Detection and Mitigation of Electrostatic Pull-in Instability in MEMS Parallel
Plate Actuators
by
Colin Stevens
A dissertation submitted to the Graduate Faculty of
Auburn University
in partial ful llment of the
requirements for the Degree of
Doctor of Philosophy
Auburn, Alabama
August 3, 2013
Keywords: MEMS, Resonator, Parallel Plate Actuator, Pull-in, Current Sensing
Copyright 2013 by Colin Stevens
Approved by
Dr. Robert N. Dean, Chair, Associate Professor of Electrical and Computer Engineering
Dr. Thaddeus Roppel, Associate Professor of Electrical and Computer Engineering
Dr. Thomas Baginski, Professor of Electrical and Computer Engineering
Dr. Vishwani Agrawal, Professor of Electrical and Computer Engineering
Dr. Darren Olsen, Assistant Professor of Architecture, Design and Construction
Abstract
Electrostatic MEMS actuators are used in a wide variety of applications including micro-
machined gyroscopes, high speed mechanical switches, variable capacitors, and vibration
isolation devices. MEMS parallel plate actuators (PPAs) are simple to realize, yet complex
nonlinear variable capacitors. If a DC voltage is applied in an attempt to move the proof
mass greater than 1/3 of the electrode rest gap distance, the device becomes unstable and
the electrodes snap into contact. Most research into this pull-in phenomenon is devoted
to extending the operational range of motion past the 1/3 instability point. This usually
involves the addition of complex external electronics. Many electronics applications, however,
only require that the actuator remain out of the pull-in region, and do not require an extended
stable range of motion. If detection of the pull-in event is all that is required, then simpler
solutions can be realized, minimizing the requirements on the driving signal. Once pull-in is
reached, the velocity of the movable plate increases rapidly until the plates make contact. The
decreasing distance causes a proportional increase in capacitance. To maintain a constant
voltage across the plates, an inrush of current must  ow into the actuator in response to the
increased capacitance. This work presents a method for detecting the inrush current using a
transimpedance ampli er circuit to convert the current to a measurable voltage. Once pull-
in is detected, the PPA is electrically shutdown to prevent damage to the actuator or the
voltage source, thus mitigating pull-in. A simulation of the expected results was performed
using a Simulink model for the actuator structure predicting the expected range of inrush
current. This result was then veri ed using a silicon micro-machined PPA connected to the
detection/mitigation circuit on a Printed Circuit Board (PCB). The experimental results
follow closely with the simulation allowing precise control in mitigating the pull-in event.
ii
Table of Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Mathematical Theory and Motivation . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Parallel Plate Actuator Overview . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.1 Derivation of Electric Field of A Parallel Plate Capacitor . . . . . . . 3
2.1.2 Derivation of Force Equations of a Parallel Plate Actuator . . . . . . 7
2.1.3 Stability Points For the PPA . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.1 Disadvantages of Pull-In . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.2 Advantages of Pull-in . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.1 MEMS Devices that Make Use of Pull-in . . . . . . . . . . . . . . . . . . . . 20
3.2 Actuator Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4 Theory Development of Pull-In Sensing Mechanism . . . . . . . . . . . . . . . . 25
4.1 Current Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.1.1 High Side Current Sensing . . . . . . . . . . . . . . . . . . . . . . . . 25
4.1.2 Low Side Current Sensing . . . . . . . . . . . . . . . . . . . . . . . . 28
4.1.3 Low-Side Transimpedance Ampli cation Sensing . . . . . . . . . . . . 29
iii
4.2 Hall E ect Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3 Optical Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.4 Piezoresistive Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.5 Secondary Capacitance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.6 Current Transformer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.7 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5 PPA Design and Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.1 Intellisuite Simulation of PPA device . . . . . . . . . . . . . . . . . . . . . . 35
5.2 Simulink Simulation Pull-in Detection Operation . . . . . . . . . . . . . . . 37
5.3 Fabrication Equipment and Processing . . . . . . . . . . . . . . . . . . . . . 42
5.3.1 Cleaning Equipment and Processes . . . . . . . . . . . . . . . . . . . 42
5.3.2 Patterning Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3.3 Silicon Processing Equipment . . . . . . . . . . . . . . . . . . . . . . 50
5.4 Plasma Assisted Dry Etching of Silicon . . . . . . . . . . . . . . . . . . . . . 51
5.4.1 Plasma Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.4.2 Reactive Ion Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.4.3 Inductive Coupling of Plasma . . . . . . . . . . . . . . . . . . . . . . 53
5.4.4 Deep Reactive Ion Etching . . . . . . . . . . . . . . . . . . . . . . . . 53
5.4.5 Electron Beam Evaporation and Sputtering System . . . . . . . . . . 57
5.4.6 Masking Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.4.7 Silicon Oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.5 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6 Experimental Veri cation of Pull-in Detection and Mitigation . . . . . . . . . . 61
6.1 Schematic and Layout of Test Platform . . . . . . . . . . . . . . . . . . . . . 62
6.2 Results of Test Platform Measurement . . . . . . . . . . . . . . . . . . . . . 68
6.3 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7 Extension of Pull-in Work to Create a Tunable Resonator . . . . . . . . . . . . 75
iv
7.1 Pull-in Resonator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.2 Pull-in Resonator Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
7.3 Pull-in resonator Experimental Veri cation . . . . . . . . . . . . . . . . . . . 80
7.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
9 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
9.1 Pull-in Detection Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . 90
9.2 Pull-in Resonator Future Work . . . . . . . . . . . . . . . . . . . . . . . . . 91
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Appendices 97
A PPA Traveller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
B Electron Beam Metal Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . 102
C Listing of Lab Equipment by Room#/Laboratory . . . . . . . . . . . . . . . . . 103
D Matlab Code used to Analyze Interferometer Data . . . . . . . . . . . . . . . . . 105
E Schematic Drawing of PPA Board Mount Fixture . . . . . . . . . . . . . . . . . 110
v
List of Figures
2.1 Electric Field Due to a Line Charge. . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Electric  eld due to an in nite sheet of charge. . . . . . . . . . . . . . . . . . . 6
2.3 Visualization of the electric  eld between two in nite sheets of charge. . . . . . 7
2.4 Voltage as a function of movable plate displacement from equation 2.33. . . . . 12
2.5 Pole Zero map of actuator at di erent points of stability . . . . . . . . . . . . . 15
4.1 Simpli ed schematic of a high side current sense circuit . . . . . . . . . . . . . . 27
4.2 Compensation circuit to account for high voltage MEMS devices . . . . . . . . . 28
4.3 Schematic of a low side current sense circuit . . . . . . . . . . . . . . . . . . . . 29
4.4 Schematic of a transimpedance ampli cation circuit . . . . . . . . . . . . . . . . 30
4.5 Polytech OFV Vibrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.1 Intellisuite simulation of PPA structure . . . . . . . . . . . . . . . . . . . . . . . 36
5.2 Transmissibility of the fabricated PPA . . . . . . . . . . . . . . . . . . . . . . . 38
5.3 Simulink Model of PPA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.4 Current and Position Results of Simulink PPA Under Ramping Voltage. . . . . 41
5.5 Spin Rinse Dryer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
vi
5.6 Vapor HMDS Application Chamber . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.7 Photoresist Spinner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.8 Matrix Oxygen Plasma Asher . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.9 [Karl Suss MA/BA6 Contact Mask Aligner . . . . . . . . . . . . . . . . . . . . 50
5.10 STS Advanced Silicon Etcher . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.11 Oxidation Furnace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.12 CHA Mark 50 dual E-beam/sputter/ion gun deposition system . . . . . . . . . 58
6.1 Electrical Schematic of Transimpedance ampli er Pull-in Detection Circuit . . . 64
6.2 Layout of Transimpedance Ampli cation Circuit. . . . . . . . . . . . . . . . . . 66
6.3 Fabricated Transimpedance Ampli cation Circuit. . . . . . . . . . . . . . . . . . 67
6.4 Pull-in Event Without Safety Shutdown. . . . . . . . . . . . . . . . . . . . . . . 69
6.5 Safety Shutdown at 4.88e-7 Amps. . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.6 PSafety Shutdown at 8.82e-7 Amps. . . . . . . . . . . . . . . . . . . . . . . . . 73
7.1 Mass Spring Damper System Diagram. . . . . . . . . . . . . . . . . . . . . . . . 76
7.2 Actuator Circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
7.3 Simulink Model Using 4M Resistance. . . . . . . . . . . . . . . . . . . . . . . . 79
7.4 Simulink Model Using 8M Resistance. . . . . . . . . . . . . . . . . . . . . . . . 80
7.5 PCB Electrode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
vii
7.6 Fabricated PPA Spring System. . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
7.7 Assembled PPA test device. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7.8 Pull-in Oscillations at Various Resistances. . . . . . . . . . . . . . . . . . . . . . 84
7.9 FFT Analysis of Pull-in Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
7.10 Frequency Trend as a Function of Resistance. . . . . . . . . . . . . . . . . . . . 86
E.1 Bottom of  xture Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
E.2 Top of  xture Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
viii
List of Tables
5.1 Intellisuite modal simulation of designed PPA device. . . . . . . . . . . . . . . . 36
5.2 Parallel Plate Actuator Device Properties . . . . . . . . . . . . . . . . . . . . . 39
5.3 Simulink Analysis of Current as a Function of PPA Position . . . . . . . . . . . 42
5.4 RCA Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.5 MORGNSOI DRIE Recipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
6.1 Experimental Analysis of Current as a Function of Position . . . . . . . . . . . 70
B.1 Electron Beam Metal Deposition Recipe . . . . . . . . . . . . . . . . . . . . . . 102
ix
Acknowledgements
I would like to thank Dr. Robert Dean for o ering his guidance in my graduate cur-
riculum, as well as his valuable input on this project, and Mr. Charles Ellis for lending his
expertise in the many areas that I had no knowledge. I would like to thank my father William
Stevens, my mother, Martha Stevens, and my sister, Kristen Fink for their encouragement
and support. I would also like to thank my co-workers in the AT&T engineering department
for their encouragement to return to college and pursue my graduate education.
x
Chapter 1
Introduction
MEMS parallel plate actuators are a commonly used type of electrostatic actuator that
have applications in a number of di erent kinds of micro-machined devices, including RF
devices [1], variable capacitors [2], accelerometers [3], micro-mirrors, and active vibration
isolators [4]. One issue that is inherent to all parallel plate actuators (PPA) is the condition
of pull-in. Pull-in is the inability of a PPA to be electrostatically actuated beyond one third
of its rest gap distance without becoming unstable. Most PPA devices in widespread use
are designed so that they only operate while in an open-loop stable range of motion. These
devices must be designed so that they are not actuated beyond this point unless additional
circuity is added to prevent pull-in from occurring.
Two of the driving factors in MEMS production today is the reduction of size and
cost. Many of the current systems in place to prevent pull-in have disadvantages that either
increase complexity of the system which can result in both a size and a cost increase, or
somehow degrade the performance of the device. If a solution to prevents pull-in from
occurring can be found that does not introduce excessive additional electronics to the system
or have adverse e ects in the operation of the device, then cost as well as size bene ts can
be realized.
This dissertation introduces the concept of the PPA, presents the theory behind its
operation, including the dynamics of motion, and points of instability. A review of the
current work being conducted into PPA?s to increase the e ciency of motion is investigated.
A simple but overlooked method of detection of PPA instability is then proposed that has the
potential to prevent damage to systems without su ering an increase in cost or complexity
of the system. By using a transimpedance ampli er to greatly amplify the current signal
1
passing through the actuator, a system can be warned of impending instability and take
steps to either mitigate or record the event. To justify this proposal, a through simulation
of the dynamics of motion is presented, and then veri ed experimentally.
In addition to the proposed transimpedance ampli cation solution to detect pull-in by
current ampli cation, introduction of a simple passive resistor to cause a modi cation of
the current can cause an interesting behavior in the dynamics of motion which can cause
the device to go into an oscillatory state. This behavior of pull-in can be utilized to create
a tunable resonator that relies only on a DC voltage and one passive component. A full
implementation of this is presented which can be used in conjunction with the detection
method previously discussed.
2
Chapter 2
Mathematical Theory and Motivation
The following chapter provides the mathematical analysis of the PPA required to un-
derstand the operation of the PPA. A derivation of the force equations covered and the
concept of pull-in instability is explained. The e ects of pull-in are then analyzed to give
an understanding of the advantages and disadvantages of the pull-in phenomena and the
motivation of this work.
2.1 Parallel Plate Actuator Overview
The parallel plate capacitor can be generally characterized as two parallel in nite sheets
of opposite charge located a set distance from each other if fringing  elds are neglected.
2.1.1 Derivation of Electric Field of A Parallel Plate Capacitor
The force acting on the capacitor can be found by deriving the equation for the electric
 eld between the two plates, which are also referred to as electrodes, from Coulomb?s law.
Using the concept of symmetry, the mathematical derivation can be greatly reduced as shown
below. By  rst  nding the equation for the electric  eld at a point due to an in nite line,
it is shown that the the components of the electric  eld vectors that are not normal to the
point cancel out. [5]
3
Figure 2.1: Electric Field Due to a Line Charge.
~E = Q
4 "R2 ^ar (Coulomb?s law point of charge) (2.1)
In 2.1, ~E is the electric  eld vector, Q is the charge in coulombs, " is the permittivity
of the material through which the electric  eld travels, R is the radius from the charge, and
^ar is the unit vector.
dQ =  ldz (2.2)
The distance from each of these points ~R has a magnitude R and a unit vector direction
that can be expressed in its component form shown in 2.3
~R = R^ar =  ^a  z^az (2.3)
4
From equation 2.3 the magnitude and unit vector direction of each dQ can be found in terms
of its component vectors shown in 2.4
R =p"2 +z2; ^ar =  ^a  z^azp"2 +z2 (2.4)
By substituting back into the Coulomb?s Law equation, the electric  eld vector at a point
due to each discrete charge density can be found. Then by summing all of the discrete charge
densities, the total electric  eld on the point due to the in nite line can be found as shown
in 2.5, 2.6, 2.7.
E =
Z  
ldz
4 "
 ^a  z^az
( 2 +z2)32
(2.5)
E =  l ^a 4 "
Z +1
 1
dz
( 2 +z2)32
(2.6)
~E =  l
2 " ba (2.7)
The equation for the electric  eld due to an in nite sheet of charge can then be found
by using the previous equation for an in nite line of charge and integrating over an in nite
number of lines as shown below. [5]
5
Figure 2.2: Electric  eld due to an in nite sheet of charge.
Now consider an in nite sheet of charge lying in the x-y plane and a point a distance z
normal to the plane. The sheet can be approximated by an in nite sheet of in nitely small
charges. Each in nitely small charge can then be approximated by a sheet density  s times
in nitely small distances dx and dy shown in equation 2.8.
dQ =  sdxdy (2.8)
To make things simpler the sheet can also be approximated by an in nite number of in nite
lines of charges with charge density  l therefore the substitution in equation 2.9.
 l =  sdy (2.9)
This reduces the equation down to one integral of in nite lines. By substituting 2.7 and
taking the integral shown in 2.10, the equation for the electric  eld vector at a point due to
6
an in nite sheet can be obtained in 2.12.
 a = y^ay +h^az (2.10)
E =  sh^az2 "
o"r
Z +1
 1
dy
(y2 +h2) (2.11)
E =  s2"
o"r
baz (2.12)
From the above equation, it is shown that the electric  eld is constant at any distance
from the surface of the sheet of charge. Because of this, the electric  eld between two
oppositely charged sheets of charge can be found by multiplying this constant value by 2 as
shown below.
Figure 2.3: Visualization of the electric  eld between two in nite sheets of charge.
E =  s"
o"r
baz (2.13)
2.1.2 Derivation of Force Equations of a Parallel Plate Actuator
Using the above equation for the electric  eld between two in nite sheet charges, the
equation for the electrostatic force resulting from this charge can be found by  nding the
7
potential energy created by the work done to bring the two sets of charges to their respective
separation distance as shown in 2.15. [5]
V = 
Z b
a
E dl =Ed = qd"A = qC (2.14)
Similarly,
FdQ = WQ (2.15)
A is the area of the capacitor, C is the capacitance of the capacitor, F is the force between
the two plates, W is work, and Q is charge on the capacitor.
UE =
Z Q
0
Q
Cdq (2.16)
By manipulating the equation Q = CV di erent forms for expressing the potential energy
in the capacitor are shown in 2.17
UE = 12Q
2
C =
1
2QV =
1
2CV
2 (2.17)
The equation for the capacitance of a parallel plate actuator as a function of distance is
as follows where "o is the permittivity of free space, "r is the relative permittivity of the
dielectric material, and d is the gap thickness in meters.
C (d) = "o"rAd (2.18)
By substituting 2.18 into 2.17 the equation for potential energy as a function of gap thickness.
UE = 12"o"rAd V2 (d = gap thickness) (2.19)
A parallel plate capacitor can be considered a parallel plate actuator with a  xed gap
distance by replacing the set gap distance d, with the variable xo x(t) in 2.21 , where xo
8
is the rest gap distance and x is the variable distance at a given moment. The electrostatic
force on the parallel plate electrodes can be found by taking the integral of the potential
energy equation shown in 2.22. [6]
d(t) = xo x(t) (2.20)
UE (t) = 12 "o"rA(x
o x(t))
V2 (2.21)
UE (t) =
Z
C
F dxdt (2.22)
Finally the force on a PPA is then
F(t) = 12 "o"rA(x
o x(t))
2V
2 (2.23)
This force is always an attractive force in which the electrodes always seek to move in a
manner that increases capacitance.
The PPA consists of a movable conductive plate suspended over another conductive
 xed plate that is clamped on at least two sides by a suspension system that maintain the
plate?s position at its rest gap distance above the  xed electrode. The suspension system
exerts a force on the movable electrode which is de ned by hooks law shown in 2.26. As the
electrostatic force due to the charge pulls the movable plate in one direction, the reactionary
force of the suspension system pushes back in the opposite direction.
Us (t) =
Z x
0
kx dxdt (2.24)
F(t) = dUs(t)dx (2.25)
F(t) = kx(t) (2.26)
9
The net force on the plates is then the sum of the electrostatic force and the spring
force.
Fnet(t) = V
2"o"rA
2xo x(t)2  kx(t) (2.27)
When the two forces are equal, the net force is zero and the system is in equilibrium;
therefore, an equation for position of the actuator can be found with respect to voltage.
The resulting equation is a quadratic expression which consists of a stable region where the
spring force balances the electrostatic force, and an unstable region where the electrostatic
force is dominant. The voltage where these two regions meet is called the pull-in voltage,
and can be found by  nding the critical point of the equation where the voltage is at its
maximum. The voltage attempting to achieve a displacement of xo3 or greater, the spring
force is no longer able to withstand the electrostatic force and the electrodes will respond
in an unstable manner and accelerate into physical contact. This unstable reaction is called
pull-in. The pull-in voltage can then be found by substituting this position into the value x
of the original expression shown in 2.30. [7]
kx(t) = V
2"o"rA
2xo x(t)2 (2.28)
xo
3 = x(t) (2.29)
Vp=
r
8
27
kx3o
"A (2.30)
2.1.3 Stability Points For the PPA
An analysis of the stability of the PPA can be performed by  nding the the pole locations
of the system as electrostatic voltage is applied to the system. To account for the dynamic
motion in one direction, the mass spring system can be modeled as a second order di erential
10
equation as shown in  gure 2.31
m x+c_x+kx = f(t) = "AV
2
2(xo x(t))2 (2.31)
At equilibrium:
kx = "AV
2
2(xo x(t))2 (2.32)
and this equation can be used to  nd an expression for V as a function of displacement, x,
which is illustrated in  gure 2.4:
V =
r
2(xo x(t))2
"A (2.33)
The graph in  gure 2.4 represents the voltage on the PPA as a function of PPA displacement
distance. As the graph shows, the voltage required to maintain a stable position increases
as the plate moves further from its nominal gap and toward the opposing plate. When the
plate reaches one-third of the full distance, the required stable voltage reaches a maximum.
Beyond this position, the force due to voltage becomes greater than the opposing spring force,
causing the equivalent voltage to decrease. Using a  xed voltage source, the system has no
way to compensate for the decreased voltage requirement to maintain a stable position, thus
causing the plates to pull together completely. [8]
11
Figure 2.4: Voltage as a function of movable plate displacement from equation 2.33.
Over the stable operating range (0  x  xo=3), equation (2.33) and a Taylor?s series
expansion for f(t) about x can be used to develop a linearized model of the electrostatic
spring e ect in order to evaluate the e ect on the poles.
For the Taylor series:
f0(f(t)) = "AV
2
(xo x(t))3 (2.34)
Let xa be a desired displacement over the stable range and Va be the voltage from (2.33).
Therefore:
f(t) 
"AV2a
2(xo xa)2
0! (x(t) xa)
0 +
"AV2a
(xo xa)3
1! (x(t) xa)
1 = N +k
ELx(t) (2.35)
12
m x+c_x+ (k kEL)x = N (2.36)
which is only valid at the particular values of Va and xa. From the quadratic formula, the
pole locations are:
poles =
 c
m +
q
c
m
2 4k kEL
m
2 (2.37)
Using (2.33), (2.35) and (2.37), an s-plane map of the poles for system between x = 0 and
x = xo/3 can be obtained. This map is shown below in Figure 2.5
13
(a) Rest Gap Position
(b) 1/2 of pull in Position
14
(c) 84% of pull in position
(d) Pull-in Position(1/3 of rest gap)(Unstable)
Figure 2.5: Pole Zero map of actuator at di erent points of stability
2.2 Motivation
MEMS PPAs have a number of advantages that make them desirable for use in micro
systems. However, one issue that must be accounted for is the pull-in instability inherent in
them. Depending on its use, the e ect of a device entering pull-in can either enhance or be
15
a detriment to operation. A number of issues can arise when the pull-in condition is reached
that can a ect the operation of MEMS devices. These issues can range in e ect from small
signal instability to complete destruction of the device. In these instances, detecting and
preventing or mitigating the pull-in event is important to ensure correct operation of the
device. The following sections list the potential undesirable e ects as well as some bene ts
associated with pull-in.
2.2.1 Disadvantages of Pull-In
1. Stiction:
In devices with a large mass, inertial forces are the predominant force explained by the
equation F = ma, but as the device scales down in size, the mass reduces by the cube
of the length. Also as devices shrink, magnetic forces are signi cantly decreased. Other
forces, however, become very important at small sizes. Two of these forces include Van
der Waals forces and capillary forces. Van der Waals forces are relatively weak forces
caused by charge imbalances between many molecules. When two surfaces come into
contact with each other, and these Van der Waals forces are stronger than the restoring
spring force of the structures, then the structures will become stuck together. Capillary
forces are characterized by the tendency of a liquid to be pulled into a small opening.
This movement is caused by attractive forces between the liquid and the material of
the wall. At large sizes, the inertial force due to mass would overcome these small
forces. But for MEMS devices these forces overcome the inertial forces, causing the
device to become stuck. This phenomena is called stiction which is a major cause of
yield failure during fabrication, but can also have e ects during device operation if two
micro-surfaces come into contact, which can happen as a result of pull-in.
2. Current Arcing through Electrode Contact:
In addition to stiction, the moving plates in MEMS devices can also become fused
together through current arcing caused by the large current density forced through a
16
small area when the plates make contact. This high current can cause the two surfaces
to permanently weld together or even vaporize. This will cause catastrophic failure of
the device.
3. Damaged Electronics:
MEMS devices often require high voltages to change plate position. This is especially
the case when pull-in is reached. When the two plates come together, the resistance of
the circuit can instantly go from very high to very low allowing large current  ow. This
current can be high enough to exceed the maximum current limits of the electronic
circuitry connected to the MEMS device. If the high current  ow is sustained long
enough, the circuitry can fail permanently.
4. Mechanical Failure:
Due to the large amount of force generated by the electrostatic voltage between the
two plates, mechanical structures in the device can be stressed beyond their fracture
limit. This will cause permanent damage to the device.
5. Erroneous sensor readings:
Many MEMS devices rely on either capacitive or resistive sensing of plate motion to
measure signals from the device. Motion of the device is normally limited to very small
displacement in order to keep the signal variation as linear as possible. Outside of a
small displacement, the signal can become nonlinear and distort the sensor reading.
The pull-in event is characterized by large displacement and nonlinear motion which
can be deleterious to a system that relies on stable linear motion.
6. Controls techniques to extend the stable range of motion are complex/costly.
2.2.2 Advantages of Pull-in
While pull-in is normally considered an undesirable condition, the phenomenon can
also be used to create sensors that respond mechanically to an input stimulus, and create
17
actuators that can generate a large amount of force. In these instances, it can be useful to
detect the occurrence of the pull-in event without inhibiting the e ect of the pull-in process.
The highly nonlinear motion of pull-in provides opportunities in which the onset of the
condition can be detected.
Much research has gone into preventing pull-in by extending the stable range of the
actuator. Many of the methods developed to extend the stability range, however, introduce
other problems that also must be taken into account such as decreased sensitivity of the
actuator, and increased cost of the controlling circuit. Sometimes the pull-in characteristic
of PPA?s is desired for correct operation. In some cases, actuators do not require the ability
of full range of motion, but only to ensure that pull-in does not occur. Also there are times
when pull-in is necessary for operation of the device. In these cases, it would be bene cial to
detect the onset of the pull-in condition at which point the device can either be shut down
when pull-in is unwanted, or when wanted, to keep track of the conditions that caused pull-in
to take place. A sensing mechanism can potentially be much simpler and less expensive to
implement than existing control systems while at the same time reducing the negative e ects
associated with other control mechanisms.
2.3 Chapter Summary
In this chapter a detailed derivation of the mathematics describing the motion of PPA?s
was covered, including the equations governing the force due to a static electric  eld and
motion that occurs during pull-in. The advantages and disadvantages were explained as they
relate to various device applications and are summarized below.
a136 Advantages
1. Applications that require a device to make mechanical electrical contact.
2. High acceleration of electrode generates high momentum for actuating objects.
a136 Disadvantages
18
1. Damage caused by stiction after pull-in occurs.
2. Current arcing caused by electrical contact of plates
3. Damaged electronics due to high voltage passing through shorted electrodes
4. Mechanical failure of the PPA caused by the excessive force of the pull-in event.
5. Erronius sensor readings from device operating outside of the linear operational
range.
6. Increased cost of extra electronics required to overcome the pull-in phenomena.
These examples provide the motivation for the research behind this work to provide
a low cost method of identifying when pull-in is about to occur and providing a way of
preventing the event or just signaling that it has occurred.
19
Chapter 3
Literature Review
An exhaustive review of prior work in the area of pull-in as well as various current
sensing methodologies has been performed. A summary of key areas of focus in regard to
pull-in follows. A number of models describing pull-in from di erent perspectives have been
developed, including frequency considerations [9] and two dimensional modeling to account
for o -axis tip-in instability [10]. These areas include device design methods intended to
reduce the e ect of pull-in, as well as open and closed loop control methods to extend the
stable range before pull-in occurs. The following sections provide a summary of the previous
work used to control and prevent pull-in.
3.1 MEMS Devices that Make Use of Pull-in
While most MEMS devices attempt to avoid pull-in instability, some devices actually
rely on it for operation. Listed below are two examples where this is the case.
1. MEMS Switches:
The MEMS switch has gained popularity for its fast switching times and very high
change in impedance between the on and o states. [11] [12] [13] [14]
2. Gas Flow Actuator:
A novel structure was Xie et al which uses the pull-in phenomena to create a peristaltic
micro-pump. A system of three membrane-based PPA?s is made from layers of para-
lyene to maintain electrostatic isolation with the pumping  uid. The membranes are
aligned in a channel and alternate between a closed and open state forcing liquid to
move between actuators with each cycle. The frequency of the state changes controls
20
the speed at which the liquid will  ow. With this pump, they were able to achieve a
1.7nL  ow rate with a pumping pressure of 1.6kPa [15].
3.2 Actuator Control
1. Open Loop Control
a136 Series Capacitor Method:
In a paper by Seeger et. al, the authors show that the stable range of a PPA can
be extended with the simple addition of a series capacitance creating a capacitor
divider with the PPA. As the distance between the plates of the PPA decreases,
the capacitance relative to the  xed capacitor increases causing a negative feed-
back where a higher percentage of the total voltage drop occurs across the  xed
capacitor. The voltage across the PPA will decrease allowing it to remain in a
stable state as the plates move closer together. This technique will allow full
range of motion for the actuator provided the correct ratio of  xed capacitor to
actuator capacitance is chosen. This method, however, also has drawbacks in that
the voltage required to move the actuator is much higher with the  xed capacitor
due to the large voltage drop across it. Also, because the voltage can be quite
high, the type capacitor required to act as the  xed capacitor in the circuit can
be quite bulky due to the required high voltage rating [16]. Additionally, this
technique is very sensitive to stray capacitance in parallel with the capacitor.
Other groups have since furthered the research into this method by applying it
toward di erent applications. XingTao Wu et. al used the method to model an
electrostatic micromirror with increased travel range using this method. [17]
Another work also used the series capacitor method but then extended the method
by closing the loop. They measured the voltage change at the actuator and
then applying a current drive to  x the plate at a certain position. They were
21
also able to reduce the voltage penalty of the series capacitor method using this
approach. [18]
a136 Fabrication Techniques:
Researcher have investigated creating novel actuator designs to mitigate pull-in.
One notable work is by Legtenberg et al. to create a cantilever spring structure
with tapered actuating electrodes positioned on the side of the beam which provide
both a vertical and lateral force on the beam. As the cantilever moves closer
to pull-in, the vertical component of the force reduces due to the positioning
and shape of the electrodes. This allows the beam to achieve full scale motion
without succumbing to pull-in. While this cantilever spring structure has a much
di erent mechanical operation than that of a PPA, the mitigation of pull-in due
to electrode positioning is novel. [19]. Zhe et. al followed a similar method in
which they modeled the pull-in e ect on various di erent shapes of electrodes,
and found that electrode with certain shapes could achieve much higher pull-in
range up to 100% of the gap distance using a power function design. [20].
a136 Negative Capacitance/MOS Capacitance:
Variations on the series capacitor method have also been shown to extend the
stable range while helping to o set the higher voltage requirement of the  xed
series capacitor. By adding a MOS capacitor in depletion as the  xed capacitor,
the same negative feedback can be created where the voltage across the MOS
capacitor increases as the PPA gap decreases. This voltage increase across the
MOS capacitor causes the depletion region to increase reducing the capacitance
of the MOS capacitor as the capacitance of the PPA increases. This ampli es the
negative feedback of the voltage across the PPA. [21]
2. Closed Loop Servo Control:
Closed loop Control systems have the ability to extend the range of motion beyond
pull-in by providing a feedback mechanism from some monitoring sensor, usually a
22
sensing capacitor, to a controller which reacts to changes in the monitoring sensor.
the controller will change the amount of voltage on the plate as needed to maintain a
particular position. This can be done faster than the ability of the actuator to react
and is an e ective way to control position past the pull-in point. A disadvantage of
this, however, is due to the large amount of electronics involved in maintaining the
feedback loop. The electronics involved can consist of an ampli er, a lowpass  lter,
a demodulator, and the controller circuitry. Control of up to 60 percent of the gap
distance has been observed, which is well above the one-third pull-in limit. [22]
a136 Sti ness Control:
Another idea, presented in a paper by by Hung and Senturia, is to use the notion
of leveraged bending to control the position of the actuator. Leveraged bending is
the idea of moving a structure by applying force only to a portion of the structure.
By positioning the electrodes only at the spring edges and not at the center of the
mass, the springs will de ect, causing the mass to move down. As more voltage is
applied, the mass will move the full gap distance, while the part being actuated,
the spring, has yet to move the one-third gap distance. This idea is similar to the
work by Legtenberg for pull-in control on a cantilever, but can also be applied to
the PPA. [23]
a136 Synthetic Voltage Division:
A work by Dean et al at Auburn University extends the idea of using the series
capacitor to increase the stable range by including a feedback controller to reduce
required input voltage, which is the main disadvantage of the series capacitor
method. [24]
a136 On-O Control:
In a work by Mol et al, a method of digital on-o input is proposed to control the
position of the actuator. A feedback controller is created by measuring change in
the capacitance of the device, if the capacitance changes past a certain reference
23
point then the input voltage is shut down. Once the spring reacts and begins
to pull back, the voltage is turned on again. The plate remains unstable with
respect to the input voltage, but because the switching mechanism occurs at a
much higher frequency than the actuator has time to respond, the plate does
not snap in. Any gap position can be maintained using this method; however,
the device will not come to rest in the unstable region. It will instead oscillate
around the control position. The oscillations can be reduced by lowering the
voltage level once oscillation begins, and increasing the switching frequency, but
complete elimination will not occur. [25]
3.3 Chapter Summary
Most research in the MEMS area on pull-in is devoted to extending the stable range or
eliminating pull-in altogether. Most of the solutions that have been discovered either add a
great deal of complexity to the circuit or change or in uence the operation of the circuit such
as causing increased operating voltage. For some purposes, the actuator does not need this
level of control, but instead only needs to detect when the event occurs and possibly then to
disable the actuator. This can be accomplished using much less circuitry while reducing the
impact on the signals controlling the actuator. This dissertation will attempt to examine the
various methods discussed earlier and use them to detect the onset of pull-in of a fabricated
test device and to disable the actuator power supply before the electrodes make contact.
24
Chapter 4
Theory Development of Pull-In Sensing Mechanism
In choosing an appropriate method for detecting pull-in in a MEMS device, a number of
di erent options were considered. Many well established sensing methods have been devel-
oped, each with their own advantages and disadvantages. Considerations such as sensitivity
of the signal, resolution, and size and cost of the sensing equipment must all be taken into
account. The following sections will explore the various methods evaluated and the reasons
for and against applying each method for this particular application.
4.1 Current Sensing
Current sensing techniques are a well-established area in power electronics and other
areas where sensing of high currents is necessary. With enough sensitivity and ampli cation,
these techniques can also be applied to characterize the dynamics of pull-in, and to identify
the moment that pull-in occurs. In dealing with smaller currents, one of the best approaches
available is resistive sensing, in which a resistor is placed in the path of the current  ow,
and the voltage drop across the resistor is measured. There are two main techniques that
are used when implementing resistive sensing, called high side sensing and low side sensing.
These techniques will be discussed in the following section.
4.1.1 High Side Current Sensing
A high side sense circuit is created by placing the resistor between the voltage source
and the load as shown in Figure 4.1. At the onset of pull-in, the motion of the movable
electrode begins to increase in speed, moving toward the  xed electrode. This movement
causes a decrease in gap separation which results in an increase in capacitance in the actuator.
25
This quick increase in capacitance causes an inrush of current into the actuator to  ll up the
extra charge storing capability. The result of this current  owing across the high side resistor
causes a voltage drop that can be measured and is related to the amount of current  owing
into the actuator. Ideally, the resistor has a low impedance in relation to the load so that
the current through the loop is not disturbed. The change in voltage is linearly proportional
to the change in current across the resistor, assuming an ideal resistor. The resistor then
acts as a constant of proportionality where increasing the value of resistance will increase
the resulting voltage change for a given current. Additional drawbacks to high side detection
are in the measurement method. Normally, the inputs of an instrumentation ampli er are
placed at each side of the resistor to amplify the voltage di erence. By doing this, the inputs
have no  xed reference, but are instead referenced to each other. The ampli er must have a
high common mode rejection ratio (CMRR) to reduce any error at the output. [26] Another
issue that arises for sensing of this type which can be especially relevant when applying to a
MEMS device is that most op amps have a maximum input rating of around ten to twenty
volts. The actuation voltage for parallel plate actuators can sometimes reach 50-100 volts,
which can easily be high enough to damage the op-amp. A compensation circuit is proposed
to solve this problem, which is shown in Figure 4.2. This circuit places voltage dividers with
precisely matched resistors on each side of the current sense resistor to create a new sense
node with a potential that is reduced from the voltage across the sense resistor by the same
ratio as the voltage divider. This allows the sense voltage to remain within the limits of the
attached operational ampli er circuit which can act as either a bu er or a gain stage that
feeds to a comparator. The comparator then servers as the controller for the detection signal.
Once the signal is tripped, the signal trips the clock of a DFF propagating through as an
output high on the DFF, remaining so until the circuit is reset. Also included is a protection
mechanism, a low Rds-on PMOS transistor is placed in the circuit path which is normally
on until the PPA snaps together, at which point the PMOS will turn o e ectively breaking
the circuit. This is added as a fail-safe to prevent circuit damage, in case the PPA snaps
26
together. [27] This circuit was tested on a breadboard as a potential solution for detecting
and mitigating pull-in. However, while this method still has potential, the initial tests found
it to be very susceptible to noise as well as the sensing voltage o set mentioned previously.
Figure 4.1: Simpli ed schematic of a high side current sense circuit
27
Figure 4.2: Compensation circuit to account for high voltage MEMS devices
4.1.2 Low Side Current Sensing
A low side current sensing is created by placing the sense resistor between the load
and ground, and measuring the current across the resistor just as in the high side current
sensing method. This is the simpler of the two alternatives because measurement across the
current sense resistor has a  xed ground reference. Also there is no need for the high side
compensation circuit discussed in the previous section. The disadvantage to this circuit,
however, is that the load is no longer attached directly to ground, but rather to a note that
has a voltage equal to IR. This can cause problems with some high frequency circuits that
require a true ground. [28]
28
Figure 4.3: Schematic of a low side current sense circuit
4.1.3 Low-Side Transimpedance Ampli cation Sensing
Low Side transimpedance ampli cation is an extension of the low-side current sensing
method, where instead of measuring the voltage across a resistor, the current is ampli ed
though a feedback resistor connected to an operational ampli er (opamp). A schematic of
this method is shown in  gure 4.4. The voltage at the output of the opamp is equal to
V = IactRSense (4.1)
Where Iact is the current though the actuator, and RSense is the feedback resistor on the
opamp which acts as the gain term for the signal. One side of the actuator is connected
to the high voltage source while the other side is connected to the negative input of an
opamp. The positive input of the opamp is connected to ground causing the negative input
to be taken to a virtual ground. To the actuator, which works at low frequencies, the opamp
should be transparent in operation. This method has the potential for very high gain sensing
of current into the picoamp levels. [29]
29
Figure 4.4: Schematic of a transimpedance ampli cation circuit
4.2 Hall E ect Sensing
Hall e ect sensors measure the voltage generated by the force created due to the inter-
action of a magnetic  eld on charged particles moving perpendicular to the magnetic  eld
called the Lorentz force shown in equation 4.2.
Florentz = q(E +v B) (4.2)
Sensors based on this principle apply a constant current through a Hall element. The
current passing through the circuit which is to be measured creates a magnetic  eld on the
current passing through the Hall element. This creates a voltage which is then ampli ed
and measured. The advantage of this technique is that this sensing mechanism is almost
completely noninvasive to the circuit allowing current to  ow unimpeded, and can sense
both ac and dc currents. The potential drawback is that this application is normally used in
high current applications to generate the magnetic  eld required to generate a measurable
voltage. It is possible however that if a large enough magnetic  eld is generated, even smaller
currents on the scale that would be required to detect pull-in could be measured. [30]
30
4.3 Optical Sensing
Another method to detect pull-in is to visually determine that pull-in is occurring using
an interferometer. Using a Polytec OFV vibrometer shown in Figure 4.5, the position of
the actuator can be converted directly into a voltage that can be measured electrically.
The pull-in condition can be detected by the increased slope of the voltage change from the
interferometer. This method should have high accuracy with no interference in the operation
of the device. The disadvantage of this technique is the large amount of external equipment
required to take measurements. [31]
Figure 4.5: Polytech OFV Vibrometer
4.4 Piezoresistive Sensing
Another method of detecting movement of the PPA is by using a piezoresistive element
to detect deformation of the spring. A piezoresistive structure changes its electrical resistivity
when mechanical strain is applied. This sensing element can be integrated into the actuator
springs, and as the spring moves toward pull-in, the resistance will change. This change can
be measured using a simple voltage divider or even a Wheatstone bridge. [32] One method
of creating a piezoresistive device is to create a di used resistor on top of the spring by
31
doping the area. The advantage of this method is that it should have almost no e ect on the
operation of the device, and external electronic circuitry is very simple. The disadvantages
are that this method of sensing is not as linear as other methods to detect motion, and
fabrication of the device becomes more complicated to implement because of the extra mask
layers required to implement a di used resistor. [33]
4.5 Secondary Capacitance
Another possible method of detection is to add a separate sense capacitor on the device
with a set voltage reference whose sole purpose is for position detection. This would require
a change in the design and fabrication of the device, but could be a very accurate method
of pull-in detection. This method was successfully used to create a tunable capacitor in a
work by Xiao et. al. [34]
4.6 Current Transformer
The current transformer is a well established method of measuring currents in power
electronics, particularly in high current applications where it is not feasible to directly mea-
sure such a large current. The current transformer consists of a primary and secondary coil
where the ratio of the currents between the primary and secondary coil is equal to the turns
ratio of the windings in the coils. Because the current transformer operates on the principle
of Faraday?s law of induction shown in 4.3, it can only measure ac signals. This should be
appropriate for measuring the pull-in e ect because it is a high frequency signal. [35]
r E =  B t (4.3)
32
4.7 Chapter Summary
As this chapter shows, a number of di erent options exist for sensing plate motion.
Each option has its own bene ts and drawbacks. There proposed sensing solutions can be
categorized into 4 main areas as follows
1. Capacitive
a136 Secondary capacitance
2. Current
a136 Hall E ect
a136 High/low side resistive measurement
a136 Current Transformer
a136 Transimpedance Ampli cation
3. Resistive
a136 Piezoresistive Measurement
4. Optical
a136 interferometric measurement
Because of time and budget constraints, not all of these options could be investigated.
Some methods were initially judged to be unsuitable for MEMS application. The Hall e ect
and transformer applications were judged to have insu cient sensitivity to achieve the stated
goal of sensing pull-in. Piezoresistive measurement, while a potentially viable solution, could
not be investigated due to the costs required to complete the additional fabrication steps
of resistor doping and patterned metal deposition. Optical sensing is very precise, and for
this reason it was used as the control for the experiments, but in real world applications,
this was not a viable option because large external equipment is required. For these reasons,
33
current sensing was selected to be the sensing method of choice, speci cally, transimpedance
ampli cation due to its ability to achieve very high gain, and little dc o set error.
34
Chapter 5
PPA Design and Fabrication
This chapter details the processes used to design and fabricate the PPA devices that
were in testing. An explanation of the design considerations is given to make the device more
suitable to the experiments conducted. These considerations include ease of measurement,
maximization of capacitance, and minimization of pull-in voltage. Also covered are the
simulations that were used to verify both the design of the device and operation of the
detection circuit. Finally, the processes used to fabricate the device are thoroughly discussed.
5.1 Intellisuite Simulation of PPA device
As shown in Figures 5.1(a) and 5.1(b), the actuator structure was designed with long,
narrow springs and a large proof mass so that stiction would not be an issue, while at
the same time having a low enough pull-in voltage to enable testing with easily accessible
equipment. Also, a problem for small silicon actuators is that if the current discharge at
snap-in is large enough, the current density through a small surface area can be enough to
fuse the surfaces together.
35
(a) Overhead View (b) Angle View
Figure 5.1: Intellisuite simulation of PPA structure
A Finite Element Analysis (FEA) was performed on a model of the device using Intel-
lisuite, a FEA software tool, to simulate the mechanical response of the device. The table
in  gure 5.1 shows the simulated resonant frequencies of the  rst 5 resonant modes. The
primary mode natural frequency though simulation was found to be 248 Hz. Once the device
was fabricated, the device was placed on a mechanical shaker to determine the transmissi-
bility curve. The resulting curve showed peak transmissibility at 227 Hz, which is relatively
close to the simulation prediction. The deviation in frequency was attributed to simulation
assumptions, variations in the material properties, and fabrication tolerances.
Mode Frequency (Hz)
1 248.844
2 463.556
3 464.521
4 523.326
5 523.52
Table 5.1: Intellisuite modal simulation of designed PPA device.
36
One consequence of designing a device in such a way that the springs wrap around the
device, as in this case, is that in deformation, the mass has a tendency to rotate. Since,
in this experiment, displacement is limited to 150  m this rotation e ect was considered
negligible. Another e ect of the extended spring length is that higher order vibration modes
have resonant frequencies only slightly higher than the primary mode. This o axis move-
ment could cause an unintended reduction in capacitance, as well as cause an imperfect
surface contact between the two electrodes when snap-in occurs causing incomplete charge
dissipation. While capacitance variation was not found to be a signi cant source of error,
contact quality had a major e ect on oscillation reliability. It was determined that o -axis
vibration was not the main contributor to the contact quality issue. Contact cleanliness and
non-uniform plating of the PCB electrode also played a signi cant role.
5.2 Simulink Simulation Pull-in Detection Operation
In order to make the parameters of the simulation as close to as accurate as possible,
a fabricated PPA was characterized so the actual parameters of the device could be entered
into Simulink. The damping coe cient was found through measuring the transmissibility
plot and reading the value at the device?s resonant frequency, which was approximately equal
to Q. A graph of this, as shown in Figure 5.2, shows the device having a natural frequency
of 227Hz with a Q value of 46. [36]
37
Figure 5.2: Transmissibility of the fabricated PPA
Table 5.2 lists the  nal properties of the designed PPA which were used to generate
models to compare the performance of the device in operation. The spring constant was
estimated using equation 5.1
keq = NlegN
Zig
Ewt3
L3 (5.1)
where keq is the equivalent spring constant of the 4 combined springs. Nleg is the number of
springs attached to the proof mass. NZig is the number of 90 degree bends in each spring. E
is Young?s modulus for silicon, w is the width of each silicon spring, t is the thickness of the
springs, and L is the total length of each spring. The damping coe cient was then inferred
from the experimental Q factor obtained through vibration testing.
38
PPA Device Properties
Spring l,w,t
55250 m
50 m
375 m
Actuator Area 0.0001 m2
Actuator Mass 0.000094125 kg
Spring Constant, k 22.0125 Nm
Rest Gap 1.50E-04m
Q-factor 46
Damping coe cient 0.000968477
Rest Capacitance 9:5pF
Natural Frequency 227Hz
Table 5.2: Parallel Plate Actuator Device Properties
A model of the parallel plate actuator was created in Simulink using the equations
presented in section 2.1. The parameters for the device were found using standard MEMS
equations. A list of the parameters of the device is shown in Table 5.2.
Figure 5.3 shows a diagram of a Simulink program that implements a simple one dimen-
sional mass spring damper system using the values shown in table 5.2. A ramping voltage
is converted to an electrostatic force value and applied to the summing node of the transfer
function. A saturation value is added with the rest gap value to show the maximum range
of travel for the PPA. The values are then fed into a function which calculates the current
at any given time through the actuator. This simulation assumes ideal properties, including
perfect voltage sources, linear spring constant, negligible resistance, and constant damping.
All of these can contribute to variations from the exponential results of the fabricated PPA.
39
Figure 5.3: Simulink Model of PPA
40
Figure 5.4: Current and Position Results of Simulink PPA Under Ramping Voltage.
The simulation shows as expected that as the voltage across the actuator increases above
the pull-in value, the movable plate increases in velocity moving rapidly toward the opposing
plate. Also, the current begins to increase exponentially as the velocity increases.
Table 5.3 shows sample current values at di erent positions after pull-in has begun. At
each position, the current increases by an order of magnitude which can be detected very
quickly.
41
Simulink Results
Position Current
1.00E-04 m 1.67e-9A
7.00E-05 m 2.5e-7A
5.00E-05 m 1.3e-6A
4.00E-05 m 2.5e-6A
1.00E-05 m 4.85e-5A
Table 5.3: Simulink Analysis of Current as a Function of PPA Position
5.3 Fabrication Equipment and Processing
The Alabama Microelectronics Science and Technology Center located at Auburn Uni-
versity is equipped with all of the tools necessary to facilitate the fabrication of MEMS
devices. This equipment, and the parameters by which theory was used, will be referenced
heavily throughout this document. In the next few sections, a brief description of the most
frequently used pieces of equipment and their operation will be discussed.
5.3.1 Cleaning Equipment and Processes
As explained in various sections of this work, cleanliness plays an important role in the
successful fabrication of microelectronic as well as MEMS devices. Wafers undergo thorough
cleaning processes between each processing step. The  rst step in silicon wafer processing is
the RCA clean. This step is needed to remove contamination on the silicon wafer that can
interfere with future steps in the process.
Before a silicon wafer can be processed, the surface must be clean of particles and
impurities which can reduce the quality of the pattern, cause interference in the operation
of electrical devices, and become a source of contamination in processing equipment such
as the oxidation furnace. To remove these particles, William Kern, in 1965 with the RCA
42
corporation, developed a cleaning process known as the RCA clean which has since become
the standard in the industry. This process involves soaking the wafer in a series of chemical
mixtures, each mixture serving a di erent cleaning purpose. Presented below is a table listing
the standard processing steps used in the RCA clean to prepare the wafer for processing. [37]
Table 5.4: RCA Process
Process Name Step Mixture Time Temperature Purpose
A IPA bath 1Min Room Temp Oil Removal
DI Water Bath 1 Min Room Temp
1 : 30% H2O2
SC-1 B 1 : 29% NH4OH 5 Min 70 Removes organic
5 : DI H2O impurities
DI Water Bath 1 Min Room Temp
1 : 49% HF RemovesC
50 : DIH2O 30 Sec Room Temp Hydrous Oxide
DI Water Bath 1 Min Room Temp
6 : DI H2O Removes
SC-2 D 1 : 30% H2O2 5-10 min 70 Heavy Metals
1 : 37% HCL
DI Water Bath 1 Min Room Temp
Standard Clean 1 (SC-1)
The SC-1 solution is a mixture of 5 parts DI water, 1 part 30% concentration H2O2, and
1 part 29% concentration NH4OH heated to 70 C. The wafer is submerged in this solution
for 10 minutes. During this time, the H2O2 creates a hydrous oxide layer on the surface of
the wafer, while the NH4OH acts to remove this layer. This process repeats continuously
and results in the removal of organic materials, light metals and other small particles on the
surface of the wafer. Once this step is complete, the remaining hydrous oxide layer must
then be removed. This is done by dipping the wafer in a mixture of 50 parts DI water, and
1 part HF acid at room temperature for approximately 30 seconds. [38]
43
Standard Clean 2 (SC-2)
The SC-2 solution is a mixture of 6 parts DI water, 1 part 30% concentration H2O2, and
1 part 37% concentration HCL heated to 70 C. The wafer is submerged in this solution for
10 minutes. This step removes heavier metals that are not removed in the SC-1 step. [38]
Spin Rinse Dryer
The STI Semitool Spin Rinse Dryer shown in Figure 5.5 is a tool designed to assist in
maintaining a particle-free wafer.
44
Figure 5.5: Spin Rinse Dryer
There are two versions available in the lab to accommodate 100mm and 125mm diameter
wafers. This device sprays deionized water across the surface of the wafer, while spinning,
to remove particles and other loosely held contaminants. The wafer then enters a drying
45
phase where the spin speed of the wafer is increased, and heated nitrogen is blown across the
wafer to remove the water from the previous rinse cycle. This step is important for cleaning
wafers during multiple steps of the fabrication process.
5.3.2 Patterning Equipment
After cleaning the wafer, a pattern can then be transferred to the wafer that will realize
the intended design. This pattern is normally created using PR. This is a material that is
sensitive to light and acts as a masking material to other processes.
Hexamethyldisilazane (HMDS) Chamber
Before PR can be applied to the wafer surface, the wafer must be primed to accept the
resist. At room temperature, water molecules from the air can attach to the wafer surface
causing PR to attach to the water molecules instead of the silicon. To remove the water,
the wafer is placed in a dehydration oven and heated to 120 C for twenty minutes. After
dehydration, a primer of HMDS is then coated over the wafer which adheres well to both
the silicon and the PR.
HMDS is normally applied in one of two ways. One way is to pour the HMDS onto the
wafer in liquid form and spin the wafer until only a very thin coating of HMDS remains. The
HMDS must then be dried completely before the resist can be applied. This method has the
advantage of being easily added to the fabrication process, but has drawbacks in that most
of the HMDS is wasted during spinning, and if the HMDS is not completely dry, the HMDS
can break down the bottom layer of PR, reducing adhesion instead of improving it. [39]
The second method is to apply the HMDS in vapor form. Because only a few monolayers
of HMDS are required to adhere to the resist, the HMDS can be allowed to evaporate and the
vapor  ows over the wafer. The molecules that attach to the wafer will provide a su cient
amount of coating needed for resist adhesion. This is the method used at Auburn. Shown
46
in  gure 5.6 is the chamber in which the HMDS is allowed to evaporate and attach to the
wafer.
Figure 5.6: Vapor HMDS Application Chamber
Photoresist Spinner
A photoresist spinner is used to accurately control the thickness of the PR  lm applied
to the wafer. A uniform thickness is achieved by spinning the wafer at varying speeds and
allowing the PR to spread over the entire surface area. Slower rpm?s will give a thicker
coating, while faster rpm?s will give a thinner coat. Longer spin times will help to increase
the uniformity of the coating over the wafer.
47
Figure 5.7: Photoresist Spinner
Matrix Oxygen Plasma Asher
The Matrix oxygen plasma asher is a type of plasma etcher that is used to remove
organic material from the surface of the wafer, usually PR. The oxygen plasma reacts with
the PR forming an ash that is then removed by a vacuum pump. The parameters that can
be modi ed include the power of the plasma, the  ow of oxygen, and the time of exposure.
This device has two main roles in MEMS processing. The  rst is called a descum step. This
is a short interval treatment meant to remove excess unwanted PR from the UV exposed
areas of a developed pattern. This step is particularly important when the process calls
for a wet chemical etch. Even a very thin  lm of PR that has not been fully removed will
completely mask the pattern from etching. The second role this device plays is to fully strip
the remaining PR after processing is complete.
48
Figure 5.8: Matrix Oxygen Plasma Asher
Karl Suss MA/BA6 Contact Mask Aligner
This device creates a pattern on a PR  lm by bringing a photolithography mask in
direct contact with the surface of the wafer and subjecting the exposed areas to a UV light
49
source. This equipment has the capability of achieving a maximum resolution of 0.7 m, and
performing backside alignment functions which are essential for creating the 3D structures
involved in this project. A high intensity Hg lamp provides a UV light source for PR
exposure, which operates at 365nm (i-line) and 436nm (g-line). Wafer sizes of up to 150mm
can be patterned using this equipment. [40]
Figure 5.9: [Karl Suss MA/BA6 Contact Mask Aligner
5.3.3 Silicon Processing Equipment
Silicon is a very strong material with compressive strength up to 80% to that of steel,
making this material very useful in MEMS devices. While having a very strong compressive
50
strength, it lacks the ability to bend plasticity, meaning that the material can fracture
easily when bent. Speci c tools have been developed that can create structures on the
desired scale, while preventing fracture. Using a DRIE system, precise etching can make
deep grooves in the silicon, and CVD systems can deposit silicon with very high accuracy.
Furthermore, oxidation of the silicon to form silicon dioxide can protect the surface from
chemical reactions, as well as forming an electrically insulating barrier.
5.4 Plasma Assisted Dry Etching of Silicon
The aspect ratio of a structure is a measure of the structure?s longer dimension compared
to its shorter dimension. The aspect ratio for the device being constructed is required to be
very high, therefore the structure is fabricated by bulk micromachining a pattern of ridges
using an ICP (Inductively Coupled Plasma) DRIE (Deep Reactive Ion Etching) system.
Plasma based etching of silicon has been around for many years, but only within the
last twenty years has the technology evolved so that plasma based etching could be utilized
for precise directional etching. The following sections will explain how plasmas are used to
etch silicon, and discuss some of the modi cations that have been made that have allowed
for the anisotropic pro les created using the DRIE process.
5.4.1 Plasma Etching
Plasma is an ionized gas that consists of a high density of free electrons. This condition
can be created by exposing the gas to a high electric  eld. As the energy from this electric
 eld is imparted to an atom in the gas, an electron in the outer shell gains enough energy
to break free of the atom, thus the atom becomes ionized. Because the electrons are much
lighter than the ions, they accelerate much faster in the presence of the electric  eld. As
the electrons move in the electric  eld toward the electrode, some electrons can become
trapped in the chamber wall creating a net negative potential, pushing away other electrons,
and attracting ions. This force creates a sheath at the edge of the plasma with a very
51
high electric potential thus accelerating the ions through the sheath, leaving a nearly charge
neutral region at the center of the plasma. The bene t of creating a plasma when used for
etching a material is that the energy imparted to the ions can cause reactions with other
materials that would normally occur only when they are thermally excited to thousands of
degrees K. By creating a plasma consisting of gas compounds that contain either  uorine
or chlorine, silicon can be nearly isotropically etched by placing the wafer inside the plasma
allowing the excited species to chemically react with the silicon to form new gaseous species
such as SiFx or SiClx which are the pumped out of the plasma chamber. This reaction
occurs at all points where the etching gas comes into contact with the silicon and is nearly
directionless. [41] [42]
5.4.2 Reactive Ion Etching
Reactive Ion Etching is a modi cation of the plasma-based etching by positioning the
silicon wafer at the bottom of the chamber below the generated plasma. This allows etching
to occur through the physical bombardment of the energetic ions exiting the sheath of the
plasma. In this method, physical sputtering and chemical reaction both occur to enhance the
silicon etch. Noble gases such as argon can cause physical bombardment that can break o 
silicon atoms and give a directionality to the etch that is perpendicular to the surface, while
remaining unreactive with the surface. Ions such as  uorine and chlorine also impact the
surface and contribute to sputtering, but upon impact they form a chemical reaction with
the silicon converting it to a gas such as SiF4. Other gasses like oxygen, when added to the
plasma, have an opposing e ect when reacting with the surface forming a passivating SiOxFy
layer. The SFx ions also work to remove this passivisation layer. The alternating generation
and removal of this layer also helps to change the pro le of the etch. Therefore, the SF6 to
oxygen ratio is very important in controlling the anisotropic nature of the etch. [43]
52
5.4.3 Inductive Coupling of Plasma
A further modi cation of the plasma generation method is to use an inductive element
that is tuned to the frequency of an RF power source. A large amount of current will  ow
through this inductive element, causing a varying magnetic  eld inside of the plasma. This
changing magnetic  eld will cause an RF electric  eld in the plasma which accelerates the
ions through the sheath. Generation using this approach can create higher plasma densities
which increase both the directionality and the rate at which the silicon is etched. [44]
5.4.4 Deep Reactive Ion Etching
Finally, one last improvement in the plasma etching process, referred to as the Bosch
Process, because it was developed by Laermer and Schilp of Bosch, was to combine separate
etching and passivating cycles to the Reactive Ion Etching process. [45] This is the process
used at Auburn for anisotropic silicon etching. First, the standard cycle using SF6 and O2
plasma etches a trench in the surface of the silicon. Then a second cycle of C4F8 plasma
reacts with the SF6 to create a Te ona174 like coating that covers the bottom and sides of the
trench to prevent further etching. The etching cycle is then repeated and the SF6 and O2
ions bombard the surface dissolving the passivisation layer at the bottom of the trench and
etching further into the silicon. Because the ions arrive mostly perpendicular to the surface,
they do not dissolve the passivisation layer on the sidewalls thereby leaving the silicon to
the side untouched. Very high aspect ratio trenches can be achieved using this process with
reports as high as 107. [46]
The STS Advanced Silicon Etcher (ASE), shown in  gure 5.10, uses all the methods
mentioned previously to perform DRIE on silicon to create a very high aspect ratio etch.
An inductively coupled plasma etches the silicon surface using chemically reactive SF6 ions.
Using an induction coil to generate the plasma has the bene t of being able to create a very
high density plasma, and because the coils are located outside of the chamber, contamination
that is associated with other methods of plasma generation does not occur. The platen, onto
53
which the wafer is attached for etching, is connected to a RF source that has two settings
at either 13.56MHz or 380KHz. A Bosch process is implemented to create an anisotropic
etching pro le by subjecting the surface to multiple SF6=O2 etch and C4F8 steps. [45]
Figure 5.10: STS Advanced Silicon Etcher
Multiple parameters can be altered which modify the etch pro le. These parameters
include the following:
a136 Gas Flow Rate
The mixture of O2 to SF6 determines how aggressively silicon is etched. The higher
the ratio of SF6, the faster and more isotropic etch is during each cycle.
a136 RF Power
54
The power of the plasma determines the amount of directionality the ions have toward
the wafer surface. An increase in the power will cause a more anisotropically etched
structure.
a136 Frequency
There are two frequency settings available for use. The  rst is 13.6 MHz and the
second is 380 KHz. Use of the 380 KHz setting can cause a phenomenon known as
footing at the interface between the Si and the SiO2 etch stop layer in which energetic
ions can become trapped in the SiO2 layer giving it a positive charge. This charge
de ects incoming ions as it approaches the surface, causing the etching species to etch
the sidewalls at a higher rate. A common issue associated with DRIE etchers that can
increase the likelihood of footing is an occurrence known as the bulls-eye e ect. This
e ect is characterized by the non uniform etching of silicon across the surface of the
wafer. The silicon has a tendency to etch at a faster rate toward the edge of the wafer
than in the center. In order to completely expose the SiO2 layer over the entire wafer,
some portions will be required to be over etched. This over exposure of the oxide layer
to ion bombardment increases the likelihood of footing. By using the low frequency
setting, the surface can be over etched without resulting in the footing e ect. [47]
a136 Etch cycle time verses passivisation cycle time
The ratio of etching to passivisation plays a role in the aspect ratio of the structures.
An increase in the etching ratio will allow more silicon etching during each cycle and
increase the amount of lateral silicon etching under the pattern. An increase in the pas-
sivisation ratio will increase the time required for the etching gas to break through the
passivisation layer, thus shortening the amount of time silicon is exposed for etching.
A number of di erent research groups have used the preceding variables to control the
sidewall pro le e ectively. One group used an alternating sequence of the Bosch process
with an isotropic dry etch to create tapered sidewalls with a precisely controlled angle. [48]
55
Another group created very high aspect ratio trenches by slowly increasing the  ow of SF6
during processing. This increase compensates for the reduced ability of the ions to reach the
surface at deeper trench depths maintaining a consistently high aspect ratio. [49]
The etch recipe called MORGNSOI was used for etching all of the silicon wafers in this
project. The parameters of this process are shown in Table 5.5.
Table 5.5: MORGNSOI DRIE Recipe
Cycles
O2 SF6 C4F8 Time
Etch Cycle 13 sccm 130 sccm 0 sccm 13Sec
Passivate Cycle 0 sccm 0 sccm 85 sccm 7 Sec
Power
Range tolerance match load match tune
13.6MHz connected to Coil 600W 99% 50% 50%
Platen connected to 13.56mHz 0-300W 99% 50% 50%
Helium Leakup
Test Time 30 Sec
Max Leakup Rate 30mTorr/min
Oxidation Furnace
The oxidation furnace, used to oxidize silicon, is a three tube horizontal furnace that
uses a controlled  ow of oxygen, hydrogen, and nitrogen gas to control oxide growth. Pure
oxygen can be used to grow extremely pure thin layers of \dry oxide", while the addition
of hydrogen allows for thicker but lower quality \wet oxide"layers. Nitrogen is also used to
both prevent oxide formation, and to purge the chamber of impurities. The furnace is a
three zone resistance heated furnace that can be set to temperatures between 400 and 1200
 C.
56
Figure 5.11: Oxidation Furnace
5.4.5 Electron Beam Evaporation and Sputtering System
The CHA Mark 50 electron beam chamber shown in Figure 5.12 is used to deposit metal
layers on the surface of a device with high accuracy. The system dposits material by using
a magnet to focus the electron beam onto a target crucible of the metal. The metal then
evaporates and is deposited onto wafers suspended in a planetary above the crucible. This
process takes place under a high vacuum which results in a very high quality  lm that can
be deposited at a high rate (50-500nm/min). [50]
57
Figure 5.12: CHA Mark 50 dual E-beam/sputter/ion gun deposition system
5.4.6 Masking Materials
When creating the masking layer for a macro-scale MEMS pattern that will undergo
plasma etching, a high selectivity between the masking material and silicon is desirable.
If the selectivity is low, a thicker mask is required to withstand the hundreds of cycles of
etching to which it is exposed. The selectivity of materials can vary with changes in the
density, power, etching time, frequency, and chemical composition of the plasma. Changes
in any of these variables to increase selectivity, however, can also have negative e ects on
the silicon etch rate, the amount of undercut, as well as the roughness of the sidewalls.
58
SiO2 can be e ectively used as an etching mask and has high selectivity at about 100-
1. Unfortunately, the time required to grow oxide increases exponentially with thickness,
therefore, growing enough oxide to withstand etching through a full wafer is not feasible.
For very deep structures, thick  lm photoresists can be used to mask the etching pro-
cess. Photoresist can have selectivities comparable to SiO2, but can be applied in much
thicker  lms. If the  lm thickness requirement becomes too high, the pattern resolution can
be a ected. Photolithography equipment can resolve patterns to a very small resolution.
Associated with any particular resolution, there is a certain vertical tolerance called depth
of focus, within which the pattern will remain in focus. Depth of focus can be de ned by
equation 5.2
DOF = k2R
2
k21NA (5.2)
Where k1 and k2 are process dependent constants, NA is the numerical aperture, and R
is the resolution to be achieved. In order to resolve a smaller feature, the depth of focus
must be lowered. If the masking photoresist is too thick, the pattern will not maintain focus
through to the surface of the resist, causing some areas to remain unexposed. [50]
5.4.7 Silicon Oxidation
It is well known that oxygen reacts with crystalline silicon to form SiO2. This property
is one of the major reasons silicon has found such widespread use in the electronics industry.
The high insulation value of SiO2, and the ability to precisely control the thickness of oxide
that is formed make it ideal for electronics devices such as MOS transistors and capacitors.
Another property of silicon dioxide is that it is a very stable molecule both thermally and
chemically, making this material useful for protecting the silicon during processing, and
chemical etch mask.
59
5.5 Chapter Summary
This chapter provides a detailed analysis all of the steps involved in the design and
fabrication of both the PPA test device and the detection and controlling circuit in which
the device will be used. An FEA analysis of the PPA device was performed to verify the
mechanical properties of the  nal device. From there a simulation of the mechanical and
electrical operation of the device was performed in Simulink to verify the design works as
expected. The fabrication tools and processes were then detailed to give an understanding
of how the device was created. The entire  nal fabrication process can be found in Appendix
A.
60
Chapter 6
Experimental Veri cation of Pull-in Detection and Mitigation
A test structure was fabricated to verify the simulations from the previous section. The
actuator created was a silicon structure fabricated using typical MEMS fabrication techniques
of photolithography and Deep Reactive Ion Etching (DRIE) on a 375 m thick p100 silicon
wafer with 0.1  cm resistivity. The photolithography mask was created using a program
called ViewMate.
The entire PPA fabrication process is relatively simple with only a few steps. First, the
wafer is cleaned using the RCA clean detailed in table 5.4. The wafer is then patterned by
exposing and developing a layer of photoresist using the photolithography mask. The wafer is
placed on a backing wafer using a temporary bonding material maintain mechanical strength
during the etching process. A thin layer of photoresist often works well as a thermally
conductive bonding material. The exposed areas of the pattern are etched all the way
through the silicon wafer using a DRIE process, speci cally the MORGNSOI program listed
in table 5.5. Once etching is complete the Individual devices are removed from the backing
wafer and gently cleaned in a bath of acetone followed by a weak phirana etch solution
consisting of 4 parts H2SO4, 1 part H2O2. These two processes should clean all of the
photoresist and passivisation residue left on the surface of the parts. Finally, the devices
are mounted on a holding wafer and suspended in an e-beam deposition chamber to deposit
a solderable metal contact of titanium and copper. The complete description of the metal
deposition parameters are listed in Appendix B. The full traveler for this process is shown
in Appendix A
The MEMS device was suspended over an electrode printed on a PCB, shown in Figure
6.3, by a non-conductive spacer of a set 150 m width, allowing the MEMS movable plate to
61
move up and down relative to the  xed plate on the PCB. When operating with resistances
greater than 1M , performing accurate measurements on electrical characteristics can be-
come problematic because the input resistance of many measurement tools is less than the
resistance of the system. For this reason, measurement of the displacement of the actuator
was performed using an interferometer which a ords a non-electrical measurement of the
system dynamics.
6.1 Schematic and Layout of Test Platform
To test the device, a transimpedance ampli cation circuit was designed in LTSpice. The
circuit consists of a transimpedance ampli er connected to a comparator which compares
to a reference voltage set to the value at which pull-in occurs. When the current reaches a
large enough pull-in value, the comparator goes high initiating the clock input of a positive
edge triggered  ip  op. The  ip  op then initiates the shut-down mechanism of the circuit.
In order to safely cut the voltage from the actuator, two operations must be performed.
First, the actuator electrodes need to be brought to the same potential. This is achieved by
connecting the drain and source of an N-channel MOSFET with a low Rds-on resistance,
and a high maximum Vds breakdown voltage, between the high side PPA electrode and
ground. The charge can then quickly  ow from the actuator to ground reducing the force on
the actuator. Second, the voltage supply must be disconnected from the circuit, preventing
the newly shorted actuator from pulling a constant current from the voltage source which
could damage the voltage source if the series resistance is low. Isolation of the voltage source
is achieved by using a P-channel MOSFET with a low Rds-on resistance and high Vds
breakdown voltage. The gate of the Pmos is connected to a resistor divider in series with an
nmos transistor. Before pull-in occurs, the nmos transistor is active causing a voltage drop
across the resistor divider providing the voltage necessary to turn the transistor on. Because
the source voltage of the resistor divider is variable, a zener diode is included to ensure that
the voltage on the pmos gate does not go above the maximum rated value. A drawback to
62
the resistor divider con guration, however, is that below a certain value of source voltage,
the drop across the resistor may not be enough to fully turn on the PMOS. This low voltage
value can be tailored by the value of the resistors used. Figure 6.1 shows the entire electrical
schematic implementing all of the previously mentioned functions.
63
Figure 6.1: Electrical Schematic of Transimpedance ampli er Pull-in Detection Circuit
64
Figure 6.2 shows the PCB layout of the schematic from Figure 6.1. This layout was
completed in a program called FREEPCB. The green areas represent copper on the top
of the board, while the red areas represent copper on the bottom of the board. The gray
areas are drill holes for vias, through hole parts, and board mount locations. The blue areas
represent silk screen on the top of the PCB, and the yellow areas represent silk screen on
the bottom of the PCB. The PPA is mounted over the two square pads in the center of the
board. The outer square pad acts as a  xed distance mount so that the entire device is equal
in height above the center contact. The inner square pad is the bottom  xed electrode. The
headers x1 and x2 were placed to allow for disconnection of the entire detection circuit from
the PPA device to connect to an external circuit for future testing. Both safety mechanisms
are located at the left side of the circuit board as shown by the three through-hole power
mosfets. The three 8-pin dip packages are the transimpedance ampli er, a small gain stage,
and a comparator. The opamps used are TI TLE2071 high speed low noise jfet input opamps,
and the comparator is a linear technologies lt1016 ultra fast precision 10ns comparator. The
output of the comparator feeds to a 74hc374 positive edge triggered  ip  op, the output of
which feeds to both an led with a reset switch to show that the circuit has been tripped,
as well as an 74hc04 inverter which provides the correct signal to shut down the safety
mechanism.
As information, the following is a list of observances that may be helpful in future work.
a136 Because such a large amount of gain is required in the ampli ers, the circuit became
very susceptible to noise. Large instantaneous changes in the DC power supply would
cause enough noise to trip the circuit. Also, physical disturbances large enough to
displace the PPA electrode would also sometimes be enough to trip the circuit.
a136 Traces had to be shortened to as short a length as possible because external interference
would be picked up on long wire runs.
65
a136 Precise tightening of the compression  xture was required to alleviate stress on the
springs which could cause the plate to tilt.
a136 More test points were needed at critical points in the circuit to make testing go more
smoothly.
Figure 6.2: Layout of Transimpedance Ampli cation Circuit.
Figure 6.3 shows the  nal populated test board used in the experiment. The design
completed in  gure 6.2 was fabricated by a PCB manufacturer called Advanced Circuits.
The PPA is mounted on the PCB using a compression acrylic compression  xture. A plastic
spacer of set thickness is placed under the PPA between the board and the device to set
66
the gap distance between the two electrodes. Another plastic  xture is location stencil to
align the device with the plate. Once the device is aligned over the  xed electrode, the
compression  xture is tightened with four screws located at each corner of the  xture. The
schematic for the acrylic  xture is located in  gures E.1 and E.2 of Appendix E Attached to
the  xture is a strip of copper conductive tape. This tape facilitated the connection between
the upper movable electrode and the detection circuit. As a side note, after a long period
of time, both the copper tape and the copper deposited became slightly oxidized which was
likely accelerated by high voltage arcing when the electrodes made contact. This oxidation
had to be cleaned to maintain reliable device operation.
Figure 6.3: Fabricated Transimpedance Ampli cation Circuit.
67
6.2 Results of Test Platform Measurement
The motion of the device was observed using an optical vibrometer and a small re ector
attached to the center of the PPA. As the actuator moves, the motion is detected and
converted to an electrical voltage with an accuracy of 20 mV . The device was  rst tested with
the shutdown mechanism of the circuit turned o and a current limiting resistor in series with
the actuator. This allowed the actuator to completely snap together to show the full motion
of the device, and the behavior of the current up until contact is made. This behavior is
shown in  gure 6.4. Similar to the Simulink simulation, the current increases when the pull-in
condition is achieved, but when the plate makes contact, the current immediately goes to the
rail indicating that the device has shorted. The current values gathered experimentally, while
similar in magnitude, di er in slope from the Simulink data, particularly when the plates
are at their closest. This discrepancy is most likely attributed to the non-idealities not taken
into account in the simulation. Squeeze  lm damping likely reduced the acceleration of the
plate at small electrode separation values causing reduced  nal current. Due to the design
of the device meant to reduce the pull-in voltage by making the springs as long as possible,
the springs may be slightly nonlinear in di erent regions of actuation.
68
Figure 6.4: Pull-in Event Without Safety Shutdown.
Table 6.1 provides a summary of currents at notable positions during the pull-in process.
As shown below, there is about an order of magnitude di erence between the simulated and
experimental values of pull-in This di erence increased as the plated moved closer to contact.
This e ect could have a number of causes. possible reasons include the following
a136 The equation 5.1 used to calculate the spring constant is only an approximation, and
may not fully account for the springs wrapping around the proof mass which also causes
a slight twisting action.
a136 Resistance in series with the circuit path can hinder current  ow causing a slowing in
velocity of the PPA.
a136 Insu cient slew rate of the transimpedance ampli er could cause the signal seen at
the output to be lower than the actual value of current passing through the capacitor.
69
a136 In order to test the transmissibility of the device, the PPA was placed on a shaker
with no bottom electrode attached so that the device would be able to move at many
times the magnitude of the input. The lack of a bottom electrode could have caused a
reduction in the constant damping that was used in the simulation in comparison to
the device when attached to the  xed bottom electrode.
a136 Similar to the previous item, as the plate moves toward the opposing plate, squeeze
 lm damping would have an increasingly large non-linear e ect causing the device to
reduce speed dramatically the close it came to contact.
Experimental Current vs. Position
Distance from Contact Current (Meas) Current (Sim)
7.00E-05 m 6.4e-8A 2.50E-07A
4.00E-05 m 2.7e-7A 2.50E-06A
1.00E-05 m 1.03e-6A 4.85E-05A
Table 6.1: Experimental Analysis of Current as a Function of Position
Next, the shutdown mechanism was turned on to test the accuracy and speed at which
the voltage could be shut down. A number of trials were run with the comparator voltage
set at di erent levels. Two examples are shown in  gures 6.5 and 6.6. In Figure 6.5,
the comparator voltage was set to 183mV. Using a gain of 375,000 on the transimpedance
ampli er, this voltage should correspond to a current value of 0.488 A. This should equate
to a plate position of around 26 m from the opposing plate. In Figure 6.6, the comparator
was set to 331mV which corresponds to a current of .882 A and a position of around 12 m.
In the two  gures, three graphs are presented. The top graph shows the position of the
70
plate shown by the interferometer. The middle graph shows the moment at which the
comparator is tripped. The bottom graph displays the voltage that is present at the output
of the transimpedance ampli er which has been converted in Matlab to display the value
of the current going through the actuator. All three graphs are synchronized with time.
The bottom graph displays a large scale ringing in the current signal at the moment the
circuit shuts down. One explanation for this is that when the shutdown mechanism circuit
is tripped the inductance in the circuit trace and mosfet that is parallel to the PPA creates
an LC tank circuit with high damping resistance that causes the circuit to oscillate while
the capacitor discharges. Because this e ect occurs after the circuit has been tripped and
the actuator returns to its rest-gap position, it can be assumed that this is predominately an
electrical e ect and has no tangible impact on the mechanical performance of the actuator.
The current graphs of both  gures verify that the shut-down circuit trips at their appropriate
values, and that the actuator immediately rises back to its rest position indicating that the
charge has been completely dissipated from the actuator.
71
Figure 6.5: Safety Shutdown at 4.88e-7 Amps.
72
Figure 6.6: PSafety Shutdown at 8.82e-7 Amps.
6.3 Chapter Summary
In this chapter, the process of verifying the simulation developed in chapter 5 is covered.
The schematic and layout of the test platform are discussed explaining the operation of the
various circuit blocks, including ampli cation stages, digital control, and safety mechanisms.
The  nal circuit was then evaluated by displaying the electrical output on an oscilloscope
and then comparing the data received with the signal from a laser vibrometer directed at
the surface of the plate. First the device was allowed to complete the full pull-in event to
record the current during the entire process and compare it with the simulated values, It was
73
found that the current did follow the a trend of exponential increase, but was o by about
an order of magnitude. The reasons for this discrepancy were discussed.
By connecting the shut-down mechanism on the circuit, it was then shown that the
circuit can indeed respond to the onset of pull-in to prevent the device from making contact.
Further, the circuit is fast enough to respond at precise locations during the pull-in event
allowing  ne grain control over the position the plate will displace beyond the 13 of the rest-
gap position. The experimental results showed fairly good correlation with the simulated
values in Simulink, with deviation at small gap values that has been predominately attributed
to squeeze  lm damping that has not been added into the simulation.
74
Chapter 7
Extension of Pull-in Work to Create a Tunable Resonator
While in most cases, the onset of pull-in has detrimental e ects, there exist some in-
stances where pull in should be allowed to occur. Some devices are designed to use the
pull-in e ect in the operation of the device. The pull-in detection method presented here
does not only have to be used to halt the onset of pull in but can be used to record that
pull-in has occurred or is about to occur. As an example, this detection method may allow
a circuit to keep track of the number of times a switch designed on the pull-in principle has
become active, and even make changes to the circuit in preparation for the contact event. In
an extension to the work on pull-in mitigation, a device was created as explained previously
to operate in just this manner. By adding one passive external resistor in series with the
PPA previously fabricated a tunable resonator was created that operates on the e ect of the
charging and discharging actuator when pull-in is completed. The device will resonate at
a frequency determined by the value of the resistor by slowing down the charging time of
the capacitor in proportion to the resistor value. The mitigation method discussed can be
directly applied to this resonator to detect each pull-in event on an on going basis with the
addition of a zener diode to protect the input of the transimpedance ampli er. This chapter
discusses the design and testing of the resonator structure to show that a stable oscillation
occurs over a wide range of frequencies. [51]
7.1 Pull-in Resonator
As discussed previously, if the voltage on the parallel plate actuator increases over a cer-
tain value, the system becomes unstable and the plates will snap together. By making both
plate surfaces of the PPA conductive and grounding the negative terminal of the actuator, a
75
short circuit is made when snap-in occurs. Once this happens, the actuator capacitor will be
allowed to discharge, thus reducing the voltage across it to zero, allowing the restoring force
of the spring to regain control. When the actuator releases, the voltage will immediately
increase back to the previous pull-in condition before the mechanical system has time to
react. This results in the actuator being locked into place. By placing a very large resistance
between the voltage source and the actuator, the current  owing from the voltage source
to the actuator can be reduced, increasing the time required to restore the charge on the
positive plate of the actuator. If the charge time is su ciently increased, due to the presence
of a large value resistor between the actuator and the power supply, the spring system will
then be able to move the actuator away from its snap in position before the voltage between
the two plates once again reaches the pull-in condition. Under non-equilibrium conditions,
a general description of the force on a parallel plate actuator can be shown by the di eren-
tial equation shown below where m is the mass of the actuator proof mass, c is a damping
coe cient which can be caused by the environment in which the actuator is operating, and
k is the spring constant of the system. Figure 7.1 shown below gives an illustration of the
mass spring damper system that is described by Figure 7.1.
Figure 7.1: Mass Spring Damper System Diagram.
d2x
dt2 +
dx
dtc+xk =
V2 o rA
2 (xo x)2 (7.1)
The circuit diagram below shows the resonator system.
76
Figure 7.2: Actuator Circuit.
By doing a simple analysis of the circuit in Figure 7.2, the current equation for the
current in the circuit can be found, and is shown in the left side of 7.1. Because the ca-
pacitance of the actuator is variable, the current into the actuator must be characterized by
two components which are shown on the right side of 7.1. The  rst component is the usual
equation for a  xed capacitor where the change in voltage over time causes a current  ow.
The second component is a result of the change in capacitance over a change in time due to
the movement of the actuator. This second component is important, because as the movable
plate approaches contact with the  xed plate, the capacitance between the two plates grows
enormously. A large series resistance added to the circuit reduces the amount of charge that
can  ow into the capacitor, which in turn requires that the voltage across the capacitor must
drop.
I = Vin VcR = VcdCadt +CadVcdt (7.2)
(7.3)
dVc
dt +Vc
  x
o x
R o rA
 
+ dxdt
 1
xo x
  
 Vin (xo x)R 
o rA
= 0 (7.4)
By substituting and reorganizing 7.2, yield 7.3, and  nally 7.4. The di erential equation
7.4 is enough to fully describe the system in Figure 7.2 with the constraint that x xo.
77
7.2 Pull-in Resonator Simulation
A MATLAB Simulink model of the actuator system was created to con rm the behavior
of a parallel plate actuator using various values of resistance to restrict charge  ow. In this
simulation, the spring system was modeled using the equations presented previously. A
constant voltage was maintained which was above the pull-in voltage of the system, while
multiple values of resistance impeding the current  ow into the actuator were inserted into
the circuit to measure the response. Because the voltage was greater than the pull-in voltage
of the actuator, the spring moved past the stability point and the plates snapped together.
Once the plates shorted, the voltage across the actuator dropped to zero and the charge on
the positive plate of the actuator quickly  owed through to the negative plate. Once fully
discharged, the two plates then released from each other and began moving apart.
If the resistance between the voltage source and the actuator was low, then the current
 ow after the plates released was large enough that the charge on the positive plate built
back up quickly before the mechanical system had time to respond. The fully charged plates
then snapped back together, resulting in an e ective permanent snap-in condition.
By increasing the resistance into the Mohm range, the charge time on the positive plate
increased enough so that the springs of the actuator have time to respond before the actuator
capacitance fully recharged. The actuator plates are able to move a small distance away from
each other before the charge built back enough to achieve pull-in voltage again. Once this
happened, the actuator returned to its original snap-in condition, creating a periodic cycle
of snap-in and release conditions at a frequency that was proportional to the resistance used.
Currently, this simulation is only a general description of the system behavior to show the
e ect of resistance on a variable capacitance, and is not highly accurate to the speci cations
of the actual device created for this experiment. A work has been performed on interfacing
a PSPICE output to Simulink which has the potential to allow a much more accurate model
of the response of the actual fabricated device [52]. In future analysis of the resonator
behavior, this technique will be used to model the system. Figure 4 and Figure 5 show
78
the resonant behavior of the actuator using two di erent high resistance values. As can be
seen in the  gure, once snap-in condition is achieved, the high resistance value added to the
system causes the actuator to move into an oscillatory state. When the value of resistance is
increased, the amplitude of oscillation increased as well, along with a corresponding decrease
in frequency due to this extended range of motion.
Figure 7.3: Simulink Model Using 4M Resistance.
79
Figure 7.4: Simulink Model Using 8M Resistance.
7.3 Pull-in resonator Experimental Veri cation
The same PPA test device used to develop the pull-in detection mechanism is used
here to create a resonant device. Similarly, the same properties that make this device well
suited to detecting the current  ow through the actuator make it an ideal device for creating
resonance at low frequencies at a suitable voltage range. As shown previously a  nite element
analysis model of the device in Intellisuite con rmed the following mechanical properties.
The primary mode natural frequency though simulation was found to be 248 Hz. Once
the device was fabricated, the device was placed on a mechanical shaker to determine the
transmissibility curve. The resulting curve showed peak transmissibility at 219.5 Hz, which
is relatively close to the simulation prediction. The deviation in frequency was attributed
80
to simulation assumptions, variations in the material properties, and fabrication tolerances.
One consequence of designing a device in such a way that the springs wrap around the device,
as in this case, is that in deformation, the mass has a tendency to rotate. Since, in this
experiment, displacement is limited to 40 m this rotation e ect was considered negligible.
Another e ect of the extended spring length is that higher order vibration modes have
resonant frequencies only slightly higher than the primary mode. This o -axis movement
could cause an unintended reduction in capacitance, as well as cause an imperfect surface
contact between the two electrodes when pull-in occurs thereby causing incomplete charge
dissipation. While capacitance variation was not found to be a signi cant source of error,
contact quality had a major e ect on oscillation reliability. It was determined that o -axis
vibration was not the main contributor to the contact quality issue. Contact cleanliness and
non-uniform plating of the PCB electrode also played a signi cant role. When operating with
resistances greater than 1M, performing accurate measurements on electrical characteristics
can become problematic because the input resistance of many measurement tools is less
than the resistance of the system. For this reason, measurement of the displacement of the
actuator was performed using an interferometer which a ords a non-electrical measurement
of the system dynamics.
Figures 7.5,7.6, and 7.7 show the completed components separately and then assembled
together. The extra circuitry on the PCB is control circuitry for the actuator that was not
used in this investigation.
81
Figure 7.5: PCB Electrode.
Figure 7.6: Fabricated PPA Spring System.
82
Figure 7.7: Assembled PPA test device.
To test the device, a 400V DC voltage was applied to the circuit. This voltage ensured
that the supply voltage remained well above the pull-in voltage of the device. Through ex-
perimentation, the measured pull-in voltage was found to be 250V. Varying series resistance
values from 21M to 81M in increments of 10M were tested in the circuit. During each
test, fairly precise audible oscillation was observed, as well as measured by the interferome-
ter. The oscillations at each frequency is displayed in Figure 7.8. A 60Hz noise signal was
also detected in the gathered interferometer data which was removed with post processing
of the data in MATLAB. In future experiments, a high-pass  lter will be added to the data
collection circuitry to reduce this noise. As resistance increased, frequency of oscillation
decreased. An FFT analysis was performed on the data in MATLAB, shown in Figure 7.9,
to show the movement of the dominant frequency. A frequency shift from 427 Hz down to
146.48 Hz was observed as resistance varied from 21M to 81 M . A displacement of about
2 m was measured at lower resistances, but only a slight increase in amplitude was observed
as the resistance was increased. It may be possible to further increase the amplitude using
larger values of resistance.
83
Figure 7.8: Pull-in Oscillations at Various Resistances.
84
Figure 7.9: FFT Analysis of Pull-in Data .
85
From the data from Figure 7.9, a de nite decreasing trend in frequency was found with
increased resistance. By graphing this data in Figure 7.10, this trend can be clearly seen.
As the resistance is reduced below 21 M a further increase in frequency is expected. This
will also, however, greatly increase the current  ow, which if too high results in large power
consumption and can damage the device. As a test, a resistance of 30 K was placed into the
circuit, and as expected, the plate became stuck in place. The current discharge, however,
caused carbon residue to buildup on the surface, degrading the contact quality, and causing
unreliable device operation.
Figure 7.10: Frequency Trend as a Function of Resistance.
7.4 Chapter Summary
By using a high value resistance in series with a parallel plate actuator under snap-
in condition, a resonant behavior results in which the device cycles in and out of snap-in.
86
This behavior occurs at a fairly precise and stable frequency which is controllable by the
amount of resistance applied to the circuit. By utilizing the nonlinear behavior inherent
in the snap-in condition, a tunable MEMS resonator can be created utilizing a minimum
amount of external circuitry. Frequencies between 146Hz and 427Hz were using values of
resistance between 21Mohm and 81Mohm. These frequency values are both well below and
well above the resonant frequency of the device. Because such high values of resistance
were used, there is very little current  ow required to actuate the device. Little more than
the current required to charge the capacitor is needed to continuously actuate the device,
although at higher frequencies, faster charging and discharging cycles will cause more power
consumption.
87
Chapter 8
Conclusions
This work showed through simulation in Simulink and experimentation with a fabri-
cated silicon MEMS structure, that by measuring the current  owing into a parallel plate
actuator, the onset of pull-in can be detected by sensing the increased current caused by
the rapid increase in capacitance using a  xed voltage source. By creating a model to ac-
curately simulate the PPA motion, the value of the current was predicted to be 1.67nA at
the onset of pull-in, then 250nA when the plate reaches approximately 70um from contact,
and increasing to 10uA just before the plates make contact. In comparison the experimental
results show that when the plate was at 70um from contact the current  owing through
the capacitor measured to be 64nA and just before contact, 1.03uA was measured. While
the data compares at the same order of magnitude, there is substantial error between the
results which increases as the plates move closer to contact. This can be explained by a
few factors. First the Simulink model used a constant value for the damping coe cient to
resist the motion of the PPA, while in reality motion at close proximity to an opposing plate
contributes a non-linear squeeze- lm damping term, which opposes the acceleration of the
pull-in e ect. Another potential source of error can be attributed to unknown resistance in
the circuit. While unlikely to cause much change, excessive resistance in the circuit could
impede the current  ow through the actuator, thus slowing the pull-in. Finally, if the slew
rate of the opamp is not su ciently high, the transimpedance ampli er may not accurately
re ect the current  owing through the actuator at high speed. This scenario cannot be ruled
out because the ampli er gain was set at or slightly above the maximum gain bandwidth
product line of 10MHz in order to produce a measurable signal from the current.
88
Further, the position of the actuator can be somewhat controlled by associating the
speed of the device with the current providing information about the position within the
pull-in region. The mitigation event can be delayed until just before the plate makes contact
thus preventing damage.
Also as an extension to the work of pull-in detection and mitigation; if the pull-in event
is allowed to complete, assuming su cient circuit protections are in place, and a high value
resistor is placed in series with the PPA thereby creating an RC time constant of the same
order of magnitude as the resonant frequency of the PPA, the device can be made to resonate
with a small displacement. Moreover, as the resistance value is increased, the displacement
distance will also increase, and the frequency of the oscillations will decrease proportionally.
This behavior was also shown both in simulation and through physical testing. The model
created in this simulation only showed a generic behavior of the device and could not, with
a high degree of accuracy determine the exact relationship between resistance and frequency
of the actual device because of a number of unknown parameters that would be both very
di cult to model, and can vary from device to device. A few of these parameters being,
motion of the PPA that is not perpendicular to the opposing contact and discharge delay
due to corrosion or non-planar electrode surface. The simulation was, however, appropriate
for showing the general e ects of the charge discharge cycle, speci cally, larger displacement
and higher frequency motion in relation to varying resistance. In testing, a frequency change
from 427Hz to 146Hz was observed with resistances ranging from 21MOhms to 81MOhms.
This type of resonator with its simplistic design could have uses in a number of di erent
applications from circuits that operate at a precise frequency, to a sensor that responds to a
large scale change in resistance, to generation of an acoustic sound source.
89
Chapter 9
Future Work
9.1 Pull-in Detection Future Work
The Simulink model can be improved to more accurately re ect the non-linearities of the
PPA. The device could be tested in di erent environments including vacuum. This should
allow for a higher Q and reduced squeeze  lm damping, which should allow the device to more
closely follow the simulated results. The device created for this experiment was designed
to magnify the motion of the PPA to achieve the largest current possible. By increasing
the sensitivity of the detection circuit, this experiment can be scaled down to smaller scale
actuators. get high side sensing to work. Fabrication of a laterally actuated PPA on an
SOI substrate would allow for a reduction in size and generally more reliable performance.
Reducing the size will decrease have o setting changes to the current signal produced. A
reduction in surface area will occur due to the smaller mass in actuation, however, The
reduced mass will also allow the device to achieve a higher velocity over a shorter distance,
which will increase the CdVdt term.
This work showed that it is possible to detect the onset of pull-in with a limited ability
to determine the position of the plate a given time over the pull-in event. More work can
be done fully examining the relationship between current and position, which will give even
more precise control over the dynamics of the phenomenon.
Further work would be to implement this system into a larger system to show reliability
over a long period of time.
90
9.2 Pull-in Resonator Future Work
The device has so far not been tested under vacuum conditions. This should result in an
increase in amplitude at the same resistance value by reducing the damping coe cient. Sec-
ondly, because the current device was created emphasizing ease of use, a frequency increase
can be achieved by creating a physically smaller version of the actuator. Because contact
quality was an issue in this experiment, utilizing an SOI design process for the miniaturized
version would minimize the amount of manual positioning currently required, which was a
major source of error.
91
Bibliography
[1] A. Sundaram, M. Maddela, R. Ramadoss, and L. Feldner, \MEMS-Based electronically
steerable antenna array fabricated using PCB technology," Microelectromechanical Sys-
tems, Journal of, vol. 17, no. 2, pp. 356 {362, Apr. 2008.
[2] C.-H. Han, D.-H. Choi, and J.-B. Yoon, \Parallel-plate MEMS variable capacitor with
superior linearity and large tuning ratio using a levering structure," Microelectrome-
chanical Systems, Journal of, vol. 20, no. 6, pp. 1345 {1354, Dec. 2011.
[3] M. Kraft, C. Lewis, T. Hesketh, and S. Szymkowiak, \A novel micromachined
accelerometer capacitive interface," Sensors and Actuators A: Physical, vol. 68, no. 13,
pp. 466{473, Jun. 1998. [Online]. Available: http://www.sciencedirect.com/science/
article/pii/S0924424798000648
[4] S. J. Kim, G. Flowers, C. Chen, and R. Dean, \Active vibration control and isolation
for micro-machined devices," in ASME 2008 Conference on Smart Materials, Adaptive
Structures and Intelligent Systems, Jan. 2008, pp. 657{664. [Online]. Available:
http://dx.doi.org/10.1115/SMASIS2008-423
[5] S. M. Wentworth, Applied Electromagnetics : Early Transmission Lines Approach,
1st ed. Wiley, Jan. 2007.
[6] V. Kaajakari, Practical MEMS: Design of microsystems, accelerometers, gyroscopes, RF
MEMS, optical MEMS, and micro uidic systems. Small Gear Publishing, Mar. 2009.
[7] Y. Nemirovsky and O. Bochobza-Degani, \A methodology and model for the pull-
in parameters of electrostatic actuators," Journal of Microelectromechanical Systems,
vol. 10, no. 4, pp. 601{615, 2001.
[8] J. I. Seeger and B. E. Boser, \Dynamics and control of parallel-plate actuators beyond
the electrostatic instability," in Transducers, vol. 99, 1999, p. 474477.
[9] A. Nayfeh, M. Younis, and E. Abdel-Rahman, \Dynamic pull-in phenomenon in
MEMS resonators," Nonlinear Dynamics, vol. 48, no. 1, pp. 153{163, 2007. [Online].
Available: http://www.springerlink.com/content/23228w234n10h5j8/abstract/
[10] J. Seeger and B. Boser, \Charge control of parallel-plate, electrostatic actuators and the
tip-in instability," Journal of Microelectromechanical Systems, vol. 12, no. 5, pp. 656 {
671, Oct. 2003.
92
[11] J. Yao and M. Chang, \A surface micromachined miniature switch for telecommunica-
tions applications with signal frequencies from DC up to 4 ghz," in Solid-State Sensors
and Actuators, 1995 and Eurosensors IX.. Transducers ?95. The 8th International Con-
ference on, vol. 2, Jun. 1995, pp. 384 {387.
[12] G. Rebeiz and J. B. Muldavin, \RF MEMS switches and switch circuits," IEEE Mi-
crowave Magazine, vol. 2, no. 4, pp. 59{71, 2001.
[13] S. Pacheco, L. P. B. Katehi, and C. Nguyen, \Design of low actuation voltage RF
MEMS switch," in Microwave Symposium Digest. 2000 IEEE MTT-S International,
vol. 1, 2000, pp. 165{168 vol.1.
[14] J. B. Muldavin and G. Rebeiz, \High-isolation CPW MEMS shunt switches. 1. mod-
eling," IEEE Transactions on Microwave Theory and Techniques, vol. 48, no. 6, pp.
1045{1052, 2000.
[15] J. Xie, J. Shih, Q. Lin, B. Yang, and Y.-C. Tai, \Surface micromachined electrostatically
actuated micro peristaltic pump," Lab on a Chip, vol. 4, no. 5, p. 495, 2004. [Online].
Available: http://pubs.rsc.org/en/Content/ArticleLanding/2004/LC/b403906h
[16] J. Seeger and S. Crary, \Stabilization of electrostatically actuated mechanical devices,"
in Solid State Sensors and Actuators, 1997. TRANSDUCERS ?97 Chicago., 1997 In-
ternational Conference on, vol. 2, Jun. 1997, pp. 1133 {1136 vol.2.
[17] X. Wu, Z. Xiao, J. Zhe, and K. Farmer, \Modeling and simulation of two passive
feedback methods to obtain large travel range of electrostatic micro mirrors," in Pro-
ceedings of 4th international conference on modeling and simulation of microsystems,
Hilton Head, SC, 2001, p. 1921.
[18] R. Nadal-Guardia, A. Dehe, R. Aigner, and L. Castaner, \Current drive methods to
extend the range of travel of electrostatic microactuators beyond the voltage pull-in
point," Journal of Microelectromechanical Systems, vol. 11, no. 3, pp. 255{263, 2002.
[19] R. Legtenberg, J. Gilbert, S. Senturia, and M. Elwenspoek, \Electrostatic curved elec-
trode actuators," Microelectromechanical Systems, Journal of, vol. 6, no. 3, pp. 257
{265, Sep. 1997.
[20] J. Zhe, X. Wu, J. Cheng, J. Wang, K. Farmer, L. Frechette, and V. Modi, \Analytic pull-
in study on non-deformable electrostatic micro actuators," in Technical Proceedings of
the 2002 International Conference on Modeling and Simulation of Microsystems, 2002.
[21] J. Seeger and B. Boser, \Negative capacitance for control of gap-closing electrostatic
actuators," in TRANSDUCERS, Solid-State Sensors, Actuators and Microsystems, 12th
International Conference on, 2003, vol. 1, Jun. 2003, pp. 484 { 487 vol.1.
[22] M. Lu and G. Fedder, \Position control of parallel-plate microactuators for probe-based
data storage," Journal of Microelectromechanical Systems, vol. 13, no. 5, pp. 759{769,
2004.
93
[23] E. Hung and S. Senturia, \Extending the travel range of analog-tuned electrostatic
actuators," Journal of Microelectromechanical Systems, vol. 8, no. 4, pp. 497 {505, Dec.
1999.
[24] R. Dean, C. Wilson, J. Brunsch, and J. Hung, \A synthetic voltage division controller
to extend the stable operating range of parallel plate actuators," in Control Applications
(CCA), 2011 IEEE International Conference on, Sep. 2011, pp. 1068 {1074.
[25] L. Mol, R. Wol enbuttel, E. Cretu, and L. Rocha, \Full-gap positioning of parallel-plate
electrostatic MEMS using on-o control," in Industrial Electronics, 2007. ISIE 2007.
IEEE International Symposium on, Jun. 2007, pp. 1464 {1468.
[26] M. Semiconductor, \Application note 746: High-side current-sense measurement:
Circuits and principles." [Online]. Available: http://pdfserv.maxim-ic.com/en/an/
AN746.pdf
[27] H. Forghani-zadeh and G. Rincon-Mora, \Current-sensing techniques for DC-DC con-
verters," in The 2002 45th Midwest Symposium on Circuits and Systems, 2002.
MWSCAS-2002, vol. 2, 2002, pp. II{577{II{580 vol.2.
[28] T. Regan, \Current sense circuit collection," Dec. 2005. [Online]. Available:
http://cds.linear.com/docs/Application%20Note/an105.pdf
[29] G. Ferrari, F. Gozzini, A. Molari, and M. Sampietro, \Transimpedance ampli er for
high sensitivity current measurements on nanodevices," IEEE Journal of Solid-State
Circuits, vol. 44, no. 5, pp. 1609{1616, 2009.
[30] C. Xiao, L. Zhao, T. Asada, W. Odendaal, and J. Van Wyk, \An overview of integrat-
able current sensor technologies," in Industry Applications Conference, 2003. 38th IAS
Annual Meeting. Conference Record of the, vol. 2, 2003, pp. 1251{1258 vol.2.
[31] P. A. Roos, M. Stephens, and C. E. Wieman, \Laser vibrometer based
on optical-feedback-induced frequency modulation of a single-mode laser diode,"
Applied Optics, vol. 35, no. 34, pp. 6754{6761, Dec. 1996. [Online]. Available:
http://ao.osa.org/abstract.cfm?URI=ao-35-34-6754
[32] A. Wisitsoraat, V. Patthanasetakul, T. Lomas, and A. Tuantranont, \Low
cost thin  lm based piezoresistive MEMS tactile sensor," Sensors and Actuators
A: Physical, vol. 139, no. 12, pp. 17{22, Sep. 2007. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0924424706006510
[33] T.-R. Hsu, MEMS & Microsystems: Design, Manufacture, and Nanoscale Engineering.
John Wiley & Sons, Mar. 2008.
[34] Z. Xiao, W. Peng, R. F. Wol enbuttel, and K. R. Farmer, \Micromachined
variable capacitors with wide tuning range," Sensors and Actuators A: Physical,
vol. 104, no. 3, pp. 299 { 305, 2003, <ce:title>Selected papers based
on contributions revised from the Technical Digest of the 2002 Solid-State
94
Sensors, Actuators and Microsystems workshop</ce:title>. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0924424703000487
[35] B. Mammano, \Current sensing solutions for power supply designers," in Unitrode Sem-
inar Notes SEM1200, 1999.
[36] S. J. Kim, \An evaluation system for mechanical and electrical characterization
of MEMS devices," Thesis, Auburn University, May 2009. [Online]. Available:
http://etd.auburn.edu/etd/handle/10415/1592
[37] R. C. Jaeger, Introduction to Microelectronic Fabrication: Volume 5 of Modular Series
on Solid State Devices, 2nd ed. Prentice Hall, Oct. 2001.
[38] R. Hull, Properties of Crystalline Silicon. IET, 1999.
[39] P. V. Zant, Microchip fabrication. McGraw-Hill Professional, 2004.
[40] S. Linder, H. Baltes, F. Gnaedinger, and E. Doering, \Photolithography in anisotrop-
ically etched grooves," in IEEE, The Ninth Annual International Workshop on Micro
Electro Mechanical Systems, 1996, MEMS ?96, Proceedings. An Investigation of Micro
Structures, Sensors, Actuators, Machines and Systems, 1996, pp. 38{43.
[41] N. Layadi, V. M. Donnelly, and J. T. C. Lee, \Cl[sub 2] plasma etching of si(100):
Nature of the chlorinated surface layer studied by angle-resolved x-ray photoelectron
spectroscopy," Journal of Applied Physics, vol. 81, no. 10, p. 6738, 1997. [Online].
Available: http://link.aip.org/link/JAPIAU/v81/i10/p6738/s1&Agg=doi
[42] D. M. Manos, Plasma etching. Academic Press, 1989.
[43] H. Zou, \Anisotropic si deep beam etching with pro le control using SF 6/O 2
plasma," Microsystem Technologies, vol. 10, no. 8, p. 603607, Nov. 2004. [Online].
Available: http://dx.doi.org/10.1007/s00542-003-0338-3
[44] J. Hopwood, \Review of inductively coupled plasmas for plasma processing," Plasma
Sources Science Technology, vol. 1, p. 109116, May 1992. [Online]. Available:
http://adsabs.harvard.edu/abs/1992PSST....1..109H
[45] F. Laermer and A. Schilp, \Method of anisotropically etch-
ing silicon," U.S. Patent 5 501 893, Mar., 1996. [Online]. Available:
http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=
PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=50&s1=
5501893.PN.&OS=PN/5501893&RS=PN/5501893
[46] F. Marty, L. Rousseau, B. Saadany, B. Mercier, O. Franais, Y. Mita, and
T. Bourouina, \Advanced etching of silicon based on deep reactive ion etching
for silicon high aspect ratio microstructures and three-dimensional micro- and
nanostructures," Microelectronics Journal, vol. 36, no. 7, p. 673677, Jul. 2005. [On-
line]. Available: http://www.sciencedirect.com/science/article/B6V44-4GBD718-1/2/
70c45f3f91afaa974b5a88fcfc274f32
95
[47] M. Wasilik and A. P. Pisano, \Low-frequency process for silicon-on-insulator
deep reactive ion etching," in Device and Process Technologies for MEMS and
Microelectronics II, J.-C. Chiao, L. Faraone, H. B. Harrison, and A. M. Shkel, Eds.,
vol. 4592. Adelaide, Australia: SPIE, Nov. 2001, p. 462472. [Online]. Available:
http://link.aip.org/link/?PSI/4592/462/1
[48] N. Roxhed, P. Griss, and G. Stemme, \A method for tapered deep reactive
ion etching using a modi ed bosch process," Journal of Micromechanics and
Microengineering, vol. 17, no. 5, p. 10871092, 2007. [Online]. Available: http:
//iopscience.iop.org/0960-1317/17/5/031
[49] A. A. Ayn, X. Zhang, and R. Khanna, \Ultra deep anisotropic silicon trenches using
deep reactive ion etching (DRIE)," in Proceedings of the 2000 Solid State Sensors and
Actuator Workshop, Hilton Head, SC, USA, Jun. 2000, p. 339342.
[50] M. J. Madou, Fundamentals of Microfabrication: The Science of Miniaturization, Sec-
ond Edition, 2nd ed. CRC Press, Mar. 2002.
[51] C. Stevens, C. Wilson, and R. Dean, \Micromachined snap-in resonators," in Inter-
national Microelectronics and Packaging Society Device Packaging Conference 2012,
Scottsdale/Fountain Hills, Arizona, 2012.
[52] C. Wilson and J. Hung, \A system simulation technique combining SPICE and
SIMULINK tools," in IECON 2010 - 36th Annual Conference on IEEE Industrial Elec-
tronics Society, 2010, pp. 41{46.
96
Appendices
97
Appendix A
PPA Traveller
I. RCA Clean
1. A Clean
(a) Acetone Bath 1 minute
(b) DI water Rinse 1 minute
(c) IPA bath 1 minute
(d) Di water Rinse 1 Minute
2. B Clean
(a) 5-1-1 mixture 5 : DI H2O 1 : 30% H2O2 1 : 29% NH4OH heated to 70 for 5
Minutes
(b) DI water Rinse 1 Minute
3. C Clean
(a) HF bath for 30 Seconds
(b) DI water Rinse 1 Minute
(c) Second DI water Rinse for 1 Minute
II. Spin Rinse Dryer
III. Dehydration Bake
IV. Photolithography
1. Deposit HMDS Vapor
98
(a) Temperature: Room Temperature
(b) Time: 5 Minutes
2. Spin Photoresist
(a) Photoresist:AZ4620
(b) Speed:2500RPM
(c) Ramp Speed: 450 R/sec
(d) Time: 45sec
(e) Expected Thickness: 7 m
3. Soft Bake
(a) Temp: 110 C
(b) Time:4 Minutes (Longer Bake Time to prevent Bubble Formation)
4. UV Exposure
(a) Channel 2 (275 Watts)
(b) Exposure Type: Hard
(c) Cycles: 3
(d) Exposure Time: 10 seconds with 20 seconds wait time
5. Pattern Development
(a) Developer: AZ400
(b) Mixture: 3 parts DI Water to 1 Part Developer
(c) Time: 2 Minutes with agitation
V. Inspect Pattern In 5 places to ensure complete development.
VI. Descum
1. Power: 300W
99
2. Time: 15 Seconds
VII. Backing wafer Application
1. Spin photoresist
(a) Photoresist:AZ5214
(b) Speed:1000RPM
(c) Ramp Speed: 300 R/sec
(d) Time: 9sec
(e) Expected Thickness: 3 m
2. Attach SOI wafer to backing wafer device layer side down
3. Soft Bake
(a) Temp:100 C
(b) Time:3 Minutes
VIII. DRIE
(a) Program:MORGNSOI
(b) 400 cycles or until backing wafer is exposed
IX. Backing Wafer and Photoresist Removal
(a) Acetone Bath
(b) Isoproponal Bath
(c) DI Water Bath
(d) Weak Piranha Clean
i. 4 parts H2SO4 1 part H2O2
ii. Let solution stand for 5 minutes to cool to 70 C
100
iii. time: 5 minutes
iv. 1 minute DI Water Bath
v. Isoproponal Bath
(e) Separate individual devices from wafer
X. Front Side Metallization
(a) Mount devices on Supporting wafer
(b) Argon Ion Abasement: 5 Minutes
(c) Metal Composition: 1000 ATi, 4000 ACu
XI. Backside Metallization
(a) remove devices from front side supporting wafer
(b)  ip devices and mount again on supporting wafer Supporting wafer
(c) Argon Ion Abasement: 5 Minutes
(d) Metal Composition: 1000 ATi, 4000 ACu
101
Appendix B
Electron Beam Metal Deposition
Table B.1: Electron Beam Metal Deposition Recipe
Ion Beam Parameters
Pressure 1 10 6 Torr
Ion Beam Duration 3 minutes
Ion Beam Cathode Current 6.46 Amps
Ion Beam Discharge Voltage 2.78 Volts
Ion Beam Current 205 Amps
Ion Acceleration Current 4 Amps
Neu emission Current 264
Electron Beam General Parameters
Discharge Voltage 55.1
Acceleration Voltage 102
Filament Current 5.12 Amps
Layer Index 3
102
Appendix C
Listing of Lab Equipment by Room#/Laboratory
a136 Room # 455-461 Microfabrication Laboratory
{ STI Semitool 200 Spin Rinse Dryer
{ Tencor Alpha-Step 200 Pro lometer
{ Imperial IV Ultra-Clean 100 Dehydration Oven
{ Blue M OTP-120 Dehydration Oven
{ Tousimis Samdry PVT-3D Critical Point Dryer
{ CHA-Industries Mark 50 E-beam/Sputter System
{ Thermco 4000 Horizontal Tube Oxidation Furnace
{ Prometrix SpectraMap FT750 and Prometrix SpectraMap SM200/e
{ Karl Suss MA6/BA6 Contact Mask Aligner
{ Nikon 203338 High Magni cation Microscope
{ Olympus SZ-PT Zoom Stereo Microscope
{ Matrix System One Oxygen Plasma Etcher Style 303
{ CEE Spin Coater
{ Hot Plate
{ YES 450-PB8-2P-CP Vacuum Bake Oven
{ YES 5 Dry Vacuum Bake Oven
{ STS Multiplex ICP Advanced Silicon Etcher
103
{ Dicing Machine
{ UV Flood Exposure System
{ Branson 5510 Ultrasonic Shaker
a136 Room# 466 MEMS Testing Lab Room#
{ Micro-manipulator Manual Probe Station
{ HP 4156A Precision Semiconductor Parameter Analyzer
a136 Room# 103 Laboratory for Electronics Assembly and Packaging
{ Sonix C mode Scanning Acoustic Microscope
{ Nikon Measurescope MM-11 with Quadra-Check 2000 Measurement System
a136 Room# 464 Advanced Packaging Print and Fire Laboratory
{ Buehler Automatic Grinder/Polisher
{ Buehler VibroMet Vibratory Polisher
{ Low Power Microscope
{ Digital Scale
104
Appendix D
Matlab Code used to Analyze Interferometer Data
clear all; clc;
M=dlmread('21meg.csv',',',2,0);
meg21=M; clear M
M=dlmread('31meg.csv',',',2,0);
meg31=M; clear M
M=dlmread('41meg.csv',',',2,0);
meg41=M; clear M
M=dlmread('51meg.csv',',',2,0);
meg51=M; clear M
M=dlmread('61meg.csv',',',2,0);
meg61=M; clear M
M=dlmread('71meg.csv',',',2,0);
meg71=M; clear M
M=dlmread('81meg.csv',',',2,0);
meg81=M; clear M
M=dlmread('91meg.csv',',',2,0);
meg91=M; clear M
%Simple Low Pass Filter
f=10000
fNorm = 70 / (f/2);
[b,a] = butter(2, fNorm, 'high');
%Apply Filter
fmeg21 = filter(b,a,meg21(:,2))*20;
fmeg31 = filter(b,a,meg31(:,2))*20;
fmeg41 = filter(b,a,meg41(:,2))*20;
105
fmeg51 = filter(b,a,meg51(:,2))*20;
fmeg61 = filter(b,a,meg61(:,2))*20;
fmeg71 = filter(b,a,meg71(:,2))*20;
fmeg81 = filter(b,a,meg81(:,2))*20;
fmeg91 = filter(b,a,meg91(:,2))*20;
%filter(a, [1 a 1], x);
fs = 10000; % Sample frequency (Hz)
t = 0:1/fs:10 1/fs; % 10 sec sample
% x = (1.3)*sin(2*pi*15*t) ... % 15 Hz component
% + (1.7)*sin(2*pi*40*(t 2)) ... % 40 Hz component
% + (2.5)*randn(size(t)); % Gaussian noise;
m = length(fmeg51); % Window length
n = pow2(nextpow2(m)); % Transform length
% DFT
f = (0:n 1)*(fs/n); % Frequency range
figure(1);subplot(2,2,1);
plot(meg21(:,1),fmeg21); title('21 MEGAOHM');xlim([0 0.05]);
ylim([ 2 2]); ylabel('distance(um)');xlabel('time(s)');
figure(2);subplot(2,2,1);
y = fft(fmeg21,n); power = y.*conj(y)/n;
[C,I] = max(power); % Power of the DFT
plot(f,power); xlim([0 500]);xlabel('Frequency (Hz)');
ylabel('Power'); title('21 MOhms');text(100,20,num2str(f(I)))
freqplot(1,:) =[21,f(I)];
figure(1);subplot(2,2,2)
%figure(2)
plot(meg31(:,1),fmeg31); title('31 MEGAOHM');xlim([0 0.05]);
106
ylim([ 2 2]);ylabel('distance(um)');xlabel('time(s)');
figure(2);subplot(2,2,2);
y = fft(fmeg31,n); power = y.*conj(y)/n;
[C,I] = max(power); % Power of the DFT
plot(f,power); xlim([0 500]);xlabel('Frequency (Hz)'); ylabel('Power');
title('31 MOhms');text(100,40,num2str(f(I)))
freqplot(2,:) =[31,f(I)];
figure(1);subplot(2,2,3)
%figure(3)
plot(meg41(:,1),fmeg41); title('41 MEGAOHM');xlim([0 0.05]); ylim([ 2 2]);
ylabel('distance(um)');xlabel('time(s)');
figure(2);subplot(2,2,3);
y = fft(fmeg41,n); power = y.*conj(y)/n;
[C,I] = max(power); % Power of the DFT
plot(f,power); xlim([0 500]);xlabel('Frequency (Hz)');
ylabel('Power'); title('41 MOhms');text(100,100,num2str(f(I)))
freqplot(3,:) =[41,f(I)];
figure(1);subplot(2,2,4)
%figure(4)
plot(meg51(:,1),fmeg51); title('51 MEGAOHM');xlim([0 0.05]);
ylim([ 2 2]);ylabel('distance(um)');xlabel('time(s)');
figure(2);subplot(2,2,4);
y = fft(fmeg51,n); power = y.*conj(y)/n;
[C,I] = max(power); % Power of the DFT
plot(f,power); xlim([0 500]);xlabel('Frequency (Hz)');
ylabel('Power'); title('51 MOhms');text(100,100,num2str(f(I)))
freqplot(4,:) =[51,f(I)];
figure(3);subplot(2,2,1)
%figure(5)
107
plot(meg61(:,1),fmeg61); title('61 MEGAOHM');xlim([0 0.05]);
ylim([ 2 2]);ylabel('distance(um)');xlabel('time(s)');
figure(4);subplot(2,2,1);
y = fft(fmeg61,n); power = y.*conj(y)/n;
[C,I] = max(power); % Power of the DFT
plot(f,power); xlim([0 500]);xlabel('Frequency (Hz)');
ylabel('Power'); title('61 MOhms');text(300,100,num2str(f(I)))
freqplot(5,:) =[61,f(I)];
figure(3);subplot(2,2,2)
%figure(6)
plot(meg71(:,1),fmeg71); title('71 MEGAOHM');xlim([0 0.05]);
ylim([ 2 2]);ylabel('distance(um)');xlabel('time(s)');
figure(4);subplot(2,2,2);
y = fft(fmeg71,n); power = y.*conj(y)/n;
[C,I] = max(power); % Power of the DFT
plot(f,power); xlim([0 500]);xlabel('Frequency (Hz)');
ylabel('Power'); title('71 MOhms');text(300,100,num2str(f(I)))
freqplot(6,:) =[71,f(I)];
figure(3);subplot(2,2,3)
%figure(7)
plot(meg81(:,1),fmeg81); title('81 MEGAOHM');xlim([0 0.05]);
ylim([ 2 2]);ylabel('distance(um)');xlabel('time(s)');
figure(4);subplot(2,2,3);
y = fft(fmeg81,n); power = y.*conj(y)/n;
[C,I] = max(power); % Power of the DFT
plot(f,power); xlim([0 500]);xlabel('Frequency (Hz)');
ylabel('Power'); title('81 MOhms');text(300,100,num2str(f(I)))
freqplot(7,:) =[81,f(I)];
figure(5);
plot(freqplot(:,1),freqplot(:,2));
108
title('Frequency vs Resistance');xlabel('Resistance (MOhms)');
ylabel('Frequency(Hz)');
xlim([21 81])
109
Appendix E
Schematic Drawing of PPA Board Mount Fixture
Figure E.1: Bottom of  xture Design.
110
Figure E.2: Top of  xture Design.
111