’s response magnitude is flat. The cut-on frequency
and resonant frequency are also shown in Figure 1-6. The vent channel dominates the low-
frequency response of the microphone, while the impedance of the vent relative to the diaphragm
dictates whether incident pressure will flow through the vent or deflect the diaphragm. The
damping and the resonant frequency of the microphone dominate the frequency response at high
frequencies. The mechanical resonance of the diaphragm is a function of its compliance and
mass. The shape of the frequency response close to the resonance frequency is determined by
damping in the microphone structure. For example, an distinct resonance peak will be evident for
an under-damped system, as shown in Figure 1-6, while no peak is found in an over-damped
system. The frequency response of a microphone can be optimized to allow it to have a larger
bandwidth in various acoustic fields. This can be accomplished several ways and is discussed in
the following section.
1.2.1 Pressure, Free and Diffuse Field Microphones
Microphones are divided into three types: free, diffuse and pressure field, determined by
their response in an acoustic field. A free field occurs when sound waves can propagate freely
20
without reflections. This type of field can be found outdoors or in an anechoic chamber as long
as the sound source is far enough away Figure 1-9(A) [4]. For this type of application, a free
field microphone should be used. When a microphone is placed into the sound field, it modifies
that field Figure 1-9(B). Pressure rises in front of the microphone due to the scattering off of the
microphone, resulting in a higher output level at certain frequencies. The maximum effect is
when the wavelength of a specific frequency is equal to the diameter of the microphone. Figure
1-9(C). Free field microphones take this effect into account by compensating for their own
disturbing presence. The damping of these microphones is increased so that the spectral shape of
the microphone response is opposite to the spectral shape due to scattering, shown in Figure 1-7.
To work correctly a free field microphone must be normally incident to the noise source.
A diffuse field exists if the field is created by sound waves arriving from all locations
simultaneously with equal probability [4]. The diffuse field microphone is designed to respond
uniformly to signals arriving simultaneously from all angles. These devices are also over-
damped like free field microphones.
A pressure field is found in enclosures which are small compared to the wavelengths of
interest [4]. A pressure field microphone should be used in this situation and should be flush
mounted on the enclosure as seen in Figure 1-8. Because the pressure field microphone has a
minimal effect on the field, no corrections are needed to account for the presence of the
microphone [4].
All three microphones can be used in any field as long as a correction factor is taken into
account. Since the devices were optimized for particular sound fields, the usable bandwidth will
decrease. The correction factors for Bruel and Kjaer microphones can be obtained on their
21
website. In addition, a nose cone corrector can be mounted onto a free field microphone to
reduce the effect of the angle of incidence.
1.2.2 Linearity and Total Harmonic Distortion
The linearity of the microphone explains how, at a fixed, flat band frequency, the
microphone output magnitude varies as a function of the amplitude of the incident pressure.
Shown in Figure 1-10 is a linear (ideal) and non-linear (real) response of a generic microphone to
a single-tone, fluctuating amplitude pressure. In the ideal case, a linear relationship between the
output voltage and the amplitude of the incident pressure is shown. In practice, however, a
variety of sources of non-linearity limit the effective maximum pressure. As seen in Figure 1-10,
the microphones actual dynamic response diverges from the linear response above a maximum
pressure, such as mechanical, electro-mechanical, and amplifier non-linearities. The maximum
pressure at which a microphone is considered linear is defined at the point where a 3% difference
between the linear and nonlinear response of the microphone is detected. The non-linearity of the
microphone is given in terms of the total harmonic distortion (THD) with respect to frequency
because the non-linear response of the microphone causes distortions in the output. The THD is
described as the ratio of the total power in the higher harmonics to the power in the fundamental
frequency, and is given as [5],
( )
( )
2
22
1
nn
pTHD
p
ω
ω
∞
==∑
. (1-1)
1.2.3 Noise Floor
The lower end of the dynamic range, called the minimum detectable signal (MDS) is
limited by the microphone noise as well as noise contributed by the interface circuitry, since it is
the output when no input is given [6]. Noise is normally given in terms of a power spectral
22
density (PSD) and the total noise power is based on the PSD integrated over the bandwidth of
interest [6]. A typical noise PSD of a piezoresistive microphone is shown in Figure 1-11. When
in thermodynamic equilibrium, the thermal noise of the system is proportional to the dissipation
of the system [6], This is also known as white noise because the PSD is constant over all
frequencies. Under non-equilibrium conditions, flicker noise arises which has an inverse
proportionality to frequency. It occurs in interface electronics and semiconductive materials, and
usually dominates at lower frequencies. The frequency when the noise PSD of the flicker noise
equals the thermal noise is known as the corner frequency [7]. The role of the individual noise
sources can be shaped by the dynamic response of the microphone. This can cause a flat thermal
noise source to have a non-flat spectral shape in the microphone output. The noise floor of a
microphone is often stated for a specific bandwidth. As an example, the noise can be specified at
a given frequency for a narrow bandwidth, or integrated over a specified bandwidth. A-weighted
noise, denoted dBA, is another common metric where the noise spectrum is passed through a
filter that approximates the response of the human ear, then integrated and converted to dB [8].
1.2.4 Requirements for an Aeroacoustic Microphone
As stated above, the performance of audio microphones is tuned to the human ear; an
instrument grade aeroacoustic microphone, however, has different requirements, shown in Table
1-1. The human ear has a minimum detectable pressure of 20 Paμ at 2kHz, also known as the
threshold of hearing. The maximum pressure that an ear can be exposed to, known as the
threshold of pain, is 20Pa . The bandwidth of the human ear is from 20Hz to 20kHz [9].
Aeroacoustic microphones must be useable in areas where the sound pressure level (SPL)
to be measured is very high, like near an aircraft jet engine, where sound pressure levels may
reach 170dB . In addition, the FAA requires certification over the frequency range of
23
45 11.2Hz f kHz≤ ≤ for full scale vehicles [1]. Aeroacoustic testing is frequently done using
1 8 scale models. To retain dynamic similarity, the frequency range of interest is enlarged by
this scale factor, meaning that acoustic testing is conducted over the range of
360 89.6Hz f kHz≤ ≤ [10]. However, the microphones should still be useable in full scale
flyover applications, which require the microphone bandwidth to extend down to 45Hz . In
addition to this requirement, the bandwidth goals for this project extend to a range of
20 120Hz f kHz≤ ≤ , specified by the sponsor of the project, The Boeing Company.
Currently, several commercial microphones are used by the aeroacoustic community. The
specifications of some of theses microphones can be seen in Table 1-2. The B&K microphones
are typically used in arrays for sound localization. The Kulite MIC-093 is used on turbulence
control screen arrays for static engine tests. The goal for the minimum detectable pressure
(MDP) of the microphone specified by the sponsor of this project is 26 dB SPL for a 1 Hz bin
centered at 1 kHz .
Acoustic arrays, which consist of many microphones arranged in a specific geometry, are
often used in aeroacoustic measurements to localize the noise source. A selective spatial
response can be determined via beam forming signal processing, which enables the acoustic
array to listen to a specific area in space [10]. This technique requires a large number of
microphones, typically in the 100’s [11], which makes MEMS microphones particularly
appealing due to the possible advantages of batch fabrication for reducing the cost of each
microphone. However, additional specifications are needed to enable the use of the microphones
in an acoustic array. For regularly spaced arrays, the microphones must be physically arranged
within one half wavelength of each other to avoid spatial aliasing. However, this is not a
requirement for logarithmically spaced spiral arrays [11]. Phase matching between microphones
24
is another necessity for acoustic arrays because beam forming algorithms use phase information
to localize sound sources. A mismatch between microphone channels can cause error in the
sound localization.
1.3 Objectives
The goal of this project is the design, fabrication, and characterization of a MEMS
piezoresistive microphone for aeroacoustic testing applications. The microphone should have a
sufficient dynamic range and bandwidth for use in scale model aeroacoustic wind tunnel tests
and full scale flyovers. The final objective is to have an accurate model to aid in the fabrication
of a second generation microphone, therefore the microphone’s characterization results will be
used to adjust the model to account for any errors. The entire project is larger then the scope of
this dissertation. The division of labor was separated into the design, fabrication and
characterization of the microphone and in addition the development of a fabrication recipe for
the piezoresistors to minimize the noise of the device is the remaining portion of the project.
1.4 Dissertation Outline
This work is organized into seven chapters. Chapter 1 presents the motivation and goals of
the project. Chapter 2 contains a literature review of MEMS microphones. Chapter 3 describes
the modeling of the microphone. Chapter 4 summarizes the optimization scheme used to decide
device dimensions. Chapter 5 outlines the process flow and device fabrication, Chapter 6 details
the characterization and model validation and Chapter 7 concludes with a summary of this work
and recommendations for future devices.
25
Table 1-1. Audio and aeroacoustic microphone specifications. Audio Aeroacoustic Max Pressure 120dB 170dB Noise Floor 23-35dBA 26 dBα Bandwidth 20 Hz - 20 kHz 20 Hz - 100 kHzα 1Hz bin centered at 1kHz
Table 1-2. Commercial microphones used in aeroacoustic testing specifications [12], [13].
Kulite MIC-093 B&K 4939 B&K 4138 Desired Boeing Specifications
Diameter 2.4 mm 6.35 mm 3.18mm Max Pressure 194 dBA 164 dB SPL 168 dB SPL 150- 160 dB SPL Noise Floor 100 dBA 5 dB SPLα 18 dB SPLα <26 dB SPLα Dynamic Range 94 dBA 159 dB SPL 150 dB SPL 124 – 134 dB SPLBandwidth ~90 kHz 4 Hz - 100 kHz 6.5 Hz - 140 kHz 100 – 120 kHz α 1Hz bin centered at 1kHz
01965 1970 1975 1980 1985 1990 1995 2000
Year
0
50
100
150
200
250
300
350
400
Figure 1-1. Number of noise restrictions at airports [1].
26
B-52
707-100
DC8-20
CV990ACV880-22
BAC-11
DC9-10DC8-61 737-100
737-200
727-100727-200
747-100
747-200DC10-10
L-1011
A300B2MD-80 747-300
747-400
737-300
A320-100A321
A330A340MD-11
777
A310-300
BAe 146-200
DC10-30
Comet 4
720
Year of initial service
Noiselevel, EPNdB(1,500 ftsideline)
1950 1960 1970 1980 1990 2000 2010 202080
90
100
110
120
Turbojet and early turbofans
First generation turbofan
Second generation turbofan
707-300B
Figure 1-2. Perceived noise levels of various aircraft [2].
Approach
Lateral
Take-off
Cutback~1,000 ft(305 m) 1,476 ft
(450 m)
6,565 ft
(2,000 m)
21,325 ft
(6,500 m)
Approach in noisiest configurationLanding gear extended, full flaps
Takeoff with maximum takeoff thrust rating
Measurement Location
Figure 1-3. FAA part FAR 36 Measurement Locations [1].
27
Figure 1-4. Noise sources of a typical commercial airplane.
Diaphragm
Cavity
Vent
Figure 1-5. Schematic cross-section of a general microphone structure.
28
10-1 100 101 102 103 104 105 106
-20
0
20
40
Mag
nitu
de [d
B]
(ref r
espo
nse
@ 1
kHz)
10-1 100 101 102 103 104 105 106
-150
-100
-50
0
50
Frequency [Hz]
Pha
se [d
eg]
Figure 1-6. Magnitude and phase of a typical aeroacoustic microphone frequency response.
Mag
. Fre
quen
cy
Res
pons
e
Figure 1-7. Example of over damping a free field microphone to increase bandwidth.
29
Figure 1-8. Pressure field microphone flush mounted in an enclosure.
A B
C
20log M
FF
PP
Dλ1
0dB
10dB
Figure 1-9. A) Traveling acoustic waves in a free-field. B) Effect of placing a microphone in the field. C) Increased pressure sensed by the microphone due to its own presence
[4].
30
Linear Solution
Nonlinear Solution
Incident Pressure
Cen
ter D
efle
ctio
n
Figure 1-10. Deviation of linear and nonlinear solutions.
10-1 100 101 102 103 104 105
10-16
10-15
10-14
10-13
10-12
Frequency [Hz]
Noi
se P
ower
Spe
ctra
l Den
sity
Sv [V
2 /Hz]
Figure 1-11. Noise power spectral density for a typical microphone.
31
CHAPTER 2 BACKGROUND
This chapter discusses the details of the transduction schemes commonly used in MEMS
microphones including piezoelectric, piezoresistive, capacitive and optical in section 2.1. A
comprehensive literature review of MEMS microphone developments is then given in section
2.2.
2.1 Transduction Schemes
This work organizes the various MEMS microphones by transduction mechanism:
piezoelectric, piezoresistive, optical, and capacitive. These transduction mechanisms are
described in the following sections. The optical scheme is broken down into intensity
modulating, polarization modulating and phase modulating. The capacitive scheme is separated
into electret and condenser. Figure 2-1 shows the various classifications of MEMS microphones.
2.1.1 Piezoelectric Transduction
Piezoelectricity is defined as the ability of some materials to generate an electric potential
in response to applied mechanical stress [14]. The Curie brothers were the first to discover that
surface charges developed on some crystals, namely crystals with a noncentrosymmetric crystal
structure, when compressed, and that the magnitude of these charges was proportional to the
applied pressure. Hankel later named this phenomenon “piezoelectricity,” and it is historically
referred to as the direct piezoelectric effect.
In addition, strain is also produced when an electrical field is applied, called the converse
piezoelectric effect [15]. This effect is caused by an internal polarization at the atomic level
[16]. The piezoelectric charge modulus, d , is the material constant relating strain and charge in
a piezoelectric material. Figure 2-2 illustrates how a piezoelectric material expands, or contracts,
when an electrical potential is placed across it.
32
2.1.2 Piezoresistive Transduction
Certain materials go through a fundamental electronic change in resistivity due to applied
stress that exceeds the resistance change that all resistors experience due to stress-induced
deformations, called the piezoresistive effect. In semiconductor piezoresistive materials, like
silicon and germanium, the change in resistivity is caused by a change in the mobility in addition
to a change in physical size. This effect was first observed by Smith in 1954 for silicon and
germanium [17]. The resistance of any given material is [18]
LRA
ρ= . (2-1)
where , , and L Aρ are the resistivity, length and cross sectional area of the resistor, respectively
(Figure 2-3). When a metal is stressed, the resistance changes due to geometric effects, however,
when a semiconductor material such as silicon is stressed, the resistivity of the material changes
as well [17]. The change in resistance is given by
dR dL dA dR L A
ρρ
= − + . (2-2)
The geometric effects can be arranged in the following manner
( )1 2dR dR
ρν ερ
= + + . (2-3)
For conductive materials such as metals where 0d ρ ρ ≈ , the maximum change in resistance is
limited to 2 times the strain if 0.5ν = . In a piezoresistive material the change in resistance is
given by
ijijkl kl
ij
Tρ
ρΔ
= ∏ , (2-4)
33
where ∏ is a rank 4 piezoresistance tensor and T is the stress tensor [19]. The magnitude of
this term can be of the order of 100 times the strain [19]. This allows for a much greater
sensitivity in piezoresistive materials compared to a standard metal strain gauge.
2.1.3 Capacitive Transduction
The capacitive transduction scheme relies on the measurement of a change in capacitance
between two electrically charged surfaces. A parallel plate capacitor is discussed here for
simplicity. The capacitance between two parallel plates is given as,
0 ACg
ε= , (2-5)
where A is the surface area, 0ε is the permittivity of the dielectric material between the plates,
and g is the distance between the plates [20]. If a force moves one of the plates, the capacitance
changes as the distance between the two plates changes. The two main classes of capacitive
sensing are electret and condenser sensing [21]. Electrets are biased with a fixed permanent
charge, which is usually implanted into a dielectric layer on the fixed portion of the microphone.
The electrical force eF in Figure 2-4 for this case is [22]
( )2
2eQF Q
Aε= , (2-6)
where Q is the charge. Condenser microphones are biased with an external voltage source. The
electrical force for this case is [22]
( ) ( )2
2,2e
AVF V xg x
ε= . (2-7)
Electret devices are not susceptible to electrostatic pull-in, however it is difficult to fabricate
devices with a stable embedded charge.
34
2.1.4 Optical Transduction
Optical microphones tend to be more complex because the mechanical energy of the
diaphragm is sensed optically first and then converted into electrical energy. One benefit of
sensing optically is that the transducer is insensitive to electromagnetic interference [23]. One
drawback is that optoelectronics, typically a photodiode, are needed to convert the optical signal
to an electrical signal. This can cause a significant amount of noise [24].
All optical microphones are light-modulating acoustic sensors rather than direct converters
of sound energy to light. They modulate light in three ways: intensity, phase and polarization
modulation [25]. Cook et al. developed the first optical transducer found in the literature in
1979, a lever displacement sensor [26]. Intensity based optical microphones use an emitter,
which shines light onto the diaphragm. The amount of light that is sensed by the receiver
changes as the diaphragm deflects. For microphones operating in the phase modulating scheme,
as the diaphragm deflects, the distance the light travels from emitter to receiver changes and is
sensed by a shift in phase. Optical microphones that operate using polarization modulation use
the fact that unpolarized light can be polarized through reflection off nonmetallic surfaces. The
degree of polarization depends on the incident angle and the material that is reflecting the light
[25]. The light is then passed through a polarizing filter and the intensity of the light is
measured.
2.2 Chosen Transduction Scheme
Of the available transduction schemes the piezoresistive scheme was chosen for this
project. The piezoresistive transduction scheme has the best attributes including:
• The ability to withstand harsh environments • The capability of being packaged with a thin profile • Compatibility with integrated circuit fabrication processes
35
Table 2-1 shows the desired traits and which transduction schemes meet each requirement.
2.3 Literature Review
This literature review discusses the important steps that progressed each microphone
transduction scheme. Select non-MEMS and pressure sensors of historical significance are
added to show how the technology developed.
2.3.1 Piezoelectric Microphones
In 1953 Medill developed the first “miniature” piezoelectric microphone for use as a
secondary standard in production testing. This microphone measured 1 1/8” in diameter and
used Rochelle salt crystals as the piezoelectric material [27]. In 1983 Royer et al. developed the
first MEMS piezoelectric microphone using zinc oxide as the piezoelectric material [28].
Between 1989 and 1991 Kim et al. developed a piezoelectric MEMS microphone using a silicon
nitride diaphragm and a zinc oxide piezoelectric material [29], [30]. Theses devices, however,
never achieved a flat frequency response. In 1992, Schellin et al. developed a new type of
microphone [31]. This work used polyurea for the piezoelectric material and achieved a high
sensitivity of 4mV/Pa, but Schellin et al., like Kim et al., could not achieve a flat frequency
response. One year later, in 1993, Ried et al. improved on Kim’s microphone by tweaking the
fabrication process to control stress in the diaphragm, achieving a flat frequency response [32].
In 2003, Ko et al. developed a device with zinc oxide as the piezoelectric material that could be
used as a microphone or a microspeaker [33]. This device performed poorly in both
applications. Also in 2003, Niu et al. improved on Kim’s design (1989) by using parylene-D for
the diaphragm, which increased the sensitivity [34]. In 2003 Zhao et al. utilized the piezoelectric
material, lead zirconate titanate (PZT), for the first time in a MEMS microphone. With this
material, Zhao et al. achieved a high sensitivity of 38 mV/Pa and a flat frequency response [35].
In 2004, Hillenbrand et al. used a cellular polypropylene material for the piezoelectric crystal,
36
but the device did not reach the sensitivity of Zhao et al. [36]. In 2007, Horowitz et al.
developed an aeroacoustic microphone using PZT [37]. This device was the first to have a large
dynamic range suitable for aeroacoustic applications. Table 2-2 summarizes the realized
performance for each of the devices.
2.3.2 Piezoresistive Transducers
The first piezoresistive microphone was developed by Burns in 1957 for use in the Bell
type 500 telephone [38]. This microphone was deemed too expensive to fabricate compared to
the standard carbon microphones and was never mass produced. Fourteen years later in 1961,
Samaun et al. [39] developed the first MEMS piezoresistive pressure sensor. A pressure sensor
is similar to a microphone except it is used to measure an absolute dc pressure instead of a
relative ac pressure. This device had a silicon nitride moisture barrier and was intended for use
in biomedical applications. It was not until 1992 that Schellin et al. developed the first MEMS
piezoresistive microphone [40]. This device was composed of a square diaphragm and four p-
type dielectrically isolated polysilicon piezoresistors. In 1994, Kalvesen et al. developed a
MEMS microphone for use in turbulent gas flows [41]. This device was the first to have an
integrated cavity and vent structure; however, the cavity was only 3 mμ deep and contributed to a
low sensitivity. In 1995, Schellin et al. developed the first microphone to use ion implanted
piezoresistors [42]. Specifically, they were p-type resistors in an n-well silicon diaphragm. In
1998, Sheplak et al. developed a silicon nitride MEMS microphone for aeroacoustic
measurements [43]. This device had a circular diaphragm and was the thinnest diaphragm
(1500A ) reported in the literature. The piezoresistors were dielectrically isolated and had two
types: arc and taper. The arc resistor was designed for current flow in the tangential direction
and the taper resistor was designed for current flow in the radial direction. This microphone had
37
a large cavity and winding vent channel integrated into the design. At the same time, Nagiub et
al. developed another MEMS microphone for use in aeroacoustic measurements [44]. The
device used a rectangular diaphragm however the dynamic range was not reported in the
literature. In 2001, Arnold et al. improved on the design of Sheplak et al. (1998) and
significantly lowered the noise floor of the device [45]. One year later, Huang et al. improved on
the original design of Nagiub (1999) with a device that had a similar minimum detectable
pressure (MDP) [46]. In 2004 Li et al. developed an audio MEMS microphone with integrated
electronics and the lowest noise floor reported in the literature of 34dB [47]. The device’s
maximum pressure was not reported. Table 2-3 outlines the various devices and shows the
reported specifications for each.
2.3.3 Capacitive Transducers
Ko et al. developed the first MEMS capacitive transducer, a condenser pressure sensor in
1982 [48]. Two years later in 1984 Hohm et al. developed the first capacitive microphone [49].
It was an electret microphone and had a rectangular diaphragm with the longer side being 8 mm
long. In 1989 Hohm et al. developed the first condenser microphone [50]. The device was 10
times smaller than the previous effort and operated from 200Hz to 20 kHz. In 1990, Bergqvist et
al. developed the first microphone built completely using microfabrication techniques [51].
Their devices were all rectangular with a size of 2 mm. In 1997 Cunningham et al. developed
the first capacitive microphone with a circular diaphragm [52]. This device was designed for
audio applications and had a 1mm diameter. In 2000, Rombach et al. developed the dual plate
capacitive microphone [53]. This microphone used two fixed plates with a diaphragm in the
middle to increase the capacitance change resulting with an increase in sensitivity. The noise
floor of the device was 23dBA and was designed for audio applications. Three years later in
2003, Scheeper et al. developed the first capacitive microphone tailored for use in aeroacoustic
38
measurements [54]. His device used a non-traditional octagonal diaphragm and had a dynamic
range of 23dB to 141dB. In 2005, Martin et al. developed the first dual backplate microphone
for aeroacoustics [55]. This device had a dynamic range of 22.5 dB to 160 dB and a bandwidth
of about 100 kHz. Table 2-4 shows the reported specifications for each of the discussed devices.
Currently the Knowles Acoustics Sisonic microphone is available for purchase. The
specification sheet reports a noise floor of 39 dBA and a bandwidth of 10 kHz. In September,
2008, Analog Devices Incorporated released the iMEMS condenser microphone. To date it has
the best performance of any commercially available MEMS microphone [56].
2.3.4 Optical Transducers
In 1991, Garthe et al. developed the first optical microphone [57]. This device used an
intensity modulating scheme and had an integrated waveguide chip fabricated using polymethyl
methacrylate (PMMA), though it was not a MEMS device. In 1992 Dziuban et al. were the first
to develop a silicon optical device [58]. This transducer had a 10 mm square diaphragm and
could only be used as a pressure on/off switch due to it’s poor performance. In 1994, Chan et al.
developed the first silicon optical pressure sensor [59]. This device had a 10mm square
diaphragm and used a phase modulation scheme. Five years later in 1999, Kots et al. developed
an intensity modulating optical microphone [60]. This was the smallest device to date with a
1.5mm circular diaphragm. In 2001, Abeysinghe et al. developed a circular pressure sensor [61].
It was a phase modulating device and had a 135 mμ diameter. In 2004, Kadirvel et al. developed
a MEMS optical microphone [62]. The microphone used an intensity modulating transduction
scheme. It was designed for aeroacoustic applications but was plagued by an inherently high
noise floor of 70dB. In 2005, Bucaro et al. developed a MEMS intensity modulating microphone
with an improved noise floor of 30.6dB [63], significantly lower then Kadirvel’s device;
39
however the maximum detectable pressure was not reported. In 2005, Hall et al. developed a
device with the lowest detectable pressure in the literature, 17.5 dBA, however this was
accomplished in the laboratory on an optical bench [64]. The device was fabricated using
Sandia’s SwIFT-Lite process but it only had a bandwidth of 4kHz. In 2006, Song et al. reported
on an optical microphone based on a reflective micro mirror diaphragm however, the dynamic
range was not reported [65]. Finally, in 2007 Hall et al. reported on a smaller microphone design
with a 24dBa noise floor [66]. Table 2-5 shows the reported specifications for each of the
discussed devices.
40
Table 2-1. Transduction schemes and desired characteristics. Characteristic Meet Fail
IC compatible fabrication Piezoresistive Optical Capacitive Piezoelectric
Table 2-2. Piezoelectric microphone specifications in the literature. Author Diaphragm Cavity Sensitivity Dynamic Bandwidth Piezoelectric Microphone/Year Dimensions Depth Range (Predicted) Material Pres. Trans.
J. Medill 1 1/8”α N/R N/R N/R ~1kHz Rochelle Microphone
[27] salt crystals
M. Royer et. al. 1.5mmα x 30μm N/R 250μV/Pa 73dBA-N/R 10 Hz - 10 kHz ZnO Microphone
[28] (0.1 Hz - 10 kHz) 1st fabricatedE. S. Kim et al. 2mmβ x 1.4μm 380μm 80μV/Pa N/R 3 kHz - 30 kHz ZnO Microphone
[29], [67] E. S. Kim et. al. 3.04mmβ x 2.0μm 380μm 1000μV/Pa 50 dBA-N/R 200 Hz - 16 kHz ZnO Microphone
[30] R. Schellin et al. 0.8mmβ x 1.0μm 280μm 4000 μV/Pa N/R 100 Hz - 20 kHz Polyurea Microphone
[40], [68] R. P. Ried et al. 2.5mmβ x 3.5μm ~500μm 920μV/Pa 57 dBA-N/R 100 Hz - 18 kHz ZnO Microphone
[32] S. S. Lee et al. 2mmγ x 4.5μm N/R 3800 μV/Pa N/R 100 Hz - 890 Hz ZnO Microphone
[69], [70] S. C. Ko et al. 3mmβ x 3.0μm N/R 30μV/Pa N/R 1 kHz - 7.3 kHz ZnO Microphone
[33] M. N. Niu et al. 3mmβ x 3.2μm N/R 520μV/Pa N/R 100 Hz - 3 kHz ZnO Microphone
[34] H. J. Zhao et al. 0.6-1mmβ x N/R 370μm 38mV/Pa N/R N/R - 20kHz PZT Microphone
[35] J. Hillenbrand et al. ~0.5cm x 55μm N/R 2-0.5mV/Pa 37-26 dBA - N/R - ~10kHz Cellular Microphone
[36] N/R polypropylene Y. Yang et al. 200-500μmβ x N/R 61-474 N/R N/R - ~30kHz PZT Microphone
[71] ~1μm μV/Pa S. Horowitz et al. 900μmα x 3.0μm 500μm 0.75μV/Pa 47.8 dBδ - 100 Hz - 6.7 kHz PZT Microphone
[37] 169 dB (100 Hz - 50 kHz) α Radius of circular diaphragm β Length of rectangular diaphragm γ Length of cantilever δ 1 Hz bin
41
Table 2-3. Piezoresistive microphone specifications in the literature. Author Diaphragm Cavity Sensitivity Dynamic Bandwidth Microphone/ Year Dimensions Depth (Predicted) Range (Predicted) Pres. Trans.
F. P. Burns type 500 tele N/R 2.0uV/Pa*mA N/R 0 - 2570 Hz Microphone [38] 1.8 cmα (0 - 2590 Hz) 1st appl. of p.r.
S. Samaun et al. 1.2mmα x 5μm N/R 9.5μV/Pa N/R N/R Pres. Trans. [39] 1st SiN H2O bar.
W. H. Ko et al. 2.29mmβ x 20μm 254μm 75nV/Pa*V N/R - 40kPa 0 - 10 kHz Pres. Trans. [72]
R. Schellin et al. 1mmβ x 1μm N/R 4.2μV/Pa*V N/R 100 Hz - 5 kHz Microphone [31], [40]
E. Kalvesten et al. 100μmβ x 0.4μm 3μm 0.09μV/Pa*V 96dBA - N/R 10 Hz - 10 kHz Microphone [41], [73] (0.10μV/Pa*V) (2 mHz - 1 MHz)
E. Kalvesten et al. 300μmβ x 0.4μm 3μm 0.03μV/Pa*V 90dBA - N/R 10 Hz - 10 kHz Microphone [74] (0.02μV/Pa*V) (10 Hz - 0.9 MHz) Cav. Stiff.
R. Schellin et al. 1mmβ x 1.3μm N/R 10μV/Pa*V 61dBA - 128dBA 50 Hz - 20 kHz Microphone [42]
M. Sheplak et al. 105μmα x 0.15μm 10μm 2.24μV/Pa*V 92dBγ - 155dB 300 Hz - 6 kHz Microphone [43], [75] (100 Hz - 300 kHz)
A. Naguib et al. 510μmβ x 0.4μm N/R .18μV/Pa*V- N/R 1 kHz - 5.5 kHz Microphone [44] 1.0μV/Pa*V
A Naguib et al. 710μmβ x 0.4μm N/R 1.0μV/Pa*V N/R 1 kHz - 5.5 kHz Microphone [76]
D. P. Arnold et al. 500μmα x 1.0μm 10μm 0.6μV/Pa*V 52dBγ - 160dB 1 kHz - 20 kHz Microphone [45] (10 Hz - 100 kHz)
C. Huang et al. 710μmβ x 0.38μm ~20μm 1.1μV/Pa*V 54dBγ – 174dB 100 Hz - 10 kHz Microphone [46]
G. Li et al. N/Rβ x 1.0μm ~400μm 10μV/Pa*V 34dBγ - N/R 100 Hz - 8 kHz Microphone [47]
α Radius of circular diaphragm β Length of rectangular diaphragm γ 1 Hz bin
42
Table 2-4. Capacitive microphone specifications in the literature. Author Diaphragm Air Capacitance Bias Sensitivity Dynamic Bandwidth Condenser/ Year Dimensions Gap Voltage Range (Predicted) Electret
W. H Ko et al. 572μmα x 25μm N/R N/R N/R 1.28μV/Pa N/R N/R Condenser [48]
D. Hohm et al. 8.0mmβ x 13μm 20μm 9 pF 350 V 3mV/Pa N/R 100 Hz - Electret [49] 7.5 kHz
A. J. Sprenkels et al. 3.0mmβ x 2.5μm 20μm N/R 300 V 25mV/Pa N/R 100 Hz - Electret [77], [78] 15 kHz
P. Murphy et al. N/R x 1.5μm 25 - 95μm N/R 200 V 4-8mV/Pa N/R 100 Hz - Electret [79] 15 kHz
D. Hohm et al. 0.8mmβ x .25μm 2μm 6 pF 28 V 0.2mV/Pa N/R 200 Hz - Condenser [50] 4.3mV/Pa 20 kHz
J. Bergqvist et al. 2mmβ x 5μm 4μm 3.5 pF N/R 13mV/Pa N/R 500 Hz - Condenser [51] 2 kHz
J. Bergqvist et al. 2mmβ x 6μm 4μm 3.5 pF N/R 6.1mV/Pa N/R 100 Hz - Condenser [51] 5 kHz
J. Bergqvist et al. 2mmβ x 8μm 4μm 3.5 pF N/R 1.4mV/Pa N/R 500 Hz - Condenser [51] 20 kHz
J. Bergqvist et al. 2mmβ x 5.1μm 2μm 5 pF 5 V 1.8mV/Pa 37 dBA - 2 Hz - Condenser [80] 120dB 20 kHz
P. R. Scheeper et al. 2mmβ x 1μm 1μm 20 pF 2 V 1.4mV/Pa N/R 40 Hz - Condenser [81] N/R
P. R. Scheeper et al. 2mmβ x 1μm 3.3μm 5-7 pF 16 V 2mV/Pa 35 dBA - 100 Hz - Condenser [82], [83] N/R 10 kHz
W. Kuhnel et al. 0.8mmβ x .25μm 2μm 1 pF 28 V 1.8mV/Pa N/R 100 Hz - Condenser [84], [85] 20 kHz
T. Bourouina et al. 500μmβ x 1μm 5μm N/R N/R 0.4mV/Pa N/R N/R - Condenser [86] 20 kHz
T. Bourouina et al. 707μmβ x 1μm 5μm N/R N/R 2mV/Pa N/R N/R - Condenser [86] 7 kHz
T. Bourouina et al. 1mmβ x 1μm 5μm N/R N/R 3.5mV/Pa N/R N/R - Condenser [86] 2.5 kHz
T. Bourouina et al. 1mmβ x 1μm 7.5μm N/R N/R 2.4mV/Pa N/R N/R - Condenser [86] 10 kHz
E. Graf et al. N/R 0.46μm N/R 15 V 38mV/Pa N/R N/R - Condenser [87] 10 kHz
J. Bergqvist et al. 1.8mmβ x 8μm 3μm 5.4 pF 28 V 1.4mV/Pa 43 dBA - 300 Hz - Condenser [88] N/R 13 kHz
J. J. Bernstein et al. 1.8mmβ x N/R N/R N/R 5-10 V 16mV/Pa 25 dBA - 300 Hz - Condenser [89] 114 dB 15 kHz
J. J. Bernstein et al. 1mmβ x N/R N/R N/R 5-10 V 16mV/Pa 25 dBA - 70 Hz - Condenser [89] 114 dB 15 kHz
Q. B Zou et al. 1mmβ x 1.2μm 2.6μm 3.6 pF 10 V 14.2mV/Pa 39 dBA - 100 Hz - Condenser [90], [91] N/R 9 kHz
43
Table 2-4. Continued. Author Diaphragm Air Capacitance Bias Sensitivity Dynamic Bandwidth Condenser/ Year Dimensions Gap Voltage Range (Predicted) Electret
Y. B. Ning et al. 2mmβ x 0.5μm 3μm 9.1 pF 6 V 3mV/Pa N/R 100 Hz - Condenser [92] 10 kHz
B. T. Cunningham et al. 1mmα x 0.5μm 2μm 5.1 pF 8 V 2.1mV/Pa N/R 200 Hz - Condenser [52] 10 kHz 1st Circular Mic
M. Pedersen et al. 1.6mmβ x 0.9μm 1.5μm 14.9 pF 15 V 5.1mV/Pa 35 dBA - 100 Hz - Condenser [93] N/R 15 kHz
M. Pedersen et al. 2.1mmβ x 0.9μm 1.5μm 18.5 pF 15 V 8.1mV/Pa 34 dBA - 100 Hz - Condenser [93] N/R 15 kHz
M. Pedersen et al. 2.2mmβ x 1.1μm 3.6μm 10.1 pF N/A 234Hz/Paγ 60 dBA - 100 Hz - Condenser [94] 120 dB 15 kHz
P. C. Hsu et al. 2.6mmβ x 2μm 4μm 16.2 pF 10 V 20mV/Pa N/R 100 Hz - Condenser [95] 10 kHz
M. Pedersen et al. 2.2mmβ x 1.1μm 3.6μm 10.1 pF 14 V 10mV/Pa 27 dBA - 100 Hz - Condenser [96] 120 dB 8 kHz
D. Schafer et al. 0.4mmα x 0.75μm 4μm 0.2 pF 12 V 14mV/Pa 27 dBA - 150 Hz - Condenser [97] N/R 10 kHz
A. Torkkeli et al. 1mmβ x 0.8μm 1.3μm 11 pF 2 V 4mV/Pa 33.5 dBA - 10 Hz - Condenser [98] N/R 12 kHz
P. Rombach et al. 2mmβ x 0.49μm 0.9μm N/R 1.5 V 13mV/Pa 23 dBA - 10 Hz - Condenser [53], [99] 118 dB 20 kHz
X. X. Li et al. 1mmβ x 1.2μm 2.6μm 1.64 pF 5 V 9.4mV/Pa N/R 100 Hz - Condenser [100] 19 kHz
R. Kressmann et al. 1mmβ x 600nm 2μm N/R N/R 2.9mV/Pa 39 dBA - N/R - Electret [101] (Corrugated) 123 dB 20 kHz
P. R. Scheeper et al. 1.95mmα x 0.5μm 20μm 3.5 pF 200 V 22mV/Pa 23 dBA - 251 Hz - Condenser [54] 141 dB 20 kHz
J. J.Neumann et al. 320μmβ x N/R N/R 1 pF N/A 1.4mV/Pa 46 dBA - 100 Hz - Condenser [102] N/R 6 kHz
S. T. Hansen et al. (70μm x 190μm) 1μm 3.56 pF N/A 7.3mV/Pa 64 dBA - 0.1 Hz - Condenser [103] x 0.4μm N/R 100 kHz
D. T. Martin et al. 0.23mmα x 2.0μm 2μm 0.74 pF 9 V 0.28mV/Pa 22.5 dBδ - 300 Hz - Condenser [55], [104] 160 dB 20 kHz
[56] 3.6V 105 dB 12kHz α Radius of circular diaphragm β Length of rectangular diaphragm γ Frequency Modulation δ 1 Hz bin Note: Bias voltage for electrets are the effective voltage
44
Table 2-5. Optical microphone specifications in the literature Author Diaphragm Cavity Sensitivity Dynamic Bandwidth Optical Year Dimensions Depth (Predicted) Range (Predicted) Modulation
J. A. Dziuban et al. 10mmβ x 0.5mm N/R 2.4/-0.8 μV/Pa (On/Off switch) N/R Intensity [58] Silicon
M. A. Chan et al. 10mmβ x 7μm 50μm 3.75mPa/fringe N/R-164dB N/R Phase [59] First MEMS Psens
A. Kots et al. 1.5mmα x 1.8μm N/R N/R N/R 0 - 15kHz Intensity [60] Non-MEMS
D. C. Abeysinghe et al. 135μmα x 7μm 0.64μm 0.016μV/Pa 0-551kPa N/R Phase [61] Psens
K. Kadirvel et al. 1mmα x 1μm 500μm 0.5mV/Pa 70dBδ-132dB 0 - 6.4kHz Intensity [62] ( 0 -20kHz)
J. A. Bucaro et al. 1.6mm x 1.5μm 200μm N/R 30.6dBδ− N/R 0.1Hz - 10kHz Intensity [63]
N. A. Hall et al. 2.1mmβx 0.80μm N/R N/R (17.5dBA) N/R - 4kHz Intensity [64]
J. H. Song et al. 800μmβ x 5μm N/R N/R N/R 0.1 - 2kHz Intensity [65]
N. A. Hall et al. 1.5mm x 2.25μm N/R N/R 24dBA N/R - 20kHz Diffraction Based [66]
α Radius of circular diaphragm β Length of rectangular diaphragm δ 1 Hz bin
Figure 2-1. Outline of the different transduction schemes of MEMS microphones.
45
A
0V =
V
B
C
V
Figure 2-2. A) A piezoelectric material in equilibrium. B) The inverse piezoelectric effect: material with an applied bias, expanding. C) The inverse piezoelectric effect:
piezoelectric material with a reverse bias, contracting [107].
46
A
B
Figure 2-3. A) An unstressed resistor with a resistance of R. B) Stressed resistor with a resistance of R + ΔR.
Figure 2-4. Variable capacitor schematic.
47
CHAPTER 3 TRANSDUCER MODELING AND DESIGN
This chapter details the design process for a piezoresistive microphone tailored to acoustic
measurements. Figure 3-1 shows the methodology of the design. A piezoresistive microphone
has three main components: diaphragm, cavity and vent. The diaphragm is a layered composite
composed of silicon, silicon dioxide and silicon nitride. The silicon dioxide layer is used to
passivate the resistors, and the silicon nitride layer is used to create a moisture barrier. These
materials were chosen because they are standard silicon processing materials. The fabrication of
these layers induces stresses in the diaphragm that must be taken into account. A sensor
mechanical model is developed to calculate pressure induced stress in the diaphragm as a
function of geometry and fabrication induced stresses. An electroacoustic model then
determines the change in resistance of the piezoresistors due to the stress determined by the
previous model. Both the electroacoustic transduction model and sensor mechanical model are
incorporated utilizing lumped element modeling (LEM) to determine the dynamics of the
multidomain system.
Design optimization, which is discussed in Chapter 4, incorporates the LEM results, design
specifications and manufacturing constraints to yield an ideal device. The mechanical model is
verified using FEA, and a cavity and vent structure is designed to accompany the diaphragm
determined during design optimization.
The mechanical model of the diaphragm and finite element analysis (FEA) verification is
described in section 3.1. The electroacoustic transduction model is derived in section 3.2. A
LEM is discussed in section 3.3. Finally, a cavity and vent structure is designed in section 3.4.
48
3.1 Composite Plate Mechanics
A brief review of previous research into plate mechanics includes the mechanical scaling
of capacitive and piezoresistive pressure transducers presented by Chau and Wise [108].
However they did not discuss the effects of in-plane forces on the stress field of the diaphragm.
Voorthuyzen and Bergveld [109] furthered the field by using a finite-difference model to
investigate the large-deflection characteristics of circular diaphragms in pressure sensors
subjected to in-plane loading over a limited dimensional domain. Sheplak and Dugundji
clarified the non-linear behavior of circular plates under tension by incorporating the structural
mechanics of thin-film diaphragms with large in-plane forces via a fundamental structural model
using von Kármán plate theory. This work shows that the deflection field is a strong function of
both in-plane loading and non-linear restoring forces [110]. The model presented in this work
extends the work of Sheplak and Dugundji [110] by incorporating a composite makeup and
compressive stresses in addition to tensile.
To determine the stresses in the diaphragm and calculate the resonant frequency, a
nonlinear composite plate model was developed. This nonlinear analysis of a circular composite
diaphragm under a static load is used to determine the behavior of the plate as a function of
geometry and fabrication induced stresses. This model describes an axisymmetric composite
diaphragm made up of transversely isotropic materials. The plate behavior was analyzed using
classical laminated theory to derive governing differential equations which were then solved
using an iterative finite difference scheme. A brief description of the derivation is given in the
next section and the details are included in Appendix A.
3.1.1 Derivation of Governing Equations
The analyzed plate is a composite structure composed of 3 layers. The base layer is
silicon, the middle layer is silicon dioxide and the top layer is silicon nitride. The piezoresistors
49
are implanted into the top of the silicon layer. To simplify the calculation, the reference plane
was put at the same layer as the resistors. This plane can be seen in Figure 3-2. The transverse
and radial deflections, ( ) ( )0 0 and w r u r , are defined at the reference plane.
The following three assumptions are stated by Kirchoff’s hypothesis and are used in the
following derivation of circular plate deflection under static loading (Figure 3-3) [111]:
• straight lines perpendicular to the neutral surface before deformation (i.e. transverse normals) remain straight after deformation,
• transverse normals do not experience elongation,
• transverse normals rotate such that they remain perpendicular to the neutral surface after deformation.
In addition, the material is assumed to be transversely isotropic and the circular diaphragm
deflection is assumed to be axisymmetric
3.1.2 Equilibrium Equations
The equilibrium equations are derived in Appendix A and are repeated here for reference
[111]:
0rr N NdNdr r
θ−+ = , (3-1)
rr r
dMrQ M r Mdr θ= + − , (3-2)
and
( )0 0r rdwd dr N rp rQ
dr dr dr⎛ ⎞ + + =⎜ ⎟⎝ ⎠
. (3-3)
where , , , , and r r rN N M M Qθ θ are the in-plane forces, moments and shear in the plate,
respectively. The subscripts r and θ refer to the radial and tangential directions.
50
3.1.3 Constitutive Relationship
Assuming that silicon is transversely isotropic, the constitutive relationships are defined
as
[ ] [ ] [ ]0
0r r rrQ Q Q zθ θ θθ
σ ε κεσ ε κε
⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎧ ⎫= + +⎨ ⎬ ⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭ ⎩ ⎭⎩ ⎭, (3-4)
where [ ]Q is defined as
[ ] 11 122
21 22
111
Q Q EQQ Q
ννν
⎡ ⎤ ⎡ ⎤= =⎢ ⎥ ⎢ ⎥− ⎣ ⎦⎣ ⎦
. (3-5)
The terms ε , 0ε and κ are the initial strain due to in-plane forces, elongation strain due to
transverse loading and curvature due to transverse loading, respectively. The forces per unit
length are found by integrating equation (3-4),
T
B
zr r
z
Ndz
Nθ θ
σσ
⎧ ⎫ ⎧ ⎫=⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭∫ . (3-6)
Substituting equation (3-4) into equation (3-6) yields
[ ] [ ] [ ]T T T
B B B
z z zor r rr
oz z z
NQ dz Q dz Q z dz
Nθ θ θθ
ε κεε κε
⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎧ ⎫= + +⎨ ⎬ ⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭ ⎩ ⎭⎩ ⎭∫ ∫ ∫ . (3-7)
It is convenient to define the extensional stiffness matrix,
[ ] [ ]T
B
z
z
A Q dz= ∫ (3-8)
and the flexural extensional matrix
[ ] [ ]T
B
z
z
B Q zdz= ∫ . (3-9)
Equation (3-7) is compactly written as,
51
[ ] [ ] [ ]o
r r rro
NA A B
Nθ θ θθ
ε κεε κε
⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎧ ⎫= + +⎨ ⎬ ⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭ ⎩ ⎭⎩ ⎭. (3-10)
The moments per unit length are determined by integrating the stress times its moment arm, z,
over the thickness:
T
B
zr r
z
Mzdz
Mθ θ
σσ
⎧ ⎫ ⎧ ⎫=⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭∫ . (3-11)
Substituting equation (3-4) into equation (3-11) yields
[ ] [ ] [ ] 2T T T
B B B
z z zor r rr
oz z z
MQ zdz Q zdz Q z dz
Mθ θ θθ
ε κεε κε
⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎧ ⎫= + +⎨ ⎬ ⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭ ⎩ ⎭⎩ ⎭∫ ∫ ∫ . (3-12)
It is now convenient to define the flexural stiffness matrix,
[ ] [ ] 2T
B
z
z
D Q z dz= ∫ . (3-13)
Equation (3-12) is compactly written as,
[ ] [ ] [ ]o
r r rro
MB B D
Mθ θ θθ
ε κεε κε
⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎧ ⎫= + +⎨ ⎬ ⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭ ⎩ ⎭⎩ ⎭. (3-14)
3.1.4 Nonlinear Solution
In the linear case it is possible to isolate the transverse deflection ( )0w and solve the
ordinary differential equation (ODE) explicitly. In the non-linear case, 0w cannot be isolated
from 0u and therefore two coupled nonlinear ODEs are derived, yielding,
2
*2 * * 22 2
1 12r
d d k P Sd d
ξξ ξ ξ ξ ξ
⎛ ⎞Θ Θ Λ+ − + Θ = − + Θ − Θ⎜ ⎟
⎝ ⎠ (3-15)
and
2 * * 2
2 22 23
2r rd S dS d d
d d d dξ ξ ξ
ξ ξ ξ ξ ξ⎛ ⎞Θ Θ Θ Χ
+ = −Λ + − − Θ⎜ ⎟⎝ ⎠
, (3-16)
52
where
*12
*
B hD
Λ = , (3-17)
2 2 2
11 12*
11
A A hA D−
Χ = , (3-18)
and
0 0
2 22 4* * * *0
* * * *
, , , ,
, , , and .2
rr
w ur dW aW Ua h h d h
N a N aN a paS S k PD D D hD
θθ
ξ φξ
= = = Θ = − =
= = = =
(3-19)
The , , and A B D matrixes are dependant on the composite makeup and are derived in
Appendix A. As the name suggests, the A matrix illustrates how the plate reacts to extensional
forces. The D matrix shows how the plate reacts to transverse forces and bending moments and
the B matrix shows the reaction of external forces to bending and transverse forces to stretching.
The *D value is a function of the composite matrices defined as
2
* 1111
11
BD DA
= − . (3-20)
Assuming a perfectly clamped plate, the boundary conditions for Θ and rS , , are
( )0 0ξΘ = = , (3-21)
( )1 0ξΘ = = , (3-22)
*
0
0rdSd ξξ
=
= , (3-23)
and
53
*
*121
11 11
1rr
dS A dSd A dξ
ξξξ ξ===
⎛ ⎞ Θ+ − = −Λ⎜ ⎟
⎝ ⎠. (3-24)
The symmetry coefficient ( )Λ is a measure of the symmetry of the composite plate. This
parameter takes into account that the reference plane is not necessarily the same as the neutral
plane. If the composite plate is symmetric about the reference plane then 0Λ = ; the more
asymmetric the plate becomes about the reference plane, the larger Λ becomes. The
approximate range for Λ is 0 0.04≤ Λ ≤ for the composite makeup considered here. The
composite coefficient ( )Χ captures the disparities between the different composite materials.
For a homogenous plate, ( )212 1 νΧ = − . For the composite makeup studied in this dissertation,
Χ deviated approximately 20% from the homogenous value. Equations (3-15) and (3-16) are
then solved using an iterative finite difference scheme discussed in Appendix A.
3.1.5 Deviation from Linearity
Ideally, a microphone should have a linear response over the entire dynamic range of the
device. Having a linear response ensures a constant sensitivity with respect to pressure, which is
essential to designing a microphone with low distortion. To ensure this, the composite plate
model is used to calculate the percent deviation from linearity of the diaphragm’s center
deflection. The result of a non-linearly deflecting diaphragm is harmonic distortion. A 5%
deviation from linearity is set as the limit for the maximum detectable pressure. Figure 3-4
shows the non-dimensional sensitivity of devices with varying in-plane forces. This figure also
shows the trade-off between having an increase in sensitivity versus an increase in the maximum
pressure to remain linear. Figure 3-5 shows the maximum pressure that can act on a plate with a
given in-plane load to remain linear.
54
As the in-plane force parameter decreases past -10 the linear range of the plate goes to
zero. This is due to the onset of buckling and this model accurately predicts the axisymmetric
buckling modes.
3.1.6 Calculation of Stresses
The stress in the plate is decomposed into initial stress due to fabrication and stress due to
pressure loading,
initial stress stress dueto loading
r r r
θ θ θ
σ σ σσ σ σ
⎧ ⎫ ⎧ ⎫ ⎧ ⎫= +⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭ ⎩ ⎭, (3-25)
where
[ ] [ ]0
0r rrQ Q zθ θθ
σ κεσ κε
⎧ ⎫⎧ ⎫ ⎧ ⎫= +⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭⎩ ⎭. (3-26)
The stress due to pressure loading is desired, therefore the non-dimensional radial and tangential
stresses due to loading are defined by
2r
r
SihEa
σΣ =
⎛ ⎞⎜ ⎟⎝ ⎠
(3-27)
and
2
SihEa
θθ
σΣ =
⎛ ⎞⎜ ⎟⎝ ⎠
. (3-28)
Solving for and r θΣ Σ yields
2
11r i i
i
dU U dd d
ν η νν ς ξ ξ ξ ξ
⎛ ⎞⎛ ⎞ ⎛ ⎞Π Θ ΘΣ = + + +⎜ ⎟⎜ ⎟ ⎜ ⎟− ⎝ ⎠ ⎝ ⎠⎝ ⎠
(3-29)
and
55
2
11 i i
i
dU U dd dθ ν η ν
ν ς ξ ξ ξ ξ⎛ ⎞⎛ ⎞ ⎛ ⎞Π Θ Θ
Σ = + + +⎜ ⎟⎜ ⎟ ⎜ ⎟− ⎝ ⎠ ⎝ ⎠⎝ ⎠, (3-30)
where
ah
ς = , (3-31)
zh
η = , (3-32)
and iν is the local Poisson’s ratio. Π is defined as
2
2
1 if
if
if
m Si
SiOm SiO
Si
SiNm SiN
Si
E EE
E EEE E EE
⎧⎪ =⎪⎪
Π = =⎨⎪⎪
=⎪⎩
, (3-33)
where mE is the Young’s modulus in the given layer.
3.1.7 Validation Using Finite Element Analysis
This section details the verification of the composite plate mechanics of the diaphragm
using FEA. The FEA was performed using the commercially available ABAQUS software
package. The elements used were 3-node quadratic, axisymmetric shell elements. This shell
theory allows for finite strains and rotations of the shell [112]. The strain measure used is
accurate to second order with regard to strain. These elements are accurate for scenarios
modeled in terms of Kirchhoff stresses with the following assumptions [112]:
• Only terms up to first order with respect to the thickness direction are included.
• The thinning of the shell due to stretching is assumed to be uniform through the thickness.
• The thinning of the shell is assumed to occur smoothly.
• All stresses except those acting parallel to the reference surface are neglected.
56
• Planar cross sections remain planar.
• Transverse shears are assumed to be small, and the material response to such deformation is assumed to be linear elastic.
The FEA model uses independent stresses defined in each layer. The geometric and
material values used can be found in Table 3-1. Figure 3-6 shows the deflection of a composite
plate calculated analytically from section 3.1 and compared to results obtained from the FEA
simulation. It is easily seen that there is excellent agreement between the two calculations.
Figure 3-7 illustrates the transition from linear to nonlinear behavior as the plate deflection
becomes large. Once again there is excellent agreement between the analytical and FEA models
well into the nonlinear regime.
3.2 Electroacoustics
With the pressure induced stress field known, an electromechanical transduction model is
coupled with the mechanical model to obtain the resulting voltage output.
3.2.1 Piezoresistors
Piezoresistivity, which is defined as the change of the resistivity of a material due to a
change in carrier mobility, as a result of applied mechanical stress. In piezoresistive transduction,
resistance modulation is a function of the applied stress and piezoresistive coefficients ( )ijπ
[113]. For the cubic crystal structure of silicon, the relationship reduces to
where , , and i i il m n are the direction cosines which are given in terms of Euler’s angles [114],
1 1 1
2 2 2
3 3 3
l m n c c c s s s c c c s s cl m n c c s s c s c s c c s sl m n c s s s c
φ θ ψ φ ψ φ θ ψ φ ψ θ ψφ θ ψ φ ψ φ θ ψ φ ψ θ ψ
φ θ φ θ θ
− + −⎡ ⎤ ⎡ ⎤⎢ ⎥ ⎢ ⎥= − − + −⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎣ ⎦ ⎣ ⎦
, (3-37)
where ( )coscθ θ= , etc. These angles are graphically illustrated in Figure 3-8 [19].
For this case, ( )100 silicon is used and the piezoresistors are implanted vertically into the wafer.
Therefore, 0θ = and 0ψ = (refer to Figure 3-8) and the matrix(3-37) reduces to
1 1 1
2 2 2
3 3 3
00
0 0 1
l m n c sl m n s cl m n
φ φφ φ
⎡ ⎤ ⎡ ⎤⎢ ⎥ ⎢ ⎥= −⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎣ ⎦ ⎣ ⎦
, (3-38)
where φ varies from 0 to 2π . Applying a transformation of coordinates to equation (3-35),
using the matrix (3-38), the longitudinal and transverse piezoresistive coefficients are [19],
58
( )( )2 2 2 2 2 211 44 12 11 1 1 1 1 1 12 l m l n m nπ π π π π= + + − + +l (3-39)
and
( )( )2 2 2 2 2 212 44 12 11 1 2 1 2 1 2l l m m n nπ π π π π= − + − + +t . (3-40)
These coefficients are plotted for ( )100 p type silicon in Figure 3-9.
The piezoresistive coefficient also depends on temperature and doping level. . This
relationship is given as a product of the low-doped room temperature value and a piezoresistive
scaling factor ( ),P N T [19] shown as,
( ) ( ), ,N T P N TπΠ = , (3-41)
where N is doping concentration and T is temperature. Many theoretical [19] and experimental
[115-117] studies have shown that the piezoresistive factor ( ),P N T is a function of doping
concentration. Kanda’s model [19] accurately predicts the effect of doping concentration and
temperature for low concentrations. Kanda’s model also shows how at higher doping
concentrations, the effect of temperature on the piezoresistive coefficients is minimized. In fact,
at concentrations above 19 310 # cm , the piezoresistance coefficient is a weak function of
temperature [117]. However, the sensitivity declines due to the reduced piezoresistive
coefficient at a high doping level [115]. Kanda’s model however, when compared to
experimental data [115-117], under predicts the decline of ( ),P N T for concentrations above
17 310 # cm . For doping concentration above 17 310 # cm , the experimentally fitted piezoresistive
factor ( ),P N T [118] is used,
( )0.201422 31.53 10,300 log cmP N K
N
−⎛ ⎞×⎜ ⎟⎝ ⎠
. (3-42)
59
The piezoresistive factor is plotted in Figure 3-10 versus concentration at room temperature.
The piezoresistors are designed to take advantage of the crystallographic dependence of p-
type silicon having almost equal and opposite piezoresistive coefficients (Figure 3-9). The
resistors are designed to isolate current either in the tangential direction or in the radial direction.
The geometry of the piezoresistors in the diaphragm can be seen in Figure 3-11. To calculate the
change in resistance, the piezoresistors are divided up into differential elements (Figure 3-12)
and their individual resistances are numerically integrated. For the arc resistor, the resistance of
an unstressed differential element is [18]
arcrddR
dzdrρ θ
= , (3-43)
and the resistance of a stressed element is given by
( )1arc arc l l t trddR d R
dzdrρ θ σ π σ π+ Δ = + + , (3-44)
where ρ is the resistivity of the material.
Summing up the unstressed differential resistors in series (equation (3-43) along the θ
direction) yields,
( ),
as arc
z rR
drdzρ θ
= . (3-45)
Summing up the differential elements in parallel yields,
0
1 jaout
ain
zr
arc ar
drdzR rρ θ
= ∫ ∫ . (3-46)
Integrating in the r direction yields,
0
ln1 j
aoutz
ain
a a
rr dz
R θ ρ
⎛ ⎞⎜ ⎟⎝ ⎠= ∫ . (3-47)
60
The resistivity is [18]
( )
1
n pq N Pρ
μ μ=
+, (3-48)
where q is the charge of an electron ( )1.6 19 q e C= − , nμ and pμ are the mobilities of electrons
and holes respectively, and N and P are the concentration of electrons and holes respectively.
The piezoresistors are doped heavily with boron which is a p-type implant ( )p n . Therefore
equation (3-48) reduces to
1 1
p p pq P q Nρ
μ μ= = , (3-49)
where pN is the doping concentration of boron [18]. The mobility is modeled after an empirical
fit [18],
0min
1p
p
ref
NN
α
μμ μ= +⎛ ⎞
+ ⎜ ⎟⎜ ⎟⎝ ⎠
, (3-50)
where min 0, , , and refNμ μ α are constants depending on the temperature and dopant [18].
Substituting in for ρ in equation (3-47) yields
( ) ( )( )
min 00 0
ln1
1
j j
aoutz z
painp
a a p
ref
rqN zr
N z dz dzR N z
N
αμ μθ
⎡ ⎤⎛ ⎞ ⎢ ⎥⎜ ⎟ ⎢ ⎥⎝ ⎠= +⎢ ⎥⎛ ⎞⎢ ⎥
+ ⎜ ⎟⎢ ⎥⎜ ⎟⎝ ⎠⎣ ⎦
∫ ∫ . (3-51)
Taking the inverse of equation (3-51) results in
61
( ) ( )( )
min 00 0
1
ln
1
j j
aa z z
aout pp
ainp
ref
Rr N zq N z dz dzr N z
N
α
θ
μ μ=
⎛ ⎞+⎜ ⎟
⎝ ⎠ ⎛ ⎞+ ⎜ ⎟⎜ ⎟
⎝ ⎠
∫ ∫. (3-52)
The ion implantation process yields a Gaussian doping profile [119],
( )
2
: 0j
zz
sp s
b
NN z N zN
⎛ ⎞−⎜ ⎟⎜ ⎟
⎝ ⎠⎛ ⎞= ≥⎜ ⎟
⎝ ⎠, (3-53)
where , z , and s j bN N are the surface concentration, junction depth and background
concentration, respectively. A sample dopant profile is shown in Figure 3-13. The fabricated
profile will differ slightly from a Gaussian profile. Performing the same procedure for the
stressed arc resistor (equation (3-44)) yields,
( ) ( ) ( ) ( ) ( )0
11 , ,
r
jaoutl
ain
a a zr
l l t tr
dR Rdzdr
r z r z r z
θ
θ
θ
ρ σ π θ σ π θ
+ Δ =
⎡ ⎤+ +⎣ ⎦
∫∫ ∫
. (3-54)
For the taper resistor, the resistance of an unstressed differential element is
2t
drdRrd dz
ρθ
= , (3-55)
and the resistance of a stressed element is given by
( )2 1t t l l t tdrdR d R
rd dzρ σ π σ πθ
+ Δ = + + . (3-56)
The factor of 2 follows from the fact that there are two legs per taper resistor. Performing the
same procedure for the unstressed taper resistor as for the unstressed arc resistor yields,
62
( ) ( )( )
min 00 0
2 ln1
1
j j
tout
tint z z
wt pp
p
ref
rr
Rq N z
N z dz dzN zN
α
θμ μ
⎛ ⎞⎜ ⎟⎝ ⎠=
+⎛ ⎞
+ ⎜ ⎟⎜ ⎟⎝ ⎠
∫ ∫, (3-57)
Similarly for the stressed taper resistor,
( ) ( ) ( ) ( ) ( )( )0
2
1 , ,
tout
jrtin
l
r
t t zr
l l t t
drR Rdzdr
z r z r z
θ
θ
θρ σ π θ σ π θ
+ Δ =
+ +
∫∫ ∫
. (3-58)
The MATLAB m-files for calculating the resistance can be found in Appendix C. 3.2.2 Wheatstone Bridge
The piezoresistors are arranged in a fully active Wheatstone bridge configuration as seen in
Figure 3-14. For a constant voltage bias ( )bV , the output of the bridge ( )0V yields,
0a a t t
ba a t t
R R R RV VR R R R
⎛ ⎞+ Δ − − Δ= ⎜ ⎟+ Δ + + Δ⎝ ⎠
. (3-59)
The arc and taper resistors are designed to have the same nominal resistance value
( )a tR R R= = . Applying this to equation (3-59) yields,
0 2a t
ba t
R RV VR R R
⎛ ⎞Δ − Δ= ⎜ ⎟+ Δ + Δ⎝ ⎠
. (3-60)
Knowing that the change in resistance is small compared to the mean resistance, the power
consumption for the circuit is
2
bVPR
≈ . (3-61)
If the device is biased with a constant current source ( )bI the response is,
63
0 2a a t t
bR R R RV I+ Δ − − Δ⎛ ⎞= ⎜ ⎟
⎝ ⎠. (3-62)
Again, assuming a balanced bridge, a tR R R= = yields,
0 2a t
bR RV IΔ − Δ⎛ ⎞= ⎜ ⎟
⎝ ⎠. (3-63)
Equation (3-63) reveals that when the device operates with an ideal constant current source
connected to the Wheatstone bridge, the output voltage does not depend on the unstressed
resistance value. The power consumption for a constant current source is
2bP I R≈ (3-64)
For a device operating with either a voltage or current supply, a power limitation would be
implemented to keep the overall power consumption below 100mW .
3.3 Lumped Element Modeling
The most accurate and complete way to mathematically describe a physical system is a
physics-based model that is correlated to an analytical expression for the system behavior. FEM
does extremely well at predicting system behavior in cases where an analytical approach is
impractical. FEM techniques can accurately predict system behavior using a numerical
approach, producing results that can accurately mimic a physical system, but the physical insight
obtained is limited. In addition, the FEM results are dependant on the convergence of the
iterative calculations as well as the numerical, and it is therefore hard to determine scaling
behavior from FEM results.
LEM is useful to gain understanding of the scaling laws of the system [120-122]. The use
of LEM reduces the complexity of a numeric or analytic expression by dividing a given
distributed system into discrete elements that are based on system interactions with energy [121].
64
Three different types of interactions are accounted for: the storage of kinetic and potential
energy, and the dissipation of energy.
The storage of kinetic and potential energy in a distributed dynamic system requires a
partial differential equation to accurately represent the physics of the problem, because the
spatial and temporal components are intrinsically coupled [123], [124]. When the wavelength of
the signal increases to the point at which it is considerably greater than the length scale of
interest, negligible variation in the distribution of energy as a function of space occurs. At this
point, the mathematical decoupling of the spatial and temporal components allows for the use of
ordinary differential equations [120]. This method assumes that the static mode shape is similar
to the dynamic mode shape up to the first resonant frequency.
Although nomenclature varies in different energy domains, the mathematics remain
constant. In lumped mechanical systems, kinetic energy is stored via mass while potential
energy is represented as the compliance of a spring and the dissipation of energy is represented
as the losses of a damper. In electrical systems, kinetic energy is represented as the magnetic
field of an inductor, while potential energy is represented as the charge across a capacitor, and
the dissipation of energy is represented as a resistor. In lumped acoustical systems, kinetic and
potential energy are represented as acoustical mass and acoustical compliance, while the
dissipation of energy is given as an acoustic resistance.
Many techniques, both graphical and analytical, have been developed to solve large
networks of interconnected elements. In these techniques, the interconnected elements are
represented using electrical circuit notation. In lumped element modeling the elements are
denoted using an equivalent circuit form for all of the energy domains, where masses are
represented as inductors, compliances are represented as capacitors, and the dissipative
65
components are represented as resistors. Once the complete equivalent circuit is constructed,
standard circuit analysis techniques can be applied to find the solution. Power flow between the
elements must be considered when managing more than one lumped element. The elements
must be represented in terms of conjugate power variables, more specifically referred to as an
effort, e, and a flow f, where the product ef is power. A table of conjugate power variables is
given in Table 3-2 for several of energy domains.
3.3.1 LEM of piezoresistive microphone
To create an analytical model for the microphone, the system is broken down into sections.
The microphone is composed of three main mechanical components: diaphragm, cavity and vent
and can be seen in Figure 3-15. The following sections discuss the impedances of each.
3.3.1.1 Diaphragm
The diaphragm is modeled as a mass, spring and a damper. The distributed diaphragm is
modeled as a clamped circular plated in order to find the lumped element representation for the
diaphragm. The plate is lumped to a piston of mass, ( )daM , a resistance of ( )daR and a
compliance, ( )daC , shown in Figure 3-16. The area of the piston and the area of the diaphragm
are not equal; rather the piston area is equated to maintain volume velocity continuity between
the physical diaphragm and the piston model. The acoustic compliance of the diaphragm is
equal to the volume of air displaced by the deflection of the diaphragm [125],
daVolCp
Δ= . (3-65)
where the change in volume is given by
( )2
0 0
a
Vol w r rdrdπ
θΔ = ∫ ∫ . (3-66)
Plugging in the non-dimensional parameters from equation (3-19) yields,
66
( )16
* *0
da
WaC dD P
ξπ ξ ξ= ∫ . (3-67)
The acoustic mass is determined by evaluating the kinetic energy expressed in acoustic conjugate
power variables to the total kinetic energy. This mass is calculated as [125],
( ) 2
0
2a
da A
w rM rdr
Volρ π
⎛ ⎞= ⎜ ⎟Δ⎝ ⎠
∫ , (3-68)
where Aρ is the areal density of the composite plate. Substituting equation (3-66) into (3-68)
and substituting for the non-dimensional variables results with,
( ) 22 15
* *02
Ada
da
WaM dD C P
ξπρ ξ ξ⎛ ⎞⎛ ⎞
= ⎜ ⎟⎜ ⎟⎝ ⎠ ⎝ ⎠
∫ . (3-69)
There is an additional effective mass that acts on the diaphragm, caused by fluid particles that
oscillate with the diaphragm. This radiation mass, radM , is given by approximating the
diaphragm as a piston in an infinite baffle [120],
2
83
airradM
aρπ
= , (3-70)
where, airρ is the density of air. The radiation mass is added to the diaphragm mass to give a
combined diaphragm mass. Assuming the diaphragm is lightly damped, the resonant frequency
of the diaphragm is
1 12dia
da da
fC Mπ
= . (3-71)
The resistance, ( )daR , is caused by damping in the diaphragm. The majority of the damping
contributions are the dissipation to the supports, the dissipation into the surrounding air and
thermomechanical dissipation within the structure. This term is difficult to analytically express
67
accurately and therefore a value of the damping ratio ( )0.03ζ = is taken from previously
fabricated devices with similar size and aspect ratio [37]. The diaphragm resistance is then
calculated[126] as
2 dada
da
MRC
ζ= . (3-72)
The total impedance from Figure 3-16C is
( ) ( ) 1dia da rad da
da
Z j M M Rj C
ω ωω
= + + + . (3-73)
3.3.1.2 Cavity
The cavity is the open area behind the diaphragm. The cavity is cylindrical in shape and
has a backplate that is assumed to be rigid. The cavity is therefore modeled as a closed cavity
with a sound hard boundary. The specific acoustic impedance in the cavity is [127],
( ) ( )0, cotZ d jZ kdω = − , (3-74)
where d is the distance from the bottom wall. For a short cavity, the cotangent function can be
expanded to yield
( ) 020
1lim , ...3kd
Z kdZ d ja kd
ωπ→
⎛ ⎞= − − +⎜ ⎟⎝ ⎠
. (3-75)
For 0.3kd ≤ all but the first term may be neglected, yielding
( ) ( ) 1,a
Z d Zj C
ω ωω
= = , (3-76)
where
20 0
aVCcρ
= , (3-77)
and V is the volume of the cavity,
68
2V a dπ= , (3-78)
a is the radius of the diaphragm, and d is the depth of the cavity. If the cavity is required to be
longer, so that 0.3kd ≤ does not hold, then the second term in the cotangent expansion of
equation (3-75) needs to be accounted for. This term represents an acoustic mass,
023a
dMa
ρπ
= , (3-79)
with the total impedance being,
( ) 1cav a
a
Z j Mj C
ω ωω
= + . (3-80).
A comparison of the full cotangent solution and the leading terms is shown in Figure 3-18.
3.3.1.3 Vent
The vent channel is designed to have a large resistive value with a small mass. To
accomplish this, the length must be as large as possible with a small cross-section. The lumped
acoustic mass for a laminar pipe flow is calculated by integrating equation (3-68) yielding [120],
2
43av
LMa
ρπ
= . (3-81)
Microfabrication processes are not capable of fabricating circular channels and therefore, the
vent channel has a square cross-section. Because of the limited space on a microphone die, the
vent channel is designed as a serpentine. This can be accommodated by calculating an effective
length of the channel by adding all of the major and minor head losses [128]. Equation (3-81) is
estimated using a hydraulic diameter and effective length yielding,
2
163
effav
H
LM
Dρ
π= , (3-82)
69
where HD is the hydraulic diameter and effL is the effective length. The hydraulic diameter is
defined as [128]
4 21H
A hD hP b= =
+, (3-83)
where A and P are the cross sectional area and perimeter, h is the depth of the vent and b is
the width. The effective length takes into account any turns in the channel by equating the
channel with turns to a longer channel without any turns. This device will only have 90 degree
turns so the effective length is equal to the physical length plus a correction factor to account for
the turns [128]
60eff HL L nD= + , (3-84)
where L is the length of the vent and n is the number of right angle turns. To determine the
resistance of the vent, volume velocity is expressed as a function of pressure yielding,
4
8aQ V dA pL
πμ
= ⋅ = Δ∫ ∫ . (3-85)
where Q is volume velocity and pΔ is pressure drop. The resistance is then taken from equation
(3-85),
4
8av
LRaμ
π= . (3-86)
Adjusting the resistance for a serpentine square channel yields,
4
128 effav
H
LR
Dμ
π= , (3-87)
The total impedance is
vent av avZ R j Mω= + . (3-88)
70
3.3.1.4 Equivalent circuit
Analyzing Figure 3-15, pressure acts on the diaphragm and vent simultaneously due to the
front side vent. There is a pressure drop across the diaphragm and through the vent and the
resulting flows converge in the compressible fluid back cavity. This results with the diaphragm
and vent impedance being in parallel and the effective impedance is in series with the back
cavity. Using this analysis, the equivalent circuit for the lumped element model is seen in Figure
3-17
The acoustic sensitivity directly relates to the voltage output of the microphone, which is
defined as ( ) ( )cq j pω ω ω , assuming sinusoidal input [75]. Using Kirchhoff circuit theory
(Figure 3-17), a transfer function for the acoustic sensitivity of the multi-domain dynamic system
is
( )( )2
c venta
dia cav vent dia cav
q j ZSj p Z Z Z Z Z
ωω ω
−= =
+ +, (3-89)
The magnitude and phase response for aS is seen in Figure 3-19 and a table outlining the
lumped element parameters is seen in Table 3-3. The piezoresistive transduction scheme is
modeled as a dependant voltage source, DSV and is given by
( ) cDS me
da
qV Sj C
ωω
⎛ ⎞= ⎜ ⎟
⎝ ⎠, (3-90)
where meS is the quasi-static mechanical sensitivity of the device in [ ]V Pa [75]. The
remaining terms represent the pressure drop across the diaphragm [ ]Pa Therefore, the total
dependant source output is in V . For a device connected to an amplifier with a very large input
impedance, the output voltage of the device is equal to the dependant source voltage. In the limit
of zero frequency, the right side of the circuit possesses (Figure 3-17) infinite impedance,
71
therefore, the microphone does not respond to slow changes in pressure. As the frequency of
interest gradually increases, the microphone response will increase linearly with frequency [75].
After a corner frequency defined by,
12c
av da
fR Cπ
= , (3-91)
the response of the microphone will be in the flat band region and will have zero phase shift until
the frequency of the incident pressure is at the resonant frequency of the device.
3.3.1.5 Cut-on frequency and cavity stiffening
To determine the cut-on frequency of the microphone and prevent cavity stiffening, a
cavity and vent structure was designed. Examining Figure 3-17, and assuming the frequency of
interest is above the cut-on frequency, the equivalent circuit reduces to Figure 3-20. The
compliance of the cavity is in series with the compliance of the diaphragm. Therefore, if the
compliances are of the same order of magnitude, the cavity will have a significant restorative
force on the back of the diaphragm, resulting in a reduction of sensitivity. Also illustrated in
Figure 3-17, the vent structure needs to have a large resistance to lower the cut-on frequency. If
the resistance of the vent is too low, all of the volume flow will pass through the vent, resulting
in no response of the diaphragm as seen in equation (3-92).
ventc
vent cav
Zq qZ Z
=+
(3-92)
3.3.2 FEA verification
A modal analysis was performed using ABAQUS and compared to the lumped element
model, equation (3-71). This was done for three different size diaphragms designated A, B and
C. The results can be seen in Table 3-4 and all results match to within 1%. An axisymmetric
72
composite plate model meshed with seventy-four 3-node quadratic thin (or thick) shell elements
(type SAX2) was used. Residual stress was supplied as an initial condition.
3.4 Electronic Noise
The dominant source of noise in a piezoresistive microphone determined by Dieme et al.
[129] is the electronic noise of the resistors. Therefore, the lowest detectable signal is
determined by the electronic noise of the Wheatstone bridge. The two dominant types of noise
in resistors are thermal noise and flicker ( )1 f noise. Thermal noise is given by [130],
4tR b KV k T R f= Δ , (3-93)
where bk is Boltzmann’s constant, kT is the temperature in Kelvin, R is the nominal
resistance, and fΔ is the frequency range over which the noise is calculated. When an external
voltage is applied to imperfect electrical conductors with interfacial or bulk defects, an excess
noise above the thermal equilibrium noise floor is observed. This excess noise exhibits an
inverse frequency 1/ f dependence. The mechanism that generates electrical 1/ f noise is still
debated, however the model used in this work follows the mechanism described by Hooge [131].
Hooge’s mechanism is described as the fluctuation in the bulk mobility of the material. He gave
an empirical formula for the noise PSD of 1/ f noise as
2
1/ fVS
Nfα
= , (3-94)
where α is the Hooge parameter (determined experimentally), N is the number of carriers in
the resistor, and V is the bias across the resistor. The voltage noise is then given by [131],
2
21/
1lnfR
V fV fNα ⎛ ⎞= ⎜ ⎟
⎝ ⎠, (3-95)
73
where 2f and 1f are the bounds of the frequency range of interest. Using equations (3-93) and
(3-95), the total noise in the Wheatstone bridge at the output is calculated to be
( ) 2 22 1
1
1 1 14 ln8N b K
arc tap
fV k T R f f V N N fα ⎛ ⎞ ⎛ ⎞= − + + ⎜ ⎟⎜ ⎟ ⎝ ⎠⎝ ⎠. (3-96)
The Analog Devices AD624, which has a noise power spectral density of 4 /nV Hz at 1kHz, is
used for amplification. The 1 f noise of the amplifier is much lower then that of the resistors
and therefore it is neglected. The total noise of the microphone coupled with the amplifier is
given by
( ) ( ) ( )22 22 1 2 1
1
1 1 14 ln 4 98N b K H
arc tap
fV k T R f f V e f fN N fα ⎛ ⎞ ⎛ ⎞= − + + + − −⎜ ⎟⎜ ⎟ ⎝ ⎠⎝ ⎠. (3-97)
For this device, the power spectral density of the noise was calculated for a 1 Hz bin centered at
( )1 kHz dB SPL .
3.5 Conclusions
All portions of the piezoresistive microphone are modeled in this chapter. A composite
plate model is derived that determines the stress in the diaphragm and the onset of nonlinearity.
An electromechanical transduction model determines the resulting change in resistance due to a
given stress field, and a lumped element model determines the overall dynamics of the
microphone including the cavity and vent structure interactions. Given a set of design variables,
the expected performance of a microphone can be calculated. This includes the frequency
response function and onset of nonlinearity. Chapter 4 implements an optimization scheme that
utilizes all of these models to generate a superior device.
74
Table 3-1. Material parameters and thicknesses used for FEA analysis. Material Thickness σ0 E ν
[μm] [MPa] [MPa] [--] Si 2.0 0 150 0.27
SiO2 0.3 -300 70 0.17 SixNy 0.1 100 270 0.24
Table 3-2. Conjugate power variables for various energy domains. Energy Domain Effort Flow Mechanical translation Force, F Velocity, v Fixed-axis rotation Torque, τ Angular velocity, ω Acoustic Pressure, P Volumetric flow, Q Electric circuits Voltage, V Current, I Magnetic circuits MMF, M Flux rate, φ Incompressible fluid flow Pressure, P Volumetric flow, Q Thermal Temperature, T Entropy flow rate, S
75
Table 3-3. Lumped element modeling parameter estimates. Acoustic
impedanceDescription
daM Equivalent acoustic mass lumped as a rigid baffle. ( ) 22 15
* *02
Ada
da
WaM dD C P
ξπρ ξ ξ⎛ ⎞⎛ ⎞
= ⎜ ⎟⎜ ⎟⎝ ⎠ ⎝ ⎠
∫
daC Volume displacement normalized by the pressure. ( )16
* *0
da
WaC dD P
ξπ ξ ξ= ∫
radM Approximating the diaphragm as a piston in an infinite baffle.
2
83
airradM
aρπ
=
Diaphragm [16], [75]
daR Do not have an accurate way of modeling the damping of the diaphragm. Damping ratios estimated from experiments of similar previously fabricated devices.
aC 2
0 0aCC
cρ∀
= , where ∀ is volume of the cavity. Cavity [127]
aM 023a
dMa
ρπ
= , where d is the depth of the cavity.
avR Assuming fully developed pressure driven flow.
4
128 effav
H
LR
Dμ
π= , where effL is the effective length and
HD is the hydraulic diameter.
Vent [128], [132]
avM Also assuming fully developed pressure driven flow.
2
163
effav
H
LM
Dρ
π=
Table 3-4. Results from FEA analysis compared to analytical results. Resonant Frequency [kHz]
Device Analytical FEA % Difference A 253.5 252.1 -0.56% B 227.2 225.7 -0.66% C 200.1 199.1 -0.50%
76
Figure 3-1. Overview of the microphone modeling process.
Figure 3-2. Schematic of composite plate.
77
Figure 3-3. Kirchoff's hypothesis showing the neutral axis and transverse normal.
10-1 100 101 102 103 104 105 106
10-4
10-3
10-2
10-1
100
Non-dimensional pressure (P*)
Cen
ter d
efle
ctio
n pe
r pre
ssur
e (W
0( ξ =
0)/P
)
Figure 3-4. Non-dimensional center deflection per unit pressure of devices with varying in-plane forces.
78
-20 -10 0 10 20 30 40 5010-2
10-1
100
101
102
In-plane force parameter (k*2)
Max
imum
pre
ssur
e to
rem
ain
linea
r (P
* max
)
Figure 3-5. Pressure that results in a 5% deviation from linearity for various inplane forces.
0 20 40 60 80 100-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
Radius (r) [μm]
Tran
sver
se d
efle
ctio
n (w
0) [ μ
m]
FEAAnalytical
Figure 3-6. Analytical deflection of clamped plate, at the onset of non-linearity (2000 Pa), compared to FEA results.
79
10-1
100
101
102
103
104
105
106
10-4
10-3
10-2
10-1
100
Non-dimensional pressure (P*)
Cen
ter d
efle
ctio
n pe
r pre
ssur
e (W
0(ξ
= 0)
/P)
AnalyticalFEA
Figure 3-7. Center deflection per non-dimensional pressure as a function of *P for various values of in-plane stresses.
φ
φ
θ
θ
ψ
z
x
y
*y
*z
*x
Figure 3-8. Description of the Euler’s angles [19].
80
2e-010
4e-010
6e-010
8e-010
30
210
60
240
90
270
120
300
150
330
180 0
tπ
lπ
110 110
l tπ π= −
Figure 3-9. Crystallographic dependence of the piezoresistive coefficients for p-type silicon [ ]1 Pa [19].
81
Figure 3-10. Piezoresistive factor dependence on doping concentration at room temperature . Blue line corresponds to Kanda’s work [19] and the pink line corresponds to the
work of Harley et al. [118]
a
,t inr
,t outr
,a outr,a inr
aθ
wtθ
tθ
Figure 3-11. Geometry of piezoresistors.
82
Taper Resistor
Arc Resistor
Figure 3-12. Differential elements of the arc and taper resistor.
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.161015
1016
1017
1018
1019
1020
Depth into substrate (z) [μm]
Car
rier c
once
ntra
tion
(Np)
[#/c
m3 ]
Figure 3-13. Sample Gaussian dopant profile.
83
t tR R+ Δ a aR R+ Δ
0V
,b bV I
t tR R+ Δa aR R+ Δ
Figure 3-14. Stressed arc and taper resistors configured in a Wheatstone bridge.
( )tpradM
,a aC M
,av avR M
, ,da da daM C R
Figure 3-15. Schematic of MEMS microphone and associated lumped elements.
84
A
p
Area, AeffMda
Cda
B
C
Mda
Cda
Rda
p(t)
Q
Figure 3-16. A) Diaphragm with distributed deflection. B) Lumped diaphragm with equivalent volume flow rate. C) Equivalent circuit for the lumped diaphragm.
aC
DS
BR
OV)(tp
aC
DSV
BR
OV
avM
avR
daCda radM M+
aM
diaZ
ventZ
cavZ qc
daR
Figure 3-17. Equivalent circuit model of the microphone.
85
0 0.5 1 1.5 2 2.5 3-10
-8
-6
-4
-2
0
2
4
6
8
10
kd
Im[Z
cavi
ty/Z
0]
SpringSpring-MassFull Sol.
Figure 3-18. Accuracy of first terms of cotangent expansion.
10-1 100 101 102 103 104 105 106
-20
0
20
40
Mag
nitu
de [d
B]
(ref r
espo
nse
@ 1
kHz)
10-1 100 101 102 103 104 105 106
-150
-100
-50
0
50
Frequency [Hz]
Pha
se [d
eg]
Figure 3-19. Magnitude and phase response of LEM normalized by the flat band response.
86
da radM M+ daCdaR
aC
aM
( )p t dq
Figure 3-20. Equivalent circuit illustrating the effect of the cavity compliance
87
CHAPTER 4 OPTIMIZATION
There are many factors that influence the behavior of the device, including composite lay-
up, aspect ratio, piezoresistor geometry, doping densities and doping profiles. Because of this,
an optimization scheme was employed to determine the variables that yielded optimal
performance specifications. This chapter begins with the methodology used to formulate the
optimization scheme, including the objective function, variables and constraints. Next, the
results are discussed for various cases including a device operating on a voltage source and
current source. In addition, an optimization was then run to determine the feasibility of
fabricating multiple devices on a single wafer. Finally, a sensitivity analysis and uncertainty
analysis was performed on the final optimized devices.
4.1 Methodology
The ultimate goal is to maximize the operational space of the device over a specified
region shown in Figure 4-1. To increase this area, the optimization scheme must minimize the
minimum detectable pressure, MDP , and simultaneously maximize the maximum detectable
pressure, maxP , and bandwidth, BW , of the device. To accomplish this multiobjective
optimization, the ε -constraint method is implemented to generate a Pareto front [133]. An
example is shown in Figure 4-2. In this figure, the utopia point is defined as the point that
possesses the best obtainable value for each objective function. Points A-D are non-dominated
solutions on the Pareto front and point E is a dominated point within the feasible region of the
design space. Since the maximum detectable pressure and bandwidth have a more specified
desired value, they are chosen to be constraints. The MDP remains the primary objective
function because the goal is to lower this value as much as possible. Mathematically this is
expressed as
88
max 1 2 and i j
MDPP BWε ε− ≤ − ≤
Minimize such that
, (4-1)
where 1ε and 2ε are the iterated constrained values varied over a desired range.
4.1.1 Objective Function
The sensitivity of the microphone is defined as the ratio of the output voltage to the input
pressure,
0me
VSP
Δ= , (4-2)
which can be seen pictorially in Figure 4-3 where 0VΔ is given by equation (3-60) or (3-63) for a
voltage or current source, respectively. The minimum detectable pressure is defined when the
output signal is the same magnitude as the noise level,
min0
N N
me
V V PPS V
= =Δ
, (4-3)
where NV is given by equation (3-97). Expressing minP in dB yields
min20 logref
PMDPP
⎛ ⎞= ⎜ ⎟⎜ ⎟
⎝ ⎠, (4-4)
where,
20refP Paμ= . (4-5)
4.1.2 Variables
As shown in sections 3.1 and 3.2, the design variables are as follows:
1 2 3 , , , ,
, , , , , , , , , , ,j a t wt a in a out t in t out
GeometricH H H a z r r r rθ θ θ−
, (4-6)
1 2 3 1 2 3
, , , , , ,Material
E E E ν ν ν α−
, (4-7)
89
and
, p B B
OtherN V or I−
. (4-8)
To satisfy equation (3-60) and balance the Wheatstone bridge, the mean resistance of the taper
resistor is set equal to the arc resistor. To accomplish this, the following relationship must hold:
, ,
2 log logwta t in a in
a ar r
θθ
⎛ ⎞ ⎛ ⎞= ⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠. (4-9)
It is important that the resistors are located at the point of highest stress. To ensure this, the outer
radii of the resistors are constrained to be 5 mμ larger then the radius of the diaphragm. This
accounts for any fabrication issues associated with the back to front side alignment and takes into
account compliant boundary conditions. FEA studies have indicated that the compliant region
scales with the diaphragm thickness. The fabricated microphone structure is similar to the work
of Chandraschakren [134] whose calculated Hooge parameter was 7 5e − . In addition the work
of Dieme [135] leads to an estimate of the value for the Hooge parameter of 5 6e − . Finally, the
material values are assumed to be constant and are not varied during the optimization. This
reduces the total optimization variables to twelve,
1 2 3 , ,, , , , , , , , , , or j a t a in t in surf B BH H H a z r r N V Iθ θ . (4-10)
4.1.3 Constraints
The constraints are comprised of fabrication and performance constraints. All of the
variables are given upper and lower bounds so that
{ } { } { }1 2 3 , ,, , , , , , , , , ,j a t a in t in surf BLB H H H a z r r N V UBθ θ≤ ≤ . (4-11)
The values for the LB and UB were determined by physical and fabrication limitations and can
be found in Table 4-1. For example, the resistors are implanted into the silicon layer. Therefore
the junction depth cannot be larger then the silicon thickness. Fabrication resolution is also finite
90
and there is a minimum line width that can be achieved. For the fabrication methods used, the
minimum line width ( )linew was estimated to be a very conservative 10 mμ . This also helps
prevent problems of punch through in the taper resistor faced by Li [136]. The features for
which the line width constraint applies are shown in Figure 4-4. The constraints are then
,, , , ,t t gap t a a linew w l w l w≥ , (4-12)
which in terms of the optimization variables are
( )
,
,
,
,
,
,,0,0,
0.
a in line
t in line
a a in line
wt t in line
t in t wt line
a r wa r w
r wr w
r w
θ
θ
θ θ
− + ≤ −
− + ≤ −
− + ≤
− + ≤
− − + ≤
(4-13)
In addition to the geometric constraints, performance constraints are also employed. For the
device response to remain linear, the onset of non-linearity must be higher than that of the
desired operational range. This results with the following constraint,
max20 logref
p PMdBp
⎛ ⎞≥⎜ ⎟⎜ ⎟
⎝ ⎠, (4-14)
where PMdB is the upper limit of the desired range. The power dissipated in the device is
constrained to be less then or equal to a specified value ( )maxW . Power is equal to
BW V I= . (4-15)
the current is related to the voltage by the resistance of the Wheatstone bridge, BR , yielding the
following constraint,
2
maxB
B
V WR
≤ , (4-16)
for a constant voltage device or
91
2maxBI R W≤ , (4-17)
for a constant current device. The 3dB+ point for the upper bandwidth limit is calculated from
the LEM in section 3.3 and is constrained to be greater than or equal to a specified BW , that is
( )3
2dB
H j f BWπ+
≥ . (4-18)
The bandwidth of the device was calculated using the lumped element model assuming a
damping ratio of 03.0=ζ . At the 3dB+ point the phase lag is 2 degrees. This value of ζ was
taken from experimentally determined damping ratios of previous similar devices [37]. Finally,
the device needs to remain in the linear regime and therefore the non-dimensional parameter *ck
is constrained to be below a value of 3.4, before buckling occurs (Recall Figure 3-5). Collecting
all of the variables and constraints with equation (4-1) yields,
{ } { } { }
( )
max 1 2
1 2 3 , ,
, ,
, ,
,
,
,
*
and
, , , , , , , , , ,
1 0
1 0
1 0
1 0
1 0
1 03.4
i j
j a t a in t in surf B
a out a in
line
t out t in
line
a a in
line
wt t in
line
t in t wt
line
c
MDPP BW
LB H H H a z r r N V UB
r rw
r rw
rw
rw
rw
k
V
ε ε
θ θ
θ
θ
θ θ
− ≤ − ≤
≤ ≤
− ++ ≤
− ++ ≤
−+ ≤
−+ ≤
− −+ ≤
− ≤
Minimize such that
2
max
1 0B
RW− ≤
(4-19)
92
To run the optimization, MATLAB’s optimization toolbox, specifically the sequential quadratic
programming function, fmincon, was utilized. All values are nondimensionalized to range from
0 to 1 which normalizes all search gradients within fmincon and provides better search directions
for the software package.
4.2 Optimization Results
The optimization was run in two modes that are discussed in the following sections. The
first mode was for a constant bias voltage applied across the Wheatstone bridge and the second
for a constant current source through the bridge.
4.2.1 Optimization with Constant Voltage
In Figure 4-5, MDP is plotted for a variety of maxP and bandwidth constraints. With
increasing bandwidth constraint, the MDP also increases, as expected. The same trend holds true
for MDP as a function of maxP , shown in Figure 4-6, also as expected. The curves corresponding
to a maximum pressure of 140 and 145 converge at a higher bandwidth because the maxP
constraint is not active. This occurs because the bandwidth constraint will not allow the
diaphragm to become more compliant. For each case the power constraint is also active at
100mW . The devices preliminarily chosen for fabrication are designated devices A, B, and C
and can be seen in Figure 4-5 and Table 4-2. All variables and constraint values can be found for
all devices in the multi-objective optimization in Appendix D.
In general, the optimization simultaneously lowers the noise floor and increases the
sensitivity. To lower the noise floor when 1/f noise dominates[131], the bias voltage needs to be
lowered and the total number of carriers needs to be increased. This is proportional to the
following variables,
93
( )2 2
bN
s j a ain
VVN z a rθ
∝−
. (4-20)
When the thermal noise is dominant the mean resistance dominates the noise floor[130]. This is
proportional to the following variables,
( )
a s
j ain
NRz a r
θ∝
−. (4-21)
The sensitivity of the device, is proportional to the compliance of the diaphragm and the
compliance of the diaphragm is proportional to the following variables,
2aSens
h∝ . (4-22)
Conversely the bandwidth constraint is proportional to
1BWMC
∝ (4-23)
and the bandwidth is therefore proportional to the following variables
222
1 1BWaaa h
h
∝ = (4-24)
The linearity constraint is inversely proportional to the compliance of the diaphragm. This
results with
max 2
hPa
∝ . (4-25)
From the proportionalities, it can be seen that the sensitivity is inversely proportional to the
bandwidth and maximum detectable pressure. This sets up the trade-off between each value and
results with the Pareto fronts in Figure 4-5 and Figure 4-6.
94
4.2.2 Optimization with a Constant Current Source
A National Instruments PXI system may be used to power the microphone. This system
runs in two modes: 4 10%mA ± and 10 15%mA ± [137]. Figure 4-7 and Figure 4-8 show MDP
plotted as a function of bandwidth constraint for a variety of maxP constraints for devices running
at 4mA and 10mA , respectively. The MDP of devices A,B and C can be seen in Table 4-3 and
Table 4-4.
Comparing Table 4-2 to Table 4-3 and Table 4-4, it can be seen that the constant current
source devices (running at 4mA ) have a higher MDP then their voltage source counterpart. The
devices running at 10mA have a lower MDP then those running at 4mA and are almost identical
to their voltage driven counterparts. A sensitivity analysis was then done on devices A, B, and C
varying the current source by the specified 10%± and 15%± . The results can be seen in Figure
4-9 and Figure 4-10. Note that a higher input current would violate the power limitation
constraint.
4.2.3 Constraining Devices to a Single Wafer
To reduce fabrication costs, the production of multiple microphone designs on one wafer is
required. However, different thicknesses, doping concentrations, and junction depths do not
allow the devices A, B and C to be processed on the same silicon wafer. To make fabrication on
the same wafer possible, the devices were optimized with an added constraint of each device
having the same thickness of silicon, oxide and nitride. The doping concentration and junction
depth were also constrained to be the same. The reformulated optimization problem is
95
{ } { } { }
( )
max 1 2
, ,
, ,
, ,
,
,
,
*
2
max
and
, , , , , or
1 0
1 0
1 0
1 0
1 0
1 03.4
1 0
a t a in t in B B
a out a in
line
t out t in
line
a a in
line
wt t in
line
t in t wt
line
c
B
MDPP BW
LB a r r V I UB
r rw
r rw
rw
rw
rw
k
VRW
ε ε
θ θ
θ
θ
θ θ
− ≤ − ≤
≤ ≤
− ++ ≤
− ++ ≤
−+ ≤
−+ ≤
− −+ ≤
− ≤
− ≤
Minimize such that
(4-26)
Table 4-5 and Table 4-6 show the results of an optimization with these added constraints
for voltage and current source (10mA) devices, respectively. The first sub-table is for a set of
devices where the parameters of device A were used as the constraints for devices B and C. The
second sub-table uses parameters from device B, constraining devices A and C, and so forth for
the remaining tables. The difference in MDP is the difference between each device’s constrained
and unconstrained MDP.
From this data, the scenario of devices run with a 10mA current source were chosen to
fabricate. Specifically, the devices from the first scenario of Table 4-6. This was chosen
because the sponsor, Boeing, ultimately would like to operate the microphones from the current
power supply from their National Instruments PXI system. Between the two modes, the devices
operating at 10mA had greater performance. Additionally, the designs to operate at 10mA
possess similar performance as devices designed with a voltage source.
96
4.2.4 Sensitivity Analysis
To determine the dependence of the results to changes in various parameters and to gain a
physical understanding of different aspects of the optimization, a sensitivity analysis was
performed. This analysis was performed on each variable in the optimization as well as the noise
figure of merit, the Hooge parameter.
The following results are all for optimized device A from Figure 4-8. As previously stated,
the optimization was performed assuming a Hooge parameter of 5 6e − . Figure 4-11 shows that
errors in the assumed Hooge parameter can have a major impact on MDP. If 1/f noise is
dominant, the voltage noise and MDP scale with the ( )1/ 2α .
The dependence of MDP with respect to all independent variables is shown in Figure 4-12.
The variable with the largest effect on the MDP is the thickness of the silicon. As the thickness
decreases, the stiffness decreases as well and the diaphragm begins to buckle resulting in the
sharp drop in MDP. The linear deflection solution however is invalid in this region. Figure 4-13
shows the effect of MDP on the thickness of silicon as well as the constraint for buckling. The
optimized point is at * 3.4ck = showing that this constraint is active and will not allow the
thickness of the silicon layer decrease.
4.2.5 Uncertainty Analysis
The uncertainty for the theoretical performance metrics is derived in this section. The
formulations presented here utilize results obtained in section 4.2.3. The MDP of design A is
analyzed. Furthermore, the predictions for the bandwidth and linearity constraints are explored.
Calculating the MDP of the microphone includes numerical integrals and therefore an
explicit analytical formula cannot be obtained. To calculate the uncertainty of the design metrics
a Monte Carlo simulation is employed. The Monte Carlo method involves assuming
97
distributions for all of the input uncertainties and then randomly perturbing each input variable
with a perturbation drawn from its uncertainty distribution [138]. The standard deviation for
each variable is estimated from manufacturing tolerances and can be found in Table 4-7. These
statistically independent values are fed into the objective function and the distribution of the
objective function is obtained. In this case, uncertainties in the design variables correspond to a
statistical distribution of the noise voltage, sensitivity and their ratio, the MDP. There will also
be a distribution for the resonant frequency and maximum detectable pressure. This process is
illustrated in Figure 4-14 where ix is the optimized design variable, iσ is the standard deviation
for each design variable and Y is a Gaussian distributed random number with mean zero,
variance one and standard deviation one.
The Monte Carlo simulation was implemented in MATLAB and the random independent
variables were generated using the randn function. The simulation was run for 100,000
iterations and the probability distribution function (PDF) for MDP can be seen in Figure 4-15.
The results highlighted in yellow and red are cases where the composite diaphragm is buckled
and close to buckling, respectively. Since the device is designed to be in the linear regime, the
optimization uses a linear solver to calculate the deflections and stresses within the diaphragm.
These solutions are invalid because the linear solution starts to deviate from the nonlinear
solution at about 3.6k ≈ . The PDFs for maxP and bandwidth can be seen in Figure 4-16 and
Figure 4-17 respectively. The distributions are not assumed to be Gaussian and all statistical
moments can be found in Table 4-9. To determine the 95% probability limits for each
specification, a numerical integration was performed. This integration starts at the mean value
and moves outwards until the value under the PDF is 47.5%. The process is done for values
above and below the mean to yield a total probability of 95%. Figure 4-18 shows this process
98
pictorially. The 95% probability for the performance of the desired specifications can be seen in
Table 4-8. It is important to note that the lower end of the confidence integral is in question
because of the error in the linear solution. The confidence integral was calculated by
numerically integrating under the PDF.
4.3 Conclusion
This chapter implemented a multi-objective optimization scheme to help design a superior
device. Devices were optimized for voltage sources as well as current sources. A secondary
optimization was also completed constraining the devices to have multiple on each silicon wafer.
A sensitivity analysis was also conducted to determine the effect of each variable on the primary
objective function, MDP. Finally an uncertainty analysis was completed using a Monte Carlo
simulation to help determine the uncertainty of the various objective functions.
99
Table 4-1. Upper and lower bounds for all variables
MOS capacitor * In addition, two contact resistances are added to each resistor to account for the current flowing to and from each resistor. * The calculated Hooge parameter is also used.
Table 6-11. Results from the radius determination experiment. BUF1A
adesign ameas 95% CI 113 108 (107.3 - 108.7)
* All measurements in μm.
Table 6-12. Measured values and standard deviation of input parameters. HSi HSi02 HSiN a Ns zj rain Θa rtin Θt Ib μm Α A μm #/cm3 μm μm deg μm deg mΑ
Figure 6-7. Boron concentration in silicon device layer determined by SIMS, the accompanying curve fit and the desired model profile.
-10 -8 -6 -4 -2 0 2 4 6 8 10-15
-10
-5
0
5
10
15
Potential [V]
Cur
rent
[mA
]
Device ADevice B
RinA = 934Ω
σ = 12.9ΩRinB = 1020Ω
σ = 15.5Ω
Figure 6-8. Input I-V curve of 12 BUF1-A and 12 BUF1-B devices.
142
-10 -8 -6 -4 -2 0 2 4 6 8 10-15
-10
-5
0
5
10
15
Potential [V]
Cur
rent
[mA
]
Device ADevice B
RoutA = 930Ω
σ = 10.2ΩRoutB = 1021Ω
σ = 16.6Ω
Figure 6-9. Output I-V curve of 12 BUF1-A and 12 BUF1-B devices.
-10 -8 -6 -4 -2 0 2 4 6 8 10-20
-10
0
10
20
Potential [V]
Cur
rent
[mA
]
-10 -8 -6 -4 -2 0 2 4 6 8 101.5
1.6
1.7
1.8
Leak
age
Cur
rent
[ μA
]
Curve Fit
data
Figure 6-10. Input I-V curve of a BUF1-A device with a linear curve fit.
143
-10 -8 -6 -4 -2 0 2 4 6 8 10-15
-10
-5
0
5
10
15
Potential [V]
Cur
rent
[mA
]
-10 -8 -6 -4 -2 0 2 4 6 8 101.8
1.9
2
2.1
2.2
2.3
Leak
age
Cur
rent
[ μA
]
Curve Fit
Data
Figure 6-11. Output I-V curve of a BUF1-A device with a linear curve fit
-20 -15 -10 -5 0 5 10-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Potential [V]
Cur
rent
[mA
]
Figure 6-12. I-V curve of diode characteristics of a BUF1-A device.
144
-12 -10 -8 -6 -4 -2 0-3
-2.5
-2
-1.5
-1
-0.5
0
Potential [V]
Cur
rent
[ μA
]
Figure 6-13. I-V curve of diode characteristics of a BUF1-A device focusing on the reverse region.
10-1
100
101
102
103
104
105
106
10-18
10-17
10-16
10-15
10-14
10-13
10-12
10-11
10-10
Frequency [Hz]
Noi
se P
SD
[V2 /
Hz]
setupV = 0.22V = 0.43V = 0.81
Figure 6-14. Noise PSD of a test taper resistor.
145
10-1 100 101 102 103 104 105 10610-18
10-17
10-16
10-15
10-14
10-13
10-12
10-11
10-10
Frequency [Hz]
Noi
se P
SD
[V2 /
Hz]
V = 0.22V = 0.43V = 0.81
Figure 6-15. Noise PSD from a test taper resistor minus the setup noise and the associated model curve fit. The horizontal line is the thermal noise floor for this device.
10-1 100 101 102 103 104 105
10-16
10-15
10-14
10-13
10-12
10-11
10-10
10-9
Frequency [Hz]
Noi
se P
SD
[V2 /
Hz]
setupV = 0.28V = 0.93V = 1.96V = 3.94V = 9.96
Figure 6-16. Noise power spectral density of a BUF1-A device.
146
10-1
100
101
102
103
104
105
106
10-18
10-17
10-16
10-15
10-14
10-13
10-12
10-11
10-10
Frequency [Hz]
Noi
se P
SD
[V2 /
Hz]
V = 0.28V = 0.93V = 1.96V = 3.94V = 9.96
Figure 6-17. Noise PSD minus the setup noise and the associated model curve fit. The horizontal line is the thermal noise floor for the device.
115 120 125 130 135 140 145 150 155 1602
3
4
5
6
7
8
9
10
Sen
sitiv
ity [n
V/P
a/V
]
Incident Pressure [dB SPL]
Figure 6-18. Sensitivity of BUF1-A devices normalized by the bias voltage.
147
115 120 125 130 135 140 145 150 155 160 165 1700
5
10
15
20
25
THD
[%]
Incident Pressure [dB SPL]
Figure 6-19. Total harmonic distortion of BUF1-A microphones.
0 1 2 3 4 5 6 7-170
-165
-160
-155
-150
-145
-140
-135
-130
Frequency [kHz]
Mag
nitu
de R
espo
nse
[dB
] (re
1V
/Pa)
Figure 6-20. Magnitude frequency response for a BUF1-A device. Vertical dotted lines mark the piecewise FRFs that were stitched together.
148
0 1 2 3 4 5 6 7-155
-150
-145
Mag
FR
F [d
B]
Frequency [kHz]
0 1 2 3 4 5 6 7-155
-150
-145
Mag
FR
F [d
B]
Frequency [kHz]
0 1 2 3 4 5 6 7-155
-150
-145
Mag
FR
F [d
B]
Frequency [kHz]
0 1 2 3 4 5 6 7-155
-150
-145
Mag
FR
F [d
B]
Frequency [kHz]
Figure 6-21. Magnitude FRF for each device with 95% CI bounds.
149
0 1 2 3 4 5 6 7-20
-15
-10
-5
0
5
10
15
20
Frequency [kHz]
Pha
se R
espo
nse
[deg
]
Figure 6-22. Phase response for each device tested.
150
0 1 2 3 4 5 6 7-20
-10
0
10
20
Pha
se [d
eg]
Frequency [kHz]
0 1 2 3 4 5 6 7-20
-10
0
10
20
Pha
se [d
eg]
Frequency [kHz]
0 1 2 3 4 5 6 7-20
-10
0
10
20
Pha
se [d
eg]
Frequency [kHz]
0 1 2 3 4 5 6 7-20
-10
0
10
20
Pha
se [d
eg]
Frequency [kHz]
Figure 6-23. Phase FRF for each device with 95% CI bounds.
151
0 1 2 3 4 5 6 70.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Frequency [kHz]
Coh
eren
ce
Figure 6-24. Coherence function between device A-5 and the reference microphone.
A B
Figure 6-25. A) Picture of a microphone die with topside lighting. B) Picture of a microphone die with backside lighting.
152
2 3 4 5 6 7 8 90
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Pro
babi
lity
Den
sity
Fun
ctio
n of
MD
P
MDP [Pa]
Figure 6-26. Minimum detectable pressure probability density function.
Multiplying by ( )2ξ ξ and applying the central difference equations (A-197) and (A-198) (where
and f x ξ→ Θ → ) to equation (A-255) yields
( ) ( ) ( )1 1n n n n n n nc a b d− +Θ + Θ + Θ = , (A-256)
where
( )( ) ( ) ( )2 2* * 12 1 ,
2nn n c r nn n na k Sξ ξ ξ ξ ξ ξ⎛ ⎞= − Ω − + + − Λ Θ⎜ ⎟
⎝ ⎠ (A-257)
201
,n n nb = Ω + Γ (A-258)
,n n nc = Ω − Γ (A-259)
and ( )3*n n
d ξ ξ= −Ρ . (A-260)
All of the remaining equations are identical to the tension case.
Non-dimensional Stresses
The stress in the composite plate is defined by equation (A-15). The stress is then
decomposed into initial stress and stress due to loading
initial stress stress dueto loading
r r r
θ θ θ
σ σ σσ σ σ
⎧ ⎫ ⎧ ⎫ ⎧ ⎫= +⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭ ⎩ ⎭, (A-261)
where
[ ] [ ]0
0r rrQ Q zθ θθ
σ κεσ κε
⎧ ⎫⎧ ⎫ ⎧ ⎫= +⎨ ⎬ ⎨ ⎬ ⎨ ⎬
⎩ ⎭ ⎩ ⎭⎩ ⎭. (A-262)
The non-dimensional radial and tangential stresses due to loading are defined by
2r
r
SihEa
σΣ =
⎛ ⎞⎜ ⎟⎝ ⎠
, (A-263)
and
2
SihEa
θθ
σΣ =
⎛ ⎞⎜ ⎟⎝ ⎠
. (A-264)
Note that SiE is Young’s modulus of the bulk silicon. To solve for the non-dimensional stress,
0 and ε κ are non-dimensionalized by
202
0 0
0 0
2 2
2
2
,
,
,
1 1 .
r r
r r
dU aEd hU aE
hd W d aKd d h
dW aKd h
θ θ
θ θ
εξ
εξ
κξ ξ
κξ ξ ξ
= =
= =
Θ= − = = Ψ =
= − = Θ =
(A-265)
Solving for and r θΣ Σ yields
2
11r i i
i
dU Ud
ν η νν ς ξ ξ ξ
⎛ ⎞⎛ ⎞ ⎛ ⎞Π ΘΣ = + + Ψ +⎜ ⎟⎜ ⎟ ⎜ ⎟− ⎝ ⎠ ⎝ ⎠⎝ ⎠
(A-266)
and
2
11 i i
i
dU Udθ ν η ν
ν ς ξ ξ ξ⎛ ⎞⎛ ⎞ ⎛ ⎞Π Θ
Σ = + + Ψ +⎜ ⎟⎜ ⎟ ⎜ ⎟− ⎝ ⎠ ⎝ ⎠⎝ ⎠ (A-267)
where iν is the local Poisson’s ratio. Π is defined as
22
1 if
if
if
m Si
SiOm SiO
Si
SiNm SiN
Si
E EE
E EEE E EE
⎧=⎪
⎪⎪Π = =⎨⎪⎪ =⎪⎩
, (A-268)
where mE is the Young’s modulus in the local region.
203
APPENDIX B PROCESS TRAVELER
Wafer: 4” n-type (100) SOI, 3000 Å thick buried oxide layer, 1.5 mμ thick silicon device layer and a 350 mμ thick handle layer, 3 5 cm− Ω . Masks: Ground strap mask (GSM) N-Well mask (NWM) Piezoresistor mask (PRM) Piezoresistor contact mask (PCM) Metallization mask (MTM) Topside vent mask (TVM) Bond pad mask (BPM) Backside vent path mask (BVP) Backside cavity mask (BCM) Process Steps: 1) Etch down to handle wafer for ground strap
a) Acetone/Methanol/DI wash b) Coat HMDS for 5 min c) Coat resist (1 mμ , AZ1512) – spin at 4000 rpm for 50 sec d) Pre-exposure bake (Hot plate) – 95°C for 60 sec e) Align (MA6) using GSM for 9.7 sec, hard contact f) Develop (AZ 300MIF) for 60 sec g) Post exposure bake (Oven) - 95°C for 60 min h) Etch Si to BOX layer (DRIE) – recipe – BUF1 Ground strap mask i) Acetone/Methanol/DI wash to remove PR j) BOE (6:1) for 7 min k) Acetone/Methanol/DI wash
2) N-Well implant
a) Acetone/Methanol/DI wash b) Coat HMDS for 5 min c) Coat resist (1 mμ , AZ1512) – spin at 4000 rpm for 50 sec d) Pre-exposure bake (Hot plate) – 95°C for 60 sec e) Align (MA6) using NWM w.r.t. GSM for 9.7 sec, hard contact f) Develop (AZ 300MIF) for 60 sec g) Post exposure bake (Oven) - 95°C for 60 min h) Light O2 ash (RF-600W, O2-400sccms) for 60 sec i) Implant (Phosphorus, 20E keV= , 24 14Q e cm= ) at MEMS exchange j) Acetone bath to remove PR for 3 hours
204
3) Piezoresistors and Oxidation a) Acetone/Methanol/DI wash b) Coat HMDS for 5 min c) Coat resist (1 mμ , AZ1512) – spin at 4000 rpm for 50 sec d) Pre-exposure bake (Hot plate) – 95°C for 60 sec e) Align (MA6) using PRM w.r.t. GSM for 9.7 sec, hard contact f) Develop (AZ 300MIF) for 60 sec g) Post exposure bake (Oven) - 95°C for 60 min h) Light O2 ash (RF-600W, O2-400sccms) for 60 sec i) Implant (Boron, 5E keV= , 26 14Q e cm= ) at MEMS exchange j) Acetone/Methanol/DI wash to remove PR k) RCA clean at MEMS exchange l) Furnace anneal ( )2 , 1050N T C= ° for 300min at MEMS exchange
m) Dry/Wet/Dry oxidation ( )950T C= ° for 65 min, 21 min, and 65 min respectively n) Chemical/Mechanical Polish to remove backside oxide at ICEMOS
4) Piezoresistor contact cut
a) Acetone/Methanol/DI wash b) Coat HMDS for 5 min c) Coat resist (1 mμ , AZ1512) – spin at 4000 rpm for 50 sec d) Pre-exposure bake (Hot plate) – 95°C for 60 sec e) Align (MA6) using PCM w.r.t. GSM for 9.7 sec, hard contact f) Develop (AZ 300MIF) for 60 sec g) Post exposure bake (Oven) - 95°C for 60 min h) BOE (6:1) for 8 min i) Acetone/Methanol/DI wash
5) Metallization
a) Sputter 1 mμ Al (1%-Si) to avoid spiking using gun 3 for best results b) Acetone/Methanol/DI wash c) Coat HMDS for 5 min d) Coat resist (1 mμ , AZ1512) – spin at 4000 rpm for 50 sec e) Pre-exposure bake (Hot plate) – 95°C for 60 sec f) Align (MA6) using MTM w.r.t. PCM for 9.7 sec, hard contact g) Develop (AZ 300MIF) for 60 sec h) Post exposure bake (Oven) - 95°C for 60 min i) Aluminum etch (Ashland 16:1:1:2) - 40°C for 2 min j) Acetone/Methanol/DI wash
6) Nitride passivation and top vent etch
a) Acetone/Methanol/DI wash
205
b) Plasma enhanced chemical vapor deposition silicon nitride – recipe – MN300A c) Acetone/Methanol/DI wash d) Coat HMDS for 5 min e) Coat resist (1 mμ , AZ1512) – spin at 4000 rpm for 50 sec f) Pre-exposure bake (Hot plate) – 95°C for 60 sec g) Align (MA6) using TVM w.r.t. PCM for 9.7 sec, hard contact h) Develop (AZ 300MIF) for 60 sec i) Post exposure bake (Oven) - 95°C for 60 min j) Etch nitride layer (Unixaxis RIE/ICP) SF6 and O2 k) Etch oxide layer (Unixaxis RIE/ICP) CHF3 and O2 l) Etch Si to BOX layer (DRIE) – recipe – BUF1 Ground strap mask m) Acetone/Methanol/DI wash
7) Bond pad contact cut
a) Acetone/Methanol/DI wash b) Coat HMDS for 5 min c) Coat resist (1 mμ , AZ1512) – spin at 4000 rpm for 50 sec d) Pre-exposure bake (Hot plate) – 95°C for 60 sec e) Align (MA6) using BPM w.r.t. PCM for 9.7 sec, hard contact f) Develop (AZ 300MIF) for 60 sec g) Post exposure bake (Oven) - 95°C for 60 min h) Etch nitride layer (Unixaxis RIE/ICP) SF6 and O2 i) Acetone/Methanol/DI wash
8) Backside vent path
a) Acetone/Methanol/DI wash b) Coat HMDS on front side for 5 min c) Coat resist on front side for protection (10 mμ , AZ9260) – spin at 2000 rpm for 50 sec d) Pre-exposure bake (Oven) – 95°C for 15 min e) Coat HMDS on back side for 5 min f) Coat resist on back side (1 mμ , AZ1512) – spin at 4000 rpm for 50 sec g) Pre-exposure bake (Oven) – 95°C for 30 min h) Align front to back (EVG) using BVP w.r.t. GSM at Georgia Tech i) Develop (AZ 400MIF) 3:1 dilution with DI j) Post exposure bake (Oven) - 95°C for 60 min k) Etch Si (DRIE) 10 mμ - recipe – BUF1VTPH l) Acetone/Methanol/DI wash to remove PR from both sides m) Coat HMDS on carrier wafer for 5 min n) Coat resist on carrier wafer top surface (10 mμ , AZ9260) – spin at 2000 rpm for 50 sec o) Attach top side of wafer to carrier wafer p) Bake (Oven) 1– 95°C for 5 min
9) Backside cavity and vent through hole
206
a) Coat HMDS for 5 min b) Coat resist on back side (10 mμ , AZ9260) – spin at 2000 rpm for 50 sec c) Pre-exposure bake (Oven) – 95°C for 30 min d) Align (MA6) using BCM w.r.t. BVP for 150 sec, hard contact e) Develop (AZ 300MIF) for 6 min f) Post exposure bake (Oven) - 95°C for 60 min g) Etch Si (DRIE, SOI kit) 350 mμ - recipe – BUF1BAO2 h) Acetone/Methanol/DI wash i) Methanol spray, DI dip and BOE (6:1) etch for 8 min j) Release carrier wafer (AZ400K PR stripper) at 40°C
10) Forming gas anneal and back plate bond
a) Forming gas anneal (N2/H2, 450T C= ° ) for 60 min at MEMS exchange b) Acetone/Methanol/DI wash c) Coat resist on front side to protect Al(10 mμ , AZ9260) – spin at 2000 rpm for 50 sec d) RCA clean SOI and Pyrex e) Acetone/Methanol/DI wash to remove PR f) Anodic bond Pyrex wafer to backside of SOI wafer
207
APPENDIX C MATLAB FUNCTIONS
The following code is the objective function used in the optimization scheme to determine MDP.
function [Obj] = sensitivitybox(X) %%note this code is assuming only p type doping!!! % note this code assumes T=300K only!!!! % %-------------------------------------------------------------------------- % constants Nb = 1e15; kb = 1.3806503e-23; %[Nm/K] T=300; %[K] f2=1000.5; % [Hz] f1 = 999.5; % [Hz] % f2=1050; % [Hz] % f1 = 950; % [Hz] global hooge COUNT mode COUNT = COUNT + 1; global E nu wline zincr tincr N sigma0 DUB DLB PM bb H(1) = X(1)*(DUB(1)-DLB(1))+DLB(1); H(2) = X(2)*(DUB(2)-DLB(2))+DLB(2); H(3) = X(3)*(DUB(3)-DLB(3))+DLB(3); H(4) = sum(H); a = X(4)*(DUB(4)-DLB(4))+DLB(4); ah = a/H(4); Nsurf = X(5)*(DUB(5)-DLB(5))+DLB(5); zj = X(6)*(DUB(6)-DLB(6))+DLB(6); thetaa = X(7)*(DUB(7)-DLB(7))+DLB(7); raout = a; rain = X(8)*(DUB(4)-DLB(4))+DLB(4); thetat = X(9)*(DUB(9)-DLB(9))+DLB(9); rtout = a; rtin = X(10)*(DUB(4)-DLB(4))+DLB(4); if mode == 0 Vb = X(11)*(DUB(11)-DLB(11))+DLB(11); else if mode == 1 cur = X(11)*(DUB(11)-DLB(11))+DLB(11); end end dlmwrite('nondimX.txt', [COUNT X], '-append')
208
if mode == 0 dlmwrite('dimX.txt', [COUNT H(1) H(2) H(3) a Nsurf zj thetaa rain thetat rtin Vb], '-append') else if mode == 1 dlmwrite('dimX.txt', [COUNT H(1) H(2) H(3) a Nsurf zj thetaa rain thetat rtin cur], '-append') end end if rain >= a rain = a - wline; % COUNT errormessage = 1; dlmwrite('errors.txt', [COUNT errormessage], '-append') end if rtin >= a rtin = a - wline; % COUNT errormessage = 2; dlmwrite('errors.txt', [COUNT errormessage], '-append') end B11=-E(1)*H(1)^2/(2*(1-nu(1)^2))+E(2)*H(2)^2/(2*(1-nu(2)^2))+E(3)/(2*(1-nu(3)^2))*(H(3)^2+2*H(3)*H(2)); A11=E(1)*H(1)/(1-nu(1)^2)+E(2)*H(2)/(1-nu(2)^2)+E(3)*H(3)/(1-nu(3)^2); D11=E(1)*H(1)^3/(3*(1-nu(1)^2))+E(2)*H(2)^3/(3*(1-nu(2)^2))+E(3)/(3*(1-nu(3)^2))*(H(3)^3+3*H(3)*H(2)*(H(3)+H(2))); Dstar=D11-B11^2/A11; k=sqrt(sigma0*H(2)/Dstar)*a; p = PM(bb); % good check. set this to say 1000 and the results should be the same P = p*a^4/(2*H(4)*Dstar); z=[0:zj/(zincr-1):zj*(1)]; % from the surface (z=0) to the junction depth % Nr = Nsurf.*ones(1,length(z)); %creates a uniform profile Nr = Nsurf*(Nsurf/Nb).^(-(z/zj).^2); % creates a gaussian profile % constant parameters global q mumin mu0 alpha q=1.6e-19; etaNref=2.4; etamumin=-0.57; etaalpha=-.146; Nref=2.35e17;
209
mumin=54.3; mu0=406.9; etamu0=-2.23; alpha=.88; pi11=6.6e-11; pi12=-1.1e-11; pi44=138.1e-11; gamma=45*pi/180; % mean diaphram orienation relative to the crystal axis % calculates T variation (Pierret) mu=mumin+mu0./(1+(Nr./Nref).^alpha); %mobility [cm^2/(V*s)] rho=1./((q.*mu.*Nr).*100); %resistivity [ohms*m] % tangents of the arc and tapered resistor with reference to the crystal structure thetawt=2*log(a/rtin)*log(a/rain)/(thetaa); % to make the resistors have the same resistance thetawt is determined by this equation phia=[(gamma-thetaa/2):(thetaa/(tincr-1)):(gamma+thetaa/2)]; phit=[(gamma+(thetat-thetawt)/2):(thetawt/(tincr-1)):(gamma+(thetat+thetawt)/2)]; % direction cosines la(:,1)=cos(phia); la(:,2)=-sin(phia); ma(:,1)=sin(phia); ma(:,2)=cos(phia); na=zeros(length(phia),2); lt(:,1)=cos(phit); lt(:,2)=-sin(phit); mt(:,1)=sin(phit); mt(:,2)=cos(phit); nt=zeros(length(phit),2); % piezoresistive coeff for both arc and tapered resistors pila=pi11-2*(pi11-pi12-pi44)*(la(:,1).^2.*ma(:,1).^2+ma(:,1).^2.*na(:,1).^2+na(:,1).^2.*la(:,1).^2); pita=pi12+(pi11-pi12-pi44)*(la(:,1).^2.*la(:,2).^2+ma(:,1).^2.*ma(:,2).^2+na(:,1).^2.*na(:,2).^2); pilt=pi11-2*(pi11-pi12-pi44)*(lt(:,1).^2.*mt(:,1).^2+mt(:,1).^2.*nt(:,1).^2+nt(:,1).^2.*lt(:,1).^2); pitt=pi12+(pi11-pi12-pi44)*(lt(:,1).^2.*lt(:,2).^2+mt(:,1).^2.*mt(:,2).^2+nt(:,1).^2.*nt(:,2).^2); Pnt=.2014*log10(1.5e22./Nr); %calculates doping dependance R = find(Pnt > 1);
210
Pnt(R) = 1; for m=1:zincr PILa(:,m) = pila.*Pnt(m); PITa(:,m) = pita.*Pnt(m); PILt(:,m) = pilt.*Pnt(m); PITt(:,m) = pitt.*Pnt(m); end eta=-z/H(4); sig=1/ah; etana=B11/(H(4)*A11); ra=[rain:(a-rain)/(N-1):a]; rt=[rtin:(a-rtin)/(N-1):a]; rand=ra/a; rtnd=rt/a; for m=1:zincr for M=1:N sigmara(M,m) = -E(1)*sig^2*P/k^2*(eta(m)-etana)./(1-nu(1)^2).*((1+nu(1))+(1-nu(1))/rand(M).*besselj(1,k*rand(M))./besselj(1,k)-k*besselj(0,k*rand(M))./besselj(1,k)); sigmata(M,m) = -E(1)*sig^2*P/k^2*(eta(m)-etana)./(1-nu(1)^2).*((1+nu(1))-(1-nu(1))/rand(M).*besselj(1,k*rand(M))./besselj(1,k)-nu(1)*k*besselj(0,k*rand(M))./besselj(1,k)); sigmart(M,m) = -E(1)*sig^2*P/k^2*(eta(m)-etana)./(1-nu(1)^2).*((1+nu(1))+(1-nu(1))/rtnd(M).*besselj(1,k*rtnd(M))./besselj(1,k)-k*besselj(0,k*rtnd(M))./besselj(1,k)); sigmatt(M,m) = -E(1)*sig^2*P/k^2*(eta(m)-etana)./(1-nu(1)^2).*((1+nu(1))-(1-nu(1))/rtnd(M).*besselj(1,k*rtnd(M))./besselj(1,k)-nu(1)*k*besselj(0,k*rtnd(M))./besselj(1,k)); % sigmara(M,m) = 0; % sigmata(M,m) = 0; % sigmart(M,m) = 0; % sigmatt(M,m) = 0; for B=1:tincr integrandt(M,B,m)=1/(rho(m)*(1+sigmart(M,m).*PILt(B,m)+sigmatt(M,m).*PITt(B,m))); integranda(M,B,m)=1/(rho(m)*(1+sigmara(M,m).*PITa(B,m)+sigmata(M,m).*PILa(B,m))); Ra(:,B)=ra(:); end end end RplusDRtea=2*trapz(rt',1./((trapz(phit,trapz(z,integrandt,3),2)).*rt')); RplusDRarc=trapz(phia,1./(trapz(ra,(trapz(z,integranda,3)./Ra),1))); Restea = 2*log(a/rtin)/(thetawt)*(1./trapz(z,1./rho));
211
Resarc = (thetaa)/log(a/rain)*(1./trapz(z,1./rho)); DRtea = RplusDRtea - Restea; DRarc = RplusDRarc - Resarc; global DynRange Resistance Sensitivity Vnoise MDP if mode == 0 DVo=((DRarc-DRtea)/(2*Resarc+DRarc+DRtea)).*(Vb*1e6); else if mode ==1 DVo= (cur*1e6)/2*(DRarc-DRtea); end end Sens=DVo/p; Sensitivity = Sens; Resistance = Resarc; Na = trapz(z,Nr)*1e6*(.5*thetaa*(raout^2-rain^2)); %carrier concentration times the volume % Na = Nr*1e6*(.5*thetaa*zj*(raout^2-rain^2)); %carrier concentration times the volume Nt = trapz(z,Nr)*1e6*(thetawt*(rtout^2-rtin^2)); %carrier concentration times the volume % Nt = Nr*1e6*(thetawt*zj*(rtout^2-rtin^2)); %carrier concentration times the volume % Vn = sqrt(4*kb*T*Resarc*(f2-f1)+1/8*hooge*Vb^2*(1/Na+1/Nt)*log(f2/f1))*1e6; if mode == 0 Vn = sqrt(4*kb*T*Resarc*(f2-f1)+1/8*hooge*Vb^2*(1/Na+1/Nt)*log(f2/f1)+(4e-9*(f2-f1))^2)*1e6; else if mode == 1 Vn = sqrt(4*kb*T*Resarc*(f2-f1)+1/8*hooge*(cur*Resarc)^2*(1/Na+1/Nt)*log(f2/f1)+(4e-9*(f2-f1))^2)*1e6; end end Vnoise=Vn; MDP = 20*log10((Vn/Sens)/(20e-6)); Obj = MDP; DynRange = 20*log10(p/(Vn/Sens)); data = [COUNT DynRange Resistance Sensitivity Vnoise MDP k thetawt]; dlmwrite('objectives.txt', data, '-append')
212
APPENDIX D OPTIMIZED DEVICES
213
Table D-1. Optimized devices operating on a current source (10mA) with a Gaussian dopant profile and Nb = 1e15 [#/cm3]. Bandwidth Pmax MDP Dyn Range H1 H2 H3 a a/h D* Nsurf zj θa rain θt θwt rtin warc larc wtap ltap wgap Ra Isupply Sens Noise FL k Power Pmaxact D. Rg. Act Act. BW
Table D-2. Optimized devices operating on a voltage source with a Gaussian dopant profile and Nb = 1e15 [#/cm3]. Bandwidth Pmax MDP Dyn Range H1 H2 H3 a a/h D* Na zj θa rain θt θwt rtin warc larc wtap ltap wgap Ra Vb Sens Noise FL k Power Pmaxact D. Rg. Act Act. BW
Since the devices are batch fabricated to yield identical devices, the devices tested for input
and output resistance values which are used to determine a mean and standard deviation. In
addition the confidence intervals for the true mean and standard deviations are calculated from
the following equations [126]
; / 2 ; / 2n nx
st stx x
N Nα αμ
⎡ ⎤− ≤ < +⎢ ⎥
⎣ ⎦, (F-1)
and 2 2
22 2; / 2 ;1 / 2
xn n
ns ns
α α
σχ χ −
⎡ ⎤≤ <⎢ ⎥
⎢ ⎥⎣ ⎦, (F-2)
where N is the number of samples, 1n N= − , t is from the t distribution table, χ is from the
χ distribution table, x is the sample mean and s is the sample standard deviation.
Frequency Response Function
The uncertainty in the frequency response function is comprised of random error in the
measurement as well as bias error due to the analog to digital conversion. For this analysis, it is
assumed that there is no error in the reference microphone used in the measurements. The
normalized random error for the magnitude frequency response function is calculated from the
following equation [126]
( )( )1/ 22ˆ. . 1ˆ
ˆ 2xy xy
r xyxy dxy
s d H fH
nH
γε
γ
⎡ ⎤ −⎣ ⎦⎡ ⎤ = ≈⎣ ⎦ . (F-3)
The standard deviation for the phase in radians is calculated using [126]
( )( )1/ 221ˆ. .2
xyxy
xy d
fs d
n
γφ
γ
−⎡ ⎤ ≈⎣ ⎦ . (F-4)
220
The 95% CI are then calculated by multiplying the standard deviation by 2 since the number of
samples is greater then 31.
In analog to digital conversion, the magnitude of each data value must be put into a
discrete digital bin. The bias error associated with this is defined as ½ of the bin size. For each
experiment it is important to note the scale for the signal input. For example with the pulse
analyzer, if the signal input was scaled to 707.1mV, the range of discretation is from -707.1mV
to 707.1mV for a range of 1.414V. This system has 16 bits so the bin with is 21.57μV.
Therefore the bias error is ½ of the bin width or 10.79μV. To determine the bias error in the
frequency response function this voltage error is divided by the known incident pressure from the
reference microphone. Figure 6-18 includes the bias error and it is shown be small relative to the
random error at lower sound pressure levels.
Hooge Parameter
The uncertainty of the Hooge parameter is propagated from the uncertainty in the
measurement, the uncertainty in the least squares fit ( )b from equation (6-9) and the uncertainty
in the total number of carriers ( )N . The error in the noise measurement is assumed negligible
due to the large number of averages in the data (2400). The uncertainty is then calculated from
the other two sources. The total number of carriers for an arc resistor is found by assuming a
Gaussian profile with the following equation,
( )2
2 2
0
12
jj
xz z
sa a aout ain s
b
NN r r N dxN
θ
⎛ ⎞−⎜ ⎟⎜ ⎟
⎝ ⎠⎛ ⎞= − ⎜ ⎟
⎝ ⎠∫ . (F-5)
Therefore the uncertainty of aN is related to the uncertainty of j, , , z , ,a aout ain sr r Nθ and
bN . To avoid calculating all of the partial differentials a Monte Carlo simulation is done to
221
calculate the standard deviation of aN . The standard deviations of the independent variables are
taken from Table 6-12. The Monte Carlo simulation was run with 1,000,000 iterations to yield
1.56 9N eσ = . (F-6)
The standard deviation of the Hooge parameter is calculated from the following equation
2 2
N bN bαα ασ σ σ∂ ∂⎛ ⎞ ⎛ ⎞= +⎜ ⎟ ⎜ ⎟∂ ∂⎝ ⎠ ⎝ ⎠
(F-7)
where
10b
Nα∂
=∂
(F-8)
and
110bbNbα −∂
=∂
(F-9)
and bσ is obtained from the least squares fit.
222
LIST OF REFERENCES
[1] FAA, "Aeronautics and space, noise standards: Aircraft type and airworthyness certification," Title 14 US Code of Federal Regulations, Part 36, 2004.
[2] J. Hawk, "The Boeing 787 Dreamliner: More Than an Airplane," in AIAA Aircraft Noise
and Emissions Reduction Symposium, Monterey, California, 2005. [3] Boeing, "Ssshh, we're flying a plane around here," in Boeing Frontiers. vol. 4, 2006. [4] B&K, "Microphone Handbook," Naerum, Denmark 1996. [5] A. V. Oppenheim, A. S. Willsky, and S. H. Nawab, Signals and Systems, 2nd ed. ed.
Upper Saddle River, NJ: Prentice Hall, 1997. [6] T. B. Gabrielson, "Mechancial-thermal noise in micromachined acoustic and vibration
sensors," IEEE Transactions on Electron Devices, vol. vol. 40, pp. 903-909, 1993. [7] A. van der Ziel, Noise in Solid State Devices and Circuits: John Wiley & Sons, Inc, 1986. [8] J. Eargle, The Microphone Handbook: Butterworth-Heinemann, 2001. [9] M. Davenport and S. J. Hannahs, Introducing phonetics and phonology. New York, NY:
Oxford University Press, 1998. [10] T. J. Mueller, Aeroacoustic Measurements, Berlin, Germany: Springer-Verlag, 2002, pp.
158-179. [11] J. R. Underbrink and R. P. Dougherty, "Array design for non-intrusive measurement of
noise sources," in National conference on Noise Control Engineering Seattle, WA: Institute of Noise Control Engineering, 1996.
[12] B&K, B&K Falcon Range of Microphone Products, Microphone Handbook, 1999. [13] Kulite, "Kulite Miniture IS Silicon Diaphragm Pressure Transducers," Leonia, NJ 1999. [14] Skoog, Holler, and Crouch, Principles of Instrumental Analysis, 6th Edition ed, 2007, p.
9. [15] W. G. Cady, Piezoelectricity. New York: Dover Publications, Inc., 1964. [16] S. Senturia, Microsystem Design. Norwell, MA.: Kluwer and Sons, Inc., 2001. [17] C. S. Smith, "Piezoresistance Effect in Germanium and Silicon," Physical Review, vol.
94, pp. 42-49, 1954. [18] R. F. Pierret, Semiconductor Device Fundamentals: Addison-Wesley, 1996.
223
[19] Y. Kanda, "A graphical representation of the piezoresistance coefficients in silicon,"
IEEE Transactions on Electron Devices, vol. 29, p. 64, 1982. [20] R. K. Wangsness, Electromagnetic Fields, 2nd ed.: John Wiley & Sons, 1986. [21] P. R. Scheeper, A. G. H. Vanderdonk, W. Olthuis, and P. Bergveld, "A Review of Silicon
Microphones," Sensors and Actuators a-Physical, vol. 44, pp. 1-11, Jul 1994. [22] S. D. Senturia, Microsystem Design. Norwell, MA: Kluwer Academic Publishers, 2001. [23] D. Garthe and J. Kobiela, "Development of an integrated optical microphone by means of
waveguide structuring on PMMA," in SPIE, 1985, pp. 618-629. [24] G. He and F. W. Cuomo, "The Analysis of Noises in a Fiber Optic Microphone," Journal
of the Acoustical Society of America, vol. 92, pp. 2521-2526, Nov 1992. [25] N. Bilaniuk, "Optical microphone transduction techniques," Applied Acoustics, vol. 50,
pp. 35-63, Jan 1997. [26] R. O. Cook and C. W. Hamm, "Fiber Optic Lever Displacement Transducer," Applied
Optics, vol. 18, pp. 3230-3241, 1979. [27] J. Medill, "A miniature piezoelectric microhpone," Journal of the Acoustical Society of
America, vol. 25, pp. 864-866, 1953. [28] M. Royer, J. O. Holmen, M. A. Wurm, O. S. Aadland, and M. Glenn, "ZnO on Si
integrated acoustic sensor," Sensors and Actuators, vol. 4, pp. 357-362, 1983. [29] E. S. Kim, R. S. Muller, and P. R. Gray, "Integrated microphone with CMOS circuits on
a single chip," in International Electron Devices Meeting, 1989, pp. 880-883. [30] E. S. Kim, J. R. Kim, and R. S. Muller, "Improved IC-compatible piezoelectric
microphone and CMOS process," in International Conference on Solid-State Sensors and Actuators, 1991, pp. 270-273.
[31] R. Schellin, G. Hess, W. Kuehnel, G. M. Sessler, and E. Fukada, "Silicon subminiature
microphones with organic piezoelectric layers-fabrication and acoustical behavior," Electrical Insulation, IEEE Transactions on [see also Dielectrics and Electrical Insulation, IEEE Transactions on], vol. 27, pp. 867-871, 1992.
[32] R. P. Ried, E. S. Kim, D. M. Hong, and R. S. Muller, "Piezoelectric microphone with on-
[33] S. C. Ko, Y. C. Kim, S. S. Lee, S. H. Choi, and S. R. Kim, "Micromachined piezoelectric membrane acoustic device," Sensors and Actuators A: Physical, vol. 103, pp. 130-134, 2003.
[34] M. N. Niu and E. S. Kim, "Piezoelectric bimorph microphone built on micromachined
parylene diaphragm," Journal of Microelectromechanical Systems, vol. 12, pp. 892-898, Dec 2003.
[35] H. J. Zhao, T. L. Ren, J. S. Liu, L. T. Liu, and Z. J. Li, "Fabrication of high-quality PZT-
based piezoelectric microphone," in 12th International Conference on Solid-State Sensors, Actuators and Microsystems, Boston, MA, 2003, pp. 234-237.
[36] J. Hillenbrand and G. M. Sessler, "High-sensitivity piezoelectric microphones based on
stacked cellular polymer films," Journal of the Acoustical Society of America, vol. 116, pp. 3267-3270, Dec 2004.
[37] S. Horowitz, T. Nishida, L. Cattafesta, and M. Sheplak, "Development of a
micromachined piezoelectric microphone for aeroacoustics applications," Journal of the Acoustical Society of America, vol. 122, pp. 3428-3436, 2007.
[38] F. P. Burns, "Piezoresistive Semiconductor Microphone," Journal of the Acoustical
Society of America, vol. 29, pp. 248-253, 1957. [39] S. Samaun, K. Wise, E. Nielsen, and J. Angell, "An IC piezoresistive pressure sensor for
biomedical instrumentation," 1971, pp. 104-105. [40] R. Schellin and G. Hess, "A silicon subminiature microphone based on piezoresistive
polysilicon strain gauges," Sensors and Actuators A: Physical, vol. 32, pp. 555-559, 1992.
[41] E. Kalvesten, L. Lofdahl, and G. Stemme, "A small-size silicon microphone for
measurements in turbulent gas flows," Sensors and Actuators A: Physical, vol. 45, pp. 103-108, 1994.
[42] R. Schellin, M. Strecker, U. Nothelfer, and G. Schuster, "Low pressure acoustic sensors
for airborne sound with piezoresistive monocrystalline silicon and electrochemically etched diaphragms," Sensors and Actuators A: Physical, vol. 46, pp. 156-160, 1995.
[43] M. Sheplak, K. S. Breuer, and M. A. Schmidt, "A wafer-bonded, silicon nitride
membrane microphone with dielectrically isolated single crystal silicon piezoresistors," in Solid-State Sensor and Actuator Workshop, Hilton Head Island, SC, 1998, pp. 23-26.
[44] A. Naguib, E. Soupos, H. Nagib, C. Huang, and K. Kajafi, "A piezoresistive mems sensor
for acoustic noise measurements," in 5th AIAA/CEAS Aeroacoustics Conference and Exhibit, AIAA Paper #99-1992 Bellevue, WA: AIAA, 1999.
225
[45] D. P. Arnold, S. Gururaj, S. Bhardwaj, T. Nishida, and M. Sheplak, "A piezoresistive microphone for aeroacoustic measurements," in 2001 ASME International Mechanical Engineering Congress and Exposition, New York, NY, 2001, pp. 281-288.
[46] C. Huang, A. Naguib, E. Soupos, and K. Najafi, "A silicon micromachined microphone
for fluid mechanics research," Journal of Micromechanics and Microengineering, vol. 12, pp. 767-774, 2002.
[47] G. Li, Y. Zohar, and W. Man, "Piezoresistive microphone with integrated amplifier
realized using metal-induced laterally crystallized polycrystalline silicon," in IEEE International Conference on Micro Electro Mechanical Systems, 2004, pp. 548-551.
[48] W. H. Ko, M. H. Bao, and Y. D. Hong, "A High-Sensitivity Integrated-Circuit Capacitive
Pressure Transducer," IEEE Transactions on Electron Devices, vol. 29, pp. 48-56, 1982. [49] D. Hohm and R. Gerh, "Silicon-Dioxide Electret Transducer," Journal of the Acoustical
Society of America, vol. 75, pp. 1297-1298, 1984. [50] D. Hohm and G. Hess, "A Subminiature Condenser Microphone with Silicon-Nitride
Membrane and Silicon Back Plate," Journal of the Acoustical Society of America, vol. 85, pp. 476-480, Jan 1989.
[51] J. Bergqvist, F. Rudolf, and J. Doe, "A New Condenser Microphone in Silicon," Sensors
and Actuators a-Physical, vol. 21, pp. 123-125, Feb 1990. [52] B. T. Cunningham, J. J. Bernstein, and J. Doe, "Wide bandwidth silicon nitride
membrane microphones," in Conference on Micromachining and Microfabrication Process Technology III, Austin, TX, 1997, pp. 56-63.
[53] P. Rombach, M. Mullenborn, U. Klein, and K. Rasmussen, "The first low voltage, low
noise differential condenser silicon microphone," in The 14th European Conference on Solid-State Transducers, EUROSENSORS XIV, Copenhagen, Denmark, 2000, pp. 213-216.
[54] P. R. Scheeper, B. Nordstrand, J. O. Gullov, B. Liu, T. Clausen, L. Midjord, and T.
Storgaard-Larsen, "A new measurement microphone based on MEMS technology," Journal of Microelectromechanical Systems, vol. 12, pp. 880-891, Dec 2003.
[55] D. T. Martin, K. Kadirvel, T. Nishida, and M. Sheplak, "An Instrumentation Grade
MEMS Condenser Microphone for Aeroacoustic Measurements," in 46th AIAA Aerospace Sciences Meeting and Exhibit Reno, NV, 2008.
[56] Analog, "ADMP410: iMEMS Microphone Data Sheet," Norwood, MA 2008. [57] D. Garthe, "A fiber-optic microphone," Sensors and Actuators A: Physical, vol. 26, pp.
341-345, 1991.
226
[58] J. A. Dziuban, A. Goreckadrzazga, and U. Lipowicz, "Silicon Optical Pressure Sensor,"
Sensors and Actuators a-Physical, vol. 32, pp. 628-631, Apr 1992. [59] M. A. Chan, S. D. Collins, and R. L. Smith, "A Micromachined Pressure Sensor with
Fiberoptic Interferometric Readout," Sensors and Actuators a-Physical, vol. 43, pp. 196-201, May 1994.
[60] A. Kots and A. Paritsky, "Fiber optic microphone for harsh environment," in SPIE,
Boston, MA, 1999. [61] D. C. Abeysinghe, S. Dasgupta, J. T. Boyd, and H. E. Jackson, "A novel MEMS pressure
sensor fabricated on an optical fiber," Photonics Technology Letters, IEEE, vol. 13, pp. 993-995, 2001.
[62] K. Kadirvel, R. Taylor, S. Horowitz, S. Hunt, M. Sheplak, and T. Nishida, "Design and
characterization of MEMS optical microphone for aeroacoustic measurements," in 42nd AIAA Aerospace Sciences Meeting and Exhibit, AIAA paper #2004-1030, Reno, NV, 2004.
[63] J. A. Bucaro, N. Lagakos, B. H. Houston, J. Jarzynski, and M. Zalalutdinov, "Miniature,
high performance, low-cost fiber optic microphone," Journal of the Acoustical Society of America, vol. 118, pp. 1406-1413, Sep 2005.
[64] N. A. Hall, B. Bicen, M. K. Jeelani, W. Lee, S. Qureshi, F. L. Degertekin, and M.
Okandan, "Micromachined microphones with diffraction-based optical displacement detection," Journal of the Acoustical Society of America, vol. 118, pp. 3000-3009, Nov 2005.
[65] J. H. Song, H. M. Gu, H. J. Park, and S. S. Lee, "Optical microphone based on a
reflective micromirror diaphragm," Microwave and Optical Technology Letters, vol. 48, pp. 707-709, 2006.
[66] N. A. Hall, M. Okandan, R. Littrell, B. Bicen, and F. L. Degertekin, "Micromachined
optical microphone structures with low thermal-mechanical noise levels," Journal of the Acoustical Society of America, vol. 122, pp. 2031-2037, Oct 2007.
[67] E. S. Kim and R. S. Muller, "IC-Processed piezoelectric microphone," Electron Device
Letters, IEEE, vol. 8, pp. 467-468, 1987. [68] R. Schellin, G. Hess, R. Kressmann, and P. Wassmuth, "Micromachined silicon
subminiature microphones with piezoelectric P(VDF/TRFE)-layers and silicon-nitride-membranes," in 8th International Symposium on Electrets, 1994, pp. 1004-1009.
[69] S. S. Lee, R. P. Ried, and R. M. White, "Piezoelectric cantilever microphone and
microspeaker," Microelectromechanical Systems, Journal of, vol. 5, pp. 238-242, 1996.
227
[70] S. S. Lee and R. M. White, "Piezoelectric cantilever acoustic transducer," Journal of
Micromechanics and Microengineering, vol. 8, pp. 230-238, Sep 1998. [71] Y. Yang, T. L. Ren, and L. T. Liu, "Design of a novel piezoelectric acoustic sensor using
surface micromachining," Integrated Ferroelectrics, vol. 80, pp. 363-372, 2006. [72] W. H. Ko, J. Hynecek, and S. F. Boettcher, "Development of a miniature pressure
transducer for biomedical applications," IEEE Transactions on Electron Devices, vol. 26, pp. 1896-1905, 1979.
[73] E. Kalvesten, L. Lofdahl, and G. Stemme, "Analytical characterization of piezoresistive
square-diaphragm silicon microphone," Sensors and Materials, vol. 8, pp. 113-136, 1996. [74] E. Kalvesten, L. Lofdahl, and G. Stemme, "Small piezoresistive silicon microphones
specially designed for the characterization of turbulent gas flows," Sensors and Actuators A: Physical, vol. 46, pp. 151-155, 1995.
[75] M. Sheplak, M. Seiner, K. S. Breuer, and M. A. Schmidt, "A MEMS microphone for
aeroacoustic measurements," in 37th AIAA Aerospace Sciences Meeting, AIAA Paper #99-0606 Reno, NV: AIAA, 1999.
[76] A. Naguib, E. Soupos, H. Nagib, C. Huang, and K. Kajafi, "Characterization of a mems
acoustic/pressure sensor," in 37th AIAA Aerospace Sciences Meeting, AIAA Paper #99-0520 Reno, NV: AIAA, 1999.
[77] A. J. Sprenkels, R. A. Groothengel, A. J. Verloop, and P. Bergveld, "Development of an
Electret Microphone in Silicon," Sensors and Actuators, vol. 17, pp. 509-512, May 1989. [78] J. A. Voorthuyzen, P. Bergveld, and A. J. Sprenkels, "Semiconductor-Based Electret
Sensors for Sound and Pressure," IEEE Transactions on Electrical Insulation, vol. 24, pp. 267-276, Apr 1989.
[79] P. Murphy, K. Hubschi, N. Derooij, and C. Racine, "Subminiature Silicon Integrated
Electret Capacitor Microphone," IEEE Transactions on Electrical Insulation, vol. 24, pp. 495-498, Jun 1989.
[80] J. Bergqvist, F. Rudolf, J. Maisano, F. Parodi, and M. Rossi, "A silicon condenser
microphone with a highly perforated backplate," in International Conference on Solid-State Sensors and Actuators, Piscataway, NJ, 1991, pp. 266-269.
[81] P. R. Scheeper, W. Olthuis, and P. Bergveld, "Fabrication of a subminiature silicon
condenser microphone using the sacrificial layer technique," in International Conference on Solid-State Sensors and Actuators, Piscataway, NJ, 1991, pp. 408-411.
228
[82] P. R. Scheeper, A. G. H. van der Donk, W. Olthuis, and P. Bergveld, "Fabrication of silicon condenser microphones using single wafer technology," IEEE Journal of Microelectromechanical Systems, vol. 1, pp. 147-154, 1992.
[83] P. R. Scheeper, W. Olthuis, and P. Bergveld, "Improvement of the Performance of
Microphones with a Silicon-Nitride Diaphragm and Backplate," Sensors and Actuators a-Physical, vol. 40, pp. 179-186, Mar 1994.
[84] W. Kuhnel and G. Hess, "A Silicon Condenser Microphone with Structured Back Plate
and Silicon-Nitride Membrane," Sensors and Actuators a-Physical, vol. 30, pp. 251-258, Feb 1992.
[85] W. Kuhnel and G. Hess, "Micromachined Subminiature Condenser Microphones in
Silicon," Sensors and Actuators a-Physical, vol. 32, pp. 560-564, Apr 1992. [86] T. Bourouina, S. Spirkovitch, F. Baillieu, and C. Vauge, "A New Condenser Microphone
with a P+ Silicon Membrane," Sensors and Actuators a-Physical, vol. 31, pp. 149-152, Mar 1992.
[87] E. Graf, W. Kronast, S. Duhring, B. Muller, and A. Stoffel, "Silicon Membrane
Condenser Microphone with Integrated Field-Effect Transistor," Sensors and Actuators a-Physical, vol. 37-8, pp. 708-711, Jun-Aug 1993.
[88] J. Bergqvist, J. Gobet, and J. Doe, "Capacitive Microphone with a Surface
Micromachined Backplate Using Electroplating Technology," Journal of Microelectromechanical Systems, vol. 3, pp. 69-75, Jun 1994.
[89] J. J. Bernstein, J. T. Borenstein, and J. Doe, "A micromachined silicon condenser
microphone with on-chip amplifier," in Solid-State Sensor and Actuator Workshop, Hilton Head, SC, 1996, pp. 239-243.
[90] Q. B. Zou, Z. J. Li, and L. T. Liu, "Design and fabrication of silicon condenser
microphone using corrugated diaphragm technique," Journal of Microelectromechanical Systems, vol. 5, pp. 197-204, Sep 1996.
[91] Q. B. Zou, Z. J. Li, and L. T. Liu, "Theoretical and experimental studies of single-chip-
processed miniature silicon condenser microphone with corrugated diaphragm," Sensors and Actuators a-Physical, vol. 63, pp. 209-215, Dec 1997.
[92] Y. B. Ning, A. W. Mitchell, and R. N. Tait, "Fabrication of a silicon micromachined
capacitive microphone using a dry-etch process," Sensors and Actuators a-Physical, vol. 53, pp. 237-242, May 1996.
[93] M. Pedersen, W. Olthuis, and P. Bergveld, "A silicon condenser microphone with
polyimide diaphragm and backplate," Sensors and Actuators a-Physical, vol. 63, pp. 97-104, Oct 1997.
229
[94] M. Pedersen, W. Olthuis, and P. Bergveld, "High-performance condenser microphone
with fully integrated CMOS amplifier and DC-DC voltage converter," Journal of Microelectromechanical Systems, vol. 7, pp. 387-394, Dec 1998.
[95] P. C. Hsu, C. H. Mastrangelo, and K. D. Wise, "A high sensitivity polysilicon diaphragm
condenser microphone," in IEEE Micro Electro Mechanical Systems (MEMS) Conference, Piscataway, NJ, 1998, pp. 56-63.
[96] M. Pedersen, W. Olthuis, and P. Bergveld, "An integrated silicon capacitive microphone
with frequency-modulated digital output," Sensors and Actuators a-Physical, vol. 69, pp. 267-275, Sep 1998.
[97] D. Schafer, S. Shoaf, and P. Loeppert, "Micromachined condenser microphone for
hearing aid use," in Solid-State Sensor and Actuator Workshop, Hilton Head, SC, 1998, pp. 27-30.
[98] A. Torkkeli, O. Rusanen, J. Saarilahti, H. Seppa, H. Sipola, and J. Hietanen, "Capacitive
microphone with low-stress polysilicon membrane and high-stress polysilicon backplate," Sensors and Actuators a-Physical, vol. 85, pp. 116-123, Aug 2000.
[99] P. Rombach, M. Mullenborn, U. Klein, and K. Rasmussen, "The first low voltage, low
noise differential silicon microphone, technology development and measurement results," Sensors and Actuators a-Physical, vol. 95, pp. 196-201, Jan 2002.
[100] X. X. Li, R. M. Lin, H. Kek, J. M. Miao, and Q. B. Zou, "Sensitivity-improved silicon
condenser microphone with a novel single deeply corrugated diaphragm," Sensors and Actuators a-Physical, vol. 92, pp. 257-262, Aug 2001.
[101] R. Kressmann, M. Klaiber, and G. Hess, "Silicon condenser microphones with corrugated
silicon oxide/nitride electret membranes," Sensors and Actuators a-Physical, vol. 100, pp. 301-309, Sep 2002.
[102] J. J. Neumann, K. J. Gabriel, and J. Doe, "A fully-integrated CMOS-MEMS audio
microphone," in 12th International Conference on Solid-State Sensors, Actuators and Microsystems, Boston, MA, 2003, pp. 230-233.
[103] S. T. Hansen, A. S. Ergun, W. Liou, B. A. Auld, and B. T. Khuri-Yakub, "Wideband
micromachined capacitive microphones with radio frequency detection," Journal of the Acoustical Society of America, vol. 116, pp. 828-842, Aug 2004.
[104] D. T. Martin, K. Kadirvel, J. Liu, R. M. Fox, M. Sheplak, and T. Nishida, "Surface and
Bulk Micromachined Dual Back-Plate Condenser Microphone," in 18th IEEE International Conference on Micro Electro Mechanical Systems (MEMS 2005), Miami, FL, 2005.
230
[105] Knolwes, "SP0102NC3-2 Product Data Sheet," Itasca, IL 2007. [106] D. Garthe, "Fiber- and integrated-optical microphones based on intensity modulation by
beam deflection at a moving membrane," Sensors and Actuators A: Physical, vol. 37-38, pp. 484-488, 1993.
[107] D. Damjanovic and G. Robert, Piezoelectric Materials in Devices: Extended Reviews on
Current and Emerging Piezoelectric Materials, Technology, and Applications, N. Setter, Ed. Lausanne, Switzerland: EPFL Swiss Federal Institute of Technology, 2002, p. 518.
[108] H. L. Chau and K. D. Wise, "Scaling Limits in Batch-Fabricated Silicon Pressure
Sensors," Ieee Transactions on Electron Devices, vol. 34, pp. 850-858, Apr 1987. [109] J. A. Voorthuyzen and P. Bergveld, "The Influence of Tensile Forces on the Deflection of
Circular Diaphragms in Pressure Sensors," Sensors and Actuators, vol. 6, pp. 201-213, 1984.
[110] M. Sheplak and J. Dugundji, "Large deflections of clamped circular plates under initial
tension and transitions to membrane behavior," Journal of Applied Mechanics-Transactions of the Asme, vol. 65, pp. 107-115, Mar 1998.
[111] J. N. Reddy, Theory and Analysis of Elastic Plates. Philadelphia: Taylor & Francis, 1999. [112] ABAQUSv6.5, "ABAQUS Analysis User's Manual Section 15.6.9: Axisymmetric Shell
Element Library.," 2002. [113] S. M. Sze, Semiconductor Sensors. New York: Wiley-Interscience, 1994. [114] A. P. Boresi and R. J. Schmidt, Advanced Mechanics of Materials, 6 ed. Hoboken, NJ:
John Wiley and Sons, Inc., 2003. [115] W. P. Mason, J. J. Forst, and L. M. Tornillo, "Recent developments in semiconductor
strain transducers," New York, 1962, pp. 110-120. [116] D. R. Kerr and A. G. Milnes, "Piezoresistance of Diffused Layers in Cubic
Semiconductors," Journal of Applied Physics, vol. 34, pp. 727-731, 1963. [117] O. N. Tufte and E. L. Stelzer, "Piezoresistive Properties of Silicon Diffused Layers,"
Journal of Applied Physics, vol. 34, pp. 313-318, 1963. [118] J. A. Harley and T. W. Kenny, "1/F Noise Considerations for the Design and Process
Optimization of Piezoresistive Cantilevers," Journal of Microelectromechanical Systems, vol. 9, pp. 226-235, 2000.
[119] R. C. Jaeger, Introduction to Microelectronic Fabrication, Second ed. vol. V. Upper
Saddle River, NJ: Prentice Hall, 2002.
231
[120] M. Rossi, Acoustics and Electroacoustics. Norwood, MA: Artech House, 1988. [121] F. A. Fischer, Fundamentals of electroacoustics. New York,: Interscience Publishers,
1955. [122] F. V. Hunt, Electroacoustics; the analysis of transduction, and its historical background.
Cambridge,: Harvard University Press, 1954. [123] H. A. Tilmans, "Equivalent circuit representation of electromechanical transducers .1.
Lumped-parameter systems," Journal of Micromechanics and Microengineering, vol. 6, pp. 157-176, Mar 1996.
[124] H. A. Tilmans, "Equivalent circuit representation of electromechanical transducers: II.
Distributed-parameter systems," Journal of Micromechanics and Microengineering, vol. 7, pp. 285-309, Dec 1997.
[125] S. Prasad, Gallas, Q., Horowitz, S. ,Homeijer, B. Sankar, B., Cattafesta, L., and Sheplak,
M., "An Analytical Electroacoustic Model of a Piezoelectric Composite Circular Plate," AIAA Journal (accepted), 2006.
[126] J. S. Bendat and A. G. Piersol, Random data : analysis and measurement procedures, 3rd
ed. New York: Wiley, 2000. [127] D. T. Blackstock, Fundamentals of Physical Acoustics: John Wiley & Sons, Inc., 2000. [128] R. W. Fox, and McDonald, A.T., Fluid Mechanics, 6th ed. New York: John Wiley &
Sons, Inc., 2001. [129] R. Dieme, G. Bosman, M. Sheplak, and T. Nishida, "Source of excess noise in silicon
piezoresistive microphones," Journal of the Acoustical Society of America, vol. 119, pp. 2710-2720, 2006.
[130] H. Nyquist, "Thermal Agitation of Electric Charge in Conductors," Physical Review, vol.
32, p. 110, 1928. [131] F. N. Hooge, "1/f Noise is no Surface Effect," Physics Letters A, vol. 29, pp. 139-140,
1969. [132] M. D. Sheplak, J., "Large Deflections of Clamped Circular Plates Under Initial Tension
and Transitions to Membrane Behavior," Journal of Applied Mechanics, vol. 65, pp. 107-115, March 1998 1998.
[133] S. A. Marglin, Public Investment Criteria. Cambridge, MA: M.I.T. Press, 1967.
232
[134] V. Chandrasekaran, "Thermoelastically Actuated Acoustic Proximity Sensor with Integrated Electrical Through-Wafer Interconnects." University of Florida, Mechanical and Aerospace Engineering, 2004.
[135] B. G. Dieme R., Sheplak M., and Toshikazu N., "Sources of excess noise in silicon
piezoresistive microphones," Journal of Acoustical Society of America, vol. 5, pp. 2710-2720, 2006.
[136] Y. Li., "Side Implanted Piezoresistive Shear Stress Sensor for Turbulent Boundary Layer
Measurements." Ph.D. diss., University of Florida, Mechanical and Aerospace Engineering, 2008.
[137] NI, "Manual for PXI system," National Instruments, Austin, TX August 2008. [138] H. W. Coleman and W. G. Steele, Experimentation and Uncertainty Analysis for
Engineers. New York: John Wiley & Sons, Inc, 1999. [139] I. PCB Piezotronics, "Precision Condenser Microphone model number 377A51," St.
Louis, MO 2003. [140] T. Schultz, M. Sheplak, and L. N. Cattafesta, "Application of multivariate uncertainty
analysis to frequency response function estimates," Journal of Sound and Vibration, vol. 305, pp. 116-133, 2007.
[141] A. Benninghoven, F. G. Rüdenauer, and H. W. Werner, Secondary ion mass
spectrometry: basic concepts, instrumental aspects, applications and trends: John Wiley & Sons, New York, 1987.
[142] K. Kadirvel, "Development of Closed-Loop Interface Circuits for Capacitive Transducers
with Applications to a MEMS Capacitive Microphone." Ph.D. diss., University of Florida, Electrical and Computer Engineering, 2007.
[143] F. J. Fransson and E. V. Jansson, "STL–Ionophone: Transducer properties and
construction," The Journal of the Acoustical Society of America, vol. 58, p. 910, 1975. [144] J. King and J. Underbrink, "Characterization of a Microelectromechanical Systems
(MEMS) Microphone," in 14th AIAA/CEAS Aeroacoustics Conference and Exhibit Vancouver, BC, 2008.
[145] J. P. Holman, "An experimental investigation of flows from zero-net mass flux
actuators." Ph.D. diss., University of Florida, Mechanical and Aerospace Engineering, 2006.
[146] J. N. Reddy, Mechanics of Laminated Composite Plates. Boca Raton, FL: CRC Press,
Inc., 1997.
233
[147] M. Madou, Fundamentals of Microfabrication. Boca Raton, FL: CRC Press, 1997. [148] W. D. Callister, Materials Science and Engineering - An Introduction. New York, NY:
John Wiley and Sons, Inc., 2000. [149] S. L. Ross, Differential Equation, Third ed. New York: John Wiley & Sons, 1984. [150] N. N. Lebedev, Special Functions & Their Applications. New York: Dover Publications,
Inc., 1972. [151] J. C. Anderson, Computational Fluid Mechanics. New York, NY: Hemisphere Publishing
Corporation, 1984. [152] M. N. Ozisik, Heat conduction. New York, NY: Wiley, 1993. [153] K. Atkinson, 1993, Second ed. New York: John Wiley & Sons, Inc., 1985.
234
BIOGRAPHICAL SKETCH
Brian grew up in Stanhope, NJ, a small town an hour west of New York City, with his
parents, Leo and Helen Homeijer, and brother Dan. After graduating from Lenape Valley
Regional High School in 1999, he attended Lehigh University in Bethehem, Pennsylvania. After
obtaining a degree in mechanical engineering in 2003, he was accepted into the mechanical
engineering graduate program at the University of Florida. During his tenure at UF, he has
worked with Dr. Mark Sheplak on the design of MEMS transducers. Upon graduation, Brian
will begin work as a research and development engineer in the Imaging and Printing Division of