You are on page 1of 8

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 24, NO.

1, FEBRUARY 2015 241

A MEMS Low-Noise Sound Pressure Gradient


Microphone With Capacitive Sensing
Ronald N. Miles, Weili Cui, Quang T. Su, and Dorel Homentcovschi

Abstract— A silicon microelectromechanical systems micro- the cost of increasing the computational burden. In addition
phone is described that detects sound pressure gradients. The to increasing the need for precision in sensitivity, the difference
diaphragm consists of a stiffened plate that rotates around a signals detected by the microphones can be so small that they
central axis in response to sound pressure gradients. The motion
of the diaphragm is converted into an electronic signal through may be buried in the noise produced by the microphones.
the use of interdigitated comb fins that enable capacitive sensing. The combination of increasingly stringent requirements on
Measured results show that the microphone achieves a substan- matched sensitivity, phase and noise floor results in increased
tially lower low-frequency sound pressure-referred noise floor cost and motivates a search for alternative acoustic sensing
than can be achieved using existing dual miniature microphone approaches. The main purpose of the present study is to
systems. Measured directivity patterns are shown to be very close
to what is expected for sound pressure gradient receivers over a explore the possibility of creating a low-cost MEMS micro-
broad range of frequencies. [2014-0094] phone that is dedicated to the task of accurately detecting these
small spatial differences in pressure.
Index Terms— Consumer electronics, audio systems, micro-
phones, acoustical engineering, microelectromechanical devices, The design approach used in creating the microphone
acoustic measurements, acoustic devices, acoustic sensors described here is motivated by investigations of auditory
sensing systems in small animals. These animals have essen-
I. I NTRODUCTION
tially the same sensing challenge that many small commercial

T HE TECHNOLOGY of miniature microphones has


advanced considerably due to the use of silicon microfab-
rication and the technology of microelectromechanical systems
products have: they must resolve very small differences in
sound pressure at their two ears. In large animals, the ears are
typically uncoupled from each other producing independent
(MEMS). These microphones are currently used in billions measures of the sound pressure, which leaves the task of
of portable electronic products. It is becoming common for examining differences in amplitude and phase to the nervous
multiple microphones to be incorporated in these small assem- system. This is much like what is done in the nondirectional
blies. By sampling the sound field at multiple spatial locations, microphones used in today’s portable electronic products.
signal processing algorithms within the device can process In contrast to what is found in large animals, many small
spatial differences in the sound field to attenuate unwanted animals employ well-developed schemes to ensure that the
sounds. Unfortunately, when the distances between the micro- spatial differences in pressure are resolved by the ears them-
phones are small relative to the wavelength of sound, the selves, prior to the signals being detected and processed by
spatial variation in a sound field can be rather small relative to the nervous system. Pre-processing the signals at the point
the spatial average of the signal. Our obsession with reducing where the sound pressure is transduced greatly reduces the
the size of these products creates significant challenges in the computational load on the nervous system. The fact that
design of the microphones. in small animals, the ears are often specialized to enhance
In addition to the requirement that the microphones them- the detection of pressure differences suggests that engineers
selves be as small as possible, when multiple microphones are should carefully consider the use of microphones that operate
used differences in their own sensitivity can adversely affect using similar principles.
their ability to resolve spatial differences in the sound field; the The challenge, of course, is to construct a miniature micro-
magnitude and phase of the sensitivities of the microphones phone that provides an accurate measure of the sound pressure
should be as closely matched as possible. Small errors in sensi- gradient. The design pursued in the present study is motivated
tivity must be corrected in the signal processing algorithm with by our earlier work on the auditory system of the parasitoid
Manuscript received March 27, 2014; revised May 19, 2014; accepted fly Ormia ochracea [1]–[3]. The tympanal structure in this
May 22, 2014. Date of publication June 25, 2014; date of current version animal enables remarkable directional sensitivity even though
January 30, 2015. This work was supported by the National Institute of the pair of ears fit in a space only 1 mm across.
Health, National Institute on Deafness and Other Communication Disorders,
under Grant R01DC009429 and Grant R01DC009859, and in part by the A number of researchers have pursued the development
State University of New York Technology Accelerator Fund. Subject Editor of microphones based on the operating principles discovered
M. Wong. in the ears of Ormia ochracea [4]–[7]. In a previous study,
The authors are with the Department of Mechanical Engineering, State
University of New York, Binghamton, NY 13902-6000 USA (e-mail: we demonstrated unprecedented low noise performance in
miles@binghamton.edu; wcui@binghamton.edu; quang.su@binghamton.edu; a miniature silicon microphone based on the fly’s ears that
homentco@binghamton.edu). utilized optical sensing to transduce the diaphragm’s motion
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. into an electronic signal [8]. The pressure sensing diaphragm
Digital Object Identifier 10.1109/JMEMS.2014.2329136 used in this microphone was comprised of a stiffened plate
1057-7157 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Vel Tech Dr. RR & Dr. SR Tech University - Chennai. Downloaded on August 27,2020 at 09:27:59 UTC from IEEE Xplore. Restrictions apply.
242 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 24, NO. 1, FEBRUARY 2015

Fig. 2. Plan view of the diaphragm. The 1 by 3 mm diaphragm is


surrounded by interdigitated comb fins and is connected to the bulk silicon
at the supporting hinge. The hinge consists of a T section. Scanning electron
micrographs of regions labeled a), b), c) and d) are shown in figure (3).

Fig. 1. Schematic representation of the differential microphone diaphragm.


The diaphragm consists of a stiffened plate that rotates about its supports in
response to sound pressure gradients. Although not shown in this schematic,
interdigitated comb fins are incorporated around the perimeter slit to enable
capacitive sensing. Sound entering the back volume through the slits tends
to be spatially uniform and hence does not drive the rocking motion of the
diaphragm. It instead provides the benefit of suppressing the nondirectional
response.

designed to rotate about a central, compliant hinge when it is


driven by sound pressure gradients.
The use of optical sensing resulted in a high-performance
design in the directional microphone described in [8] and elim-
inated certain design challenges that exist in the much more
conventional, capacitive sensing approach. It can, however,
introduce its own, entirely different challenges and has not
yet been incorporated in large volume microphone products.
As will be described below, several of the design challenges
Fig. 3. Scanning electron micrographs (SEM) of the interdigitated comb fins.
in the use of capacitive sensing for this diaphragm have been a) and b) Show the upper left corner of the diaphragm from two angles viewed
overcome through the use of a new fabrication process to through the back side hole. c) Shows the central hinge of the diaphragm
construct interdigitated comb fin electrodes. The main purpose viewed from the top and d) shows the central hinge of the diaphragm viewed
through the back side hole. The general locations of a, b, c, and d are indicated
of the present study is to demonstrate the use of conventional in figure (2). The stiffeners are approximately 30 microns thick and 2 microns
capacitive sensing in a miniature silicon microphone that wide. The comb fins are approximately 5 microns thick and 2 microns wide.
detects pressure gradients over a substantial portion of the The skin of the diaphragm is approximately 1.6 microns thick.
audible range of frequencies with substantially lower low
frequency noise than that of currently available miniature with the rotational moment of inertia, I , determines the natural
microphones. frequency of rotation about the flexible hinge. The details
of the design considered here have been determined through
II. T HE D IAPHRAGM FOR D ETECTING extensive design optimization using finite element analysis and
P RESSURE G RADIENTS experiments. This design optimization has considered both
The differential microphone diaphragm used here consists the acoustic performance of the diaphragm and the structural
of a lightweight plate that is designed so that its primary integrity during fabrication and service.
motion consists of rotation about a hinged axis as depicted As shown in figure (2), the design incorporates interdigi-
schematically in figure (1) [9]–[11]. The diaphragm is attached tated fin electrodes around the diaphragm perimeter to enable
to the substrate only at the hinge supports; it is separated at all capacitive sensing. These structures are shown in the scanning
other locations around its perimeter by a slit as shown in the electron micrographs in figure (3).
figure. The silicon chip that supports the diaphragm is attached In the design of conventional capacitive microphones that
to a circuit board which closes the bottom side of the air space utilize a parallel back plate electrode, it is well known that
beneath the diaphragm. the force associated with the biasing electric field can have
In the design considered here, the hinge that supports the a marked influence on the diaphragm’s dynamic response to
diaphragm is created by constructing a T-section support that is sound. Because of the nature of the interdigitated fin electrodes
anchored in the supporting substrate and traverses the width of used here, the electric force has significantly less impact on the
the diaphragm at its centerline. The geometry of the T-section response. These forces are, however, included in the simplified
support sets the equivalent torsional stiffness, kt , which along lumped parameter model described below.

Authorized licensed use limited to: Vel Tech Dr. RR & Dr. SR Tech University - Chennai. Downloaded on August 27,2020 at 09:27:59 UTC from IEEE Xplore. Restrictions apply.
MILES et al.: MEMS LOW-NOISE SOUND PRESSURE GRADIENT MICROPHONE 243

While many details of the structural design have been where the equilibrium equation (2) was used to simplify the
carefully considered in our finite element models, it is very result. Equation (5) may be written in the form
helpful to consider the simplest possible lumped-parameter
I δ̈ + ke δ + ct δ̇ = M (6)
model of the diaphragm’s response to sound. This guides
the design and provides essential insight in the selection of where
the dominant design parameters. Again, our detailed finite V12 d 2 C1 V22 d 2 C2
element model of the acoustic response allows us to account ke = kt − | θ 0 − |θ (7)
2 dθ 2 2 dθ 2 0
for many other important effects such as the response to
is the equivalent torsional spring constant including the effects
high frequency resonant modes, which we have minimized as
of the electric forces. Of course, as the bias voltages V1 and V2
much as is feasible in our design optimization process. In the
are increased, the effective stiffness of the system will either
following, we will then assume that the dominant motion of 2
the diaphragm consists of pure rotation about the hinge axis increase or decrease, depending on the signs of ddθC21 |θ0 and
d 2 C2
shown in figure (1). | .
dθ 2 θ0
In any case, it is very desirable that these terms be as
It is assumed that bias voltages V1 and V2 are applied to the small as possible to minimize the effects of the bias voltage
electrodes adjacent to each end of the diaphragm and that the on the diaphragm’s response.
diaphragm will be electrically connected to a charge ampli- From equation (6), the transfer function between the incre-
fier, as will be described below. The capacitances between mental diaphragm rotation, δ, and the applied moment, M is
the diaphragm and the fixed electrodes will be denoted by 1 1/I
C1 and C2 . The governing equation for the rotation, θ , of the H Mδ (ω) = = 2 (8)
ke − I ω + ı̂ωct
2 ω0 − ω + 2ı̂ωζ ω0
2
diaphragm is then √
where ω02 = ke /I and ζ = ct /(2 ke I ).
V12 dC1 V 2 dC2 Since the microphone diaphragm is small relative to the
I θ̈ + kt θ + ct θ̇ = M + (θ ) + 2 (θ ) (1)
2 dθ 2 dθ wavelength of sound the spatial variation in pressure is small
where again, I is the mass moment of inertia about the rotation over its surface in comparison to the average pressure. Let
axis, kt is the equivalent torsional stiffness, ct is the equivalent the origin of our coordinate system be at the center of the
torsional dashpot constant, and M is the applied moment due diaphragm with the rotation oriented in the y direction and
to the sound field about the rotation axis. the total diaphragm length L measured in the x direction. The
Depending on details of the geometry and whether there sound pressure as a function of x and t can be expanded in a
is any nonzero static deflection of the diaphragm (due two-term Taylor’s series [8]
to fabrication-induced imperfections), the two capacitances ∂p
C1 (θ ) and C2 (θ ) may depend on θ in a nonlinear manner. p(x, t) ≈ p(0, t) + x|x=0 (9)
∂x
The motion of the diaphragm due to sound pressure, however, The moment applied by the sound field about the diaphragm’s
is small, allowing the system to be expressed as a linear model
axis of rotation is then given by [8]
for small rotations about a fixed position. When DC bias
voltages are applied to the microphone, the static equilibrium ∂p
M(t) ≈ |x=0 I A (10)
position θ0 can be determined by solving ∂x
where, for this rectangular diaphragm, of length L, and width
V12 dC1 V 2 dC2 b, I A = bL
3
k t θ0 = |θ0 + 2 |θ (2) 12 is the area moment of inertia about the axis of
2 dθ 2 dθ 0 rotation.
As discussed above, let the diaphragm rotation be expressed Equation (10) shows that the moment applied to the
as θ = θ0 + δ, where δ is the small rotation due to sound diaphragm is the product of the pressure gradient in the
pressure. Substitution into equation (1) yields x direction and the area moment of inertia. For a plane,
harmonic sound wave at a frequency ω, incident at an angle
I δ̈ + kt · (θ0 + δ) + ct δ̇ φ relative to the x axis, the sound pressure may be expressed
V 2 dC1 V 2 dC2 as
=M+ 1 (θ0 + δ) + 2 (θ0 + δ) (3)
2 dθ 2 dθ
p(x, t) = Peı̂ (ωt −ω cos(φ)x/c0 ) (11)
Linear approximations for the capacitance terms can be used
for small deflections. The first two terms of the Taylor series where c0 is the nominal speed of sound propagation and
about θ = θ0 are P is the amplitude. Equations (8) through (11) enable us to
calculate the transfer function between the rotation δ and the
dC1 dC1 d 2 C1 amplitude of the plane wave, P
(θ0 + δ) ≈ |θ0 + δ |θ ,
dθ dθ dθ 2 0 (ı̂ ω cos(φ)/c0 )I A /I
dC2 dC2 d 2 C2 H Pδ (ω) = (12)
(θ0 + δ) ≈ |θ0 + δ |θ (4) ω02 − ω2 + 2ı̂ωζ ω0
dθ dθ dθ 2 0
This clearly has the frequency response of a band-pass filter.
Equation (3) can then be approximated as
The response is proportional to cos(φ) as expected for a
V12 d 2 C1 V22 d 2 C2 diaphragm that detects pressure gradients. Since the mechan-
I δ̈ + (kt − | θ 0 − |θ )δ + ct δ̇ = M (5) ical sensitivity is proportional to I A /I , the ratio of the area
2 dθ 2 2 dθ 2 0

Authorized licensed use limited to: Vel Tech Dr. RR & Dr. SR Tech University - Chennai. Downloaded on August 27,2020 at 09:27:59 UTC from IEEE Xplore. Restrictions apply.
244 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 24, NO. 1, FEBRUARY 2015

assembled on the opposite side of the circuit board from the


MEMS chip. Figure (6) shows a prototype package for the
assembly to facilitate handling and shielding during testing.
The operational amplifier shown in figure (4) is intended
to function as a charge amplifier with a capacitor, C f in the
feedback loop. A parallel resistor, R f is used to minimize the
gain at very low frequencies and reduce the DC offset voltage
at the output. We will assume this resistor has sufficient
resistance to not affect the passband gain. The change in
output, v o (volts), for a given rotation of the diaphragm may
then be shown to be
dC 1
Fig. 4. Charge amplifier circuit used for capacitive sensing. The bias voltages dv o dθ V1 + dC 2
dθ V2
were supplied by AA 1.5 volt batteries. R f = 10G, C f = 1 pF. The =− (13)
operational amplifier is the OPA657 from Texas Instruments. dθ Cf
In the frequency domain, the electrical sensitivity describing
the transduction from the perturbed rotation δ to the electrical
output voltage v o may then be expressed as
dC 1
dθ |θ0 V1 + dC
dθ |θ0 V2
2
Hδv 0 (ω) = − (14)
Cf
Since the bias voltages V1 and V2 are assumed to be constant,
this approximate result is independent of frequency, ω.
If the derivatives in equation (14) are known, the electrical
output of the microphone for a given amplitude of a plane
acoustic wave may be determined from the product of the
results of equations (12) and (14),

Fig. 5. The circuit board for the read-out circuit. a) and c) show photographs H Pv o (ω) = Hδv 0 (ω) × H Pδ (ω) (15)
and b) and d) show CAD renderings of the design. The components for the
read-out circuit are placed on the opposite side of the circuit board from the To evaluate equations (14) and (15), we must estimate the
MEMS chip. External connections to the circuit board were made using a changes in the capacitors C1 and C2 for a given rotation about
micro-HDMI connector shown in the upper panels a) and b).
the equilibrium rotation θ0 . A model for the capacitance is
described below.
Note that because of fabrication-induced stresses, the two
ends of the diaphragm will typically be deflected slightly in
the same direction (on the order of a couple of microns). A
rotation of the diaphragm will result in one of the capacitors
to increase while the other decreases. The two derivatives in
equations (13) and (14) will then have opposite sign. In this
case, it is desirable to have V1 and V2 to also have opposite
signs. In addition, if the magnitudes of the derivatives are
not well matched, this mismatch can be compensated for by
adjusting V1 and V2 to maximize the output signal.

IV. I NTERDIGITATED C OMB F INS


Fig. 6. Microphone prototype package to protect the MEMS chip during FOR C APACITIVE S ENSING
handling and testing. The upper panels show computer aided design renderings
and the lower panels are photographs of the finished prototype. Upper panel
A. Fabrication
a) shows the complete assembly while b) and c) are cutaway views showing In order to create a capacitive sensing scheme that provides
the locations of the spacer and MEMS chip.
adequate performance for the differential microphone, a new
interdigitated comb fin design has been created as described in
moment of inertia to the mass moment of inertia, one should
detail in [12]. This consists of fins having length 75 microns,
seek designs that maximize the surface area relative to the
thickness of five microns, width of two microns, and a gap
mass.
between neighboring fins of two microns. The period of the
repeating sets of fixed and moving fins is 8 microns.
III. T HE C IRCUIT The fabrication process is illustrated in figure (7) and begins
The circuit used to obtain an electronic output is shown in with a deep reactive ion etch of a bare silicon wafer to
the schematic in figure (4). Figure (5) shows a prototype of create trenches that are approximately 3 microns wide and
the circuit where the electronic surface mount components are 30 microns deep (step 1). These trenches are used to form

Authorized licensed use limited to: Vel Tech Dr. RR & Dr. SR Tech University - Chennai. Downloaded on August 27,2020 at 09:27:59 UTC from IEEE Xplore. Restrictions apply.
MILES et al.: MEMS LOW-NOISE SOUND PRESSURE GRADIENT MICROPHONE 245

Fig. 7. Fabrication process for interdigitated comb fins (see text). 1) through Fig. 8. General cross sectional geometry for interdigitated comb fins.
6) show the six major steps in the fabrication process.

the stiffeners shown in figures (2) and (3). A second deep


reactive ion etch is performed to create trenches that are
approximately 3 microns wide and 5 microns deep. These
are used to create the interdigitated fins around the perimeter
of the diaphragm. A thermal oxide film is then grown in Fig. 9. Mapped geometry for interdigitated comb fins.
step 2) over the entire wafer surface. Phosphorous doped
silicon is then deposited in step 3) onto the thermal oxide
through low-pressure chemical vapor deposition (LPCVD) The electrostatic field (E X , E Y ) in the case of the two
which is then annealed to form polycrystalline silicon. Slits aligned electrodes in figure (9) can be written as shown in [15]
are then etched around the perimeter of the diaphragm so that −α
E X − i EY = , (16)
it will eventually be supported only on the flexible hinges R1 (Z ) R2 (Z )
in step 4). A through-wafer, deep reactive ion etch is then where α is a real constant and
performed in step 5) to free the back side of the diaphragm.   
This etch stops on the thermal oxide film grown in step 2). R j (Z ) = Z − a j Z − b j , j = 1, 2
Finally, the diaphragm is released in step 6) using buffered and R j (X) > 0 for X > b j > a j . The constant α results by
hydrofluoric acid to remove the thermal oxide. imposing the potential condition
∂V α
B. Analytical Model for Comb Fin Capacitance E X (X, 0) = − = , X ∈ (b1 , a2 ) . (17)
∂X R1 (X) R2 (X)
In typical capacitive microphones that employ charged
Integrating this relationship yields
electrodes that are parallel to the pressure sensing diaphragm,  a2
it is usually feasible to use very simple formulas to estimate the dX
α = V1 − V2 . (18)
capacitance as a function of the diaphragm deflection. As will b1 |R1 (X) R2 (X)|
be shown in the following, in the present design, fringing fields Also, the charge Q 2 on the electrode (a2 , b2 ) , by taking
may cause the simplified formulas to lead to errors on the order into account the formula (16) for the electric field, can be
of 30% in capacitance for electrode geometries of interest. obtained as
In this section we develop an analytical method for efficient  b2
dX
calculation of capacitances and forces in a general asymmetric Q 2 = −εα , (19)
comb drive system taking into consideration the main fringing a2 |R1 (X) R2 (X)|
fields. The approach is based on conformal mappings and the ε = 8.86 × 10−12 [F/m] being the permittivity of air.
determination of the electrostatic field in the canonical domain By eliminating the constant α between equations
(upper half-plane). (18) and (19) and expressing the definite integrals by
Consider one periodic unit of the asymmetric comb drive means of complete elliptical integrals [16] we can write the
shown in cross section in figure (8) where the left fin has the capacitance per unit length of the condenser system in the
potential V1 and the right fin is charged at potential V2 . The Z -plane C Z as
symmetry lines M∞ A1 , B1 N∞ , N∞ A2 and B2 M∞ shown  
in figure (8) are field lines. K k
CZ = ε , (20)
The conformal mapping of the upper half-plane Im (Z ) > 0, K (k)
Z = X + i Y (Figure (9)) into the physical domain can be where the modulus k and the complementary modulus k  are
obtained by using the Schwartz-Christoffel formula [13]. This 
(a2 − b1 ) (b2 − a1 ) 
mapping determines the coordinates of the electrodes in the k= , k = 1 − k2. (21)
mapped domain a1 , b1 , a2 , b2 . (See [14, p. 240]). (b2 − b1 ) (a2 − a1 )

Authorized licensed use limited to: Vel Tech Dr. RR & Dr. SR Tech University - Chennai. Downloaded on August 27,2020 at 09:27:59 UTC from IEEE Xplore. Restrictions apply.
246 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 24, NO. 1, FEBRUARY 2015

Fig. 10. Predicted capacitance and derivative with respect to displacement


of each capacitor C1 and C2 . Note that the derivative of the capacitance is
Fig. 11. Predicted and measured acoustic sensitivity in volts/pascal. The
reasonably constant over a wide range of deflections. This is a very desirable
predicted and measured results are in very close agreement. The measurements
feature of the present design approach since it leads to highly linear response were performed in the anechoic chamber at SUNY Binghamton, which ensures
and excellent dynamic range.
that the sound input closely approximates an ideal plane acoustic wave at
frequencies above approximately 80 Hz. The results are obtained for the most
The total energy E of the system can be written as sensitive direction (x). The parameters used in the model are: I = 2.38 ×
10−14 [kgm 2 ], kt = 1.90 × 10−7 [Nm/rad], ct = 2.69 × 10−11 [N m s/rad],
1 and Ia = 2.59 × 10−12 [m 4 ].
E = − C Z (V2 − V1 )2 . (22)
2
Since the capacitance of a 2-D condenser is invariant under
a conformal mapping (see [13]) the final expression for the of the fins from the pivot axis, the relationship between the
capacitance C of a pair of fixed and mobile fins of length L f rotation angle and the average deflection between the fins
(in the physical domain), by neglecting the end effects is is d = θlave . In the present design, because the change in
  capacitance is small for the fins near the pivot, the fins less than
K k a distance l = ±L/4 are not connected to the read-out circuit.
C = Lfε [F]. (23)
K (k) The average distance of the active fins from the pivot axis may
be estimated from lave ≈ (L 2 (1/2−1/8)/2+bL/2)/(L/2+b).
C. Predicted Capacitance In the present design L = .003 m, b = .001 m so that
The analytical model above can be used to calculate the lave ≈ 0.001275 m.
capacitance of a wide range of geometries of interdigitated A very rough estimate of the capacitance can be obtained
comb fins. Fortunately, the fabrication process described above by using the well-known parallel plate formula, C parallel =
permits a rather wide range of geometries to be constructed. Aslit /w where Aslit = L f ∗ h 1 ∗ Nslit is the total area of the
In the present study, we seek a design that provides a sufficient gaps with Nslit being the total number of gaps. w = w3 −w1 −
source capacitance and change in capacitance with diaphragm w2 = 2μm is the gap between the fins. This gives C parallel ≈
displacement to facilitate electronic readout, minimal added 1.04 pF, 30% less than the value of approximately 1.5 pF
mass, minimal added viscous damping (due to the air between in figure (10). The nominal capacitance will then be on the
the fins) and, reasonably linear variation in capacitance with order of 1 pF, which is approximately the input capacitance
diaphragm deflection. of typical high impedance preamplifiers.
The present design consists of identical fins with the Small stress gradients in the microphone diaphragm cause
following dimensions: w1 = w2 = 1μm, w3 = 4μm, initial deflections of the diaphragm ends to be between 2 and
h 1 = h 2 = 5μm, the length of the fins L f = 75μm and 5 μm after release. Because both ends are typically displaced
vertical distance between fins d in the range [0μm, 5μm] . in the same direction, when the diaphragm rotates, one of the
The right fin in figure (8) is charged at potential V0 the other two capacitors C1 or C2 will increase while the other decreases
fin being grounded. in capacitance. To obtain maximum voltage output, the bias
The predicted total capacitance of each capacitor C1 and C2 voltages V1 and V2 in equation (14) should then have opposite
is shown in figure (10) as a function of the relative vertical polarity.
displacement between the fins. When the displacement is zero,
the fins are perfectly aligned and the model indicates that the V. M EASURED M ICROPHONE O UTPUT
capacitance is approximately 1.5 pF. The derivative of the
capacitance with respect to displacement is zero for zero The predicted and measured output of the microphone as a
displacement, as expected. For displacements of approximately function of frequency is shown in figure (11). It clearly shows
2 μm and higher, the capacitance varies fairly linearly and the the bandpass filter characteristic as predicted in equation (12).
derivative is approximately −2 × 10−7 F/m. The measurements were obtained for the microphone oriented
Having the predicted capacitance as a function of the deflec- in its most sensitive direction in an anechoic chamber to obtain
tion d enables us to estimate the derivatives of the capacitance the response to a plane wave. The measured output voltage
as a function of the rotation θ in equations (1) through (7). is shown relative to the sound pressure obtained using a
If we consider lave to be the average distance in the x direction Brüel & Kjaer 4138 1/8 inch microphone.

Authorized licensed use limited to: Vel Tech Dr. RR & Dr. SR Tech University - Chennai. Downloaded on August 27,2020 at 09:27:59 UTC from IEEE Xplore. Restrictions apply.
MILES et al.: MEMS LOW-NOISE SOUND PRESSURE GRADIENT MICROPHONE 247


Fig. 12. Measured circuit output noise (volts/ Hz). Results are shown with
and without the microphone chip installed on the circuit board. The increased
output noise at approximately 400 Hz with the chip present is likely due to
the thermal noise of the microphone diaphragm.
Fig. 14. Comparison of the one-third octave band equivalent plane-wave
sound pressure noise for the present capacitive pressure gradient microphone
with those presented in [8]. The present capacitive readout scheme produces
output noise levels that are comparable to those obtained using the optical
readout of [8] at frequencies below about 2 kHz, and are as much as
30 dB below those obtained using currently available low-noise hearing aid
microphones. The noise floor of the capacitive microphone examined here was
measured in the anechoic chamber at SUNY Binghamton. The 1/3 octave band
noise levels of the chamber are below the scale of the vertical axis (typically
−15 dB). The A-weighted noise of the chamber is 0.5 dBA.

Fig. 13. Measured equivalent input pressure noise and estimated


√ thermal
noise of the microphone for a plane acoustic wave (pascal/ Hz). It is
estimated that diaphragm thermal noise dominates between approximately
100 Hz to 2 kHz.

VI. N OISE F LOOR OF THE M ICROPHONE

Because of the need to measure minute pressure differences


(or pressure gradients), the noise floor of any miniature
directional microphone is a very challenging design parameter.
In the present design, the diaphragm has excellent sensitivity
to sound pressure gradients, which greatly facilitates the
electronic readout. The measured output noise of the simple
readout circuit of figure (4) is shown in figure (12). The circuit
output noise was measured both with and without the MEMS
microphone chip connected to the circuit. The figure shows
that the output noise is increased by the attachment of the
microphone chip, particularly at 60-70 Hz (likely due to shield-
ing/grounding troubles) and at approximately 400 Hz, near the
resonant frequency of the diaphragm (as shown in figure (11)).
The data of figures (11) and (12) may be used to obtain the
equivalent input plane-wave pressure noise of the microphone
Fig. 15. Measured directivity patterns of the pressure gradient microphone
as in [8]. The result is shown in figure (13), which also shows at four frequencies. The results are normalized relative to the most sensitive
the predicted input pressure noise due to thermal excitation direction. The measured results closely resemble the expected figure eight
of the diaphragm, as described in [8]. Figure (14) shows the pattern over a broad frequency range. Directivity indexes are also shown for
each frequency. An ideal pressure gradient microphone will have a directivity
measured noise floor in 1/3rd octave band levels along with the index of about 4.8 dB, which is close to the results found here. The ratio of
results shown in [8] for comparison. The input-referred noise the maximum response to the minimum response varies from 12 to 53.
floor of the capacitive microphone of the present design is
comparable to that of the optical microphone discussed in [8] VII. M EASURED M ICROPHONE D IRECTIVITY
and are substantially below the two-microphone system used Because the microphone is designed to respond primar-
in hearing aids at frequencies below about 2 kHz. ily to pressure gradients, it is expected that the directivity

Authorized licensed use limited to: Vel Tech Dr. RR & Dr. SR Tech University - Chennai. Downloaded on August 27,2020 at 09:27:59 UTC from IEEE Xplore. Restrictions apply.
248 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 24, NO. 1, FEBRUARY 2015

pattern should correspond to the figure eight that is well- [13] P. Henrici, Applied and Computational Complex Analysis, Discrete
known for first order gradient detectors. To examine the Fourier Analysis, Cauchy Integrals, Construction of Conformal Maps,
Univalent Functions. vol. 3. Hoboken, NJ, USA: Wiley, 1993.
directional response of this design, the microphone was placed [14] S. G. Krantz, Complex Variables: A Physical Approach With Applica-
on a motorized rotation stage in the anechoic chamber at tions and MATLAB. Boca Raton, FL, USA: CRC Press, 2008.
SUNY Binghamton and the electronic output was measured at [15] D. Homentcovschi, “Electrostatic field of a system of aligned elec-
trodes,” J. Electrostat., vol. 26, no. 2, pp. 187–200, 1991.
1 degree increments around a full circle. Broadband random [16] I. S. Gradshtein and I. M. Ryzhik, Tables of Integrals, Series, and
noise was used as the sound source from loudspeakers that Products. New York, NY, USA: Academic, 1980.
were placed approximately three meters from the microphone.
Power spectra were measured with a bandwidth of approxi-
mately 8 Hz.
The results are shown in figure (15). The polar patterns
closely resemble the ideal figure eight expected for a pressure Ronald N. Miles received the B.S. degree in elec-
gradient receiver. trical engineering from the University of California
at Berkeley, Berkeley, CA, USA, and the M.S.
and Ph.D. degrees in mechanical engineering from
VIII. C ONCLUSIONS the University of Washington. He has been with
the Department of Mechanical Engineering, SUNY
The results presented here demonstrate the feasibility of Binghamton, Binghamton, NY, USA, since 1989,
constructing a miniature MEMS microphone that is sensitive to and has served as the Director of Graduate Studies,
sound pressure gradients and that utilizes capacitive sensing. Director of Undergraduate Studies, Associate Chair,
Department Chair, Professor, Distinguished Profes-
The simple capacitive sensing scheme is shown to achieve sor, and currently serves as the Associate Dean for
low sound input-referred noise, particularly at low frequencies Research for the Watson School of Engineering and Applied Science.
where conventional approaches exhibit relatively poor noise
performance. Measured directivity patterns are shown to be
very similar to the figure eight pattern expected for a pressure
gradient microphone over a broad frequency range.
Weili Cui received the B.S. degree in mechanical
engineering from Southwest Jiaotong University, and
ACKNOWLEDGEMENTS the Ph.D. degree in mechanical engineering from
Binghamton University, Binghamton, NY, USA. He
Carl Capello provided assistance with the electronic assem- worked as a Research Professor at the Department
bly. We fabricated the silicon chips at the Cornell NanoScale of Mechanical Engineering, Binghamtom University,
since 2004 for a decade and currently teaches in
Facility. the Bioengineering Department, Watson School of
Engineering and Applied Science.

R EFERENCES
[1] R. Miles, D. Robert, and R. Hoy, “Directional hearing by mechanical
coupling in the parasitoid fly Ormia ochracea,” J. Acoust. Soc. Amer.,
vol. 98, no. 6, pp. 3059–3070, 1995.
[2] D. Robert, R. Miles, and R. Hoy, “Directional hearing by mechanical
coupling in the parasitoid fly Ormia ochracea,” J. Comparative Physiol. Quang T. Su received the B.S. and Ph.D. degrees
A, Neuroethol., Sensory, Neural, Behav. Physiol., vol. 179, no. 1, pp. 29– in mechanical engineering from Binghamton Uni-
44, 1996. versity SUNY, Binghamtom, NY, USA, in 1998 and
[3] R. Miles and R. Hoy, “The development of a biologically-inspired 2005. He is currently an Assistant Professor in the
directional microphone for hearing aids,” Audiol. Neurotol., vol. 11, Mechanical Engineering Department. His research
no. 2, pp. 86–94, 2006. interests include acoustic and vibration characteri-
[4] P.-H. Sung, J.-Y. Chen, K.-H. Yen, and C.-Y. Wu, “CMOS compatible zation methods for microsensors and actuators, and
directional microphone,” in Proc. IMPACT, Oct. 2007, pp. 149–152. shock and vibration reliability of electronics.
[5] S. Ando, T. Kurihara, K. Watanabe, Y. Yamanishi, and T. Ooasa,
“Novel theoretical design and fabrication test of biomimicry directional
microphone,” in Proc. Int. Solid-State Sens., Actuators, Microsyst. Conf.,
2009, pp. 1932–1935.
[6] A. Lisiewski, H. Liu, M. Yu, L. Currano, and D. Gee, “Fly-ear
inspired micro-sensor for sound source localization in two dimensions,”
J. Acoust. Soc. Amer., vol. 129, no. 5, pp. EL166–EL171, 2011.
[7] M. Touse, J. Sinibaldi, K. Simsek, J. Catterlin, S. Harrison, and Dorel Homentcovschi received the B.S. and M.S.
G. Karunasiri, “Fabrication of a microelectromechanical directional degrees from the University of Bucharest, Depart-
sound sensor with electronic readout using comb fingers,” Appl. Phys. ment of Mathematics, in 1965, and the Ph.D. degree
Lett., vol. 96, no. 17, p. 173701, 2010. from the same university in 1970. He has been with
[8] R. N. Miles et al., “A low-noise differential microphone inspired by the Polytechnica University of Bucharest, Depart-
the ears of the parasitoid fly Ormia ochracea,” J. Acoust. Soc. Amer., ment of Mathematics, in the period 1970-2001,
vol. 125, no. 4, pp. 2013–2026, 2009. where he served as Assistant Professor, Associate
[9] R. Miles, “Comb sense microphone,” U.S. Patent 7 545 945, Jun. 9, 2009. Professor, Professor, Distinguished Professor, and
[10] R. N. Miles, “Surface micromachined differential microphone,” U.S. Chairman of the 2nd Mathematics Department. In
Patent 7 992 283, Aug. 9, 2011. the period 1995-2001, he was also a 1st Degree
[11] R. Miles, S. Sundermurthy, C. Gibbons, R. Hoy, and D. Robert, Researcher and Director of the Institute of Applied
“Differential microphone,” U.S. Patent 6 788 796, Sep. 7, 2004. Mathematics of the Romanian Academy. Since 2001, he has worked as
[12] R. Miles, and W. Cui “Hinged MEMS diaphragm and method of a Research Professor with SUNY Binghamton, Department of Mechanical
manufacturing thereof,” U.S. Patent 14 039 149, Jun. 9, 2013. Engineering, Binghamtom, NY, USA.

Authorized licensed use limited to: Vel Tech Dr. RR & Dr. SR Tech University - Chennai. Downloaded on August 27,2020 at 09:27:59 UTC from IEEE Xplore. Restrictions apply.

You might also like