CONFIDENTIAL

† Confidential

Electrical Characterization of MEMS

Microphones

Gokulraj Chandramohan

Under the guidance of : Prof.dr.P.M.(Lina) Sarro

Electronic Components, Technology and Materials (ECTM)
Delft Institute of Microsystems and Nanoelectronics (DIMES)
Faculty of Electrical Engineering, Mathematics and Computer Science
Science
Delft University of Technology, The Netherlands

Supervisors : Dr. Twan van Lippen, Dr. Klaus Reimann and Iris Bominaar-
Bominaar-Silkens
Corporate I&T / Research
Microsystems Technology Group,
NXP Semiconductors, Eindhoven, The Netherlands
Netherlands

August 2010
 Abstract
A reliable characterization of a MEMS microphone is essential for a better un-
derstanding of the device physics, for estimating the device performance and
for guidance during the research, for better microphones. The work presented
in this report focuses on the investigation and comparison of dierent mea-
surement principles and techniques used to measure the key device parameters
resonance frequency and compliance. The rst generation MEMS microphones
that are developed by NXP are inuenced by mechanical vibrations that re-
duces the device performance. Therefore, new techniques will be implemented
to reduce this noise and these techniques are based upon matching the fre-
quencies of two exible plates in the microphone sensor, we investigate these
resonance frequencies in this thesis. This investigation is performed with two
dierent measurement techniques : electrical impedance measurements and laser
vibrometer experiments. These measurements are performed in air and vacuum
conditions and the inuence of ambient pressure, bias voltage and back volume
on the resonance frequency is investigated. The vacuum measurements are in
good agreement with the nite element simulation results. Additionally the
results obtained from dierent measurements are compared and the dierence
between the results are analyzed. Finally, recommendations are made for future
measurement conditions and setups.
Acknowledgement
I owe my great thanks to many people who helped and supported me during
my thesis and graduate studies.
First I would like to thank Professor dr. Lina Sarro for giving me a won-
derful opportunity to do my master thesis in collaboration with the Electronic
Components, Technology and Materials (ECTM) group at TU Delft and the Mi-
crosystem Technology group of NXP Semiconductors, Eindhoven. I'm especially
grateful for her support, thoughtful guidance and encouragement throughout my
academic life in The Netherlands.
Secondly, my sincere thanks to dr. Twan Van Lippen, Iris Bominaar-Silkens
and dr. Klaus Reimann for their constant support, encouragement and guidance
in spite of their busy schedules. They were always available to resolve my
doubts. Also I want to express my gratitude to Remco Pijnenburg and Mieke
Botermans for their suggestions and comments during our bi-weekly meetings
at NXP. Additionally my deepest thanks to dr. ir. Martijn Goossens for his
constant help and patience in the measurement lab.
I would like to thank my friends Andrés Vásquez, Shuang Song, Onur Kaya,
Mahidhar, Karthik, Venkatasubramaniam, Aditya, Aadithya, Rajat Bharatwaj,
Madhavan, Shilesh, Chockalingam, Supriya, Vinoth and my house mates San-
tiago, Cesar, Eric Periot, Kevin Brands, Sundeep for their support and being
there for me when I needed them most.
I would like to express my love to my parents Chandramohan and Umarani,
and my sister Divya for giving me the freedom to dream and chase my dreams.
Finally my sincere and heartfelt thanks to the Almighty for all the shine of
health and wealth bestowed upon me, throughout my life.

1
Contents

1 Introduction 5
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 MEMS Microphone Concept 9

2.1 MEMS Capacitive Microphones . . . . . . . . . . . . . . . . . . . 9
2.2 NXP MEMS Microphone . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Microphone Sensitivity . . . . . . . . . . . . . . . . . . . . 12
2.3 Body-Noise Cancellation . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.1 Frequency Matching . . . . . . . . . . . . . . . . . . . . . 15
2.4 Compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3 Laser Vibrometer Measurement 19

3.1 Measurement Setup . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Test Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.1 2X and 50X Magnication Measurement . . . . . . . . . . 23
3.3 Backplate Frequency Measurement Results . . . . . . . . . . . . 26
3.4 Measurement of membrane at Low Frequency . . . . . . . . . . . 27
3.4.1 Wafer sample measurement . . . . . . . . . . . . . . . . . 27
3.4.2 Packaged sample measurement . . . . . . . . . . . . . . . 28
3.5 Extraction of Electrostatic Force . . . . . . . . . . . . . . . . . . 29
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4 Electrical Impedance Measurement 31

4.1 Frequency Response Measurement . . . . . . . . . . . . . . . . . 31
4.1.1 Measurement Set-ups . . . . . . . . . . . . . . . . . . . . 32
4.1.2 Measurement setup Calibration . . . . . . . . . . . . . . . 33
4.2 Resonance Frequency . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2.1 Extraction of Resonance frequency . . . . . . . . . . . . . 36
4.2.2 Air and Vacuum Measurement Results . . . . . . . . . . . 39
4.2.2.1 Wafer and Packaged Sample Measurement Results 40
4.2.2.2 Backplate resonance frequency and stress . . . . 42
4.3 Capacitance  Voltage Measurement . . . . . . . . . . . . . . . . 43
4.3.1 Extraction of Membrane Compliance . . . . . . . . . . . . 44
4.3.2 Optimal Frequency selection . . . . . . . . . . . . . . . . 47
4.3.3 Inuence of AC voltage . . . . . . . . . . . . . . . . . . . 48
4.3.4 Membrane compliance in vacuum . . . . . . . . . . . . . . 48

2
 4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5 Measurement Results Comparison and Discussion 50

6 Conclusions and Recommendations 54
6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . 55

Bibliography 56
A Inuence of calibration techniques on the extracted results 57
B 10 cycles with same conditions (2X) 59

3
List of Symbols

2
Am Area of membrane [m ]
2
a Acceleration [m/s ]
C Compliance [m/N]
Cm Membrane Compliance [m/N]
Cef f Eective Compliance [m/N]
c Speed of sound [m/s]
Cs Series Capacitance [F]
C0 Capacitance at 0V [F]
CP Parasitic Capacitance [F]
D Dissipation Factor [-]
F Force [N]
fres Resonance Frequency [Hz]
Fe Electrostatic Force [N]
f00 Fundamental Eigen mode frequency [Hz]
g Air Gap [m]
hm Thickness of membrane [m]
hbp Thickness of backplate [m]
k Spring Constant [N/m]
kef f Eective Spring constant [N/m]
kel Electrical Spring constant [N/m]
M Mass [kg]
mef f Eective mass [kg]
Rm Radius of membrane [m]
Rbp Radius of backplate [m]
Se Electrical Sensitivity [V/Pa]
Vbias Bias Voltage [V]
Vpull−in Pull-In Voltage [V]
x Displacement [m]
3
ρa Density of Air [kg/m ]
ε Permittivity [F/m]
3
ρ Density of Silicon [kg/m ]
σfres Error in resonance frequency [Hz]
0
ωnm Eigen value [-]
σ Stress [Pa]
σCm Error in membrane compliance [m/N]

4
Chapter 1

Introduction
This chapter is structured into three sections. The rst two sections focus on
the project's motivation and the objectives of the work. The nal topic gives
an overview about the organization of this report.

1.1 Motivation
Humans have always tried to extend their capabilities. Firstly, they extended
their mechanical powers. They invented the steam engine, the combustion en-
gine, the electric motor and the jet engine. Mechanization thoroughly changed
society. Secondly, they extended their brains. They invented means for arti-
cial logic and communication: the computer and the internet. This information
phase is changing society again, where we cannot yet fully predict the end re-
sult. However, this is not all. By inventing sensors, humans are now learning to
articially expand their senses. Sensorization together with mechanization and
informatization will bring about the third industrial revolution [1].
Sensors transform signals from dierent energy domains to the electrical
domain. Fig. 1.1 shows the dierent signal domains[2].

Figure 1.1: Sensor Classication according to six signal domains[2]

This gure explains the conversion from dierent signal domains into the
electrical domain. In our case, for example, the microphone converts the acoustic

5
sound, which is a mechanical signal, into an electrical signal by using capacitive
principle.
There is always a need for improvement and the demand for faster, cheaper,
more reliable quality products. Our thesis deals with NXP MEMS (Micro Elec-
tro Mechanical Systems) microphones. Nowadays microphones are present in
all electronic gadgets such as mobile phones, personal digital assistants (PDA),
digital cameras and camcorders. This provides a big market space for the mi-
crophone business (Fig.1.2).

Figure 1.2: MEMS Microphones market prediction[3]

NXP produces microphones and speakers for mobile phones. In order to give
the customers the best in technology microphones, NXP chooses to use MEMS
technology based capacitive microphones. Other microphones used are conven-
tional condenser or electret microphones, piezoelectric, piezoresistive, magnetic
or optical microphones [4]. Out of these, the most commonly used microphones
are condenser microphones consisting of a exible diaphragm and a rigid counter
electrode. The diaphragm acts as one plate of capacitor, and the sound vibra-
tions produce changes in the capacitance between the plates. The most popular
microphones used in most of the present electronic devices are the electret mi-
crophones. It is expected that the electret microphones will soon be replaced
by the MEMS based microphones [5].
The advantages of MEMS microphones are their reproducibility, smaller size,
integration with electronic circuitry in the same package, volume production and
automated pick-and-place compatibility. When considering MEMS microphones
in mobile phones they have several functional advantages like improved sound
quality, greater range and smaller size [6] which leaves space for other MEMS
devices such as gyroscopes and accelerometers. These advantages made NXP

6
to opt for the MEMS microphones.
NXP has developed the rst generation microphone sensor and wants to
improve its performance for better sensitivity and less noise. To accomplish
those tasks, research has to be done to understand the device characteristics
and to come up with solutions to reduce the noise.

1.2 Objective
The NXP microphone is a parallel plate capacitive microphone and its simple
structure is shown in Fig.1.3. The main components of a MEMS capacitive
microphone are a thin conductive membrane and a thick perforated conductive
backplate, separated by an air gap. Hence, the plates form a parallel plate
capacitor. These plates are biased with a constant charge Q.
The operation of the sensor is based on the capacitive principle. When
the sound vibrates the membrane, it changes the distance between the plates.
Therefore, the capacitance changes, which is translated into an electrical signal.
The sensor is sealed by a back chamber so that sound can only reach one side
of the membrane.

Figure 1.3: MEMS Capacitive microphone[7]

The current NXP MEMS microphones are subject to body-noise, which is the
noise signal generated by the external mechanical vibrations. The desired behav-
ior of the microphone is to be sensitive for sound only, but body noise implies
that the microphone is sensitive for both sound and acceleration. Therefore,
new techniques will be implemented in the next generation sensor to cancel this
body-noise. One such technique is intrinsic compensation, which is also known
as frequency matching.
Frequency matching is the technique that cancels the body-noise by match-
ing the resonance frequencies of membrane and backplate. This can be realized
by the present MEMS technology. The matching is achieved by making the
backplate more exible, so that its resonance frequency shifts closer to the mem-
brane resonance frequency. When the resonance frequencies are matched, both

7
the membrane and backplate exhibit equal deection for mechanical vibrations,
thus avoiding the noise output.
Therefore, the main topic of this thesis is to investigate the resonance fre-
quency of both membrane and backplate. Furthermore, we study the membrane
compliance. The compliance is directly related to the sensitivity of the micro-
phone sensor.
Dierent measurement principles are applied and compared. As it is known
that the resonance frequency of the membrane depends on the bias voltage, air
damping and back-volume. We investigated the inuence of ambient pressure,
bias voltage and back volume on the resonance frequency.

1.3 Thesis Organization

The theory and background information necessary to understand this thesis
is given in chapter 2. Furthermore, we explain body noise and discuss the
operating principle of the MEMS capacitive microphone.
In chapter 3 the laser vibrometer technique is discussed, along with the key
results obtained from the measurement. Chapter 4 is completely focused on
the electrical impedance measurements, their setups and the results obtained.
Chapter 5 is dedicated to comparing the dierent measurement principles and
results. Finally, Chapter 6 concludes with suggestions for future work and
recommendations.

8
Chapter 2

MEMS Microphone Concept

This chapter focuses on the theory and information necessary to understand
the work done in this thesis. First an introduction about the NXP MEMS
microphone, its structure and operating principle is described. Then the body-
noise problem is introduced with the possible measures to avoid it.

2.1 MEMS Capacitive Microphones

A microphone is an acoustic-to-electric transducer that converts sound into an
electrical signal. Our main topic of interest is on MEMS capacitive microphones;
MEMS (Micro Electro Mechanical Systems) are made up of components between
1 to 100 micrometers in size. The main purpose of MEMS is to use semiconduc-
tor fabrication technologies and processes to manufacture miniature mechanical
elements which can be used as sensors and actuators on a silicon substrate [8].
Both the electrical and mechanical components can be combined to build an
electro-mechanical system with a specic functionality. MEMS fabrication uses
the integrated circuit batch processing technology for mass manufacturing and
low device costs, additionally allowing system-on-chip options. Fig.2.1 shows a
complicated MEMS device A Ratchet with dimensions in micrometer range.
A MEMS capacitive microphone combines both MEMS mechanical charac-
teristics and capacitive microphone's characteristics. In general, a capacitive
microphone consists of a exible plate, named membrane, and a perforated
rigid plate named backplate, separated by an air gap. Fig.2.2 shows the MEMS
microphone operating principle. The membrane and backplate act as parallel
plates of a capacitor and the air in-between the plates provides the dielectric
of the capacitor. The back chamber is necessary to make sure that the sound
reaches the membrane only from one side. This back chamber must be large
enough to let the membrane move freely. The membrane is sensitive to acoustic
pressure, and thus vibrates for the incoming acoustic signal. The backplate is
perforated and thus allows the airow to pass through. Hence the backplate
is insensitive to the acoustic signal. The movement of the membrane changes
the capacitance between the plates. This can be translated into electrical signal
through the external circuitry.

9
 Figure 2.1: MEMS realization of mechanical component A Ratchet [9]

Figure 2.2: Capacitive microphone cross sectional view and its operation[10]

2.2 NXP MEMS Microphone

The MEMS microphone developed by NXP Semiconductors is based on the
capacitive microphone principle. The cross section view of the NXP MEMS
microphone with its dimensions is shown in Fig.2.3. Fig.2.4 shows the MEMS
microphone in the wafer. The four contact pads are connected to the backplate,
membrane and the substrate (2X).

10
Figure 2.3: NXP MEMS microphone cross section of the device with dimensions

(a)

(b)
Figure 2.4: NXP MEMS microphones in (a) on a wafer. (b) Single device

The packaged NXP MEMS microphone is shown in Fig.2.5. It is packaged

in a case of 4.72 mm x 3.76 mm x 1.1 mm with an opening for sound in the
bottom. The Application Specic Integrated Circuit (ASIC) for the read-out of
the audio signal is placed inside the same package. There are advantages and
disadvantages of this package. The disadvantage is that the air inside the pack-
age will damp the moving membrane. The metal cap denes the back volume
and in principle an innite back volume is required for maximum sensitivity,
which means no metal cap. In reality that is not possible since no back cham-
ber means no reference pressure, so then the microphone cannot operate. Also
the package is necessary to protect the device from rough environments in the
real working environments.

11
 Figure 2.5: Packaged NXP MEMS microphone

2.2.1 Microphone Sensitivity

Before getting into the discussion of electrical sensitivity we will introduce the
term called compliance. The elastic behavior of membrane is commonly quan-
tied by this term instead of spring constant . The compliance (C) provides a
measure of exibility of the membrane and it is expressed as

1
C= (2.1)
k
where k is the spring constant in N/m. Hence, the unit of compliance is m/N.
The equation that describes the electrical sensitivity of the membrane is
given by [11],

Vbias
Se = Am Cef f (2.2)
g
where Se denotes the Electrical sensitivity, Cef f the eective compliance
and Am the area of membrane. Vbias is the bias voltage of the device and g is
the distance between the membrane and backplate, also known as the air-gap
thickness.
The eective compliance of the membrane depends on the mechanical mem-
brane compliance extracted from C-V measurement Cm,cv , the back volume
compliance Cbv

Vbv
Cbv = (2.3)
ρa c2 A2m
and the electrical spring softening compliance Ckt,el :

2g 3
Ckt,el = 2 V2
(2.4)
επRm bias
Cef f depends on these terms according to :

1 1 1 1
= + − (2.5)
Cef f Cm,CV Cbv Ckt,el

12
 where Vbv  Back volume, c  Speed of sound, ρa  Density of air, Rm -
radius of membrane, ε- permittivity in F/m.
The spring softening compliance is negative since it opposes the spring restor-
ing force. The equations describe how the membrane compliance and the back
volume inuence the sensitivity of the device. The back volume reduces the
membrane movement which will aect the device performance. Therefore an
optimal back volume for better performance has to be identied. Thus the
measurement of membrane compliance and the analysis of back volume eects
are necessary.
From Eqn.2.2 it is clear that the sensitivity depends on dierent parameters
such as area of the membrane, eective compliance, bias voltage and the air
gap. Therefore it is always a trade o between these terms to get the maximum
sensitivity. For example, increasing the area of membrane in order to enhance
the sensitivity is limited by the desired small device dimensions.

2.3 Body-Noise Cancellation

Body-noise is the undesired eect of a MEMS microphone to behave as an ac-
celerometer. In other words, body-noise means that structure-borne sound is
converted by the microphone into an electrical output, which cannot be distin-
guished by the receiver from the sound signal. Body-noise is a performance-
limiting factor of microphones in a vibration environment. Body movement due
to external vibrations is compared with normal operation in Fig.2.6.

Figure 2.6: Body movement in comparison with normal operation[10]

Fig.2.6 (a) shows the normal operation of the microphone for the acoustic
signal. In Fig.2.6 (b) the movement of membrane due to external mechanical
vibrations can be identied. This movement causes the microphone membrane
to vibrate, so that the capacitance between the plates changes. This change in
capacitance produces a noise signal, which aects the performance of the micro-
phone. Potential sources of body-noise are the vibrations in the environment,
such as a working machine where the microphone is mounted or the propagation
of sound of the speaker through the mobile phone circuits. The operational fre-
quency range of the microphone is from 20Hz to 20kHz. This is well above the
frequency (12Hz) of the vibrations produced by human body, so human-body
movements are not important since they are intrinsically ltered [10].

13
 NXP came up with several techniques to cancel this body noise [12]. At
present, two methods are under consideration :

1. An intrinsic solution, based on matching the frequency response of the

backplate to that of the membrane. This technique is called frequency
matching. It will be discussed in detail in the next subsection.

2. The second method uses a separate accelerometer, placed next to the

microphone sensor. This accelerometer acts as a deaf microphone, which
senses only the body-noise signals [12]. The output of the accelerometer
is subtracted from the output of the microphone. The accelerometer can
be integrated in the same wafer as the microphone. The advantage of this
technique is that the nal device can be used as a microphone as well as
an accelerometer, oering an additional functionality for mobile devices.

14
2.3.1 Frequency Matching

Figure 2.7: Frequency matching technique (a) Acoustic signal membrane move-
ment (b) For acceleration signal membrane moves much more than the back-
plate (No Frequency matching) (C) After frequency matching both membrane
and backplate move similarly in response to an external acceleration.

The intrinsic solution thus removes the body noise by making the backplate
equally sensitive for the mechanical vibrations as the membrane. In other words,
it is to match the membrane and backplate movement for the mechanical vibra-
tions. This way, both backplate and membrane deect the same when mechan-
ically actuated, making the capacitance constant due to external vibrations.
Since the backplate is perforated it is not sensitive for the acoustic signal so
that the microphone is still sensitive to sound. We visualized the body noise
cancellation procedure in Fig.2.7.
Fig.2.7(a) shows the normal operation of the device for acoustic signal where
the membrane moves and the backplate not. In Fig.2.7(b) the membrane moves
much more than the backplate due to the external acceleration as the membrane

15
is much more exible than the backplate. Hence the body noise is generated.
Fig.2.7(c) shows that after frequency matching the backplate is moving similarly
to the membrane due to the external acceleration, thus avoiding the noise signal.
Hence, the backplate has to be designed in such a way that it is as sensitive
to the external acceleration as the membrane. This can be explained in a simple
approach. The membrane and backplate can be represented as a spring-mass
system with a mass `M' and a spring constant `k' for the external acceleration
as illustrated in Fig.2.8.

Figure 2.8: Mechanical model of the sensor for body acceleration[11]

Consider two simple mechanical formula's, the rst being Newton's second
law of motion

F =M ∗a (2.6)

with F the force on the mass, M the mass and a the acceleration. The second
equation is Hooke's law

F =k∗x (2.7)

for an extension x of the spring due to a force F with k the spring constant.
When a mass is connected to a spring, these two forces are equal and we nd

F =k∗x=M ∗a

M
x(a) = ∗a (2.8)
k
stating that the displacement of the mass (or the extension of the spring) is
proportional to the applied acceleration a. The proportionality is given by the
spring constant k and the mass M.
The displacement `x' of the mass due to acceleration `a' has a sensitivity

dx M
=
da k

16
 according to Eqn.2.8. This expression is valid for a static situation, i.e.
for frequencies considerably below the fundamental resonance frequency of the
system. The same ratio M/k appears in the resonance frequency (fres ) of the
spring mass system :

r
1 k
fres = (2.9)
2π M
dx
Therefore, we may conclude that the sensitivity to acceleration,
da , and the
resonance frequency (fres ) are coupled inherently. By reducing the resonance
frequency, we can thus improve the sensitivity to acceleration.
M
From Eqn.2.8 and Eqn.2.9 it follows that if the ratio
k is equal for both
membrane and backplate, the displacement due to acceleration will be equal.
This is exactly the desired behavior. Hence, body-noise compensation can be
achieved by matching the resonance frequencies of membrane and backplate.

Figure 2.9: Comparison of backplate designs for freq. matching

In the nominal microphone design the resonance frequency of the backplate

is higher than the membrane resonance frequency. In a new design to achieve
frequency matching the backplate is suspended by micromachined springs in-
stead of the common rigid connection. Fig.2.9 shows the comparison between
the nominal, rst-generation and the suspended designs. The stress in the
backplate layer is reduced by the use of springs, thus making it more exible,
resulting in lower resonance frequency.

17
2.4 Compliance

Figure 2.10: Frequency spectrum of the microphone with both membrane and
backplate resonance frequencies, indicating that sensitivity increases with re-
duction in resonance frequency

As stated earlier, the elastic behavior of membranes is quantied by the term

called compliance (Eqn.2.1) instead of the spring constant. Hence, we can ex-
press the resonance frequency as

r
1 1
fres = (2.10)
2π C ∗M
This expression shows the relation between the resonance frequency and the
compliance. By reducing the resonance frequency, the compliance increases,
leading to increase in sensitivity to acceleration. An increased sensitivity to
acceleration implies that the membrane is more exible, and thus also allows an
increased acoustical sensitivity.
Fig.2.10 shows the frequency spectrum of the microphone obtained based
on an approximate lumped element model. Both the membrane and backplate
resonance frequencies are visible in the spectrum. Fig.2.10, shows that the
acoustical sensitivity increases with reduction in membrane resonance frequency
from 120 kHz to 80 kHz. In the lumped element model only the resonance
frequency of the membrane is tuned and the backplate resonance frequency is
not altered.

18
Chapter 3

Laser Vibrometer
Measurement
This chapter focuses on the laser vibrometer measurement technique and the
set-up used. Additionally, the results of the measurements are discussed.

3.1 Measurement Setup

Laser vibrometer measurement allows us to visualize the fundamental as well
as the higher mode frequencies. Furthermore, it allows us to determine the
displacement of the backplate due to an applied bias voltage. The operating
principle of this setup is shown in Fig.3.1.

Figure 3.1: Polytec Laser Vibrometer working principle[13]

The laser beam is directed at the surface of the device, which is placed on
the chuck, and the vibration amplitude and frequency are extracted from the
Doppler shift of the laser beam frequency due to the motion of the surface.
The beam from the laser is divided into a reference beam and a test beam

19
with a beamsplitter. The test beam then passes through the Bragg cell, which
modulates the frequency. This beam is then directed to the target. The motion
of the target adds a Doppler shift to the beam. The captured image is processed
in the computer using the built-in algorithms. The result is visualized in the
computer screen by the Polytec Laser Vibrometer software. The graphical user
interface (GUI) of this software is shown in Fig.3.2, showing the movement of
the backplate in the top half and the respective frequency spectrum with all the
resonance peaks in the bottom half.

Figure 3.2: Polytec software interface and the measured resonance peaks of the
backplate

We used this equipment to measure the resonance frequencies of the circular

membrane and backplate. We note that we can only determine the dierent
resonance peaks, (i.e) the fundamental resonance frequency and the higher mode
frequencies, if the entire surface of the plate can be scanned with the laser.
In the case of the NXP MEMS microphone sensor, the membrane is located
beneath the backplate, as was illustrated in Fig.2.3. Hence it is dicult to scan
the surface of the membrane, while it is easy to investigate the backplate. We
will, therefore, rst discuss measurements on the backplate. At the end of this
section we will show how the membrane can be investigated.

20
Figure 3.3: Measured fundamental and higher-order resonance modes of the
backplate

Several higher modes of the circular backplate are shown in Fig.3.3. We per-
formed this measurement in the laser vibrometer by selecting a specic scanning
grid. This scanning grid consists of an array of measurement points on which
the laser beam is consecutively focused during the measurement. Fig.3.4 shows
the scanning grid formed over the backplate.

Figure 3.4: Scanning grid with scan points superposed on an optical image of
the backplate

Two types of measurements are possible, a non-contact vibration measure-

ment and a measurement that involves electrical actuation of the device. We
will only discuss measurements on electrically actuated devices.

21
 Figure 3.5: Polytec Laser Vibrometer Setup[10]

The Polytec laser vibrometer setup is shown in Fig.3.5. This setup is used
for the electrical measurement in which the device is placed on the chuck and
the probes are connected to the contacts. A constant DC voltage is applied with
an actuation AC voltage from the signal generator to the microphone. The fre-
quency is swept over a wide range from 1 kHz to 3.5 MHz. The frequency range
limitations are based on the selection of specic vibrometer controller. This vi-
brometer controller allows us to select the correct decoder for the measurements.
We can select the Vibrometer controller as either VD-06 (350 kHz) or VD-09 (2
MHz) based on the frequency range we need to measure. This laser vibrometer
has a maximum frequency range of about 20 MHz and the resolution is down
to tens of pico meters.

Figure 3.6: Dierent measurements performed on the laser vibrometer, (a) fo-
cusing laser beam on the backplate, (b) focusing laser beam on the membrane,
(c) focusing laser beam through the acoustic hole on the membrane in a pack-
aged sample

22
 The observation of membrane movement from the bottom of the wafer is
not possible with the current setup, since the laser beam as well as the electri-
cal connections are made from the top. Fig.3.6 illustrates how we can perform
measurements on the membrane, even though this is located below the back-
plate. As shown in Fig.3.6(b) the laser beam must then be focused through
the backplate hole on the membrane. It is in this case not possible to scan the
entire surface of the membrane. Fig.3.6(c) illustrates another membrane mea-
surements. A packaged sample can be positioned such that the membrane is on
top, so that we do not have to focus through the backplate hole. Commonly,
we use a microscope objective with 2X magnication. In order to analyse the
movement of the backplate and the membrane more in detail, we also use 50X
magnication. We present the results in the next sections.

3.2 Test Measurements

Laser vibrometer measurements provide us with details about the operation of
the device. Especially the deection proles can be measured and the resonances
can be identied. Before presenting the actual measurement results in Sec.3.3,
we will now discuss some test measurements.
The measurements are repeated several times to check the reproducibility of
the measurement. These repeated measurements are done for the same condi-
tions, (1) by repeating the experiment for several times without lifting the probe
needles and (2) by lifting and re-connecting the needles after a single measure-
ment. The repeated measurements revealed that the results do not change with
repetition.
Furthermore, we performed a measurement to check the inuence of vibrom-
eter range of the Polytec setup on the measured readings. This is analyzed by
performing a measurement for changing the vibrometer range for same condi-
tions. The results for the varying vibrometer range prove that the choice of
range does not inuence the measured results.
The measurement plots supporting these conclusions are presented in Ap-
pendix B.

3.2.1 2X and 50X Magnication Measurement

We compared results for dierent magnication by performing measurements
on the backplate of the device in wafer. We varied the AC voltages at dierent
DC voltages for both 2X and 50X magnication. The advantage of 50X mea-
surement is that we can focus the laser beam on the backplate perfectly and the
real movement of the backplate can be analyzed.
The advantage of 2X magnication is that we have access to the entire
backplate surface, allowing us to specify a grid (see Fig.3.4) and investigate the
dierent resonance modes. Fig.3.7 shows the frequency spectrum of the device
at 2X magnication. The backplate displacement is shown on the y-axis and it
can be observed that the displacement peaks at resonance frequency (180 kHz).
The displacements values plotted in Fig.3.8 are obtained at 40 kHz.

23
Figure 3.7: Backplate displacement at 2X normal magnication at 5Vdc and
0.3 Vac (Measurement)

Fig.3.8 shows that the displacement increases with the AC as well as the DC
voltage.

Figure 3.8: Backplate displacement at 2X normal magnication with respect to

dierent DC and AC voltages (Measurement)

Fig.3.9 shows the frequency spectrum of the device at 50X magnication

along with the backplate displacement in y-axis. Fig.3.10 shows the relation be-
tween the backplate displacement measured at 40 kHz with DC and AC voltages
at 50X.

24
Figure 3.9: Backplate displacement at 50X magnication at 5Vdc and 0.5 Vac
(Measurement)

Figure 3.10: Backplate displacement at 50X magnication with respect to dif-

ferent DC and AC voltages (Measurement)

The 2X and 50X measurement results of the backplate show that the dis-
placement increases for increasing AC and DC voltages. We note that the
2X magnication measurement is inuenced by both membrane and backplate
movement, because the laser beam can scatter at the backplate holes or it might
even pass through the holes. The 50X magnication measurement, on the other
hand, only measures the backplate deection. Since it allows us to focus the
laser beam on the region between the holes, it gives the pure displacement of
the backplate. This gives us an idea that the reading obtained from 50X oers
the real backplate movement. However, the 50X magnication does not allow
to scan the displacement prole over the full backplate. The 2X magnication
should therefore be used to measure qualitative mode shapes. The measured
mode-shapes agree well with the expectations that the next higher order has
one nodal line (zero amplitude) more than the previous.

25
3.3 Backplate Frequency Measurement Results

Figure 3.11: Frequency spectrum of backplate measured in laser vibrometer.

Also higher mode frequencies are visible

This measurement is performed on the wafer sample shown in Fig.2.4. The wafer
is placed on the chuck below the laser scanning head of the measurement setup.
The backplate is on top and can thus be easily scanned. The backplate resonance
frequencies are measured at the normal 2X magnication. We actuated the
backplate by applying an AC voltage of 50mV. Due to the electrostatic force
between the backplate and the membrane, the plates will oscillate at the applied
frequency. In addition, we applied a DC voltage and performed measurements
for VDC = 5V. The dierent resonance modes observed from the backplate are
shown in Fig.3.3 and the corresponding frequency spectrum is shown in Fig.3.11.
Table.3.1 shows the comparison between measured and calculated backplate
resonance frequencies. The higher-order resonance frequencies are calculated
using the Eqn.4.11. First from the measured f(0,0) frequency, the stress value is
calculated using the Eqn.4.11 and by using the stress value and corresponding
Eigen values, the higher-order frequencies are calculated.

Backplate Resonance Measured (kHz) Calculated (kHz)

Modes
f(0,0) 162 162 (Measured)
f(0,1) 402 368
f(0,2) 695 577
f(0,3) 1060 787

Table 3.1: Comparison of measured and calculated higher mode frequencies

It is observed that the calculated values for the higher modes are dierent
from the measured frequencies (see Table.3.1). The reason for this dierence
is the bending stiness of the backplate. The bending stiness of the tension
dominated backplate is related to the stress and thickness. Here the equation
(Eqn.4.11) used to calculate the higher modes are based only on the stress,
and the thickness is neglected. This thickness term and additionally the mate-
rial properties like Young's modulus will have some inuence on the resonance

26
frequency which leads to the dierence between the calculated and measured
values.
Both stress and resonance frequencies are related to each other through the
Eqn.4.11. It is to be noted that the initial stress in the backplate is 160MPa after
deposition, while the stress calculated using the measured resonance frequency
is 64MPa. If we use 160MPa for calculating the rst resonance frequency using
the Eqn.4.11 we end up with 215 kHz, while in reality the measured frequency
is 162 kHz. Finite element method (FEM) simulations show that the initial
stress of the backplate layer relaxes to 65 MPa as the plate is patterned [7].
This indicates that the measurement results are in good agreement with the
simulations.

3.4 Measurement of membrane at Low Frequency

3.4.1 Wafer sample measurement
We investigated the membrane as was illustrated in Fig.3.6(b). This measure-
ment is performed to analyze the frequency response below the membrane reso-
nance frequency. These measurements are performed on the microphone samples
on the wafer, without package.
In order to analyze the displacement, dierent measurements are performed
such as varying the AC voltage at dierent DC voltages. These measurements
are performed at higher magnication 50X in order to focus the laser beam
through the backplate hole on the membrane. Fig.3.6(b) shows this measure-
ment set-up.
Fig.3.12 shows the frequency spectrum of the device along with the mem-
brane displacement in y-axis. The measured results show that the displacement
values are high when compared to backplate displacement (see Fig.3.10), since
the membrane is more exible. Fig.3.13 shows the dependence of the membrane
displacement on DC and AC voltages. The membrane displacement increases as
expected. We note that this measurement is complicated as it is hard to focus
the laser beam through the holes. The results are therefore not so accurate.

Figure 3.12: Membrane displacement seen through the backplate holes at 5Vdc
and 0.5 Vac (Measurement)

27
Figure 3.13: Membrane displacement with respect to dierent DC and AC volt-
ages (Measurement)

3.4.2 Packaged sample measurement

This measurement is performed to scan the membrane surface to identify the
inuence of the package on the resonance modes of the membrane. The packaged
sample is used in the laser vibrometer measurement as shown in Fig.3.6(c).
In this measurement the laser beam is focused on the membrane through the
acoustic hole.
The acoustic hole is not placed above the center of the membrane and it
only allows optical access to a section of the membrane. The frequency spectra
of the measured device for two dierent frequency ranges are shown in Fig.3.14
and Fig.3.15.

Figure 3.14: Measurement of membrane frequency spectrum of a packaged sam-

ple till 100 kHz in laser vibrometer (Measurement)

28
Figure 3.15: Membrane frequency spectrum of a packaged sample till 350 kHz
(Measurement)

Fig.3.14 clearly shows the presence of an extra resonance at approximately

25 kHz, which is as strong as the membrane resonance. We attribute this extra
resonance to the package. In Fig.3.15 it is impossible to identify the resonance
frequency of backplate because of the interference with resonances caused by
the air enclosed in the package.

3.5 Extraction of Electrostatic Force

The low frequency measurements are performed to nd a way to extract the
electrostatic force acting on the membrane and backplate. A simple electro-
static force equation used from displacement (x) and spring constant (k) the
electrostatic force (Fe ) can be written as (Hookes Law),

Fe
x= (3.1)
k
the spring constant in this equation can be used either from the membrane
or backplate to extract the force acting on them. The displacement obtained
for the backplate at 50X magnication at 5V and 1 Vac is around 38pm (see
Fig.3.10) and the spring constant calculated for the backplate using Eqn.3.2 is
2406 N/m.

1
Cm = (3.2)
2ρhπ 3 Rm
2 f 2 (0)
res
Finally the AC electrostatic force extracted using Eqn.3.1 is 91nN. This is a
rough and very simple calculation, it gives an idea of the amount of electrostatic
force acting on the backplate and we did not derive a theoretical expression to
compare this value.

3.6 Summary
In this chapter, we rst discussed about the laser vibrometer operating principle
and its measurement setup. Then the test measurements performed such as,
repeatability measurement, changing vibrometer range measurement and 2X

29
and 50X magnication measurement were shown. Later the measurements on
backplate and membrane in both the wafer and package sample were discussed
and nally an introduction about electrostatic force extraction was shown.
From the backplate frequency measurement, both the theoretical and mea-
sured frequencies are compared and the dierence between the values is ana-
lyzed. It is identied that the dierence is due to the bending stiness of the
backplate and stress relaxation due to backplate perforation. The stress values
extracted from the measured frequencies are in good agreement with the FEM
simulations.
The low frequency measurements are performed to analyze the displacement
of backplate and membrane. These measurements are performed at both 2X
and 50X magnications. From the results, the trend of the displacement with
respect to AC and DC voltages are identied. We note that the results from 2X
measurements are not accurate because the laser beam can scatter through the
backplate holes or it can pass through the holes at certain scanning region of
the backplate. This disadvantage is avoided in 50X magnication measurement
since it focuses only on the no hole region of the backplate very accurately and
it gives the pure displacement of the backplate. However, the 50X magnication
does not allow to scan the displacement prole over the full backplate. The 2X
magnication should therefore be used to measure qualitative mode shapes.
A 50X measurement is performed to analyze the displacement of membrane.
This measurement is complicated as it is hard to focus the laser beam through
the holes and the results are not so accurate because of scattering caused by
the edge of holes or the moving membrane. This displacement measurement is
used to extract the electrostatic force measurement using the Hookes law. The
simple calculations provided the amount of force acting on the backplate.
Finally the packaged sample measurements provided an overview of the air
damping inside the package. This results in shifting the resonance frequency
of the membrane. This packaged measurement results (Fig.3.15) looks similar
to the results (Fig.4.11(c)) obtained from electrical impedance measurement of
packaged sample under air.

30
Chapter 4

Electrical Impedance
Measurement
This chapter focuses on the electrical impedance measurement. We used two
setups for this experiment. Also the extraction of key parameters such as reso-
nance frequency and compliance along with the results obtained are discussed.

4.1 Frequency Response Measurement

As stated earlier in Chapter 2 the measurement of the device frequency response
is necessary to analyze the device performance. Here the same frequency re-
sponse is measured and the mechanical resonances of membrane and backplate
show up in the electrical impedance of the sensor.
Impedance is the hindrance to the ow of alternating current . Impedance
(Z) is expressed as a combination of Resistance (R) and Reactance (X) and is
measured in ohms (Ω) [14]. It can be expressed as a complex quantity:

Z = R + j X

This impedance measurement oers access to the dissipation. Dissipation

is also denoted as losses and this can be calculated from the measured real
and imaginary part of the impedance (Eqn.4.1). This dissipation will be high
at resonance frequency fres and it can be seen in Fig.4.1. This dissipation
peaks at the resonance frequency because of the large change in capacitance
at resonance and it dissipates more energy. This helps us to determine the
resonance frequency of the membrane and backplate.

Re(Z) R
D= = (4.1)
|Im(Z)| |X|
Another electrical measurement is the Capacitance  Voltage (C-V) measure-
ment. As the name states, this measurement is to determine the dependency
of capacitance between the microphone membrane and backplate with respect
to the DC voltage. It helps to determine the mechanical compliance of the
membrane (Cm ).

31
 Figure 4.1: Dissipation as a function of frequency (Measured)

4.1.1 Measurement Set-ups

The measurement set-up used is a HP Impedance Analyzer (HP 4194A) probe
station. Fig.4.2 shows the experimental set-up for the measurement. The gure
shows the set-up in vacuum conditions, the same set-up can be used for the
air measurements without the vacuum chamber. The advantage of vacuum
measurement is that clear and sharp peaks can be seen at resonance whereas
the air damping shifts and broadens the resonance peaks.

Figure 4.2: Electrical Impedance measurement setup

The MEMS microphone is placed on the chuck and the probe needles are
connected to the membrane and backplate contacts. The high potential is con-
nected to the membrane and the low one is connected to the backplate contact.
Since this is an electrical measurement we need to apply an external potential
to actuate the device. This setup can be used for both frequency-dependent and
voltage-dependent measurements.
Other setup used is the HP Multifrequency RLC  meter (HP 4275A). Fig.4.3
shows the measurement setup used for this impedance measurement. It is mostly
used for capacitance voltage measurement. MEMS microphone is placed on the

32
chuck and the probes are connected same as for the Impedance measurement.

Figure 4.3: Capacitance - Voltage measurement setup

4.1.2 Measurement setup Calibration

Both these measurement setups are calibrated in a dierent way. The HP
impedance analyzer is calibrated using an external standard substrate, which
consists of dierent resistors of standard (50Ω) values and short circuit connec-
tions.

Figure 4.4: Comparison of measurement setups C-V curve measured (Before)

Those standard and short circuit connections are used to calibrate the impedance
analyzer together with an open circuit calibration. On the other hand just an
open calibration is performed for the RLC Multifrequency meter by just pressing
a button in the setup.
We performed tests to check whether the measurement setups provide sim-
ilar results. Capacitance voltage measurement is performed on HP Impedance
analyzer and HP RLC Multifrequency meter for ve MEMS microphone sam-
ples under same conditions. It is observed that the capacitance voltage curves
measured for ve samples using the two setups with same measurement con-
ditions are dierent. It shouldn't be the case, since the samples have to give
the same capacitance values for both setups. This dierence can be seen in the
capacitance voltage curve in Fig.4.4 and it shows the comparison of C-V curve
for one sample between two setups.

33
 In order to analyse the dierence between the setups a xed capacitance
measurement is performed on both setups. The xed capacitor is a ceramic
capacitor with a value of 3.9 ± 0.1 pF, whose value is stable over a wide range
of voltage and frequency. After measuring the capacitor in both setups for
a variety of frequency and voltages, the result obtained from HP Impedance
analyzer is 4.46pF and from the RLC Multifrequency meter is 3.87pF. This
gives us an idea that there is a dierence of around 0.5pF measured from the
Impedance analyzer and it caused the dierence in measured capacitance results.

Figure 4.5: Comparison of measurement setups C-V curve measured(After)

We found that the dierences between both setups are because of the calibra-
tion techniques used. The cause for the problem is identied as the 50Ω resistors
in the external standard substrate. Since the device is old and the repeated use
before every measurement caused the device to degrade and it is no more equal
to 50Ω (Standard value) but was between 53Ω and 60Ω. Using the degraded
device to calibrate the impedance analyzer caused dierence in the measure-
ments. Finally the open calibration method used in the RLC Multifrequency
meter is followed in the HP Impedance analyzer. Since this is the most robust
calibration for the low frequency and small capacitances which gives the least
calibration uncertainties. After this change in calibration method, capacitance
values measured agrees each other quite well in both setups. Fig.4.5 shows the
C-V curve with new calibration technique.
From Fig.4.5 we can see that the dierence between the two setups measured
capacitance voltage curves are now reduced from 0.5pF to around 0.02pF. This
dierence is reasonable and it could be from the parasitic capacitances of the
probes to the microphone substrate. The inuence of calibration techniques on
the membrane compliance extracted are listed in Appendix A.

4.2 Resonance Frequency

First we will show how the pure mechanical resonance frequency is extracted
from the electrical impedance measurement. Pure mechanical resonance fre-
quency is the value obtained at zero bias voltage. At zero bias voltage the

34
resonance is not visible in the electrical signal, but the 0V frequency can be
extrapolated.
In general the mechanical resonance frequency of the spring mass system is
given by the equation [11]

s
1 Kef f
fres = (4.2)
2π mef f
where Kef f is the eective spring constant in N/m and mef f is the eective
mass in kg, ρ is the density of silicon in kg/m3 , hm is the thickness of membrane
in m and Rm is the radius of the membrane in m. Eective parameters are used
here because the membrane is not rigid and therefore cannot be regarded as
a point-mass. Assuming a parabolic deection prole, the eective membrane
mass equals half the actual mass [11].

m 1 2
mef f = = ρhm πRm (4.3)
2 2
In order to use this Eqn.4.2 for our resonance frequency measurement, a
slight modication has to be done. In our case Kef f is a combination of both
mechanical and electrical spring constant in which the mechanical part (Km )
is independent of the bias voltage and the electrical part (Kel ) is completely
dependent on the bias voltage. This can be given by the equation

s
1 Km − Kel (Vbias )
fres (Vbias ) = (4.4)
2π mef f
Here the electrical spring constant takes the negative sign, since it acts
against the mechanical spring force. The electrical spring constant for center
deection is given by the equation [11]

2 2
επRm Vbias
Kel (Vbias ) = (4.5)
2g 3
where ε is the dielectric permittivity of air, Rm is the radius of the circu-
lar membrane or backplate and g is the air gap between the membrane and
backplate. Replacing Kel in Eqn.4.4 [11]

s
1 Km επRm2 V2
bias
fres (Vbias ) = − 3
(4.6)
2π mef f 2g mef f
From this equation it is clear that the resonance frequency depends on the
bias voltage. With increase in the bias voltage the resonance frequency shifts
down. This eect is called the Electrostatic Spring Softening. We need to correct
for this eect in order to determine the pure mechanical resonance frequency.
Additionally the higher harmonics and the resonances of the backplate can
be measured. This can be useful to compare and validate with nite element
simulations [7].

35
4.2.1 Extraction of Resonance frequency
As stated earlier this measurement is performed in the HP Impedance analyzer
and this is done in both air and vacuum conditions. This measurement is
performed to measure the complex impedance and the main task is to identify
the mechanical resonances of both membrane and backplate. This measurement
is done at dierent bias voltages Vbias superimposed with an actuation voltage
Vac and by sweeping the frequency over a wide range. This AC voltage acts as
an actuation voltage for the oscillating membrane.
The complex impedance is represented by series capacitance Cs and dissi-
pation factor D. Alternatively, parallel capacitance Cp can be selected but the
problem with this Cp is that the value drops at higher frequencies due to the
series resistance of the electrodes. This drop can be seen during the measure-
ments that the value of series capacitor at 2 MHz is 3.3 pF and the parallel
capacitor is 2.9 pF (Fig.4.6).

Figure 4.6: Series capacitance Vs parallel capacitance at 2MHz (Measured)

The dissipation (D) is calculated using the Eqn.4.1 and it is plotted as a

function of frequency. Fig.4.7(a) shows the dissipation (D) as a function of
frequency in air. It can be seen that there are two peaks indicating the funda-
mental resonance of membrane around 75 kHz and the fundamental resonance
of backplate around 170 kHz. Since this measurement is performed in air the
resonance peaks are strongly damped due to the friction with the air. The
same dissipation as a function of frequency measured in vacuum can be seen
in Fig.4.7(b). Here the peaks are sharp and its easy to identify the resonances.
Additionally the peak seen at the 200 kHz in vacuum is the rst higher mode
of the membrane which will be discussed in the later section.

36
Figure 4.7: Dissipation as a function of Frequency as extracted from impedance
measurement performed in (a) air (b) vacuum

Also the series capacitance is calculated from the measured impedance us-
ing the Eqn.4.7. This helps us to identify the resonance frequency since the
capacitance drops at the resonance.

−1
Cs = (4.7)
Im(z) ∗ ω

ω = 2πf (4.8)

Fig.4.8 shows the series capacitance as a function of frequency in both air

and vacuum. It is observed that the capacitance change at resonance in vacuum
is sharp and the values are large indicating the resonance frequencies.

Figure 4.8: Series Capacitance as a function of Frequency as extracted from

impedance measurement performed in (a) air (b) vacuum

37
 As said earlier this resonance frequency includes the electrical part. From the
Fig.4.9, we can see that the dissipation increases with increase in bias voltage
indicating that the peaks are getting steeper and the frequency shifts down.
This eect is used for extraction of the mechanical resonance frequency. In
order to extract the mechanical resonance, the Eqn.4.6 can be rewritten as,

Figure 4.9: Dissipation as a function of frequency measured for several bias

voltages

2
επRm
fres (Vbias )2 = fres (Vbias = 0)2 − V2 (4.9)
8π 2 g 3 mef f bias
2 2
This shows that there is a linear relation between Vbias and fres . The fre-
quency at which the dissipation is high is collected for each bias voltage and
2 2
fres is plotted against Vbias (Fig.4.10). Since this relation is linear we obtain a
linear t, which can be related to Eqn.4.9 to extract the mechanical resonance
frequency at zero bias voltage.
The linear equation obtained from the plot is

kHz 2
 
f 2 [kHz 2 ] = −64.2 + 6830 kHz 2
 
V 2

Comparing with Eqn.4.9 at Vbias = 0V,

√
2
fres (Vbias = 0) = 6830kHz 2 = 83kHz

Mechanical resonance frequency obtained for the membrane is 83kHz and

for the backplate is 170kHz in air for the wafer sample.

38
 Figure 4.10: Linear plot of Frequency square Vs Voltage Square

The error associated with the resonance frequency can be calculated using
the following procedure. A function calledLINEST (known_y's, known_x's,
const, stats) is available in the Microsoft Excel program, in which
- known_y's is the frequency square value
- known_x's is the voltage square value
- const is a logical value specifying whether to force the Y-axis crossing to
equal 0. In our case we don't want to force our constant to zero. Thus it is
selected as TRUE or 1.
- stats is a logical value specifying whether to return additional regression
statistics. If this term is TRUE or 1, the function will return us the errors
associated with the obtained values.
For example, the LINEST function result for the linear t in Fig.4.10 is given
in Table4.1 in which, fres (0)2 is the constant and σfres
is the error associated
2 (0)
0
0 επR2 2
with the constant and the term − 64.2[kHz ] = −
8π 2 g 3 mef f which is related
to the slope 'm' and the error associated with the slope is σm .

m[ kHz
2

V 2 ] -64168933.12 fres (0)2 [kHz 2 ] 6833429194

2
σvm [ kHz
V2 ] 5819521.533 σfres 2
2 (0) [kHz ] 182159506.7

Table 4.1: LINEST function data of the linear t

The error equation obtained for resonance frequency using error linearisa-
tion is given in Eqn.4.10 and the statistical error in the membrane resonance
frequency is 1 kHz.

σfres
2 (0)
σfres (0) = (4.10)
2fres (0)

4.2.2 Air and Vacuum Measurement Results

As said earlier, this impedance measurement was measured under both air and
vacuum conditions. The measurements are performed on both wafer (Fig.2.4)

39
and packaged samples (Fig.2.5). As was shown in Fig.4.7 it is easy to identify the
sharp peaks of the resonance frequency of both membrane and the backplate.
Additionally the higher mode frequencies are visible in the vacuum measure-
ment, which can be compared with the laser vibrometer measurement. This air
and vacuum measurement comparison is used to check the inuence of the air
damping and the back volume on the moving membrane. Damping is the eect
that reduces the amplitude of the oscillation. In our case, the resonance peaks
are damped which will shift the resonance frequency of the membrane. This
eect can be seen clearly in the measurement results obtained.

4.2.2.1 Wafer and Packaged Sample Measurement Results

Figure 4.11: Comparison of Dissipation Vs Frequency for both wafer and pack-
aged sample (Measurement)

First the impedance measurement results obtained from the wafer and packaged
sample in both air and vacuum conditions are discussed. Fig.4.11 shows the
dissipation vs. frequency plot obtained for both wafer and packaged sample in
air and vacuum.

40
 The dierence between the two measurements can be clearly seen from the
Fig.4.11(a) and (b). The air damping of the resonance frequencies are under-
stood from the broad and sharp peaks. The mechanical resonance frequencies
extracted from dierent samples both in air and vacuum conditions are listed
in the Table.4.2.
From the values listed in the Table.4.2 it is observed that the membrane
resonance frequencies extracted from the wafer samples in air are almost in
the same range as that of vacuum. It shows that the air damping in case of
wafer measurements are not aecting the resonance frequencies of the membrane
much. The reason could be that the wafer is placed over the chuck and the back
volume is nearly innite and the air passage is very smooth without aecting
the oscillating membrane much.

Wafer Resonance Resonance Packaged Resonance Resonance

Sample in air in Sample in air in
No (kHz) vacuum No (kHz) vacuum
(kHz) (kHz)
1 82 ± 0.8 84 ± 1.0 1 74 ± 1.0 83 ± 0.7
2 79 ± 0.8 80 ± 0.8 2 75 ± 0.9 81 ± 0.8
3 82 ± 0.4 84 ± 1.0
4 81 ± 0.6 81 ± 0.9
5 83 ± 1.0 82 ± 0.8

Table 4.2: Membrane resonance frequencies in air and vacuum for both wafer
and vacuum samples (Measurement)

On the other hand the results obtained for the package samples in both air
and vacuum conditions are dierent. This can be observed in the Fig.4.11(c)
that the inuence of air in the package is quite high on the moving membrane.
It is almost impossible to identify the backplate resonance frequency in air
measurement. Additionally we see a peak at low frequency range at around
23 kHz. This peak is also visible in the Laser Vibrometer packaged sample
measurement in Fig.3.14. This peak is assumed as the resonance of the package,
where the air in the package is dominating the membrane resonance and it shifts
its frequency.
If we compare the membrane frequency results extracted from both wafer
and package samples in vacuum measurements they are almost matching with
each other. This gives a good sign that the results obtained from vacuum
measurements give info about the membrane instead of the package. The com-
parison between Fig.4.11(c) and (d) gives an idea what's the eect of damping
on the moving membrane. It seems to be quite high. Hence an optimal back
volume that is the volume inside the package has to be selected which gives a
good sensitivity for the present device parameters such as 5V bias, 2 μm air
gap, area of membrane and eective compliance. The back volume selected for
3
the device is 4mm , the reason to select this back volume is that, increasing
the back volume for increased sensitivity could shift the rst package resonance
into the audio range and more obvious reason is that the customer will not be
interested in big size packaged device.

41
4.2.2.2 Backplate resonance frequency and stress
The resonance frequency of the backplate can be extracted in both cases in the
same way as explained for the membrane in Sec.4.2.1 and the values obtained
are around 180 kHz these values are listed in Table.4.3.
The backplate resonance frequencies listed in Table.4.3 for both air and
vacuum conditions are almost same. This indicates that the backplate resonance
frequencies are not aected much by the damping in case of wafer samples. But
the packaged samples are heavily damped and its hard to see the backplate
resonance peaks (see Fig.4.11(c)). This is other proof for the inuence of air in
package. Here also the vacuum measurements provide reliable data about the
backplate.

Wafer Resonance Resonance Packaged Resonance Resonance

Sample in air in Sample in air in
No (kHz) vacuum No (kHz) vacuum
(kHz) (kHz)
1 175 181 1 Hard to 180
see
2 180 176 2 Hard to 182
see
3 181 175

Table 4.3: Backplate resonance frequencies in air and vacuum conditions mea-
sured on wafer and packaged samples (Measurement)

The mechanical resonance frequency extracted can be used to calculate the

residual stress in the membrane and the backplate using the Eqn.4.11 [11].

0 r
ωnm σ
fnm = (4.11)
2πR ρ
where f nm is the frequencies of the Eigen modes, σv is the Stress in Pa, ρ is the
0
3
density of Si in kg/m , R is the radius of membrane/backplate in m and ωnm
is the Eigen-value corresponding to the respective Eigen-mode of the tension
dominated plate in which  n and m represent the number of nodal diameter
and nodal circles, respectively. The Eigen values for the lowest twelve Eigen
modes are listed in Table.4.4.

(n,m) (0,0) (1,0) (2,0) (0,1) (3,0) (1,1)

ω'nm 2.41 3.83 5.14 5.52 6.38 7.02

(n,m) (4,0) (2,1) (0,2) (5,0) (3,1) (6,0)
ω'nm 7.59 8.42 8.65 8.77 9.76 9.94

Table 4.4: First 12 Eigen modes and Eigen values for a tension dominated plate
[11]

The values used in Eqn.4.11 to extract the residual stress in membrane and
backplate are, f00,mem = 83 kHz, f00,bp = 170 kHz, Rm = 460μm, Rbp = 460μm,
3
ρ = 2301 kg/m and by using the Eigen value for the rst Eigen mode (0,0),

42
the stress calculated in membrane is 23MPa and in backplate 96MPa. These
extracted values can be compared with the stress values obtained from nite
element simulations [7] to check how well our simulations and reality agrees
with each other.
As discussed in Sec.3.3, the theoretical value calculated for the backplate
frequency using the initial 180 MPa stress is 215 kHz. The dierence between the
theoretical and measured frequencies are observed in this electrical measurement
as well. This dierence is because of the stress variation during processing.
Extensive research is going on related to the stress relaxation to identify the
exact reason for the mismatch [7].
The membrane stress value (23 MPa) calculated and the Eigen-mode value of
5.52 are used in the same Eqn.4.11 to calculate the rst higher mode frequency
of the membrane. The rst higher mode of the membrane calculated is around
190 kHz. This frequency can be seen in the dissipation Vs frequency vacuum
plot at around 200 kHz. This gives a good sign that our theoretical calculations
are agreeing with the real device.
Additionally the term Compliance, which is the inverse of Spring Constant,
can be calculated from the resonance frequency. Equation4.12 is used for the
calculation of membrane compliance.

1
Cm = (4.12)
2ρhπ 3 Rm
2 f 2 (0)
res

4.3 Capacitance  Voltage Measurement

Figure 4.12: (a) Biased stage (5V) of both membrane and backplate and (b)
Pull-in voltage stage

As the name states, this measurement is to determine the dependency of the

capacitance between the microphone membrane and backplate with respect to
the DC bias voltage. This measurement is done at dierent frequencies by
sweeping the bias voltage. In principle this measurement has to be performed
when the membrane is stable and its not oscillating. This is possible only

43
at zero bias voltage, but at zero bias voltage we can't measure the change in
capacitance with respect to voltage. In order to perform this measurement an
optimal frequency has to be selected where the membrane is stable and there
should not be any interference from the resonance frequencies of the backplate.
The DC bias voltage between the membrane and backplate is varied and
an AC actuation voltage is superimposed with the bias voltage. Due to the
electrostatic force (Eqn.3.1) the membrane is attracted towards the backplate.
During the measurement the distance between the membrane and backplate
decreases with respect to voltage thus increasing the capacitance. At certain
voltage the membrane collapses with the backplate and that voltage is called
the Pull-in voltage. When the applied bias voltage is subsequently reduced,
the membrane will release eventually from the backplate. The voltage at which
the membrane releases back from the backplate is called the Pull-out or release
voltage. Fig.4.12 shows the normal stage of the device and the Pull-in stage.
The Pull-in and Pull-out voltages are indicated in the capacitance voltage
curve of our device (Fig.4.13). The curve shows the bias voltage is swept from
10V to 10V and back. As stated earlier we can observe the change in capaci-
tance with respect to the bias voltage. Also in the curve the value of capacitance
suddenly increases to a value greater than 10pF. The voltage at which the ca-
pacitance increases suddenly is the Pull-in voltage and at certain voltage the
membrane releases back from the backplate indicating that it is coming out of
the Pull-in stage. The voltage at which the membrane releases is called the
Pull-out voltage.

Figure 4.13: Capacitance Vs Voltage plot (Measurement)

4.3.1 Extraction of Membrane Compliance

The extraction of membrane compliance is described here. The Pull  in volt-
age of NXP MEMS microphones are typically in the order of 7.5V. Eqn.4.13
shows the equation for the Pull  in voltage [11]. This equation can be used to
calculate the membrane compliance (Cm ) value from the Pull  in voltage and
the membrane compliance calculated for 7.5V pull-in is 12 mm/N.

44
 s
g3
Vpull−in = 2 C
(4.13)
2επRm m

Another procedure to extract the membrane compliance will be discussed

now. The Eqn.4.14 shows the relation between the capacitance and the bias
voltage at low voltages. This relation is used to extract the membrane compli-
ance.

Co2 Vbias
2
C(Vbias ) = Co + Cm (4.14)
4g 2
C0 is the membrane - backplate capacitance without parasitic capacitances
Cp which add another constant capacitance to this Eqn.4.14. C0 is calculated
2
ε0 πRm
from the plate capacitance formula C0 = = 3.09pF . The value obtained
g
for the constant from the t (Fig.4.15) is dierent from the theoretical value
calculated. This is because, the value obtained from measurement has parasitic
capacitance included in it.
From the equation it is understood that the relation between capacitance
and voltage is a polynomial. This equation is valid only for 60% of Pull  in
voltage. In our case 60% of Pull  in is 4.5V. Thus a polynomial plot can be
drawn if we zoom into the range of 4.5V to 4.5V. Fig.4.14 shows the plot of
that valid range.

Figure 4.14: Capacitance Vs Voltage plot for the valid range

As expected we have a parabolic behavior in the valid range. In order to

extract the compliance value from the data, we need to plot the valid range
2
(60%) Vbias Vs Cs . The plot obtained is shown in Fig.4.15. The linear equation
obtained from the plot can be related to Eqn.4.14 and the membrane compliance
is extracted. The membrane compliance extracted is 6.5 mm/N.

45
Figure 4.15: Capacitance Vs voltage square plot for the valid range (Measure-
ment)

The error associated with the compliance value is calculated in the same way
explained in Sec.4.2.1. The values obtained after applying LINEST function for
C02
linear t is given in Table4.5. Here a= 4g 2 Cm .

a 4.95 x 10
−15
Co + Cp 3.73 x 10
−12

−16 −15
σva 1.41 x 10 σvCo +Cp 1.39 x 10

Table 4.5: LINEST function data of the polynomial t

The error obtained for membrane compliance using this theory is given in
Eqn.4.15 and the statistical error in the membrane compliance is approximately
0.2 mm/N.

4g 3
 
σCm = 2
σa (4.15)
επRm
Additionally, the membrane compliance calculated from the pull-in can be
compared with the results extracted from the Eqn.4.14. The compliance calcu-
lated using the pull-in voltage is 12 mm/N and the extracted values are around
6.5 mm/N. In other words, a factor of two dierent between the values. The
reason for the dierence could be that the approximations used for compliance
extraction is not much accurate, since it is based on the lumped element model.
Also in our Eqn.4.14 the capacitance C0 is used for a solid plate but in reality
the fringing elds due to the perforation in backplate will aect the capacitance
between the plates. The drop of capacitance with respect to this perforation
level is analyzed by NXP researchers and it is found that a drop of 15% in ca-
pacitance for 30% perforation [15]. More analysis has to be done to check and
prove this dierence since not many literature's are available to cross verify it.

46
4.3.2 Optimal Frequency selection
As stated earlier, this capacitance voltage measurement is done by sweeping
the DC bias voltage at certain frequency. Here the measurement and analysis
performed to identify the optimal frequency is discussed. Capacitance voltage
measurements are performed on ve samples on wafer for dierent frequency
values such as 20 kHz, 200 kHz, 1 MHz and 2 MHz.

Figure 4.16: Membrane compliance comparison for 2 samples for dierent fre-
quencies (Extracted)

Fig.4.16 shows the comparison of compliance values extracted for 5 samples

at dierent frequencies. It can be observed from the extracted compliance val-
ues, they are high at low frequency and it gradually decreases with increase in
frequency and nally it is almost stabilized at high frequencies. This gives us
a clear picture of the membrane compliance trend with respect to frequency.
The reason for this behavior is that at low frequencies such as 20 kHz, the in-
ertial force of the membrane is small and it will move in accordance with the
frequency, which makes the membrane oscillate more, and with increase in fre-
quency it is unable to follow such high frequency and nally it comes to almost
stable position. This gives us a situation where the membrane is almost stable
and it is the desired condition for the capacitance voltage measurement, which
made us to select 1 MHz as the optimal frequency for all the measurements. A
further advantage of 1 MHz is that this frequency does not coincide with a reso-
nance of the sensor. At a resonance, the membrane will not be sti, as desired.
Hence, 200 kHz would not be a suitable choice, since this is in near vicinity
of both membrane and backplate resonance frequencies Fig.4.11(a). It is not a
good idea to rely on the values obtained close to the resonance frequencies.

47
4.3.3 Inuence of AC voltage

Figure 4.17: Inuence of AC voltage on compliance values (Extracted)

This measurement is performed to nd the inuence of AC voltage superimposed

with the DC voltage. This eect has to be analyzed since it is necessary to
identify the parameters inuencing the resonance frequency and the membrane
compliance. Capacitance voltage measurement is performed for dierent AC
voltages from 3mV to 50mV for two dierent samples and the compliance values
obtained are shown Fig.4.17. From the gure it is observed that the inuence
of AC voltage on the compliance values is negligible and the values agree with
each other within the t error range. This gives us an idea about the choice
of AC voltage and now it proves that any value of AC can be selected for our
measurements. The AC voltage selected for our measurements is 200 mV.

4.3.4 Membrane compliance in vacuum

Capacitance voltage measurement is performed on packaged MEMS micro-
phones in both air and vacuum conditions by sweeping the DC bias voltage
at 1 MHz frequency and the compliance values are extracted. These extracted
values are listed in the Table.4.6.

Packaged Compliance Compliance

Sample in air in vacuum
No (mm/N) (mm/N)
1 8.6 7.9

Table 4.6: Membrane compliance values in air and vacuum conditions measured
on wafer and packaged samples (Measurement)

From the result listed in Table.4.6 it is observed that the values are dierent
for both air and vacuum measurements. In general, the value should be same in
both cases, since we are measuring at 1 MHz and the membrane is assumed to
be stable. The reason is unclear for this change in compliance values. In order

48
to identify the trend, more measurements has to be performed on dierent
packaged samples. With more results, it will be easy to predict or identify the
device behavior.

4.4 Summary
In this chapter, the electrical impedance measurement techniques and their
measurement setups were discussed. The extraction of resonance frequency and
compliance values were shown. Also the results of wafer and packaged samples
for both air and vacuum conditions were listed.
First the dierence in calibration techniques between dierent measurement
setups were identied. The reason for the dierence is because of the degrada-
tion of the external standard substrate used for calibration. Finally the open
calibration is followed in all the setups and the dierence in readings were elim-
inated.
In case of resonance frequency measurement, series capacitance is used in-
stead of parallel capacitance because of drop in parallel capacitance value at
higher frequencies.
From the extracted membrane resonance frequency measurement results (Ta-
ble.4.2) on both wafer and packaged samples under air and vacuum conditions,
the inuence of air in the package is identied. The air and vacuum measure-
ment results on wafer matches well but on the other hand the air and vacuum
measurements on packaged sample is dierent. The resonance frequencies mea-
sured and the stress values extracted from the resonance frequencies under vac-
uum conditions matches perfectly with the FEM simulation results. This proves
that the vacuum measurement results are reliable and those are the ones to be
considered for comparing with FEM simulations.
Finally from the C-V measurements the membrane compliance is extracted
and the pull-in voltage of the device is identied. The optimal frequency to
perform C-V measurement is identied as 1 MHz. Also the inuence of AC ac-
tuation voltage over the compliance is found to be none. The membrane compli-
ance values extracted from C-V, calculated from pull-in voltage and membrane
resonance frequencies are compared with each other and the dierences between
the values are analyzed in Chapter.5.

49
Chapter 5

Measurement Results
Comparison and Discussion
In this chapter, results from resonance frequency, compliance and laser vibrom-
eter measurements are compared and discussed.
First the impedance and capacitance - voltage measurements are related to
each other by the relation given by Eqn.4.12. This shows us that the membrane
compliance values can be calculated from the extracted membrane resonance
frequency. This calculated membrane compliance can be compared with the
membrane compliance extracted from capacitance - voltage measurement and
the membrane compliance calculated from the pull-in voltage. Table.5.1 lists
the values obtained from all these techniques on the very same sample.

Membrane Membrane Membrane

compliance from compliance from compliance from
Resonance pull-in voltage C-V
frequency (80 kHz) (mm/N) (mm/N)
(mm/N)
13 (Eqn.4.12) 12 (Eqn.4.13) 6.5 (Eqn.4.14)
9.4 (Eqn.5.3)

Table 5.1: Membrane compliance comparison from dierent measurements

From the values listed, it can be observed that the membrane compliances
from resonance frequency and pull-in voltage are in the same range and the
compliance extracted from C-V is dierent. One reason for this dierence is
that the approximation used for resonance frequency extraction is not accurate
since it is based on a quasi static model (lumped element model); But the shape
of the deection prole changes with respect to frequency. The compliance
value calculated from the resonance frequency is static membrane compliance.
In order to compare these two values we need to convert the static membrane
compliance into dynamic membrane compliance. This can be done by using the
conversion factor obtained from the equations described below.
Static membrane compliance is given by [11],

50
 1
Cm w (5.1)
4πσm hm
and the dynamic membrane compliance is given by [11],

0 1
Cm w (5.2)
2.89πσm hm
The ratio of static and dynamic obtained is,

0
Cm
w 0.72 (5.3)
Cm
and this factor is used to convert the static compliance into dynamic com-
pliance. These converted values are listed in Table.5.1 and still the values are
dierent. Furthermore, the extraction of membrane compliance from C-V is at
the condition where the membrane is assumed to be stable. But we have seen
in Sec.4.3.2 that at lower frequencies or close to resonances the C-V curve is
inuenced. From Sec.4.3.2 it follows that the error is small for the chosen fre-
quency of 1 MHz. Hence the compliance extracted from C-V is static membrane
compliance.
Another factor that could account for this dierence is the electrical eld
distortion (Fringing eld) due to the perforation in backplate. As stated earlier,
the perforation in the backplate leads to drop in capacitance [15]. This means
that the electrostatic force is smaller for a perforated backplate than for a solid
one as assumed in the C-V equation (Eqn.4.14). The compliance extracted from
C-V is therefore too low.
Another assumption which could have an inuence is the tting range in
case of C-V measurement. Now it is stated that the 60% of pull-in is the valid
range for compliance extraction. In case if the tting range is 40% then the
value extracted will be dierent. More analysis and perfect modelling has to be
done to understand this dierence in depth.
The compliance from pull-in is higher than the other values. The reasons
are unclear for the dierence in the compliance values. May be the fringing eld
play a role, since the analytical model assumes a solid plate. In reality the plate
is perforated and it deforms with respect to voltage.
From the measurement results listed in Table.5.2 it is identied that the
results from vacuum measurements are the ones to be trusted. It matches very
well with the FEM simulations (Table.5.2) and provides a good data to com-
pare the stress values extracted from vacuum measurement with the simulation
results (Table.5.3).

51
 Resonance Resonance Resonance
in vacuum in vacuum in
(kHz) (kHz) calculated
Wafer Packaged (kHz)
Sample Sample FEM
Simulation
[7]
84 ± 1.0 83 ± 0.7 80
80 ± 0.8 81 ± 0.8 80

Table 5.2: Resonance Frequency extracted from wafer and package samples in
vacuum compared with FEM simulation results

Membrane Membrane Membrane

stress stress Stress
extracted extracted calculated
(MPa) (MPa) (MPa)
Wafer Packaged FEM
Sample Sample Simulation
(Vacuum) (Vacuum) [7]
22 24 20
23 23 20

Table 5.3: Membrane stress extracted from membrane resonance frequency for
wafer and package samples in vacuum compared with FEM simulation results

In case of stress value comparison between simulation and measurement re-

sults, vacuum results gave a good match and this shows that the values obtained
are more realistic. But vacuum setup is complicated and it won't be available
in all labs. The most available measurement is the C-V measurement. In case
of sensitivity it is directly related to compliance and this is measured at the
condition where the membrane is stable. It has no inuence from resonance
frequency and its deection proles only if frequency is chosen well.
The laser vibrometer (LV) measurement results can be compared with the
resonance frequencies and their higher modes extracted/calculated from impedance
measurements. Laser vibrometer gives us a clear picture to visualize the oscilla-
tions of the backplate and its fundamental and higher mode frequencies can be
measured. Table.5.4 lists the values of backplate resonance frequencies measured
in laser vibrometer and the rst resonance of backplate measured in electrical
impedance and the calculated higher modes using Eqn.4.11 and we can observe
that the values deviate from the measured values for the higher modes. One
reason for this dierence is the bending stiness of the backplate. Here the
equation (Eqn.4.11) used to calculate the higher modes is based only on the
stress and the thickness and bending stiness are neglected. This thickness
term and additionally the material properties like Young's modulus will have
some inuence on the resonance frequency which leads to increased higher mode
frequencies due to the added forces from the bending.

52
 Production From laser From
sample vibrometer impedance
measure- measure-
ment ment
(kHz) (kHz)
f00 189 182
(Measured)
f01 460 415
(Calculated)
f02 780 650
(Calculated)

Table 5.4: Backplate resonance frequency and higher mode frequencies compar-
ison between Laser vibrometer and electrical impedance

Another measurement performed using laser vibrometer on packaged sample

can be compared with the impedance measurements of packaged samples in air.
When we compare the laser vibrometer results (see Fig.3.14 and Fig.3.15) with
Fig.4.11(c), the presence of the peak at around 25 kHz is visible in both results
and the complete frequency spectrum agree with each other quiet well.
It can be hoped that the laser vibrometer enables new measurements in the
future to gain more understanding of the device mechanics. For example, the
low frequency measurements performed to analyze the displacement (Fig.3.13)
can be extended to extract the electrostatic force.

53
Chapter 6

Conclusions and
Recommendations

6.1 Conclusions
The work presented in this report focused on the investigation and comparison
of dierent measurement principles and techniques used to measure the key de-
vice parameters, resonance frequency and compliance. The inuence of ambient
pressure, bias voltage and back volume on the resonance frequencies were an-
alyzed. The performance limiting factor called body noise was introduced and
the techniques to avoid it was shown. One key technique to eliminate body
noise is the frequency matching. Hence it becomes obvious to measure and
analyze the resonance frequency of the membrane and backplate. Additionally
the compliance term which is related to the sensitivity of the device was also
measured and analyzed.
Electrical impedance measurements provide us details about the inuence of
bias voltage over the resonance frequency and the inuence of air in the package
by comparing the air and vaccum results. Also the inuence of the electrode
resistance at higher frequencies is proved by the drop in Cp . Cs is then a better
choice. Additionally from the comparison of two dierent setups used for elec-
trical impedance measurements, it is identied that the calibration techniques
were dierent which results for the mismatch between them. With the common
open calibration in both setups resulted for the agreement of readings. Hence
open calibration is followed for all the electrical impedance measurement as the
least error-prone calibration. From the results of the electrical impedance mea-
surements it is identied that the vacuum measurement results are more reliable
and it gives info purely related to membrane and backplate. The capacitance
voltage measurements provided the pull-in voltage of the device and the mem-
brane compliance extracted was compared with other measurement readings
and the mismatch was analyzed. Also the optimal frequency for C-V measure-
ment is identied as 1 MHz and the inuence of AC actuation voltage over the
membrane compliance is found to be none.
Laser vibrometer measurements provided us with details about the oper-
ation of the device especially the deection proles and the resonances were
easily identied. From the dierence between measured and calculated higher

54
mode resonance frequencies, it is concluded that the inuence of the bending
stiness of the backplate and other material properties like Young's modulus
are not negligible. The stress values extracted from the measured fundamental
resonance frequencies are in good agreement with FEM simulations. The nor-
mal 2X and 50X magnication measurements provides us with details about the
dierence between the magnications. The problem with the 2X magnication
measurement is identied as being inuenced by both membrane and backplate
movement. The 2X magnication should therefore be used to measure quali-
tative mode shapes. In case of 50X magnication measurement it focuses the
scanning surface very accurately and it gives the pure readings of the subject.
This shows that the reading obtained from 50X is the one to trust. However,
the 50X magnication does not allow to scan the displacement prole over the
full backplate.
It has been shown that laser vibrometer and electrical impedance measure-
ments are related to each other and the results obtained are in the same range.
This is important as electrical measurements take less eort and equipment,
while the laser vibrometer measurements are direct measurements of the me-
chanical movement.

6.2 Recommendations
Some recommendations can be pointed out to extend the work performed here.
In case of laser vibrometer measurement only the electrical measurements were
performed. Additionally mechanical actuation is possible by placing the sample
over a shaker. This helps to visualize/determine the asymmetric mode shapes of
the circular membrane/backplate which can then be compared with the calcu-
lated values. Also the measurement performed for displacement can be extended
to extract the electrostatic force, if proper equations are derived.
In case of electrical impedance measurement, the equations used for com-
pliance extraction should be modied including the eects of fringing eld.
In spring suspended backplate design, the resonance frequency is tunable by
the number and properties of springs and the desired frequency range can
be reached. The most important next step is to process a full sensor with a
spring suspended backplate. Measurement of resonance frequencies of spring
suspended backplates can be performed to check the reduction in the body
noise. This is the most important value of the sensor and a detailed analy-
sis and measurements have to be performed. The mechanical stability of the
suspension is an important point which needs to be investigated.

55
Bibliography
[1] G. Meijer, Smart Sensor Systems. WILEY, 2008.
[2] P. French, Silicon Sensors. Delft University of Technology, 2009.

[3] Electronics Advocate, Mems microphone market, February 2010.

[Online]. Available: http://www.electronicsadvocate.com/wp-content/
uploads/2010/02/isupplimemsmarket.jpg

[4] Wikipedia, Microphone, April 2010. [Online]. Available: http:

//en.wikipedia.org/wiki/microphone

[5] F. S. Consulting, Silicon mems microphone market set to grow

strong, says the information network, March 2010. [Online]. Available:
http://futuresource.trackandmonitormedia.com/?p=32638

[6] Design News, Tiny mems microphone boosts audio quality, Septem-
ber 2009. [Online]. Available: http://www.designnews.com/article/
339420-Tiny_MEMS_Microphone_Boosts_Audio_Quality.php

[7] A. F. V. Quintero, "master thesis - stress evolution of functional layers

inmems devices throughout the processing", 2010.

[8] R. Jaeger, Introduction to Microelectronic Fabrication, 2nd ed. Prentice

Hall, 2002.

[9] Paul Mcwhorter, About mems, May 2009. [Online]. Available:

http://www.memx.com/

[10] F. Felberer, Master thesis - characterisation of mems microphones for body

noise, 2009.

[11] I. Bominaar-Silkens, NXP Semiconductors, Tech. Rep., 2009, in prepara-

tion.

[12] G. Langereis, Microphone models for body-noise cancellation, NXP Semi-

conductors, Tech. Rep. NXP-R-TN 2008/00313, 2008.

[13] M. Goossens, The polytec laservibrometer:measuring mems movements on

the nano scale, February 2010. [Online]. Available: http://nww.nxp.com

[14] Wikipedia, Electrical impedance, May 2010. [Online]. Available:

http://en.wikipedia.org/wiki/Electrical_impedance

[15] G. Langereis, A lumped element model for mems microphones, NXP

Semiconductors, Tech. Rep. PR-TN-2005/00209, 2005.

56
Appendix A

Inuence of calibration
techniques on the extracted
results
Membrane compliance values extracted for 5 samples measured in both Mul-
tifrequency RLC meter and Impedance Analyzer are listed in Table.A.1. This
measurement is performed with old calibration technique and the inuence of
the calibration can be observed in the results. The results are dierent for same
samples measured in dierent setups under same conditions.

Sample No Multifrequency RLC Impedance Analyzer

meter (mm/N) (mm/N)
1 6.4 8.5
2 6.2 8.7
3 7.0 8.4
4 6.5 8.9
5 6.7 9.1

Table A.1: Compliance values for 5 samples from dierent measurement setups

Table.A.2 shows the membrane compliance values extracted for 5 samples in

dierent setups under new open calibration technique. It can be observed that
the results are matching with each other and this open calibration technique is
followed for all other measurements.

57
 Sample No Multifrequency RLC Impedance Analyzer
meter (mm/N) (mm/N)
1 6.4 6.6
2 6.2 6.4
3 7.0 6.8
4 6.5 6.6
5 6.7 7.0

Table A.2: Compliance values for 5 samples from dierent measurement setups
after new calibration

58
Appendix B

10 cycles with same

conditions (2X)
DC Voltage = 5V, AC voltage = 0.5 V, Vibrometer range = 500mV, Freq.
Range = 0 to 200 kHz

Figure B.1: Ten measurements for the same conditions for same sample to check
the reproducibility of the measurement

59
Measurement for changing vibrometer range

Figure B.2: Measurements for the same conditions for same sample to check the
inuence of the vibrometer range

60