You are on page 1of 13

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO.

3, JUNE 2009 597

A Study for the Effect of the PCB Motion


on the Dynamics of MEMS Devices
Under Mechanical Shock
Fadi M. Alsaleem, Student Member, ASME, Mohammad I. Younis, Member, ASME, and
Mahmoud I. Ibrahim, Student Member, ASME

Abstract—We present a theoretical and experimental investiga-


tion into the effect of the motion of a printed circuit board (PCB)
on the response of microelectromechanical systems (MEMS) de-
vices to shock loading. For the theoretical part, a 2-DOF model
is used, where the first degree of freedom accounts for the PCB.
The second degree of freedom represents the motion of the MEMS
microstructure. Low-g acceleration pulses are applied to the
MEMS–PCB assembly base to simulate shock pulses generated
from a drop-table test. Simulation data are presented to show the
effects of the natural frequency of the PCB, the natural frequency Fig. 1. Schematic of a microstructure mounted on a PCB.
of the microstructure, and the shock pulse duration. Universal 3-D
spectra representing the effect of these parameters are presented.
It is found that neglecting the PCB effect on the design of MEMS to facilitate the electrical interconnections and the structural
devices under shock loads can lead to undesirable motion of their support [3]–[5]. MEMS microstructures, prior to getting
microstructures. The effects of electrostatic force and squeeze film assembled on a PCB, are housed in a mold compound that can
damping are investigated. It is found that the amplification of
motion due to the PCB can cause early pull-in instability for be, for example, ceramic or plastic. However, recently, in some
MEMS devices implementing electrostatic forces. The effect of MEMS applications, such as RF MEMS switches [6] and low-g
higher order modes of a microbeam is studied through a contin- accelerometers [7], the MEMS device is directly placed on the
uous beam model coupled with a lumped model of the PCB. The PCB in order to reduce the system complexity and cost. Mount-
limitations of the 2-DOF model are discussed. An experimental ing the microstructure of a MEMS device (directly or within the
investigation is conducted to verify the theoretical results using a
capacitive accelerometer. Experimental data for the response of mold compound) on a PCB may raise concerns about the role of
the accelerometer while it is mounted on two representative PCBs the PCB in amplifying the shock effects on the microstructure
due to different low-g shock conditions are shown. [2008-0026] and distorting its sensing/actuating signal. This is particularly
Index Terms—Electrostatic force, mechanical shock, micro- important in applications, such as micro-g accelerometers,
electromechanical systems (MEMS), microstructure, printed cir- where very high sensitivity to motion is required [7].
cuit board (PCB), reliability, squeeze film damping (SQFD). Fig. 1 shows a schematic for the assembly (packaging) levels
of a typical MEMS device [3]–[5]. As seen from the figure, four
I. I NTRODUCTION levels of assembly can be identified. These represent the con-
nection between a microstructure and a substrate, the bonding

T HE reliability of microelectromechanical systems


(MEMS) devices in shock environment represents one
of the key factors to their commercialization [1], [2]. Unlike
between the substrate and the chip, the connection of the leads
from the chip carrier to the PCB, and the connection of the PCB
to the supported structure. These four connections are almost
that in macroscale devices, where the failure is primarily due rigid to ensure an undistorted signal from the microstructure
to fatigue, mechanical shock loads in microscale systems can while sensing or actuating. However, because the microstruc-
cause different failure mechanisms such as mechanical failure, ture and the PCB are flexible structures (beams, plates, and
stiction, and short circuits. In commercial devices and electron- diaphragms), the interaction of their motion can be significant.
ics, MEMS devices are placed on a printed circuit board (PCB) Therefore, whether a microstructure is attached directly to a
PCB [6], [7] or is packaged in a standard assembly process as
Manuscript received February 1, 2008; revised July 29, 2008 and January 2, in Fig. 1, the interaction of its motion with the PCB should be
2009. First published April 21, 2009; current version published June 3, 2009. analyzed. Other sources of significant effect on the motion of
This work was supported by the National Science Foundation under Grant a microstructure are electrostatic forces in electrostatic MEMS
0700683. Subject Editor L. Spangler.
The authors are with the Department of Mechanical Engineering, State and squeeze film damping (SQFD). Both introduce nonlinear
University of New York (SUNY), Binghamton, NY 13902 USA (e-mail: effects that also need to be analyzed in depth, combined with
myounis@binghamton.edu). the PCB effect.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. The current trend in the IC and MEMS industry is to make
Digital Object Identifier 10.1109/JMEMS.2009.2016278 the PCB very thin and flexible. This means that the PCB is now
1057-7157/$25.00 © 2009 IEEE
598 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 3, JUNE 2009

more susceptible to experience considerable bending motion results and experimental data showing the effects of different
when subjected to mechanical shock and vibration. This motion PCBs on the response of MEMS devices under base-excitation
can be transmitted to the microstructure, leading to either its shock loads will be presented.
collapse or false function. These new PCBs are characterized The organization of this paper is as follows. In Section II, a
by having low natural frequencies. A coupling in the response brief background on the modeling of shock and impact load is
between such PCBs with MEMS devices that have low natural presented. In Section III, we present simulation results on the
frequencies, such as microcantilever beams, may occur. On the effects of PCB alone on a MEMS device. Section IV presents
other hand, thick or small PCBs of high natural frequencies are details of the parameter extraction for the studied capacitive
expected to cause coupling problems with stiffer microstruc- accelerometers and PCBs. Section V provides simulation and
tures, such as clamped–clamped microbeams and prestressed experimental results on the effects of PCBs. Following that, in
diaphragms. Section VI, we discuss the effects of electrostatic forces and
Many authors have studied the response of MEMS devices SQFD on the response of a MEMS device. In Section VII, we
under shock loads without incorporating the PCB effects [1], analyze the impact of electrostatic forces and SQFD combined
[9]–[15]. Some of these works are reviewed in [15]. The with that of a PCB on the response of a MEMS device.
response of MEMS devices to shock loads, including the Section VIII presents a microbeam model coupled with a
PCB (assembly) effects, has not been investigated thoroughly. lumped-mass model for the PCB and addresses the limitations
Gogoi et al. [16] remarked that the fundamental frequency of the 2-DOF model. Finally, in Section IX, we summarize and
of the system represented by the chip attached to the PCB conclude this paper.
should lie outside the intended operating frequency range of
the MEMS structure. Fan and Shaw [3] evaluated the response
of a microstructure used for acceleration measurement under II. A PPROACHES TO M ODEL S HOCK AND I MPACT L OAD
severe shock loads. They indicated that the PCB, on which As mentioned earlier, we assume here that both the substrate–
the microstructure is mounted, introduces undesirable effects chip and the chip–PCB assembly are rigidly connected [2],
on the acceleration measurements. To alleviate this factor, they [21]. Hence, the assembly effect reduces to that of the PCB
proposed to stiffen the PCB. motion only. We use a 2-DOF model (Fig. 2) to study the
Srikar et al. [2] studied the reliability of MEMS devices PCB effect on the response of a MEMS device under shock
subjected to shock loads. They indicated that most MEMS load. The first degree of freedom accounts for the PCB motion,
devices experience shock loads as quasi-static loads since their and the second degree of freedom represents the motion of the
natural periods are much smaller than the duration of shock microstructure, such as a beam or a plate, which is mounted
loads. Srikar et al. [2] pointed out that the MEMS package on the PCB. Shown in Fig. 2 are km : microstructure stiffness,
reduces the shock load applied to the microstructure, and the kPC : PCB stiffness, cm : microstructure damping, cPC : PCB
worst case scenario is to transfer this shock pulse without damping, mm : microstructure mass, and mPC : PCB mass. In
significantly altering its shape or intensity. In this paper, how- this model, we assume that the MEMS device is placed on the
ever, we show that, for certain package designs and for certain center of the PCB, which represents a worst case scenario. This
shock duration values, the microstructure response might be is because the center of the PCB is expected to have the largest
amplified significantly because of the motion of the package motion due to shock [as will be shown later in Fig. 12(b)].
over a PCB. Impact forces and base-excitation pulses are two methods
In a previous work [17], the effect of the PCB motion on the to model shock forces affecting a 2-DOF system [22], similar
response of a microstructure was studied theoretically using a to the case of the MEMS–PCB assembly. Impact shock force
2-DOF model. The mechanical shock was modeled as a single- occurs, in the MEMS–PCB assembly case, when a sudden
point force impacting the PCB. In [18], a continuous–lumped- force is applied directly to the PCB, such as that of a hammer
mass model was used to simulate the dynamic response of hit [Fig. 2(a)]. On the other hand, base-excitation acceleration
the PCB MEMS assembly under the effect of a point force pulse occurs when the base of the assembly undergoes a sud-
impacting the PCB. In [19], we studied theoretically and ex- den acceleration change [Fig. 2(b)]. This change happens, for
perimentally the response of a capacitive accelerometer under example, due to the drop of the assembly to the ground or due
the combined effects of shock and electrostatic forces. to a drop-table test. In this paper, we study the latter case of
The objective of this paper is to present an accurate dy- Fig. 2(b). Fig. 3(a) and (b) shows shock pluses due to a base-
namical model for the mechanical shock problem on the PCB excitation pulse and an impact force.
MEMS assembly. This paper aims to understand the vibration To model a shock pulse due to a drop test [20], [23], it
and motion consequences due to shock, which may cause stic- is assumed to be a half-sine pulse of period T [Fig. 3(a)].
tion and short-circuit problems due to the impact of microstruc- Under this assumption, the governing equations of motion of
tures among themselves and with the substrate. This paper does the system are given by
not consider the effect of shock on the functionality of the
MEMS as a sensor or actuator. Here, we model the mechanical
shock as a base-excitation acceleration pulse affecting the PCB mm ẍ1 + km (x1 − x2 ) + cm (ẋ1 − ẋ2 ) = 0 (1)
MEMS assembly, which is an accurate way to model shock
mp ẍ2 − km (x1 − x2 ) − cm (ẋ1 − ẋ2 )
pulses due to drop-table or drop tests [20]. A comparison
between this case and the model of [17] is made. Simulation + kPC x2 + cPC ẋ2 = kPC y + cPC ẏ (2)
ALSALEEM et al.: STUDY FOR THE EFFECT OF THE PCB MOTION ON THE DYNAMICS OF MEMS DEVICES 599

motion can be written as

mm ẍ1 + km (x1 − x2 ) + cm (ẋ1 − ẋ2 ) + fsqueeze


2
εAVdc
= (3)
2 [d − (x1 − x2 )]2
mp ẍ2 − km (x1 − x2 ) − cm (ẋ1 − ẋ2 ) + kPC x2 + cPC ẋ2
= kPC y + cPC ẏ (4)

where ε is the dielectric constant of the gap medium, A is


the electrode area of the microstructure, d is the capacitor gap
width, and Vdc is the dc polarization voltage. The squeeze film
term fsqueeze can be expressed using the Blech model [24], [25],
which is a series expression, as fsqueeze = Csqu (ẋ1 − ẋ2 ), where

64σPa A   m2 + n2

Csqu =
π 6 ωd (mn) {[m2 + n2 ]2 + σ 2 /π 4 }
2
m,n,odd
(5)

and σ is the squeeze number, σ = (12Aωm η/Pa [d − (x1 −


x2 )]2 ). Here, Pa is the ambient pressure, ωMEMS is the natural
frequency of the microstructure, and η is the viscosity coeffi-
cient of air. In [19], we found that using one term in the series
yields sufficient accuracy. By considering only the first term of
(5), where m = 1 and n = 1, the squeeze damping coefficient
becomes
 
64σPa A 2
Csqu = 6 . (6)
π ωm [d − (x1 − x2 )] [4 + σ 2 /π 4

The Blech model in (6) is modified to allow variation in the gap


distance with the motion [19].
Fig. 2. Two approaches to model shock and impact. Shown is the MEMS–
PCB assembly, (a) subjected to a single impact force [17] and (b) a base III. S IMULATION R ESULTS ON THE PCB E FFECTS
acceleration due to a drop-table test.
In this section, (1) and (2) are used to model the influence
of the PCB motion on the response of a microstructure in the
absence of electrostatic forces or SQFD. Using modal analysis,
(1) and (2) can be solved analytically [23].
As a case study, a MEMS diaphragm of an omnidirectional
pressure microphone [26] is considered. The diaphragm has
a fundamental natural frequency of 24 kHz. Fig. 4(a) and (b)
shows the maximum relative amplitude of the diaphragm with
respect to the PCB (x1 − x2 ). The response in this figure and
all subsequent ones is normalized with respect to the static
deflection of the microstructure to an equivalent static load.
Fig. 3. Schematic of a half-sine pulse used to model the actual shock loads,
(a) base shock acceleration, and (b) impact shock force. For comparison Results are shown for different PCB natural frequencies ωPCB ’s
purposes of the two shock load types, it is assumed that Fo = Ao mPC . with T = 1.0 ms and T = 60 μs. In the figures, the damping
ratio is assumed zero, considering the worst case scenario for
where y and ẏ are the displacement and velocity of the assembly a shock response. The figures show that the MEMS response
base, respectively. The base acceleration ÿ is assumed to be is always amplified due to the presence of the PCB, except for
a half-sine pulse of amplitude Ao , which is expressed as ÿ = the zones where the natural frequency of the PCB is very low
Ao [sin(ωPulse t)u(t) + sin(ωPulse (t − T ))u(t − T )], where (the response is attenuated) and very high (the response is the
ωPulse is the pulse frequency = π/TPulse , t is the time, and same as if there is no PCB). Moreover, as the shock duration
u(t) is the unit step function. decreases, the amplification amplitude increases.
Many MEMS devices are actuated by electrostatic forces and For comparison purposes, we show in Fig. 5 the maximum
operated in the presence of SQFD. In this case, the equations of normalized relative amplitude of the response for the same
600 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 3, JUNE 2009

Fig. 4. Undamped response of a microdiaphragm to a base shock load. (a)


T = 1.0 ms. (b) T = 60 μs.

shock conditions in Fig. 4 due to a single impact load on the Fig. 5. Comparison of the undamped response of a microdiaphragm due to a
PCB [17]. It is clear from the figures that the diaphragm exhibits single force and a base-excitation load. (a) T = 1.0 ms. (b) T = 60.0 μs.
similar amplification zones for the base shock load case as in the
single force impact. The main differences, however, are that, as
IV. E XPERIMENTAL S ETUP AND P ARAMETER E XTRACTION
the natural frequency of the PCB increases, the diaphragm starts
to experience the shock load as if there is no PCB in the case In this section, we discuss the parameter extraction for two
of the base shock load. On the other hand, the response dies out capacitive accelerometers and two PCBs, which we will use as
in the case of a single impact load. Furthermore, the response case studies for the simulation and experimental work of this
of the diaphragm dies out as the natural frequency of the PCB paper.
decreases in the base shock load case, whereas it experiences
the shock load as if there is no PCB in the impact shock
load case. A. MEMS Characterization
Next, we show in Fig. 6 a universal 3-D plot for the nor- Two different samples of a commercial off-the-shelf
malized deflection of a microstructure as a function of the capacitive accelerometer (Fig. 7), fabricated by Sensata
shock frequency, the PCB natural frequency, and the natural Technologies,1 are used for simulation and experimental inves-
frequency of the microstructure. In the figure, we assume a tigation. We label them as sample “a” and sample “b.” The sam-
damping ratio for the PCB of 0.05 and a damping ratio of the ples are made up of two alloy-42 cantilever beams of 150-μm
microstructure of 0.05 in Fig. 6(a) and 0.5 in Fig. 6(b). These thickness and a proof mass (approximately of 9.0-mm length
plots can serve as a guideline for MEMS designers to ensure and 5.32-mm width) attached to their tips. The proof mass
reliable operation of their various MEMS devices on a PCB. forms one side of the capacitive electrode used for detection.
For example, the pressure microphone has a natural frequency The samples are identical in their geometrical properties. The
of 24 kHz. The shock duration that this microstructure could differences between them are the stiffness of their cantilever
experience in a drop test is between 0.1 and 1.0 ms. This beams and the gap separating the proof mass from the substrate.
shock range corresponds to ωMEMS /ωPulse = 2.4 − 24. Based The experimental setup shown in Fig. 8(a) and (b) was used for
on Fig. 6, a designer can select a suitable PCB, which has the
minimum effect on amplifying the unwanted response of the
microphone due to shock load. 1 www.sensata.com.
ALSALEEM et al.: STUDY FOR THE EFFECT OF THE PCB MOTION ON THE DYNAMICS OF MEMS DEVICES 601

Fig. 7. (a) Photograph of a capacitive accelerometer, fabricated by Sensata


Technologies.1 (b) Capacitive accelerometer taken apart.

linear damping ratio were found. Table I gives a summary of the


different parameters for each sample. Although Table I shows
that the two samples have low natural frequencies, they both
share the main characteristics of a MEMS device. The gap and
the thickness of their structures are in the micrometer range,
their responses are dominated by the SQFD effect, and they are
actuated electrostatically. We conducted modal analysis using
white noise, which revealed that the torsional modes of the
Fig. 6. Universal 3-D plots of the normalized relative amplitude of a mi- device get activated at very large frequencies. Hence, they will
crostructure for different ωPCB and ωPulse values. (b) Effect of damping in be neglected in this paper.
reducing the amplification in the MEMS response due to the presence of a PCB.
(a) ζMEMS = 0.05. (b) ζMEMS = 0.5.
B. PCB Characterization
both the experimental investigation and extracting the mechan-
ical parameters of the samples. The first parameter extracted Two PCBs of representative fundamental natural frequencies
was the effective stiffness of the cantilevers. The device was were chosen. The first one (PCB1) is a circular disk of a fun-
actuated using various dc voltages, and the transient response damental natural frequency that is very large compared to the
of the proof mass was monitored using a laser vibrometer. natural frequency of the accelerometer [Fig. 11(a)]. The other
Fig. 9 shows an example of the data obtained experimentally one (PCB2) is a rectangular plate with a fundamental natural
for sample b. Using an S-DOF model for the capacitive sensor, frequency close to the natural frequency of the accelerometer
the equilibrium solution can be written as [Fig. 11(b)]. The second PCB is designed to adhere to the
JEDEC standard [27] for electronic component shock testing.
2
εAVdc We began our characterization for the two PCBs by mounting
x3 − 2dx2 + d2 x − = 0. (7) them on the head of a shaker. A laser vibrometer was used to
2k
monitor the velocity response of each PCB. A reference ac-
The stable solution of (7) was used to curve-fit the experi- celerometer was mounted on the head of the shaker to measure
mental data. To measure the gap width d, which separates the its acceleration. A random noise voltage was applied as an input
proof mass from the stationary electrode, a high dc voltage to the electromagnetic shaker to obtain the frequency response
beyond pull was applied. Fig. 10 shows the measured transient of the two PCBs. Fig. 12 shows an example of the obtained
response of the proof mass of sample b. Finally, the vibration frequency response and the corresponding mode shape of PCB1
response of the device was measured to identify the natural fre- at its first fundamental natural frequency. The figure shows that
quency and linear damping ratio. A continuous random signal PCB1 has a first natural frequency around 6.0 kHz. Similarly,
was used to drive the shaker. Using a curve fitting technique, it was found that PCB2 has a fundamental natural frequency of
both the fundamental natural frequency of the device and the 230 Hz.
602 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 3, JUNE 2009

Fig. 10. Measured transient response of the proof mass for a voltage beyond
the pull-in voltage.

TABLE I
SUMMARY OF THE EXTRACTED PARAMETERS FOR EACH SAMPLE
OF THE S TUDIED C APACITIVE A CCELEROMETER

Fig. 8. (a) Photograph of the experimental setup used for testing.


Fig. 11. Photograph of the two PCBs used for testing. (b) PCB2 placed over
(b) Schematic of the experimental setup and the data acquisition system. In the
a shaker head. (a) PCB1 (ωPCB = 6 kHz). (b) PCB2 (ωPCB = 230 Hz).
figure, the resistor R is used as a current-limiting resistor to protect the device
in the case of pull-in from the high current of the closed circuit.
V. E XPERIMENTAL D ATA AND C OMPARISON W ITH
S IMULATION R ESULTS ON THE PCB E FFECTS
This section presents a comparison between experimental
and theoretical results for the responses of the capacitive ac-
celerometer, including the PCB effect, due to base shock loads.
Sample “a” was used in this investigation. First, the response
of sample “a” without a PCB was obtained. Toward this, the
capacitive accelerometer was mounted directly on the top of
the shaker, and its response was monitored. Two displacement
measurements were obtained using the laser vibrometer, i.e.,
one for the absolute motion of the accelerometer and the
other for the substrate motion, which is the same as the PCB
motion assuming a rigid connection between the PCB and the
accelerometer substrate. By subtracting these two measure-
ments, the relative displacement of the capacitive accelerometer
relative to the substrate (PCB) was obtained. A positive value
Fig. 9. Variation of the proof-mass displacement normalized by the gap
spacing underneath the proof mass d for various values of Vdc . Shown in the of the relative displacement means that the microstructure
figure are both simulation and experimental data of sample b. deflects toward the substrate. In this case, SQFD becomes
ALSALEEM et al.: STUDY FOR THE EFFECT OF THE PCB MOTION ON THE DYNAMICS OF MEMS DEVICES 603

Fig. 13. Shock spectrum of sample a obtained theoretically and experimen-


tally, (a) including and (b) without including the effect of PCB1.

ωMEMS = 6 kHz/187 Hz = 32, it is expected, according to


Fig. 6(b), that the PCB will transfer the shock load to the
accelerometer microstructure without any alteration. Fig. 13(b)
compares the simulation and experimental results for the max-
imum normalized relative amplitude of the tested accelerom-
eter for different shock durations when the accelerometer is
mounted on PCB1. We assumed a damping ratio that is equal
to 0.05 for PCB1 in the simulations. A comparison between
Fig. 12. (a) Frequency response of PCB1. (b) Three-dimensional plot for the
first mode shape of PCB1, which was obtained experimentally by measuring
Fig. 13(a) and (b) shows that the shock spectra of the two
the PCB response at different locations using the laser vibrometer. The figure cases are similar. This agrees with the conclusion of Section III,
shows that the PCB moves purely up and down with its maximum deflection at i.e., for a base shock load, as the ratio between the natural
its center.
frequencies of a PCB and a MEMS device increases, the PCB
transfers the shock load to the MEMS device without any
significant. A negative value indicates that the microstructure alteration (i.e., the shock spectrum of the MEMS device when
deflects away from the substrate. In this case, SQFD be- mounted on a PCB should be similar to the one when the
comes weak. MEMS is not on a PCB).
Fig. 13(a) shows the normalized maximum response of the Next, the accelerometer was mounted on PCB2. The ratio of
accelerometer for different shock durations (shock spectrum) the natural frequency of PCB2 to the accelerometer’s natural
while it is not mounted on any PCB. In this figure, the experi- frequency is 1.23. Fig. 6(a) shows that a high amplification for
mental data are compared with S-DOF shock spectrum. In the the MEMS response occurs at this ratio. However, since the
simulation, only a linear damping ratio that is equal to 0.5 is damping ratio of the accelerometer is very high, we expect less
assumed. Experimentally, the response of the proof mass while amplification [Fig. 6(b)]. Fig. 14 compares the simulation and
it is deflecting away from the substrate is measured so that experimental results for the maximum normalized relative am-
the effect of SQFD is weak. The figure shows good agreement plitude of the tested accelerometer for different shock durations
between simulation and experimental results. when the accelerometer is mounted on PCB2. We assumed a
Next, the accelerometer was mounted on PCB1 shown in damping ratio that is equal to 0.05 for PCB2 in the simulations.
Fig. 11(a). The PCB was fixed using four screws to the head The figure shows a good agreement between the simulation
of the shaker. Since the ratio of the PCB’s natural frequency to and experimental results. A comparison between Figs. 13 and
the accelerometer’s natural frequency is very high, i.e., ωPCB / 14 shows that, at this specific ratio between the PCB and the
604 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 3, JUNE 2009

Fig. 14. Simulation-versus-experimental results for the shock spectrum of


sample a, including the effect of PCB2.
Fig. 16. (Stars and diamonds) Simulation results and (circles and squares)
experimental data of the proof-mass deflection with shock amplitude when T =
5.0 ms [19].

Fig. 15. Transient response of the proof mass with and without PCB2 when
subjected to a mechanical shock, as monitored through a laser Doppler vibrom-
eter (T = 5.0 ms). Fig. 17. S-DOF model of a parallel-plate electrostatic MEMS device sub-
jected to base excitation.

MEMS device, the response of the microstructure is amplified.


This is because the natural frequency of the PCB is near that of VI. E FFECT OF E LECTROSTATIC F ORCE
the microstructure. W ITHOUT I NCLUDING THE PCB
Fig. 15 shows a sample of the measured transient response We investigated in the previous section the effect of the
of the device with PCB2 (dynamic response amplification) and PCB alone on the motion of a MEMS device. In order to fully
without PCB2 (quasi-static response with no amplification) due understand the behavior of an electrostatically actuated MEMS
to a shock load of T = 5.0 ms. This figure shows a significant device, one needs to include the nonlinearities introduced to the
amplification for the response of the device due to the dynamic problem in a real application. The first source of nonlinearity
effect of the poorly designed PCB. Note that, without a PCB, affecting the motion of a microstructure is the electrostatic
the device responds quasi-statically to shock load. Hence, the force, which acts as a negative quadratic nonlinearity. Another
PCB here has adverse effect on the performance of the device. source of nonlinearity is introduced through SQFD. This effect
We can observe from the figure a nonlinear behavior in the was ignored in the previous section since the proof mass was
accelerometer response. This behavior is characterized by a studied while moving away from the substrate.
large deflection of the proof mass (near −100 μm) when it The model shown in Fig. 17 is used to investigate these
moves away from the substrate while the response is suppressed nonlinearity effects on the microstructure without the presence
going toward the substrate. This is due to the effect of SQFD, of the PCB. The equation of motion, in this case, is given by
which increases toward the substrate. Fig. 16 [19] shows more
clearly the nonlinear effect of SQFD on the deflection of the 2
εAVdc
proof mass. It should also be mentioned that the figure shows mz̈ + cż + kz + fsqueeze = − F0 mÿ(t) (8)
2(d − z)2
a drift in the measurement where the proof mass appears to
oscillate steadily around an equilibrium position near −20 μm where z is the relative deflection of the proof mass, which
instead of zero. This is just due to the laser vibrometer reading is the absolute motion of the proof mass x minus that of the
drift over time. substrate y (z = x − y), and ÿ(t) denotes the base acceleration
ALSALEEM et al.: STUDY FOR THE EFFECT OF THE PCB MOTION ON THE DYNAMICS OF MEMS DEVICES 605

Fig. 19. Simulation-versus-experimental results for the accelerometer shock


spectrum of sample b when subjected to a 3-g shock at Vdc = 0 V.
Fig. 18. Shock spectra of sample b when subjected to a shock load of 1 g and
various electrostatic loads for the case of SQFD and high linear damping.

that is provided to the device in the form of a half-sine shock


pulse given in Section II. The squeeze film term fsqueeze can
be expressed using the Blech model [24], [25] as fsqueeze =
Csqu ẋ, where Csqu is defined in (6).
Next, we investigate the electrostatic force combined with
SQFD and their effects on the shock spectrum of sample b
(see Table I). Here, we assume that the accelerometer works
under ambient pressure. We use a linear damping ratio of 0.4,
which is obtained experimentally for sample b of the capacitive
accelerometer. Since (8) is nonlinear, an analytical solution
is not available. Hence, we use direct time integration for
the equation of motion to obtain the shock spectrum using
Runge–Kutta method. It is important to note that applying
the electrostatic force causes an initial static deflection of the Fig. 20. Simulation-versus-experimental results for the accelerometer shock
spectrum of sample b when subjected to a 3-g shock at Vdc = 60 V.
proof mass and applying the shock load thereafter causes the
proof mass to oscillate about its new equilibrium position due
to the electrostatic force. For the purpose of comparing with using the shaker. Figs. 19 and 20 compare the simulation
experimental data, all simulated shock spectra illustrated in this and experimental results, where the device was subjected to a
section account for this new equilibrium position by subtracting mechanical shock of 3 g and dc voltages of 0 and 60 V, respec-
the static deflection of the proof mass due to the electrostatic tively. The figures show good agreement between simulations
force from the total response to obtain the dynamic component and experimental data within the tested range. The experimental
of the response only. data were obtained only for a small range of shock durations
Fig. 18 shows the shock spectrum of the MEMS device when due to experimental limitations from both the shaker and laser
subjected to acceleration pulses of amplitude of 1.0 g. We can vibrometer. For large shock durations, the response was quasi-
note from the figure that, as we increase the electrostatic force, static when subtracting the two measurements, which was not
the deflection of the microstructure increases significantly. This well captured by the laser vibrometer. When the shock duration
behavior shows that the electrostatic force acts as a softener was too small, it was close to the natural period of the shaker,
to the microstructure. Young et al. [28] observed the same be- which caused disturbances in the signal.
havior on a spring–mass–damper model with a cubic-softening
spring term in the equation of motion. The figure also clearly
VII. E FFECT OF E LECTROSTATIC F ORCE
shows the effect of SQFD in suppressing the motion of the
I NCLUDING THE PCB M OTION
microstructure in the dynamic zone causing no amplification,
unlike the amplification observed in Fig. 13(a) and (b), where Here, we study the combined effect of the PCB and the elec-
SQFD was not present. It is also observed that the microstruc- trostatic force on the response of the capacitive accelerometer
ture experiences pull-in near the quasi-static region of the shock of sample a to base shock load. We used the model shown in
spectrum where SQFD has small effect. Fig. 21 with (3) and (4).
Comparison With Experiment: The same procedure ex- Fig. 22 compares the calculated pull-in voltage of the capac-
plained in Section V was used here, where the device was itive device against the shock amplitude when modeling the de-
actuated by a dc voltage and subjected to mechanical shock vice alone without the PCB and with the PCB for the two cases
606 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 3, JUNE 2009

Fig. 23. Schematic of the coupled model used to represent the effects of the
Fig. 21. Schematic for a 2-DOF model of a parallel-plate MEMS device motion of the PCB on the response of a clamped–clamped microbeam under
mounted on a PCB undergoing base shock load excitation. base load excitation.

VIII. C ONTINUOUS –L UMPED M ODEL


In this section, we present a continuous model of a MEMS
microstructure that is coupled with a lumped-mass model of the
PCB (Fig. 23). We adopt this model to study the limitations of
considering the MEMS device in the aforementioned analysis
as a lumped-mass model. For illustration, a clamped–clamped
microbeam is used in this investigation. Clamped–clamped mi-
crobeams play an important role in many MEMS devices, such
RF switches and thermal actuators. Using Hamiltonian’s princi-
ple, the equations of motion for the microbeam deflection w and
the PCB when subjected to a base excitation can be written as
⎡ ⎤
4 2 L 2
∂ w ∂w ∂w ⎣Ebh ∂w ∂2w d2X
EI + ρbh 2 + c̃ = dx⎦ 2 −ρbh 2
∂x 4 ∂t ∂t 2L ∂x ∂x dt
Fig. 22. Plot of the pull-in voltage of sample a against the shock load ampli- 0
tude of a half-sine pulse using the 1-DOF and 2-DOF models. Results are shown (9a)
with and without including the effect of SQFD for the case of T = 5.0 ms
and ωPCB = 230 Hz (PCB2). L
d2X dX ∂2w dy
(m+ρbhL) 2 + kp X + cp + ρbh 2 dx = kp y+ cp
TABLE II dt dt ∂t dt
0
PULL-IN VOLTAGE UNDER SHOCK FOR THE SELECTED POINTS
OF F IG . 22 IN THE P RESENCE OF SQFD (9b)

where x is defined as the position along the microbeam length;


X is the PCB displacement; I is the moment of inertia of the
cross section; E is Young’s modulus; ρ is the material density;
t is the time; c̃ is the damping coefficient of the microbeam
with and without SQFD. The linear damping ratio is assumed per unit length; y is the base motion; kp , cp , and m are the
to be 0.5. It is noted here that the 2-DOF model (accounting effective stiffness, damping coefficient, and the mass of the
for PCB2) shows an earlier pull-in compared to the 1-DOF PCB, respectively; and L, h, and b are the microbeam length,
model. The results indicate the importance of modeling the thickness, and width, respectively. Because the mass of the
effect of the PCB motion in electrostatic MEMS. By neglecting microbeam is much smaller than that of the PCB, its effect on
the effect of the PCB, there is a possibility that the device the motion of the PCB is neglected. We assume no damping
will fail to function properly and that it might collapse and fail acting on the PCB, which represents a worst case scenario.
mechanically or electrically. When comparing the cases of with Therefore, the equation of motion of the PCB is decoupled
and without SQFD, we note that SQFD has significant impact from the microbeam. Hence, (9b) is reduced to
in raising the threshold of dynamic pull-in due to mechanical d2 X
shock. Table II shows the selected experimental points for the m + kp X = kp y. (10)
dt2
results of Fig. 22. The table compares the pull-in threshold
under shock with and without PCB2 in the presence of SQFD. Next, we solve for X in (10) and substitute the result in
The experimental data agree with the simulation results. (9a). We then use a reduced-order model [29] to discretize the
ALSALEEM et al.: STUDY FOR THE EFFECT OF THE PCB MOTION ON THE DYNAMICS OF MEMS DEVICES 607

Fig. 24. Results for the MEMS–PCB assembly response to base shock Fig. 25. Results for the MEMS–PCB assembly response to base shock load
load of shock amplitude of 400 g and duration T = 1 ms using (dashed) of shock amplitude of 400 g and T = 1 ms using (dashed) the 2-DOF model
the 2-DOF model, (solid) the continuous–lumped model, and (dotted) the and (solid) the continuous–lumped model. The natural frequency of the PCB is
continuous–lumped model, including midplane stretching. The natural fre- 87.57 kHz. Here, w refers to the maximum deflection of the microbeam in the
quency of the PCB is 1.0 kHz. Here, w refers to the maximum deflection of continuous–lumped model and to the relative deflection of the microbeam for
the microbeam in the continuous–lumped model and to the relative deflection the case of the lumped-mass model.
of the microbeam for the case of the lumped-mass model.

resulting equation into a finite-degree-of-freedom system con-


sisting of ordinary differential equations in time. The odd un-
damped linear mode shapes of the straight microbeam are used
as basis functions in the Galerkin procedure. The even modes
are not included in the solution since they are asymmetric
modes, and hence, their influence on the beam dynamics is
negligible.
The material properties of the investigated microbeam are
assumed to be E = 169.0 GPa and ρ = 2332 kg/m3 . The
microbeam is assumed to have L = 900 μm, h = 1.5 μm,
b = 100 μm, and d = 2.0 μm.
Fig. 24 compares the continuous–lumped model results [in-
cluding and not including the nonlinear midplane stretching
term, the integral term in (9a)] with the 2-DOF model results
when the shock load amplitude is 400 g and its duration value
is 1.0 ms. The PCB’s natural frequency, in this case, is assumed
to be 1.0 kHz. In the 2-DOF model, the effective stiffness of the
clamped–clamped beam is calculated as

32Ebh3
k= . (11)
L3
The figure shows that the results of the 2-DOF model and
the continuous–lumped model (without including midplane
stretching) are in good agreement. However, the figure shows
that midplane stretching has the effect of stiffening the beam.
Because the focus of this paper is on the participation of higher
order modes, not on the effect of midplane stretching, we will
neglect midplane stretching in the rest of the paper to allow for
a more quantitative comparison with the 2-DOF linear model.
Next, we show in Fig. 25 the response of the microbeam Fig. 26. Results for the MEMS–PCB assembly response to base shock load
of shock amplitude of 400 g and duration T = 27 μs using (dashed) the 2-DOF
using the continuous–lumped model and the 2-DOF models model and (solid) the continuous–lumped model. The natural frequency of the
for the case of a natural frequency of the PCB that is equal PCB is 87.57 kHz.
to 87.57 kHz. This natural frequency was chosen to be near
the second symmetric natural frequency of the microbeam. We models for shock load amplitude of 400 g and duration of 27 μs.
notice from the figure that the two plots are almost identical. In The natural frequency value of the PCB is the same as in
Fig. 26, we show the response of the microbeam using the two Fig. 25. The shock duration value was chosen here to be near
608 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 18, NO. 3, JUNE 2009

the natural period of the second symmetric natural frequency [3] M. S. Fan and H. C. Shaw, “Dynamic response assessment for the MEMS
of the microbeam. The figure shows that the 2-DOF model accelerometer under severe shock loads,” NASA, Washington, DC, Rep.
TP—2001-209978, 2001.
fails to predict the microbeam response for this specific shock [4] R. Ghaffarian, D. Sutton, P. Chaffee, N. Marquez, A. Sharma, and
loading case (when the natural frequency of the PCB and the A. Teverovsky, “Thermal and mechanical reliability of five COTS
shock pulse frequency are near the second symmetric natural MEMS accelerometers,” NASA Electronic Parts and Packaging Program,
2002. [Online]. Available: http://nepp.nasa.gov/eeelinks/February2002/
frequency of the microbeam). Thermal_and_Mechanical_Reliability.pdf
In summary, we conclude that the higher order modes of [5] G. Ken, “MEMS PCB assembly challenge,” Circuits Assem., vol. 11,
vibration of a microbeam may contribute to its response only no. 3, p. 62, 2002.
[6] H. Chang, J. Qian, B. Cetiner, F. Flaviis, M. Bachman, and G. Li, “Design
when one of those modes has a natural frequency that is near and processes consideration for fabrication RF MEMS switches on printed
both the natural frequency of the PCB and the shock pulse circuit boards,” J. Microelectromech. Syst., vol. 14, no. 6, pp. 1311–1322,
frequency. Dec. 2005.
[7] N. Yazdi and K. Najafi, “An all-silicon single-wafer micro-g
accelerometer with a combined surface and bulk micromachining
IX. C ONCLUSION AND S UMMARY process,” J. Microelectromech. Syst., vol. 9, no. 4, pp. 544–550,
Dec. 2000.
We investigated the response of MEMS devices, including [8] U. Wagner, J. Franz, M. Schweiker, W. Bernhard, R. Muller-Fiedler,
B. Michel, and O. Paul, “Mechanical reliability of MEMS-structures un-
the effect of the PCB motion, for different conditions of shock der shock load,” Microelectron. Reliab., vol. 41, no. 9, pp. 1657–1662,
pulses. It was found that neglecting the PCB effect on the mod- Sep. 2001.
eling of a microstructure of a MEMS device could underesti- [9] J. De Coster, H. Tilmans, J. Van Beek, G. Rijks, and R. Puers, “The
influence of mechanical shock on the operation of electrostatically
mate the microstructure motion. Universal 3-D plots accounting driven RF-MEMS switches,” J. Micromech. Microeng., vol. 14, no. 9,
for the natural frequency of the PCB, the natural frequency pp. S49–S54, Aug. 2004.
of the microstructure, and the shock duration are generated. [10] T. Brown, “Harsh military environments and microelectromechanical
(MEMS) devices,” in Proc. IEEE Sensors, 2003, vol. 2, pp. 753–760.
These can be used to define the amplification zones in which [11] N. Tas, T. Sonnenberg, H. Jansen, R. Legtenberg, and M. Elwenspoek,
the normalized response of a microstructure is amplified. Such “Stiction in surface micromachining,” J. Micromech. Microeng., vol. 6,
plots can be used by MEMS designers to help ensure safe no. 4, pp. 385–397, Dec. 1996.
[12] D. Tanner, J. Walraven, K. Helgesen, L. Irwin, N. Smith, and N. Masters,
operation of their devices. “MEMS reliability in shock environments,” in Proc. IEEE Int. Reliab.
For MEMS devices actuated electrostatically, it is found Phys. Symp., San Jose, CA, 2000, pp. 129–138.
that a poor design of the PCB can lead to an early dynamic [13] X. Fang, Q. Huang, and J. Tang, “Modeling of MEMS reliability in shock
environments,” in Proc. 7th Int. Conf. Solid-State Integr. Circuits Technol.,
instability (dynamic pull-in) under shock load. In these devices, Beijing, China, 2004, pp. 860–863.
SQFD is found to have the effect of suppressing the shock [14] O. Millet, D. Collard, and L. Buchaillot, “Reliability of packaged
effects and stabilizing the device. MEMS in shock environments: Crack and stiction modeling,” in Proc.
Des., Test, Integr. Packag. MEMS/MOEMS, Cannes, France, 2002,
An experimental investigation has been conducted to char- pp. 696–703.
acterize a capacitive accelerometer that is mounted on PCBs. [15] M. I. Younis, R. Miles, and D. Jordy, “Investigation of the response
Experimental results for the capacitive accelerometer when it of microstructures under the combined effect of mechanical shock
and electrostatic forces,” J. Micromech. Microeng., vol. 16, no. 11,
is mounted on two different PCBs due to different base shock pp. 2463–2474, Oct. 2006.
load conditions are shown. It is found that these results are in [16] B. Gogoi, M. Vujosevic, and S. Petrovic, “Challenges in MEMS packag-
good agreement with the simulation results. ing,” in Proc. SMIT Int. Conf., Rosemont, Cook, IL, 2000, pp. 775–783.
[17] F. M. Alsaleem, M. I. Younis, and R. Miles, “An investigation into the
Finally, a continuous model for the microbeam coupled with effect of the PCB motion on the dynamic response of MEMS devices
a lumped model for the PCB is used to study the limitations of under mechanical shock loads,” J. Electron. Packag., vol. 130, no. 3,
considering a microstructure as a lumped mass. It is found that pp. 031 002-1–031 002-10, Sep. 2008.
[18] M. I. Younis, F. M. Alsaleem, and D. Jordy, “The response of clamped–
higher modes of vibration of the microstructure have negligible clamped microbeams under mechanical shock,” Int. J. Non-Linear Mech.,
effects on its response, except for the case when one of those vol. 42, no. 4, pp. 643–657, May 2007.
modes has a natural frequency that is near both the natural [19] M. I. Younis, F. M. Alsaleem, R. Miles, and Q. Su, “Characteriza-
tion of the performance of capacitive switches activated by mechanical
frequency of the PCB and the shock pulse frequency. shock,” J. Micromech. Microeng., vol. 17, no. 7, pp. 1360–1370,
As a future work, we plan to investigate the effect of higher Jun. 2007.
order modes of a microstructure when they interact with those [20] D. S. Steinberg, Vibration Analysis for Electronic Equipment, 3rd ed.
New York: Wiley–Interscience, 2000.
of a PCB. This problem requires a continuous–continuous [21] J. M. Pitarresi and A. Primavera, “Comparison of modeling techniques for
model for both the microstructure and the PCB. the vibration analysis of printed circuit cards,” Trans. ASME, J. Electron.
Packag., vol. 114, no. 4, pp. 378–383, 1992.
[22] A. G. Hernried and J. L. Sackman, “The two-degree-of-freedom equip-
ACKNOWLEDGMENT ment structure system,” J. Eng. Mech., vol. 112, no. 6, pp. 621–628,
Jun. 1986.
The authors would like to thank Sensata Technologies for [23] B. Balachandran and E. Magrab, Vibrations. Columbus, IN: Thomson
Eng., 2003.
providing the parts used for testing. [24] J. J. Blech, “On isothermal squeeze films,” J. Lubr. Technol., vol. 105,
pp. 615–620, Oct. 1983.
[25] M. Andrews, I. Harris, and G. Turner, “A comparison of squeeze-film
R EFERENCES theory with measurements on a microstructure,” Sens. Actuators A, Phys.,
[1] G. Li and J. Shemansky, “Drop test analysis on micro-machined struc- vol. 36, no. 1, pp. 79–87, Mar. 1993.
tures,” Sens. Actuators A, Phys., vol. 85, no. 1–3, pp. 280–286, Aug. 2000. [26] W. Cui, “Analysis, design and fabrication of a novel silicon micro-
[2] V. Srikar and S. Senturia, “The reliability of microelectromechanical phone,” Ph.D. dissertation, Mech. Eng. Dept., State Univ. New York at
systems (MEMS) in shock environments,” J. Microelectromech. Syst., Binghamton, Vestal, NY, 2005.
vol. 11, no. 3, pp. 206–214, Jun. 2002. [27] JEDEC Mechanical Shock, Standard JESD22-B104-B, 2001.
ALSALEEM et al.: STUDY FOR THE EFFECT OF THE PCB MOTION ON THE DYNAMICS OF MEMS DEVICES 609

[28] D. Young, M. V. Barton, and Y. C. Fung, “Shock spectra for nonlinear Mohammad I. Younis received the B.S. degree in
spring–mass systems and their applications to design,” AIAA J., vol. 1, mechanical engineering from Jordan University of
no. 7, pp. 1597–1602, 1963. Science and Technology, Irbid, Jordan, in 1999, and
[29] M. I. Younis, E. M. Abdel-Rahman, and A. H. Nayfeh, “A reduced- the M.S. and Ph.D. degrees in engineering mechanics
order model for electrically actuated microbeam-based MEMS,” from Virginia Polytechnic Institute and State Univer-
J. Microelectromech. Syst., vol. 12, no. 5, pp. 672–680, Oct. 2003. sity, Blacksburg, in 2001 and 2004, respectively.
Since then, he has been an Assistant Professor
with the Department of Mechanical Engineering,
State University of New York, Binghamton, where
he is currently the Director of the MEMS Motion
and Characterization Laboratory. He serves as an
Associate Editor for Mathematical Problems in Engineering. He is the holder
of a U.S. patent on a MEMS switch triggered by shock. His research interests
are in the area of dynamics and vibration of MEMS and NEMS.
Fadi M. Alsaleem received the B.S. degree in Dr. Younis is a member of the American Society of Mechanical Engineers.
mechatronics engineering in 2003 from Hashemite
University, Zarqa, Jordan, a graduate certificate in
mechatronics engineering from the American Uni-
versity of Sharjah, Sharjah, United Arab Emirates,
and the M.S. degree in mechanical engineering, with Mahmoud I. Ibrahim received the B.S. degree
a thesis on the reliability of MEMS devices, in 2007 in mechanical engineering in 2007 from the State
from the State University of New York, Binghamton, University of New York, Binghamton, where he is
where he is currently working toward the Ph.D. de- currently working toward the M.S. degree in the De-
gree in the Department of Mechanical Engineering. partment of Mechanical Engineering. His M.S. thesis
His research is focused on utilizing the nonlinear includes work on the reliability of MEMS devices
dynamic response of MEMS devices to design new mass and acceleration under mechanical shock, electrostatic actuation, and
sensors. squeeze-film damping.
Mr. Alsaleem is a Student Member of the American Society of Mechanical Mr. Ibrahim is a Student Member of the American
Engineers. Society of Mechanical Engineers.

You might also like