1 11

Chapter 2 Acce lerometer Theory & Design

2.1

Introduction
An accelerometer is a sensor that measures the physical acceleration

experienced by an object due to inertial excitation. In aerospace applications

forces or due to mechanic al

accelerometers are used along with an

gyroscopes for navigation guidance and flight control. Conceptually, accelerometer behaves as a damped acceleration, mass on a spring.

When the

accelerometer experiences

the mass is displaced and the

displacement is then measured to give the acceleration [17] . In these devices, piezoelectri c, piezoresistive and capacitive into an

techniques are commonly used to conve rt the mechanical motion

electrical signal. Piezoelectric accelerom eters rely on piezoceramics (e.g. lead zirconate titanat e) or single crystals (e.g. quartz , tourmaline ). They are unmatched weight in terms of their upper frequency range, low packaged temperature shock range. Piezoresistive applications. accelerometers are accelerometers

and high in

preferred

high

Capacitive

performance is superior in low frequency range and they can be operated in servo mode to achieve high stability and linearity.

12

Modern

accelerometers

are

often

sma ll

micro

electro-

mechanical systems

(MEMS), consisting of litt le more than a

cant ilever beam with a proof-mass (also known as seismic-mass) realized in single crystal silicon using surface micromachining or bulk microm achining processes.

2.2

Working principle of accelerometer

Casing

Fig. 2.1 Schematic of an accelerometer The principle of working of an accelerometer can be explained by a simple mass (m) attached to a spring of stiffness (k) that in turn is attached to a casing, as illustrated in fig 2.1. The mass used in

accelerometers is often called the seism ic-mass or proof-mass. In most cases the system also includes a dashpot to provide a desirable damping effect.

13

The dashpot with damping coeff icient (c) is normally attached to the mass in parallel with the spring. When the spring mass system is subjected to linear acceleration, a force equal to mass times acceleration acts on the proo f-mass, causing it to deflect. This deflection is sensed by a suitable means and converted into an equivalent electrical signal. Some form of damping is required, otherwise the system would not stabilize

quickly under applied acceleration. To derive the motion equation of the system Newton second law is s

used, where all real forces acting on the proo f-mass are equal to the inertia force on the proo f-mass. Accord ingly a dynamic problem can be treated as a problem of static equilibrium be obtained by direct formulation and the equation of motion can This

of the equations of equilibrium.

damped mas s-spring system with applied force constitutes second order mechanical system. From the stationary direction is,

a classical

observer s point of view, the sum of all forces in the z

= +

+ =

14

15

(2.1)

Where m = mass of the proo f-mass x = relative movement of the proo f-mass with respect to frame c = damping coefficient k = spring constant F = force applied The equation of motion is a second order linear differential equation with constant coef ficients. The general sol ution x (t) is the sum of the

complementary (2.2) =

function XC (t) and the particular integral Xp (t) [18]. +

The complementary

function satisfies the homogeneous equation (2.3)

+ to = 0 + is The solution

(2.4) Substituting (2.4) in (2.3)

(ms2 + cs + k) C e st = 0

16

As cannot be zero for all values of t, then ( 2 + called as the auxiliary or characteristic

+ )=0

equation of the system. The solution to

this equation for values of S is 1,2 =

2 1

( 4 )

(2.5)

From the above equation 2.5, the foll owing useful formulae are derived k m

(2.6)

c/m = 2 n

(2.7) (2.8)

= c/2
Where

km

n = undamped resonance frequency k = spring constant

m = mass of proo f-mass c = damping coefficient

= damping factor
Steady state performance In the steady state condition, amplitude a and frequency that is, with excitation accelerat ion

, the amplitude of the response is consta nt . Thus for

and is a function

of exc itation amplit ude and frequency

static response

=0, the deflection

17

amplitude

X = 0 = F/k. X = ma / k Here the sensitivity S = X / a = m/k Dynamic performance For the dynamic performance transform of eqn (2.1) it is easier to consider the Laplace S of an accelerometer is defined by, (2.10) (2.9)

x(s) = a(s)

1 s +c s k + m m
2

(2.11)

It can be seen by comparing eqn (2.6) and (2.10) that the bandwidth

of

an accelerometer sensing element has to be traded off with its sensitivity since S 1/ n2 (this trade off can be partly overcome by applying

feedback , i.e closed loop scheme). The sensor response is determin ed by system. A damping damping present in the in high

factor ( ) between 0.6 to 1.2 results

response time, fast settling time, good bandwidth and linearity.

2.3 Specifications of the accelerometer

The accelerometer that is to be designed shall have a measuring range of 30 g with a resolution of 50 milli g i.e a dynamic range of 600. The total non-linearity from the sensor element, electronics and

from other sources shall not be more than 1% of full scale output (FSO). The sensor shall have a bandwidth ( 3dB) of >100 Hz and the cross-axis sensitivity shall be limited to a maximum and hysteresis of 1% of FSO. The sensor s as 0.15% of FSO

bias stability

values are specified

each. Finally the sensor shall have a response time of less than 1msec and it shall perform over a temperature range of -20 to 80C.

2.4 Configuration of accelerometer

Various aspects are taken in to consideration before finalizing the configuration available of the accelerometer. Special attention conditioning is paid electronic to the circuit,

fabrication

processes, signal

material selection, electrical routing and packaging. The configuration of the accelerometer is as shown in fig 2.2. (i) The sensor is configured to have a three wafer (glass-silicon -glass) configuration. (ii) Differential capacitance transduction method is selected as it offers

the advantages of low temperature sensi tivity, higher transduction

efficiency and the method can be readily adopted for closed loop operation. (iii) Glass wafers are used for the top and bottom plates and a thin film of aluminum material is deposited on the inner side of glass

wafers using E-beam metal evaporation process. The central wafer consists of the active proo f-mass which moves as a function of the applied acceleration thereby causing change in capacitance. (iv) The proo f-mass is supported on all four sides by L-shaped beams. The proof -mass exhibits piston like mov ement and remain parallel to electrodes at all accelerations. Also any geometrical change in the beam length due to temperature variation, limits the proo f-

mass to in-plane rotation only and it does not experience any out of plane bending. This configuration reduces the overall sensor

chip size thereby improving the per wafer yield and also reduces the non -linearity associated with cantilever type support

structures. (v) The mechanical central support for the proo f-mass is provided at the

plane of the proo f-mass. Posi tioning of the beams at

central plane of proo f-mass will reduce the cros s-axis sensitivity of the sensor. (vi) The proo f-mass to electrode gap is selected as 22 microns, this eliminates the requirement of complicated device level vacuum

sealing and also need for perforations simplifying the process. (vii) Bulk micr o-machining process using

on the proo f-mass thus

KoH

is

considered

for

realizing the micro structures. (viii) Modular concept is used for realizing the final device. MEMS chip and the signal conditioning electronics are realized separately and packaged on a signal platform.

Top electrode

Top glass wafer

Support beams

Si wafer

Proof -mass

Bottom Glass wafer

Bottom Electrode

Fig 2.2 Accelerometer configuration

2.5

Mater ial selection

Single crystal silicon (100) mater ial is selected for accelerometer

structure.

Silicon is almost an ideal str uctural material, it has about the The melting

same youn gs modulus as steel but is as light as aluminum.

point of silicon is 140 0 0 C and the ther mal expansion coefficient is much

less than

steel which

makes

it dim ensionally

stable

even at high

temperatures. Silicon additional

wafers are extre mely flat,

can accept coatings & Silicon exhibits features can be

thi n-film layers for building and

microstructures. geometrical

no mechanic al hysteresis realized using standard

precise

photolithogr aphy and

etching

techniques.

Electrically conductive proof-mass. Similarly

silicon with resist ivity 0.1 -cm is selected for the Pyrex 7740 glass is chosen for top and bottom

wafers to reduce stray capacitance and to provide required sealing. The glass wafers are bonded to silicon wafer using anodic bonding process. Electrodes and electrical contact pads are realized by depositing sub micron thickness Aluminum coating, using E-beam evaporation proces s. The material properties of silicon and Pyrex glass are shown in Table 2.1 Material Property Silicon 7 1.69 0.28 2.5 2.3 Pyrex glass 0.5-0.7 400 0.17 0.5 2.225

y (yield strength) 109 N/m 2

E (Young s modulus)1 011 N/ m2

(Poisson ratio) s expansion coefficient) 10-6 m t/m t C g/c m3

(thermal (density)

Table 2.1: Material properties of silicon and Pyrex glass.

2.6 Analytical design The following assumptions are made to begin the design work (i) The proo f-mass size and spring stiffness are selected in such a way that there shall be a capacitance change of around 1fF for 50 milli g (minimum signal resolution). that This limitation comes from the

capacitance

can be handled comfortably

by the

electronics scheme. (ii) For calculating the proo f-mass to electrode gap and dampi ng structure is consi dered as working This is due to fabrication und er faci lity

aspects, the micro

ambient pressure conditions.

limitation in chip level sealing under vacuum.

The structural

parameters

of the

proo f-mass

are optimized

to

achieve the required sensitivity and bandwidth. The proo f-mass length, width, thickness are represented as l1, b1, h1. The L- Beam dimensions

are represented by l2, l3, b2, and h2 respectively, which are shown in fig.2.3. The optimized device dimension are given below Proof-mass size (l1 X b1 X h1) Length of beam (l2) Length of beam (l3) : : : 2500 X 2500X 300 m 3200 m 640 m

Beam Width (b2) Beam thickness (h2) Air gap (i)

: : :

150 m 55 m 22 m

Area of the proo f-mass (A) = l1 X b1 = 6.25 x10-6 m2

(ii) Mass of proo f-mass (m) = V. (Density of the silicon

= (A x h1) x = 4.313 x10 -6 Kg

= 2300 kg/m 3, V = Volume of proof -mass)

Mass of the beams (mb) = 4(l2+l3) x b2 x h2 x

= 2.91456 x10 -7 Kg.

All dimensions are in microns

Fig 2.3 Accelerometer geometrical details

(iii)

Maximum force on the proof-mass

Let F be the force acting on the proo f-mass, due to 1 g acceleration, which is given by the relation. F=m xa Where m is the m ass of the proo f-mass and a is the acceleration F = 4.313 e-6 X 9. 81 X 1 = 0.0423 mN This force is shared by all the beams equally. Force acting on each beam W = F/4 = 0.01057mN (vi) Moment of inertia of beams

b2 h2 I=( ) = 2.07 e-18 m4 12

(v)

Deflection of the beam The L beam is rigidly fixed to the frame on one side and other side is

attached to the Proof-mass. This can be considered as two cantilever springs in series, rigidly fixed on one side and guided on the other side. The numerical equation for the above is given as
Wl 3 12EI

Deflection =

Where E = Modulus of elasticity for silic on E=1.69 x 10 I= moment of inertia of beam.

11

N/ m2

For beam having length l2 1 =

Wl 2
3

12EI Wl 3 12EI
3

For beam having length l3 2 =

W l3 +l3

= 1+ 2 =

12EI

= 8.3166 10 8 m /g
At 30g the maximum deflection =2.49 x10-6 m (vi) Bending stress in the beam The formula for determining the bend ing stress in a beam

under simple bending is given by

M y I

=
b

Where b = Bending stress in the beam M = Bending moment acting on the beam y = the perpendicular distance to the neutral axis. At 30 g M = W l2 30 M = 0.01057 10 3200 10
3 6

30

M = 1.014 10 Nm

For maximum bending stress

y=

h2

= 13.4MPa

(vii)

Factor of safety Since silicon is a brittle mater ial, the UTS value is taken for

calculating the factor of safety design margin over the theoretical design capacity. Calculating factor of safety at 30g FOS = ultimate strength / maximum stress = 7000 / 13.4 = 522

(viii) Natural frequency The natural frequ ency fn of the spring mass system is given by fn = 1 2

k/ m
of the spring

(2.12)

Where k is the stiffness

= 508.62 N/m

From equation 2.12

fn = 1728Hz.

2.7 Electr ical design

Capacitance is the ability of a body to hold an electric charge. The capacitance between two parallel plate conductors the geometry of the conductor can be calculated if properties of the

and the dielectric

insulator between the conductors are known. (i) Nominal Capacitance (Co) The Nominal Capacitance of a parallel plate capacitor with overlapping area `a separated by a distance `d is equal to

=
Where

(2.13 )

= Relative permittivity =1) 0 =

of the diel ectric medium

(for Air

Permittivity of free space = 8.85 x 10 -12 F/m

Co= 2.514 pF (ii) Accelerometer Sensitivity The initial gap between the proo f-mass and the electrode is 22m. Let

C1 and C2 are capacitances between top electrode and proo f-mass and bottom electrode and proo f-mass respe ctively under the application 1 g. Since the system is a differential gravitation decreases. of

capacitor, under the influence of

force, as one side capacit ance increases the other side it

0 r a d+ C1 = 2.5047 pF a C2 = 0 r d C 2 = 2.5237 pF C1 = Change in capacitance is therefore calculated as, C = C2 C1 = 19 fF Sensitivity = C / applied acceleration = 19 fF/g (2.14)

2.8 FEM modelling and simul ation

The design and development of a MEMS device is highly challenging task involving simulation of micro structure behavior under coupled

environmental

load conditions.

FEM is an essential tool for MEMS design of the static and dynamic behavior of

and it provides accur ate stimulation complex structures mrket for MEMS

at micro scale. Several FEM tools are available in the modelling and simulation. However in this case,

Coventorware

software

is used for accelerometer

mode lling and FEM

analysis, Intellisuite

software is used for wet etching process simulation

and SABER for system level simulation.

In Coven torware solid models are built from 2-D layout tool with process information and meshes are created on solid models in the

preprocessor.

Coventorware

analyzer

module

uses various numerical

approaches such as 3-D FEM of MEMMECH, 3-D BEM of MEMELECTRO modules physics. Since our accelerometer structure an eight node manhattan has regular shaped beams and plate s, for solving the partial differ ential equations of mathematical

brick element is used for meshing the mode l.

The element has an orthogonal geometry i.e all model faces are planar and join at right angles. The model is meshed with uniform mesh densi ty through the model to reduce errors.

Fig 2.4 Meshed model of accelerometer structure

2.9 Mesh convergence study

The accuracy of a discrete solution of a partial differential equation depends on the density of the mesh. Therefore, the accuracy of an analysis can only be judged by the comparison of results on meshes of increasing degrees of freedom i.e. study ing mesh convergence. To study

mesh

convergence

multiple

mesh

models

of different densities increased

are by

created.

The mesh density

on the beams is gradually

reducing the element size while keeping the aspect ratio same. The proo fmass is meshed with a mesh den sity of 600 throughout the study. The

results of the proo f-mass displacement (microns) with 1 g acceleration as a function of mesh density of the beams are plotted in fig 2.5.

Fig 2.5 Mesh convergence simulation result From fig 2.5, it can be seen that for a mesh density of 800 elements and more on the beams, the change in varia tion of proo f-mass displacement is less than 1%, hence convergence is achieved. All further analysis is

done considering the optimized mesh density. Elements on the L-beam

joining

face to the frame are comple tely constrained

and all other

elements have 6 degrees of freedom.

2.10 FEM simulation results

The model is subjected to accele ration load upto 30 g and the response of the sensor for displacement, change in capacitance, bending stresses and cros s-axis sensitivity are studied.

2.10.1 Acceleration vs. displacement

The structure is subjected to 0 g to 30 g acceleration in +Z The

direction, in steps of 3 g to analyze the proo f-mass displacement. analysis results are shown in table 2.2. The maximum the proof-mass in Z-direction is 3.48 microns.

displacement of

Acceleration 0 3 6 9 12 15 18 21 24 27 30 (g) Displacement 0 0.348 0.696 1.044 1.392 1.74 2.088 2.436 2.784 3.132 3.48 (m)

Table 2.2 Acceleration Vs displaceme nt of proof-mass

As shown in table 2.2 & fig 2.6, the proof-mass displacement is linear with respect to the applied acceleration.

Fig 2.6 Applied acceleration Vs Proof- mass displacement

2.10.2 Acceleration vs. capacitance

Coupled electr o-mechanical simulat ion is done for the accelerometer by applying a voltage of 5V on the top and bottom electrodes and

grounding the proo f-mass. The accelero meter is subjected to accelerati on varying from 0 g to 30 g in both + Z and

direction.

The

capacitances across top electrode and proof-mass, bottom electrode and proof-mass are obtained. The nominal capacitance of the accelerometer is 2.514 pF. The resultant change in capacitance value in steps of 3 g is given in Table 2.3 and the sensitivity of the accelerometer is 27.4 fF/g.

ACCELERATION (g) 0 3 6 9 12 15 18 21 24 27 30

CHANGE IN CAP. FOR POSITIVE ACC. (pF) 0 0.080249 0.16055 0.24119 0.321998 0.403377 0.485374 0.568108 0.651715 0.73633 0.822097

CHANGE IN CAP. FOR NEGATIVE ACC. (pF) 0 -0.08011 -0.16041 -0.24095 -0.32185 -0.40324 -0.48524 -0.56797 -0.65158 -0.73619 -0.82196

Table 2.3 Acceleration Vs change in capacitance

CHANGE IN CAPACITANCE

1 0.8 0.6 0.4 (pF) 0.2 0 -0.2 0 -0.4 -0.6 -0.8 -1

12

15

18

21

24

27

30

ACCELRATION (g)

Positive Acceleration Negative Acceleration

Fig 2.7 Change in capacitance with acceleration

From the change in capacitance equation 2.14, it can be seen that the change in capacitance is nonlinear with applied acceleration. The

above data is analyzed for non-linearity using least square method curve fitting technique. The equation of the best fit line is y = 27.3684x +

(-3.5698) and the maximum non-linearity is 0.764% which is well within the design requirement.

2.10.3 Bending stresses in the beams

Maximum bending stress occurs in L-beams due to the applied

acceleration in the Z directio n. The max imum bending stress is occurring at the point where the L-beam is anchor ed to the frame and at the sharp corner of the L beam, fig 2.8 gives the stress distribution The magnitude of maximum in the model. is

von Mises stress at 30 g acceleration

11 MPa, which is much less than the UTS value of silicon which is 7000 MPa.

Fig 2.8 Stress distribution in the accelerometer at 30 g

2.10 .4 Modal analysis

Modal analysis is done to obta in the first and second mode

frequencies of the accelerometer. The first mode resonant frequency of the structure is, out of plane vibration along Z axis which is at 1470 Hz. The thumb rule is that the first mode resonant frequency of the structure shall be at least three times the requir ed bandwidth. The second mode

frequency is, in X-Y plane at 3323 Hz. fig. 2.9 shows the modes of vibration. maximum Both the frequency modes of operation are far away from the operational bandwidth of 100Hz, hence no interference on

accelerometer functioning is expected.

Mode 1:1470Hz

Mode 2: 3323Hz

Fig 2.9 Modal analysis results of the accelerometer

structure

2.11 Cross - axis sensitivity

The accelerometer shall sense acceleration along the Z-axis only and it shall be insensitive to the applied acceleration along other two

axes. The cros s-axis sensitivity specifica tion of the accelerometer is < 1% of FSO. As already presented earlier the proo f-mass thickness is 300 m and the supporting L-beam thickness is 55 m. From fabrication process point of view, it is very easy to realize the beams aligned to the top or bottom plane of the proo f-mass by etching the beams to the required thickness from one side. The accelerome ter is symmetrical about x and y axes. The meshed accelerometer model is subject ed to cross acceleration along X-axis for three diff erent beam pos itions along the thickness of the Proof-mass. The results are shown in fig 2.10. The cros s-axis sensitivity increases as function of the distance from central plane of the proo f-mass. For the beam position, at the central plane of the proof-mass, the cross-axis

sensitivity obtained is 0.01% of FSO, hence the position of the support beams is finalized at the center of the proo f-mass.

Fig 2.10 Effect of beam position on the cross axis sensitivity

2.12

System level simulation of the sensor

Traditionally MEMS devices have been simulated using field

solvers, such as finite element method method differential (BEM) analysis tools. These from the

(FEM) and boundar y-eleme nt tools solve complex parti al of the

equations

derived

detailed

description

physical design, but those equations are far from simple and take a lot of time to solve.

However, in system level or hig h-level simulation method, simulations are done based on the behavior of a device as expressed by reduce dorder equations [19]. It simulates the overall behavior of complete model instead of the interactive behavior of many finite elements that comprise the model. The complex mathematical description used with hig h-level

models leads to a much smaller number of degrees of freedom that reduces the number of computations in much faster simulation physical analogy performed by the solver, resulting and the

runs. The higher level of abstraction trans lation, permits rotation,

between

electroni cs, of a

thermodynamics,

and other

entities

the interconnection

number of physical domains. The system level simulation module tool offered by Coventorware design. architect

called SABER is used in the present model libra ries. the behavior

SABER uses code that to

behavioral expresses

The models include underlying indivi dual

of the

components subjected

electrical, mechanical or other domain stimuli.

2.12.1 System level modelling

The system level accelerometer model (fig 2.11) is built using

standard library components and consists of four support beam elements and a rectangular proo f-mass plate. Two electrode elements at the top

and bottom of the proo f-mass allow elect rostatic excitation and capacitive detection due to vibrat ion.

Fig.2.11 System level behavior model of the accelerometer

2.12.2 Small signal AC analysis

To determine the resonant behavior of the accelerometer structure a small signal AC analysis is performed. The model is excited over a

frequency response

range of 1 Hz to 10 kHz. Fig 2.12 shows the maximum and phase angle details as a function of the appl ied

acceleration.

It can be seen that the natural frequency of the system is

1432 Hz where the phase angle is 90 deg.

Fig 2.12 Small signal AC analysis results

2.12.3 Pull in analysis

During anodic bonding process of acceleromete r, for bonding

silicon wafers with glass wafers, upto 1000 VDC is applied between the wafers. At such high voltages due to electrostatic attraction the proo f-

mass may come in complete area cont act with electrodes on the glass wafers and may not revert back to norm al position due to stiction arising out of large area of contact. Pull in analysis is done to know the safe working voltage that can be applied between the proo f-mass and

electrodes.

The analysis

is done by grounding

the proo f-mass and

varying bottom electrode voltage from 25 to 160 VDC.

Fig 2.13 Pull in analysis result As shown in fig 2.13 pull in anal ysis result, the electrostatic force overcomes spring force of the structure at pull in voltage of 147.97 V.

Hence during fabrication suitable bumps are provided on the proo f-mass to overcome stiction problem.

2.12.4 Transient analysis

Transient analysis is done to estimate the device sensitivity and

response time to the applied acceleration input. The aerospace senso rs need to have quick response and fast settling times. 1 g is applied on the structure An acceleration of after

in 0.1 micr o-sec and withdrawn

1m-sec in 1micr o-sec. Fig 2.14 shows the result of transient

analysis.

The result shows that the displacement and capacita nce change are very closely following the input 1 g signal. The system has response time of

less than 1 m-sec which is well within the design requirements sec.

of 1m-

Fig 2.14 Transient analysis result with 1 g input signal

2.13

Dynamic analysis
accelerometer design the sensing proo f-mass is

In the current

capped on both sides using glass wafers at atmospheric pressure. As the proof-mass moves towards the statio nary electrodes, pressure between

the two layers increases developing damping forces. This pressure drives out the entrapped air between the parallel plates. On the contrary, when the proof-mass is moving away from the electrode the pressure in the gap is reduced causing surrounding air to flow into the gap. In both cases the

force on the proo f-mass caused by buil t-up pressure is always again st the movement of the plate. The work done by the plate is consumed by the viscous flow of the air and transfor med into heat. In other words, the air film acts as a damper and this type of dampi ng is called squeeze film damping. The damping phenomenon is shown in fig.2.1 5. In the past, considerable research was done in cha racterizing squeeze film damping behavior in MEMS structures circuit model of using squeeze [20-29]. Veijola et al developed equivalent film damping applicable to MEMS

accelerometers

R-L elements

[23]. Sadd et al considered the

incompressible effects of gas at low squeeze numbers [27].

Movement of Proofmass

Air Moving Inside Proof-mass

Air Moving Outside

Fig.2.15 Damping phenomenon

2.13.1 Squeeze film damping

Starr proposed the following condit ions which are to be satisfied to obtain satisfactory behavior of damping in micro accelerometers, [20].

1. The behavior of squeeze film is governed by both viscous and inertial effects. For small geometries inertial effects are neglected. The specific condition of validity is:

Where

d 2 / << 1.0

(2.15)

= frequency of oscillation of proo f-mass d = air gap = density of air 1.16 e-18 kg /m 3 = dynamic viscosity of the damping media i.e. air 1.86 x 10 -11 N-sec/m 1470 2 2 22 1.16 10
11 18

= 0.2

1.86 10

2. The air flow is assumed as continuu m, slip flow condition at the boundaries may reduce the effectiveness of damping. To overcome this problem the film thic kness shall be > (100 times mean free path of air) [20] [29]. At room temperatur e, the mean free

path of air is 0.065 microns. Hence, the film thickness shall 6.5 microns.

be >

3. Squeeze no. which is dimensionle ss number, is a measure of compression of fluid in the gap. If is low (close to zero implying

low speed), the gas film follows nearly incompressible viscous flow. If is large ( > 0.3), the film essentially acts as an incompressible air spring and exhibits little energy dissipation. 2 12 b 2
2 h 0 Pa

= 0.26

(2.16)

Where

= frequency of oscillation of proo f-mass b = half width of the plate h0 = nominal film thickness Pa = ambient gas pressure

4. For squeeze film damping conditio n, the damping coefficient (C) is given by
w 3 2f ( )n l l C= 3 d0

(2.17)

Where

f(

w ) = Shape function l

w, l = width and length of moving plate n = Natural Frequency of the structure do = Initial gap between moving and stationary plate C = 0.05545 and

Cc (critical damping coefficient) = 2 m n = 0.1033

=
C = 0.0554 5 = 0.54 0.1033 Cc

Where is damping factor. As accelerometer the damping factor is less For than one, designed

is unde r-damped

system.

unde r-damped system

response time is given by T(r) =

1 4

= 4.25 10-5 sec

2.13.2 Simulat ion results

To study the dynamic behavior of the acceleromete r, a squeeze film damper model is made with moving proof-mass, electrodes, air gap and air column as shown in fig 2.16. The air is assumed to be moving in and out of the gaps at the four side edges of the proo f-mass .

Fig. 2.16 Squeeze film damping model

2.13.3 Squeeze film damping coefficient var iation with frequency

To realize frequency operational a sensor the with required linearity forces shall over the operational be linear within the

range,

damping

frequency

range and the spring

force shall

be low and

damping coefficient shall be constant. Fig 2.17 shows the variation of damping force and spring force over large frequency range. At low

frequencies, air can escape with little resistance and the spring force is small. At high frequencie s, the air is held captive by its own inertia, there is not enough time for the air to move out of the way as the structure oscillates. force. The air gets compressed, resulting in an increased spring to the

The damping is caused by viscous forces and proportional

velocity of the oscillating structure.

But, if the gas is compressed and

does not move much, then the damping force will be lower. This explains why the damping gets smaller as the frequency increases above about 1 MHz.

Fig 2.17 Damping analysis results over large frequency range

Fig 2.18 Damping analysis results over small frequency range

Fig.2.19 Damping coefficient variati on with frequency However, as shown in fig 2.18, within the operational range of accelerometer damping force is proportional operation and spring force is negligible. frequency

to the frequency of

It can be seen from fig 2.19,

within the operational range of frequen cy damp ing co-efficient variation is negligible.

2.13.4 Harmo nic analysis

Harmonic analysis is performed using MEMMECH module of the

coventorware software to find the linearity of the accelerometer response over the required bandwidth of operation under applied acceleration. The accelerometer is subjected to a harm onic load of 30g at frequen cy coeffi cient for the

ranging from 1 Hz to 100Hz. The modal-damping analysis is 0.055.

Displacement (m)

Fig.2.20. Harmonic analysis result

Fig 2.20 gives the displacement response of accelerometer at 30g input up to requir ed bandwidth frequency of 100 Hz. The frequency sensitivity

of displacement within the operation bandwidth is negligible.

2.14 Interface electronics

The MEMS Accelerometer is electri cally equivalent to a differential parallel plate capacitor structure with the capacitance change occurring

due to the change in the gap between the parallel plates. The switche dcapacitor charge integration method has been widely used in MEMS [30-38]. The interface circuit sha ll

capacitive sensor interface circuits have the following features

Interface with the sensor with given nominal capacitance of 2.5pF and also nullify any capacitance offset that is present. The required output swing of the circuit is 0 to 5V with a DC bias output voltage at 0g The circuit should be able to handle a total capacitance change of 0.6pF The circuit should be able to resolve a minimum capacitance

change of 1fF The bandwidth should be more than 100Hz

2.14.1 Capacitance to voltage conversion scheme

The interface circuit for converting the variation in capacitance into voltage is implemented using a standard capacitance to voltage

conversion ASIC, MS3110 from Irvine Sensors. MS3110 [39] is a general purpo se, ultra noise CMOS IC that

requires only a single +5V DC supply and some decoupling components. The ASIC is capable of sensing capacitance changes down to 4aF/rtHz. It

can interface with either a differential capacitor pair or a single capacitive sensor. The salient features of the IC that enables its suitability for

integration with the accelerometer chips are listed below: Capacitance resolution: 4aF/rtHz Differential variable capacitance sensor interfacing Gain and DC offset trim Programmable bandwidth adjustment On chip EEPROM for storage of program coefficients

The ASIC functional block diagram is shown in fig 2.21 .

Fig 2.21 Block diagram of MS3110 ASIC

The fundamental charge amplifier

operation of the conversion scheme is, that of by the tran s-conductance operational The sense

amplification interfacing

the sensor capac itance bridge.

nodes of the capacitance bridge are fed by a square wave signal generated internally and whose amplit ude oscillates between zero

g bias voltage and 0V. The amplified signal is then low pass filtered in the next section. The bandwidth of the LPF is programma ble.

 The last block is a gain amplifier with the provision of gain and offset programmability. The various coefficients for capacitance

bridge balancing, bandwidth, in an EEPROM.

sensitivity,

gain and offset are stor ed

The various timing signals for EEPROM read, write, and square wave are generated internally.

The ASIC senses the change in capacitance between two capacitors and provides a voltage output proporti onal to the change. The transfer function of the ASIC is given as: Vout = GAIN * V2P25 * 1.14 * (CS2T-CS1 T)/CF + VREF Where Vout is the output voltage Gain = 2 or 4V/V V2P25 = 2.25 VDC CS2T = CS2IN + CS2 CS1T = CS1IN + CS1

CF is selected to obtain the required sensitivity

VREF can be set to 0.5V or 2.25V DC for CS = CS2T CS1T =0

The pin diagram of the ASIC and pin description are given in Fig 2.22 and Table 2.4 respectivel y.

Fig 2.22 MS3110 ASIC pin diagram Pin No 1 2 3 CHPRST V2P25 TESTSEL IC reset. Internally pulled up. 2.25VDC reference Enables the user to bypass on-chip EEPROM and program the IC directly 4 5 6 7 CS2IN CSCOM CS1IN SDATA Capacitor sense input 2 Capacitor sense common Capacitor sense input 1 Serial Data input, used for serial data input the EEPROM or the IC Name Description

port for programming

registers directly. 8 9 10 SCLK NC HV16 Serial clock input No Connection 16VDC input port, tied to 16V when wr iting to EEPROM, grounded otherwise 11 12 13 14 15 16 WRT NC -V VO +V NC Write select No connection Negative voltage rail, usually 0V IC Voltage output Positive voltage rail, usually 5V No connection Table 2.4 MS 3110 pin diagram descr iption

2.14.2 Features of the inter face electronics

The circuit gives a DC bias of 2.25V at 0g, which is also equal to the reference voltage of the ASIC The sensitivity of the sensor can be programmed from 20mV/g to 66mV /g. Thus the maximum output swing is 2V for 30g The fabricated accelerometer chips show a deviation in nominal capacitance values from the designed values.

 The differential

capaci tance bridge is not balanced at zero g. As a

result there is an initial offset in the output voltage when at zero g. The offset is nullified by using the internal capacitances in the ASIC. In cases where the offset is much more than the limit of the internal capacitances, provision is made to add an extern al

capacitor of suitable value in parallel with the lower capacitance in the bridge. Provision is made in the interface circu it board to make possible the tuning of the ASIC coefficients after the final assembly of the components on the PCB Board to cater for packaging effects also.

2.15 Results & discussion

An accelerometer with a range of 30 g and with linearity & cross-axis sensitivity less than 1% of full scale output (FSO) is configured. Detailed mechanical simulation techniques and electrical design is done using FEA show that the design

and the results

meets required specifications of the sensor. Comparison of analytical in table 2.5. and simulati on results are presented

Results

Analytical

FEM simulation 0.116 11 1470 26.0 -

System level simulation 0.121 -1432.5 28.07 0.2

Proof-mass displacement (m/g ) Stress at 30 g (MPa) Natural frequency( Hz) Sensitivity (fF/g) Response time (msec)

0.083 13.4 1728 19.0 0.042

Table 2.5 Comparison of results

From the comparison of results presented above it can be seen that the analytically estimated maximum displacement of the proo f-

mass is almost 30% less than the simula ted FEM results under the applied 1 g acceleration. This is due to the fact that during analytical calculations it is

estimated that the L-beams are fixed on one side and guided on the other side. However, the smaller length which is 20% of larger length (l3) of the L- beam (l2) contributes to

(640microns)

twisting in addition to bending, hence the discrepancy.