Micromachined Inertial Sensors

NAVID YAZDI, FARROKH AYAZI, AND KHALIL NAJAFI, SENIOR MEMBER, IEEE

Invited Paper

This paper presents a review of silicon micromachined ac- industrial applications such as robotics and machine and
celerometers and gyroscopes. Following a brief introduction to vibration monitoring; in many other applications, such as
their operating principles and specifications, various device struc- tracking and monitoring mechanical shock and vibration
tures, fabrication technologies, device designs, packaging, and
interface electronics issues, along with the present status in the during transportation and handling of a variety of equipment
commercialization of micromachined inertial sensors, are dis- and goods; and in several military applications, including
cussed. Inertial sensors have seen a steady improvement in their impact and void detection and safing and arming in mis-
performance, and today, microaccelerometers can resolve acceler- siles and other ordnance. High-sensitivity accelerometers
ations in the micro-g range, while the performance of gyroscopes
has improved by a factor of 102 every two years during the past
are crucial components in self-contained navigation and
eight years. This impressive drive to higher performance, lower guidance systems, seismometry for oil exploration and
cost, greater functionality, higher levels of integration, and higher earthquake prediction, and microgravity measurements and
volume will continue as new fabrication, circuit, and packaging platform stabilization in space. The impact of low-cost,
techniques are developed to meet the ever increasing demand for small, high-performance, micromachined accelerometers in
inertial sensors.
these applications is not just limited to reducing overall
Keywords— Accelerometer, gyroscope, inertial sensors, micro- size, cost, and weight. It opens up new market opportunities
fabrication technologies, micromachined sensors, micromachin-
ing, rate sensor, silicon sensors. such as personal navigators for consumer applications, or
it enhances the overall accuracy and performance of the
systems by making formation of large arrays of devices
I. INTRODUCTION feasible.
Micromachined inertial sensors, consisting of accelerom- Micromachined gyroscopes for measuring rate or angle
eters and gyroscopes, are one of the most important types of rotation have also attracted a lot of attention during
of silicon-based sensors. Microaccelerometers alone have the past few years for several applications. They can
the second largest sales volume after pressure sensors, be used either as a low-cost miniature companion with
and it is believed that gyroscopes will soon be mass micromachined accelerometers to provide heading infor-
produced at similar volumes. The large volume demand mation for inertial navigation purposes or in other areas
for accelerometers is due to their automotive applications, [1], including automotive applications for ride stabilization
where they are used to activate safety systems, including air and rollover detection; some consumer electronic applica-
bags, to implement vehicle stability systems and electronic tions, such as video-camera stabilization, virtual reality,
suspension. However, the application of accelerometers and inertial mouse for computers; robotics applications;
covers a much broader spectrum where their small size and a wide range of military applications. Conventional
and low cost have even a larger impact. They are used in rotating wheel as well as precision fiber-optic and ring laser
biomedical applications for activity monitoring; in numer- gyroscopes are all too expensive and too large for use in
ous consumer applications, such as active stabilization of most emerging applications. Micromachining can shrink the
picture in camcorders, head-mounted displays and virtual sensor size by orders of magnitude, reduce the fabrication
reality, three-dimensional mouse, and sport equipment; in cost significantly, and allow the electronics to be integrated
on the same silicon chip.
Manuscript received March 9, 1998; revised April 29, 1998. The work This paper will review both micromachined accelerome-
at the University of Michigan on inertial sensors was supported by the ters and gyroscopes by providing an introduction to their ba-
Defense Advanced Research Project Agency under Contracts JFBI 92-
149 and DABT63-C-0111 and by the National Science Foundation under sic operation, reviewing different types of devices reported
NSF-NYI Grant ECS-925-7400. in the literature and their performance, discussing important
The authors are with the Center for Integrated Sensors and Circuits, issues in their design and operation, and presenting a
Department of Electrical Engineering and Computer Science, University
of Michigan, Ann Arbor, MI 48109-2122 USA. summary and conclusion along with a discussion on the
Publisher Item Identifier S 0018-9219(98)05106-8. future trends in this important category of micromachined

0018–9219/98$10.00  1998 IEEE

1640 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

 acceleration (TNEA) m/s ) is [2]

TNEA (3)

where is the Boltzmann constant and is the temper-

ature in kelvin. Equation (3) clearly shows that to reduce
Fig. 1. General accelerometer structure and its mechanical
mechanical noise, the quality factor and proof mass have
lumped model. to be increased.
In the most general case, the proof-mass motion can have
six degrees of freedom. But typically in a unidirectional
sensors. Due to space constraints, interface electronics accelerometer, the geometrical design of the suspension is
and packaging issues for inertial sensors are only briefly such that one of these is dominant [3] and the device has
described. Also, other topics such as temperature compen- low off-axis sensitivity. The cantilever support has been
sation, calibration, and self-testing are not discussed in this one of the early popular suspension support designs [4],
paper. [5] due to its simplicity, lower spring constant, and internal
stress relief of the beams. However, this configuration
II. MICROMACHINED ACCELEROMETERS results in a larger off-axis sensitivity unless the device
is fully symmetric. Also, symmetric full-bridge supports
A. Structure, Operation, and Specifications result in a very low off-axis sensitivity [6]–[8], and by
using a crab-leg or folded-beam configuration [3] in a
An accelerometer generally consists of a proof mass
full-bridge support, the residual stress of the beams can
suspended by compliant beams anchored to a fixed frame.
also be relieved. The general design of accelerometers
The proof mass has a mass of , the suspension beams
can be performed using the above equations, as well as
have an effective spring constant of , and there is a
mechanical relations describing the spring constant [9],
damping factor ( ) affecting the dynamic movement of the
[10] and damping factor as a function of device geometry
mass. The accelerometer can be modeled by a second-order
and ambient pressure [11], [12]. Further, the device first-
mass-damper-spring system, as shown in Fig. 1. External
order design optimization can be obtained using the same
acceleration displaces the support frame relative to the
equations [4], [5], while the final accelerometer design can
proof mass, which in turn changes the internal stress in
be simulated and optimized using commercially available
the suspension spring. Both this relative displacement and
finite element method or dedicated microelectromechanical
the suspension-beam stress can be used as a measure of the
systems (MEMS) software packages [13].
external acceleration.
Accelerometers are typically specified by their sensitivity,
By using Newton’s second law and the accelerometer
maximum operation range, frequency response, resolution,
model, the mechanical transfer function can be obtained
full-scale nonlinearity, offset, off-axis sensitivity, and shock
survival. Since micromachined accelerometers are used in
(1) a wide range of applications, their required specifications
are also application dependent and cover a rather broad
spectrum. For instance, for microgravity measurements
where is the external acceleration, is the proof mass devices with a range of operation greater than 0.1 g,
displacement, is the natural resonance a resolution of less than 1 g in a frequency range of
frequency, and is the quality factor. The zero frequency to 1 Hz are desired, while in ballistic and
static sensitivity of the accelerometer is shown to be impact sensing applications, a range of over 10 000 g with
a resolution of less than 1 g in a 50 kHz bandwidth is re-
quired. Table 1 summarizes typical performance parameters
(2)
of accelerometers with medium resolution for automotive
applications and high performance for inertial navigation
As evident, the resonance frequency of the structure applications.
can be increased by increasing the spring constant and
decreasing the proof mass, while the quality factor of
the device can be increased by reducing damping and by B. Device Types
increasing proof mass and spring constant. Last, the static A variety of transduction mechanisms have been used in
response of the device can be improved by reducing its microaccelerometers. Some of the more relevant and useful
resonant frequency. approaches will be reviewed here.
The primary mechanical noise source for the device is 1) Piezoresistive Devices: The first micromachined [4],
due to Brownian motion of the gas molecules surrounding and one of the first commercialized, microaccelerometers
the proof mass and the Brownian motion of the proof- [14] were piezoresistive. These accelerometers incorporate
mass suspension or anchors. The total noise equivalent silicon piezoresistors in their suspension beam. As the

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1641

Table 1 Typical Specifications of Accelerometers for 2) Capacitive Devices: In the presence of external
Automotive and Inertial Navigation Applications acceleration, the support frame of an accelerometer moves
from its rest position, thus changing the capacitance
between the proof mass and a fixed conductive electrode
separated from it with a narrow gap. This capacitance can
be measured using electronic circuitry. Silicon capacitive
accelerometers have several advantages that make them
very attractive for numerous applications ranging from
low-cost, large-volume automotive accelerometers [19],
[20] to high-precision inertial-grade microgravity devices
[21]–[26]. They have high sensitivity, good dc response
and noise performance, low drift, low temperature
sensitivity, low-power dissipation, and a simple structure.
However, capacitive accelerometers can be susceptible to
electromagnetic interference (EMI), as their sense node has
high impedance. This issue can be addressed by proper
packaging and by shielding the accelerometer and its
support frame moves relative to the proof mass, the suspen- interface circuit.
sion beams will elongate or shorten, which changes their Some of the most widely used structures for capaci-
stress profile and hence the resistivity of their embedded tive accelerometers are vertical and lateral structures, as
piezoresistors. These piezoresistors are generally placed at shown in Fig. 2. Many capacitive accelerometers utilize
the edge of the support rim and proof mass, where the the vertical structure, where the proof mass is separated
stress variation is maximum. Therefore, a resistive half- by a narrow air gap from a fixed plate, forming a parallel
bridge or full bridge can be formed by employing two or plate sense capacitance [7], [8], [20]–[27]. In these devices,
four piezoresistors. the proof mass moves in the direction perpendicular to
The main advantage of piezoresistive accelerometers is its plane ( -axis) and changes the air gap. In a lateral
the simplicity of their structure and fabrication process, as accelerometer, a number of moving sense fingers are at-
well as their readout circuitry, since the resistive bridge tached to the proof mass, and the sense capacitance is
generates a low output-impedance voltage. However, formed between these and the fixed fingers parallel to
piezoresistive accelerometers have larger temperature them. The sense direction in lateral accelerometers is in the
sensitivity, and smaller overall sensitivity compared to proof-mass plane ( – directions) [19], [29]–[31]. Some
capacitive devices, and hence a larger proof mass is designs use a “see-saw” structure, shown in Fig. 3, where
preferred for them. The early development of bulk- a proof mass is suspended by torsional beams so that
micromachining technology, and the experience gained one side is heavier than the other side and in response
by several companies in the development and commer- to acceleration in the -axis, the proof mass moves out
cialization of piezoresistive pressure sensors, helped the of its plane [32]–[34]. The advantages of this structure
initial development of piezoresistive microaccelerometers over conventional parallel-plate -axis devices are built-in
using bulk-micromachining and wafer-bonding technology overrange protection, larger sensitivity, and higher pull-in
[4], [14]–[18]. The device reported in [4] uses a silicon voltage [34].
middle wafer to form the proof mass and the beams, The open-loop sensitivity of a capacitive accelerometer
while two glass wafer caps are bonded on top and is proportional to the proof-mass size and capacitance
bottom to cover the structure and provide shock stop and overlap area and inversely proportional to the spring con-
damping. Other devices [14], [15] use a lower bonded stant and air gap squared. Early micromachined capacitive
glass base and a silicon overhang on the top for shock accelerometers [7], [21]–[23] utilized bulk silicon micro-
stop, which is formed by silicon fusion bonding and machining and wafer bonding to achieve a thick, large
etch back. Recently, monolithic implementation of a proof mass and high sensitivity. One of the first reported
piezoresistive microaccelerometer with its interface com- devices [21] used a silicon middle wafer anodically bonded
plementary metal–oxide–semiconductor (CMOS) circuitry to two glass wafers on top and bottom to form a -axis
was demonstrated [17], [18], which is based on using accelerometer. The device had two differential sense ca-
a slight modification of a standard CMOS process to pacitors, with the proof mass forming the middle electrode
implement the accelerometer readout and temperature and metal on the glass wafers forming the top/bottom
compensation circuitry and bulk etching of the silicon fixed electrodes. The air gap was formed by recessing
wafer from backside to form the device structure. Self- the silicon or glass wafers. This device with a proof-mass
testing of the piezoresistive accelerometers is possible by size of 4.6 mgr and air gap of 2 m provided g-level
using thermal actuation [15] or electrostatic forces [16]. performance. The second generation of this device [22] had
The sensitivity of all these devices is typically around a resolution of better than 1 g/ Hz in a bandwidth of
1–2 mV/g in a 20–50-g range with an uncompensated zero frequency to 100 Hz, with a temperature coefficient
temperature coefficient of sensitivity (TCS) of 0.2%/ C. of offset (TCO) of 30 g/ C and TCS of 150 ppm/ C. To

1642 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

 (a)

Fig. 4. SEM view of Analog Device’s ADXL05

surface-micromachined polysilicon accelerometer [39].

(b) pull-in voltage, and lower damping. The device consists of

Fig. 2. Structure of (a) vertical and (b) lateral capacitive ac- a 12- m-thick boron-doped silicon inertial mass suspended
celerometers [69]. 7.5 m above the glass substrate by two narrow, high-
aspect-ratio, 12 3 m torsion beams. A large number
of 300- m-long capacitive sense fingers are attached to
the end of the proof mass and are separated by a 2 m
air gap from fixed fingers anchored to the substrate. The
accelerometer provides a sensitivity of 20 fF/g over a range
of 4 g and a bandwidth of 30 Hz in atmosphere.
Surface-micromachined accelerometers [19], [20], [29],
[30], [36]–[40] offer the opportunity to integrate the sensor
and interface circuitry on a single chip. These devices uti-
lize deposited polysilicon layers to form the sense element
and are well suited for both vertical and lateral capacitive
accelerometers. Integration of the interface circuitry with
the sensor enables detection of very small capacitance
variation ( 1aF [38]), while the accelerometer and all its
interface electronics can be implemented in a small area.
Fig. 4 shows a scanning electron microscope (SEM) view
of the ADXL05, a polysilicon accelerometer developed and
Fig. 3. A z -axis capacitive torsional accelerometer [34]. in production by Analog Devices, which has 0.5 mg/ Hz
noise floor with a range of 5 g and shock survivability
reduce its temperature sensitivity and long-term drift, the of 1000 g [39]. Also, by employing a vertical and two
later generation of this device was fabricated using three lateral accelerometers, an integrated three-axis accelerom-
silicon wafers [24], [35]. Another significant early design etry system has been designed by researchers at Berke-
with g performance was fabricated using glass–silicon ley and fabricated through Sandia National Laboratory’s
bonding and bulk micromachining and utilized a closed- process [38]. The same group has also developed a three-
loop readout and control circuit to achieve a 120 dB axis accelerometer with a single sense element, as shown
dynamic range [23]. in [40], (Fig. 5). Typically, surface microaccelerometers
A number of capacitive microaccelerometers with achieve several 100 g resolution in 100 Hz bandwidth
medium resolution have been fabricated using the bulk [37]–[40]. All of these devices have small proof mass and
silicon dissolved wafer process [27], [33], [34]. Among hence high mechanical noise (unless the device is vacuum
these, the device reported in [33] is in high-volume packaged).
production by Ford. The -axis accelerometer presented While bulk-micromachined devices can attain higher res-
in [27] incorporates damping holes in the proof mass to olution due to their large proof mass, they generally require
control damping and uses a second silicon wafer, bonded wafer bonding. These devices, if not formed using only
on top, to provide overrange protection. The torsional silicon wafers, could have large temperature coefficients.
accelerometer shown in [34, Fig. 3] is fabricated using Moreover, forming damping holes in their thick structural
a three-mask dissolved wafer process. It uses multiple layers is not easy, and they typically require packaging
interdigitated fingers and a varying overlap area method at a specified pressure to control the damping [35]. To
of capacitance change to achieve higher linearity, larger address these shortcomings, an all-silicon, fully symmetric,

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1643

 constant tunneling current. Acceleration can be measured
by reading out the bottom deflection voltage in this closed-
loop system. Tunneling accelerometers can achieve very
high sensitivity with a small size since the tunneling current
is highly sensitive to displacement, typically changing by
a factor of two for each angstrom of displacement [48].
However, these devices have larger low-frequency noise
levels [49].
Micromachined tunneling accelerometers were first in-
troduced by researchers with the Jet Propulsion Laboratory
(JPL), Pasadena, CA [41]–[43]. One of their more recent
devices provides a bandwidth of several kilohertz, with a
noise floor of less than 10 g/ Hz in the 10–200-Hz
frequency range [43]. However, a high supply voltage (tens
to hundreds of volts) is required for these devices, thus lim-
iting their application. Using the dissolved wafer process,
a low-voltage micromachined tunneling accelerometer has
been developed [44], [46], [47]. This accelerometer has
a noise spectral density of the sensor-circuit module of
4 mg/ Hz (at 0.5 Hz) and 0.1 mg/ Hz (at 2.5 kHz)
Fig. 5. Photograph of a monolithic three-axis polysilicon surface and has a minimum detectable acceleration of 8 mg in
micromachined accelerometer with integrated sigma–delta readout a 2.5 kHz bandwidth [47]. In continuous operation over
and control circuit [40]. 720 h, the accelerometer shows an offset and sensitivity
variation of 0.5% [47]. Since these sensors are operated in
high-precision accelerometer has been developed [26], as a closed-loop mode, and since the operating current levels
shown in Fig. 6. This device uses a combined surface are typically very small, the drift of the tunneling barrier
and bulk micromachining process to obtain a large proof usually does not have a large impact on the performance
mass, controllable/small damping, and a small air gap of the sensor [46].
for large capacitance variation—all by using a single sil- 4) Resonant Devices: The main advantage of resonant
icon wafer. Sense electrodes are created by depositing sensors is their direct digital output. The first resonant ac-
polysilicon on the wafer. These electrodes, while thin, celerometers were fabricated using quartz micromachining
are made very stiff by embedding thick vertical stiffeners [50], [51]. Silicon resonant accelerometers are generally
(using a trench refill technology) in them so that force based on transferring the proof-mass inertial force to ax-
rebalancing of the proof mass becomes possible. There ial force on the resonant beams and hence shifting their
are eight suspension beams, which are symmetric with frequency [52]. To cancel device thermal mismatches and
respect to the proof-mass centerline and result in low nonlinearities, a differential matched resonator configu-
cross-axis sensitivity. An SEM view is shown in Fig. 6(b). ration can be used [53]. Recently, two high-sensitivity
This device is operated closed loop using an oversampled resonant accelerometers have been reported [54], [55].
modulator. It achieves a high sensitivity of 2 pF/g The devices use wafer-thick proof mass and achieve high
with a full-bridge support, a low noise level, and low resolution (700 Hz/g with 524 kHz center frequency [55])
temperature sensitivity, which enable it to attain g and and very good stability (2 g in more than several days
sub- g performance. [54]). However, these devices typically have small band-
3) Tunneling Devices: Some high-resolution physical width (less than a few hertz). Also recently, surface-
sensors, including microaccelerometers [41]–[47], use micromachined resonant accelerometers are developed [56],
a constant tunneling current between one tunneling [57]. The resonator reported in [57] consists of parallel
tip (attached to a movable microstructure) and its beams, and its operation is based on rigidity change of the
counterelectrode to sense displacement. Fig. 7 [44] shows resonator due to its cross-sectional shape change, which is
the general operating principal of a micromachined induced by the external acceleration. In this manner, the
tunneling accelerometer. As the tip is brought sufficiently device is expected to achieve high sensitivity of 10%/g
close to its counter-electrode (within a few angstroms) [57].
using electrostatic force generated by the bottom deflection 5) Thermal Devices: Another class of accelerometers is
electrode, a tunneling current ( ) is established and based on thermal transduction. One of the first thermal
remains constant if the tunneling voltage ( ) and dis- accelerometers used the principle that the temperature flux
tance between the tip and counterelectrode are unchanged. from a heater to a heat-sink plate is inversely propor-
Once the proof mass is displaced due to acceleration, tional to their separation [58]. Hence, by measuring the
the readout circuit responds to the change of current and temperature using thermopiles, the change in separation
adjusts the bottom deflection voltage to move the proof between the plates (which is representative of acceleration)
mass back to its original position, thus maintaining a can be measured. Devices with a moving thermopile array

1644 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

 (a)

(b)

Fig. 6. (a) g z -axis microaccelerometer structure. (b) SEM of a fabricated device [26].

and fixed heater, and vice versa, can be fabricated [58]. 6) Other Devices: In addition to the aforementioned de-
Recently, a novel thermal accelerometer was reported that vice types, accelerometers also use many other princi-
does not have any moving mechanical parts. Its operation ples, including optical [60], [61], electromagnetic [62],
is based on free-convection heat transfer of a small hot and piezoelectric [63], [64]. The motivation for the de-
air bubble in a sealed chamber [59]. The device consists velopment of optical accelerometers has been combining
of a thermally isolated heater that forms a hot air bubble. optics and silicon micromachining to exploit advantages
The heat distribution of this bubble changes in the presence of both, as well as achieving miniature devices with very
of an acceleration and becomes asymmetric with respect high EMI noise immunity [60] or good linearity [61].
to the heater. This heat profile can be sensed by two The electromagnetic accelerometer reported in [62] utilizes
symmetrically placed temperature sensors and is a measure two coils, one on top of the proof mass and the other
of the acceleration. The initial prototypes achieved a 0.6 separated by an air gap at the bottom, where the proof-
mg sensitivity, and future devices are expected to achieve mass displacement changes the mutual inductance of the
sub- g performance [59]. two coils. By using a simple readout circuit with a 2.5

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1645

 tilayer ceramic hermetic packages [72] have been utilized
to house the sensor and its interface circuit. The overall
packaging cost can be reduced, and the performance can be
improved, by packaging the sensor/circuit at the wafer level
using capping glass or silicon wafers that are bonded to the
device wafer, then using plastic injection molding for the
final external package, as demonstrated by Motorola [73]
and Ford [74].
Sensor readout and signal-processing circuit design and
development is also an important aspect of micromachined
accelerometers. In a piezoresistive accelerometer, the sensor
piezoresistors form a half-bridge or full-bridge, and the
Fig. 7. Structure of a tunneling microaccelerometer [44].
output voltage of the bridge can be directly amplified
by the interface circuit. Tunneling sensors also require
mW power dissipation, a very linear response over a 50 g simple readout circuitry that convert the tunneling current
range was achieved [62]. Piezoelectric materials, mainly to voltage, buffer and amplify this voltage, and feed it
ZnO, have also been used in accelerometers to directly back to the sensor [47]. Resonant sensors are operated in
convert the force affecting the proof mass to an electrical closed loop, and their output is their frequency, which can
signal [63], [64]. The piezoelectric charge generated by be measured by a digital counter. Their interface circuit
acceleration can be directly coupled to the gate of an MOS consists of sense and drive circuitry. A sustaining amplifier
transistor and amplified [63]. One of the problems with in the loop is needed to compensate for the losses and
piezoelectric materials is their leakage that deteriorates the maintain the resonance [76]. Capacitive readout circuits
dc response of the device. By isolating the piezoelectric mostly use capacitance-to-frequency converters (oscillators)
film from the leakage paths, a more flat response at near [77]–[79], capacitive ac-bridges [30], [36], or switched-
dc frequencies can be obtained [63]. capacitor circuits [24], [35], [37], [38], [80]–[82].
Also, many researchers [65]–[67] have fabricated arrays Accelerometers can be operated open loop or closed loop.
of accelerometer switches for use as a threshold or digital In the open-loop mode, the overall linearity, bandwidth,
accelerometer. The main advantage of these devices is a and dynamic range cannot be better than the parameters of
digital output and simple and low-power interface circuitry the sensor structure by itself. However, open-loop sensors
[67]. require simpler interface circuitry and are inherently stable
In addition to bulk and surface micromachining, electro- in the frequency range below their resonance.
forming and mixed processes [25], [26], [28], [69] have Closed-loop operation improves the overall sensor linear-
been used to improve performance further. Electroforming ity, dynamic range, and bandwidth. The interface circuitry
can produce structures with large proof mass. For instance, in a closed-loop device reads the sensor signal and uses
a lateral accelerometer fabricated by LIGA produces a 1 feedback to maintain the proof mass in a null position
g/ Hz resolution in air and a temperature-compensated in a stable manner. Loop stability can be ensured if the
TCO of 150 ppm/ C [28], [68]. The device presented accelerometer is overdamped with a dominant pole [35]
in [25] uses a combination of thick epigrowth, SIMOX or is provided using a lead compensator in the feedback
formation, and wafer bonding to attain high precision. The path [36], or a lead-lag [23] or lead compensator [38] in
70 m proof mass is formed by two bonded epilayers, while the forward path. Furthermore, the dynamics of the loop
the thin uniform air gap is created by etching the sacrificial can be controlled by adding an electronic pole in the loop
buried oxide layer. A family of trench refill technologies [75]. Closed-loop accelerometers using different schemes,
for MEMS has been developed [69] and applied to fab- including phase locked loop [79], analog [82], and pulse
ricate integrated three-axis accelerometers with a medium width modulation [83], have been developed. Recently,
resolution. The small mass size in surface-micromachined electromechanical, oversampled, sigma–delta modulators
devices [29], [30], [36]–[40] can be somewhat circum- have become more attractive, since they provide direct
vented by using silicon-on-insulator (SOI) wafers, where digital output and force-feedback control of the proof mass
the structure height is defined by the top silicon layer, or simultaneously over a wide dynamic range [23], [24], [33],
by depositing a thick polysilicon layer using an epitaxial [35]–[38].
process [70].
D. Commercialization
C. Packaging and Interface Circuits The application of micromachined accelerometers in au-
The accelerometer package has to protect the sensor tomotive industry has been a major driver for their initial
structure without inducing significant stress or drift. Mount- commercialization. EG&G IC Sensors and Lucas Nova
ing errors or misalignment directly affect the sense direction Sensors were among the first companies in the United
of the device and its overall performance. Also, the package States that commercialized mainly piezoresistive, and later
should not adversely affect the sensor frequency response or capacitive, microaccelerometers [14]–[16], [31]. However,
temperature sensitivity. Both metal can [22], [71] and mul- one of the most successful micromachined accelerometers

1646 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

Fig. 8. Ford’s accelerometer consisting of a sensor chip and a CMOS interface chip [33]. A
low-cost SMD plastic package is used for this hybrid module [74].

in the market has been the ADXL50 by Analog Devices, E. Challenges and Future Trends
which is a surface-micromachined, closed-loop, fully mono- Today, microaccelerometers, with low and medium sensi-
lithic accelerometer integrated with its interface circuitry tivities, are in large volume production with prices under $5.
[71]. This accelerometer has been mainly developed for This is a good indication that their technology is maturing
air-bag deployment and has a range of 50 g with 6.6 and that numerous challenges involved in their design
mg/ Hz noise floor and shock survivability of 2000 and mass production are resolved. Although a number of
g. Motorola has also been mass producing a capacitive high-precision accelerometers have been already developed
accelerometer for air-bag applications [84]. This sensor has [21]–[25], most are not evaluated completely as inertial
a range of 40 g, a sensitivity of 40 mV/g, and a peak grade yet, and they are expensive for several applica-
noise level of 15 mV in 400 Hz bandwidth, and utilizes tions and still need improvement to overcome long-term
a plastic dual in pin integrated circuit (IC) package to stability and temperature-sensitivity problems. Therefore,
reduce cost. Other companies, including Delco and Ford, the development of low-cost, inertial-grade accelerometers
mass produce accelerometers for automotive applications. with sub- g noise levels, good long-term stability, and
Fig. 8 shows Ford’s hybrid packaged accelerometer, which low temperature sensitivity still remains a challenge. The
consists of a sensor chip, fabricated using bulk silicon development of low-stress, low-drift packaging technolo-
micromachining, and a CMOS sigma–delta interface chip gies for inertial-grade devices without affecting device
[33]. This accelerometer is designed for automotive pas- performance and stability is also being pursued. A major
sive restraint systems, has a range of 50 to 30 g challenge is the interface circuit, where low-drift read-
with 400 Hz bandwidth, provides serial digital output out/control circuitry with high sensitivity, low noise level,
with sensitivity of 3 kHz/g, has a self-test feature, and and large dynamic range is needed. It is believed that
is available with a low-cost plastic surface mount device highly stable (inertial-grade) multiaxis devices with sub-
(SMD) package [74]. Endevco produces a large variety of g resolution and over 120 dB dynamic range in several
piezoelectric, piezoresistive, and capacitive accelerometers hundred hertz bandwidth will be developed in the near
with full-scale ranges from 2 to over 4000 g for a wide future.
range of applications, including impact and crash sensing
and vibration analysis [85]. In Europe, silicon accelerom-
III. MICROMACHINED GYROSCOPES
eters are in volume production by Bosch, and in Japan by
Denso.
CSEM in Europe has produced one of the first high- A. Operating Principles and Specifications
precision silicon microaccelerometers [21], [22], [24]. Almost all reported micromachined gyroscopes use vi-
These devices are capacitive bulk micromachined. One brating mechanical elements to sense rotation. They have
of the latest versions, MS6100, is targeted for low-power no rotating parts that require bearings, and hence they can
applications and achieves sub-mg resolution with a TCO be easily miniaturized and batch fabricated using micro-
of 50 g/ C and stability of 1 mg for a 2 g device with machining techniques. All vibratory gyroscopes are based
less than 0.5 mW power dissipation [86]. In the United on the transfer of energy between two vibration modes
States, some companies active in navigation and guidance, of a structure caused by Coriolis acceleration. Coriolis
such as Litton, have also been pursuing development of acceleration, named after the French scientist and engineer
g accelerometers [25]. G. G. de Coriolis (1792–1843), is an apparent acceleration

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1647

 Table 2 Performance Requirements for
Different Classes of Gyroscopes

(a) (b)
Fig. 9. (a) The Coriolis effect. (b) Tuning-fork vibratory gyro- scale range of at least 50 /s and a resolution of about
scope. The tines are differentially driven to a fixed amplitude. 0.1 /s in a bandwidth of 50 Hz, all at a cost of $10–20
Coriolis force is detected either as differential bending of the
tuning-fork tines or as a torsional vibration of the tuning-fork stem. [89]. The operating temperature is in the range from 40
to 85 C. There are also several other applications that
require improved performance, including inertial naviga-
that arises in a rotating reference frame and is proportional
tion, guidance, robotics, and some consumer electronics.
to the rate of rotation. To understand the Coriolis effect,
Today, optical gyroscopes are the most accurate gyroscopes
imagine a particle traveling in space with a velocity vector
available in the market. Among these, ring laser gyroscopes
. An observer sitting on the -axis of the coordinate
have demonstrated inertial-grade performance, while fiber-
system, shown in Fig. 9(a), is watching this particle. If the
optic gyroscopes are mainly used in tactical-grade applica-
coordinate system along with the observer starts rotating
tions. Delco’s hemispherical resonator gyroscope (HRG)
around the -axis with an angular velocity , the observer
is a vibratory gyroscope that has achieved impressive
thinks that the particle is changing its trajectory toward the
inertial-grade performance [90]. Although highly accurate,
-axis with an acceleration equal to . Although no
these devices are too expensive and bulky for many low-
real force has been exerted on the particle, to an observer,
cost applications. Achieving “tactical- and inertial-grade”
attached to the rotating reference frame an apparent force
performance levels has proven to be a tough challenge
has resulted that is directly proportional to the rate of rota-
for micromachined gyroscopes, and new technologies and
tion. This effect is the basic operating principle underlying
approaches are being developed. Because of their greater
all vibratory structure gyroscopes.
compatibility with batch fabrication technologies, this paper
Resolution, drift, zero-rate output (ZRO), and scale factor
will only review silicon micromachined vibratory gyro-
are important factors that determine the performance of a
scopes.
gyroscope. In the absence of rotation, the output signal
of a gyroscope is a random function that is the sum of
white noise and a slowly varying function [87]. The white B. Review of Micromachined Vibratory Gyroscopes
noise defines the resolution of the sensor and is expressed A number of vibratory gyroscopes have been demon-
in terms of the standard deviation of equivalent rotation strated, including tuning forks [91]–[94], vibrating beams
rate per square root of bandwidth of detection [( /s)/ Hz [95], and vibrating shells [89]. Tuning forks are a clas-
or ( /h)/ Hz]. The so-called “angle random walk” in h sical example of vibratory gyroscopes. The tuning fork,
may be used instead. The peak-to-peak value of the slowly as illustrated in Fig. 9(b), consists of two tines that are
varying function defines the short- or long-term drift of the connected to a junction bar. In operation, the tines are
gyroscope and is usually expressed in /s or /h [87]. Scale differentially resonated to a fixed amplitude, and when
factor is defined as the amount of change in the output rotated, Coriolis force causes a differential sinusoidal force
signal per unit change of rotation rate and is expressed in to develop on the individual tines, orthogonal to the main
V/( /s). Last, an important factor for any gyroscope that is vibration. This force is detected either as differential bend-
primarily defined by device imbalances is the ZRO, which ing of the tuning fork tines or as a torsional vibration
represents the output of the device in the absence of a of the tuning fork stem. The actuation mechanisms used
rotation rate. for driving the vibrating structure into resonance are pri-
In general, gyroscopes can be classified into three differ- marily electrostatic, electromagnetic, or piezoelectric. To
ent categories based on their performance: inertial-grade, sense the Coriolis-induced vibrations in the second mode,
tactical-grade, and rate-grade devices. Table 2 summarizes capacitive, piezoresistive, or piezoelectric detection mech-
the requirements for each of these categories [87], [88]. anisms can be used. Optical detection is also feasible,
Over the past few years, much of the effort in developing but it is too expensive to implement. In general, silicon
micromachined silicon gyroscopes has concentrated on micromachining processes for fabrication of vibratory gy-
“rate-grade” devices, primarily because of their use in roscopes fall into one of four categories: 1) silicon bulk
automotive applications. This application requires a full- micromachining and wafer bonding; 2) polysilicon surface

1648 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

Fig. 10. Draper’s first silicon micromachined double-gimbal vi- Fig. 11. SEM view of Draper’s single-crystal silicon-on-glass
bratory gyroscope (1991) [100]. tuning-fork gyroscope [103].

micromachining; 3) metal electroforming and LIGA; and 4) which is capacitively monitored. Fig. 11 shows an SEM
combined bulk-surface micromachining or so-called mixed view of the device with a perforated mass to minimize
processes. damping. The in-plane motion of the structure is lightly
Piezoelectric vibratory gyroscopes were demonstrated in damped by air, while out-of-plane motion is strongly
the early 1980’s. Examples of these devices are fused- damped due to squeeze film effects. Therefore, for out-
quartz HRG by Delco [90], quartz tuning forks [96] like of-plane modes, rises rapidly as pressure is reduced, in
the Quartz Rate Sensor by Systron Donner [97], [98], and contrast to the in-plane , which shows a small increase
a piezoelectric vibrating disc gyro [99]. Although quartz as the pressure drops. At pressures of 100 mTorr, a
vibratory gyroscopes can yield very high quality factors of 40 000 was observed for the drive mode and 5000 for
at atmospheric pressure with improved level of perfor- the sense mode. The silicon-on-glass technology used in
mance, their batch processing is not compatible with IC this device has the advantage of low stray capacitance.
fabrication technology. In the late 1980’s, after successful The noise equivalent rate observed by this structure was
demonstration of batch-fabricated silicon accelerometers, 470 /h in a 60 Hz bandwidth, equivalent to 0.02 /s in a 1
some efforts were initiated to replace quartz with silicon Hz bandwidth or angle random walk of 0.72 h [103].
in micromachined vibratory gyroscopes. The Charles Stark The scale-factor accuracy was better than 0.1%, and bias
Draper Laboratory demonstrated one of the first batch- stability was 55 /h overnight. The projected performance
fabricated silicon micromachined rate gyroscopes in 1991. was 10–100 /h for bias stability and resolution in a 60 Hz
This bulk silicon device was a double gimbal vibratory bandwidth.
gyroscope supported by torsional flexures, with the vibrat- If the sense and drive resonant modes of a tuning fork
ing mechanical element made from p silicon [100]. have equal frequencies, the output signal will be amplified
As illustrated in Fig. 10, the outer gimbal was electro- by the quality factor of the sense mode, resulting in
statically driven at a constant amplitude using the drive much higher sensitivity. However, this involves extreme
electrodes, and this oscillatory motion was transferred to control of device dimensions and may lead to temperature
the inner gimbal along the stiff axis of the inner flex- drift problems if these natural frequencies do not track
ures. When exposed to a rotation normal to the plane with temperature. Because of these difficulties, most tuning-
of the device, Coriolis force causes the inner gimbal to fork designs are not based on matched vibration mode
oscillate about its weak axis with a frequency equal to frequencies.
the drive frequency. Therefore, maximum resolution is Other tuning-fork designs have used electromagnetic
obtained when the outer gimbal is driven at the resonant excitation to obtain a large amplitude of motion [92],
frequency of the inner gimbal, causing the sensitivity to [93], [104]. Bosch’s silicon yaw rate sensor [93] achieves
be amplified by the mechanical quality factor of the sense vibration amplitudes as large as 50 m using a permanent
resonance mode of the structure. A rotation rate resolution magnet mounted inside a metal package. This device was
of 4 /s in a 1 Hz bandwidth was realized using this fabricated through a combination of bulk- and surface-
structure. micromachining processes, and it consists of two bulk-
Later in 1993, Draper reported an improved 1 mm micromachined oscillating masses, each of which supports
silicon-on-glass tuning fork gyroscope [91] fabricated two surface-micromachined accelerometers for detection of
through the dissolved wafer process [101]. This gyroscope Coriolis force. The sensor chip is anodically bonded to a
was electrostatically vibrated in its plane using a set supporting glass wafer and is covered by another silicon
of interdigitated comb drives [102] to achieve a large cap wafer. Operating at atmospheric pressure, the device
amplitude of motion (10 m). Any rotation in the plane has shown a resolution of 0.3 /s in a 100 Hz bandwidth,
of the substrate perpendicular to the drive mode will then thanks to its large amplitude of vibration. Although such
excite the out-of-plane rocking mode of the structure, a large amplitude of oscillation (50 m) can increase the

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1649

output signal level, it increases the total power consumption
and may cause fatigue problems over long-term operation.
Cross talk between the sense and drive modes was mini-
mized through mechanical decoupling of these modes by
separating the oscillator and sense proof masses, resulting
in a stable ZRO.
Piezoresistive detection has also been used in some gy-
roscope designs. Daimler Benz has demonstrated a tuning-
fork angular rate sensor for automotive applications that
piezoresistively measures the rotation-induced shear stress
in the stem of the tuning-fork device [94]. In this de-
vice, a piezoelectric actuation mechanism was used by
depositing a piezoelectric aluminum nitride (AlN) thin-
film layer on one of the tines. The use of piezoelectric
thin films such as AlN and ZnO on silicon degrades Fig. 12. SEM view of a comb-driven polysilicon surface micro-
and causes large temperature variation of offset and machined z -axis vibratory gyroscope [113].
sensitivity [105]. This device was fabricated through a
combination of bulk micromachining and bonding of SOI
electrostatically driven into resonance using comb drives.
wafers. Researchers at the University of Neuchatel, Switzer-
Any deflections that result from Coriolis acceleration are
land, have demonstrated a tuning-fork design based on two
detected differentially in the sense mode using interdigitated
isolated vibrating proof masses, each supported by a four-
beam bridge-type suspension [104]. These proof masses are comb fingers. This device, 1 mm across, was integrated
electromagnetically vibrated in plane and antiphase, and with a transresistance amplifier on a single die using the
the rotation-induced out-of-plane motion is then detected Analog Devices BiMEMS process. The remaining control
by means of four piezoresistors connected in a Wheatsone and signal-processing electronics were implemented off
bridge configuration, showing a sensitivity of 4 nV/ /s chip. Quadrature error nulling and sense-mode resonant
with excellent linearity up to 750 /s. This device was frequency tuning can be accomplished in this design by
fabricated through silicon bulk micromachining and was applying a control dc bias voltage to the position sense
wafer-level vacuum packaged by anodic bonding of the fingers. The dc bias voltage generates an electrostatic
silicon wafer to encapsulating glass wafers. In general, negative stiffness, which reduces the resonant frequency of
package-induced stress on the sensor structure can be the sense mode. By slightly changing this dc bias voltage on
lowered by low-temperature anodic bonding of glass wafers the differential comb fingers ( V), a lateral electrostatic
with silicon [106]. Although piezoresistive devices are field arises that can be used to align the drive mode
easier to fabricate and require a simpler electronic interface oscillations and reduce the quadrature error. This device
due to their lower output impedance compared to capacitive demonstrated a resolution of 1 /s/ Hz with performance
devices, they have large temperature sensitivity and poor projected to improve to 0.1 /s/ Hz in a second-generation
resolution. device.
Also reported in the literature are capacitive bulk- Samsung has also reported a very similar surface-
micromachined silicon-on-glass vibrating beams [107], micromachined -axis device, shown in Fig. 12, with a
vibrating membranes [108], and double-gimbaled structures 7- m-thick polysilicon resonating mass supported by four
[109]. Since the Young’s modulus of single-crystal silicon fishhook-shaped springs [113]. This device, though not
changes with crystallographic orientation, symmetric integrated with electronics, has demonstrated a resolution
vibrating structures made of single-crystal silicon may of 0.1 /s at 2 Hz, an operating bandwidth of 100 Hz, and
show excessive mechanical coupling between drive and a linearity of 1% full scale in a range of 90 /s. Hybrid
sense modes (due to this anisotropy), resulting in a large attachment of the sensor chip to a CMOS application-
ZRO with unacceptable drift characteristics [107]. specific integrated circuit (ASIC) chip used for readout
Surface-micromachined vibratory gyroscopes have also and closed-loop operation of the gyro was done in a
been demonstrated. Some have been integrated with the vacuum-packaged ceramic case.
readout electronic circuitry on a single silicon chip, Murata has presented a surface-micromachined polysil-
reducing parasitic capacitances and hence increasing the icon gyroscope that is sensitive to lateral ( - or -axis)
signal-to-noise ratio. In addition, the vibrating structure angular rate [115]. The sense electrode was made un-
is made of polysilicon, which has a high quality factor derneath the perforated polysilicon resonator by diffusing
and an orientation-independent Young’s modulus. Single- phosphorous into the silicon substrate (junction isolation).
and dual-axis polysilicon surface-micromachined gyro- This device showed an open-loop noise-equivalent rate
scopes have been realized by researchers at Berkeley of 2 /s/ Hz. The junction-isolation scheme used in this
[110], [111] and Samsung [112]–[114]. Berkeley’s - device, although simple, has the disadvantage of relatively
axis vibratory rate gyroscope [110] resembles a vibrating large parasitic capacitance and large amount of shot noise
beam design and consists of an oscillating mass that is associated with the existing pn junction, which in turn

1650 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

Fig. 13. Closeup die shot of Berkeley’s dual-axis rate gyroscope,
integrated with sense and drive electronics using Analog Devices,
Inc., surface-micromachining process [116].

degrade the resolution. Later in 1997, Samsung reported Fig. 14. SEM of HSG-IMIT’s surface-micromachined x-axis vi-
a similar device that used a 3000-Å-thick polysilicon sense bratory gyroscope (MARS-RR) fabricated through Bosch’s process
with a 10-m-thick structural polysilicon layer [120].
electrode underneath a 7.5- m-thick low-pressure chemical
vapor deposition (LPCVD) polysilicon resonating mass
[112]. Since the detection mode is highly damped by standard Bosch foundry process featuring a 10- m-thick
squeeze film damping, these devices have to operate under structural polysilicon layer in addition to the buried
vacuum. Samsung’s device, vacuum packaged in an Al O polysilicon layer, which defines the sense electrodes. The
case, showed an improved open-loop noise-equivalent rate reported performance of this device is quite impressive: a
of 0.1 /s/ Hz with a good linearity up to 100 /s. random angle walk of 0.27 h, a bias stability of /h,
Berkeley has reported a surface-micromachined dual-axis and a scale-factor nonlinearity of 0.2%.
gyroscope based on rotational resonance of a 2- m-thick Researchers at General Motors and the University of
polysilicon rotor disk, as shown in [116] (Fig. 13). Since Michigan have developed a vibrating ring gyroscope [89],
the disk is symmetric in two orthogonal axes, the sensor can schematically shown in Fig. 15. This device consists of a
sense rotation equally about these two axes. This device, ring, semicircular support springs, and drive, sense, and
integrated with electronics, yielded a random walk as balance electrodes, which are located around the structure.
low as 10 h with cross-axis sensitivity ranging 3–16%. Symmetry considerations require at least eight springs to
Resolution can be further improved to 2 h by frequency result in a balanced device with two identical flexural
matching at the cost of excessive cross-axis sensitivity. Also modes that have equal natural frequencies [121]. The ring is
reported in the literature is a cross-shaped nickel-on-glass electrostatically vibrated into an in-plane elliptically shaped
two-axis micromachined gyroscope [117], which has shown primary flexural mode with a fixed amplitude. When it
a rate sensitivity of 0.1 mV/ /s. is subjected to rotation around its normal axis, Coriolis
The JPL, in collaboration with the University of Califor- force causes energy to be transferred from the primary
nia, Los Angeles, has demonstrated a bulk-micromachined, mode to the secondary flexural mode, which is located
precision silicon MEMS vibratory gyroscope for space ap- 45 apart from the primary mode, causing amplitude to
plications [118], [119]. This clover-leaf-shaped gyroscope build up proportionally in the latter mode; this buildup
consists of three major components: a silicon clover-leaf is capacitively monitored. The vibrating ring structure has
vibrating structure; a silicon baseplate, which is bonded to some important features compared to other types of vi-
the clover-leaf structure; and a metal post, which is epoxied bratory gyroscopes. First, the inherent symmetry of the
inside a hole on the silicon resonator. A hermetically sealed structure makes it less sensitive to spurious vibrations.
package, 1 1 0.7 in, houses the microgyroscope and Only when the ring has mass or stiffness asymmetries
most of its control electronics. This packaged gyroscope can environmental vibrations induce a spurious response.
has a 7 Hz split between its drive and sense mode ( Second, since two identical flexural modes of the structure
1.44 kHz), a scale factor of 24 mV/ /s, a bias stability of “with nominally equal resonant frequencies” are used to
70 /h, and an angle random walk of 6.3 h. sense rotation, the sensitivity of the sensor is amplified
Recently, researchers at HSG-IMIT, Germany, have by the quality factor of the structure, resulting in higher
demonstrated and reported a surface-micromachined sensitivity. Third, the vibrating ring is less temperature
precision -axis vibratory gyroscope (MARS-RR) with sensitive since the vibration modes are affected equally
a very small ZRO achieved by mechanical decoupling of by temperature. Last, electronic balancing of the struc-
the drive and sense vibration modes [120]. This device ture is possible. Any frequency mismatch due to mass
(6 mm ), shown in Fig. 14, was fabricated through the or stiffness asymmetries that occurs during the fabrica-

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1651

 Fig. 16. SEM view of a PRG [123]. The polysilicon ring is 1
Fig. 15. Structure of a vibrating ring gyroscope. mm in diameter, 3 m wide, and 35 m tall.

tion process can be electronically compensated by use

of the balancing electrodes that are located around the with a reported root-mean-square noise floor of 0.15 /s in
structure. a 30 Hz bandwidth and an in-run drift of approximately
The first micromachined version of the vibrating ring 0.05 /s [125]. This device was fabricated through deep dry
gyroscope was fabricated by electroforming nickel into a etching of a 100- m-thick silicon wafer, which was then
thick polyimide (or photoresist) mold on a silicon substrate anodically bonded to a glass support wafer.
in a post circuit process [89], [121], [122]. The gyroscope Levitated micromachined spinning-disc gyroscopes have
demonstrated a resolution of 0.5 /s in a 25 Hz bandwidth also been investigated [126], [127]. The concept was based
limited by the readout electronic noise. The sensor was on a rotor disc, levitated using electromagnetic or electro-
integrated with a low-input capacitance source-follower static means and spun at a very high rate by means of
buffer and the amplifier on a silicon chip. The zero bias a motor to produce angular momentum. With additional
drift was 10 /s over the temperature range 40 to 85 C, electrostatic fields, the rotor can be held in equilibrium
and the sensitivity of the device varied by less than 3% even if the sensor is tilted or inverted. It is predicted that
over the same temperature range. Scale-factor nonlinearity spinning microgyroscopes can yield a lower drift than a
in a 100 /s rate range was 0.2% [122]. vibrating structure gyroscope [127]. The performance of
To improve performance further, a new polysilicon ring these devices is yet to be demonstrated.
gyroscope (PRG) [123] was recently fabricated through a
single-wafer, all-silicon, high-aspect-ratio p /polysilicon C. Design Issues and Considerations
trench-refill technology [124] at the University of Michigan. Vibratory gyroscopes can be operated open or closed loop
In this new process, the vibrating ring and support springs to measure rate of rotation (angular velocity). In the open-
are created by refilling deep dry-etched trenches with loop mode, the response to a change in rotation rate is not
polysilicon deposited over a sacrificial LPCVD oxide layer. instantaneous, as time is required for the amplitude of the
Each sense electrode is made from a p silicon island (12 sense mode to reach its steady-state value. With matched
m deep) hanging over an ethylenediamine-pyrocatechol sense and drive resonant modes, the time constant associ-
(EDP)-etched pit. Fig. 16 shows a SEM picture of a 1.7 ated with this amplitude buildup is approximately equal to
1.7 mm PRG. This device provides several impor- [121]. This response time limits the bandwidth of the
tant features required for high-performance gyroscopes, sensor to a few hertz. To obtain larger open-loop bandwidth,
including small ring-to-electrode gap spacing ( 1 m) for gyroscopes are sometimes operated with a slight mismatch
increasing the sense capacitance; large structural height for in the sense and drive mode resonant frequencies; however,
increasing the radius and sense capacitance and reducing the as theoretically shown in [128], this is done at the cost of
resonant frequency; and a better structural material (polysil- reduced sensitivity. In the closed-loop mode of operation,
icon) for increasing with an orientation-independent the sense mode amplitude is continuously monitored and
Young’s modulus. By taking advantage of these features, a driven to zero, and hence the bandwidth and dynamic range
tactical-grade ring microgyroscope with projected random of the sensor can be increased beyond the open-loop values
walk as small as 0.05 / h can be potentially realized, even with matched resonant modes. The bandwidth is then
providing orders of magnitude improvement in perfor- limited by the readout and control electronics and can be
mance. increased to values approaching the resonant frequency of
British Aerospace Systems and Equipment, in collabora- the structure.
tion with Sumitomo Precision Products, has also developed With matched sense and drive resonant modes, assuming
a micromachined single-crystalline silicon ring gyroscope that the readout electronic noise has a white spectrum

1652 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

around the resonant frequency of the structure and that the zero bias error should then be further reduced electronically.
detection circuit has a bandwidth of BW, the minimum For instance, the nickel ring gyroscope can be electronically
detectable electronic signal for a capacitive device can be trimmed to reduce the zero bias from 60 /s to less than
expressed by 0.5 /s.
A high-performance gyroscope should have an accurate
BW (4) scale factor over a wide dynamic range as large as 140
dB. The scale factor should have a small temperature
This equation shows that reducing the noise of the readout sensitivity. Special attention should be paid to the structural
circuit, increasing the Coriolis-induced capacitance change material. Metal structures used on silicon substrates or
of the device, lowering the resonant frequency, increasing piezoelectric layers cemented on silicon structures can
the mechanical quality factor, and minimizing the parasitic cause large variation of scale factor with temperature.
capacitances are determining factors in improving the reso- All-silicon devices are suitable for high-performance appli-
lution. If the closed-loop bandwidth is not extended beyond cations. Variation of resonant frequency over temperature
the open-loop bandwidth ( ), then the resolution in (due to temperature dependence of Young’s modulus) can
the closed-loop mode is also expressed by the theoretical also affect the scale factor of the sensor, and therefore, some
maximum open-loop value (4) [121]. Although the resonant temperature compensation of the scale factor is required
frequency of the structure should be lowered for improved [121].
sensitivity, it must be kept above environmental noise To obtain a high mechanical quality factor, both external
( 2 kHz). Stronger Coriolis forces will be obtained by and internal energy losses should be minimized. Balanced,
increasing the amplitude of vibration in the drive mode symmetrical, isolated structures that do not radiate acoustic
( ), thus improving the resolution. energy out of their anchors are therefore required. In
It should be noted that the fundamental limiting noise balanced symmetrical structures, the dependence of the
component of the mechanical structure is due to the Brown- output signal on rigid body motion of the structure caused
ian motion of the sense vibration mode, and its total noise by linear acceleration can be electronically rejected. An
equivalent acceleration is given by (3). This acceleration example of a balanced system is a tuning-fork design,
noise corresponds to an input rotation rate noise whose made from two symmetrical tines oscillating antiphase,
spectral density is determined using the transfer function which has no net motion at the junction and therefore
(from input rotation rate to the sense mode acceleration) of can be mounted inertially stable at this junction. Energy
each particular device. At high quality factors (operation in losses due to squeeze film damping by air molecules will
vacuum), resolution is usually not limited by Brownian mo- be significantly reduced if the resonant device is operated
tion of the structure; it is limited by the readout electronic in vacuum. This calls for hermetically sealed, low-stress,
noise. robust vacuum-packaging techniques that are capable of
An important performance parameter for a vibratory holding vacuum levels (10 –10 Torr) for extended pe-
gyroscope is its zero rate output (or zero bias). Geometrical riods of time. Silicon or glass wafers [93], [104] bonded to
imperfections in the vibrating mechanical structure and/or the sensor substrate can provide hermetically sealed, chip-
the sense and drive electrodes as well as electrical coupling level encapsulation [129]. Internal energy loss is reduced if
between these electrodes can cause an output signal in the the structure is fabricated from low-loss materials such as
absence of rotation. For instance, in tuning-fork designs, single- and polycrystalline silicon. Another concern about
if due to fabrication flaws the tines’ centers of mass a vibrating mechanical element is the long-term drift and
are not precisely aligned in the plane of vibration, their fatigue problem. Small vibration amplitude can help reduce
inertial forces produce a vibrating torque about the stem these problems.
just as the Coriolis torque does. This error, often called Last, readout, signal-processing, and control electron-
the quadrature error, can be distinguished from the rate ics are needed for open- and closed-loop operation of
signal simply because it is in phase quadrature with the a vibratory gyroscope. Part of the electronics drives the
Coriolis-induced signal. If too large, it may cause errors in first mode of the structure into resonance and keeps the
sensing the rotation rate and can even saturate the amplifier amplitude of vibration constant. Extremely small Coriolis-
[110]. In the case of shell gyroscope designs, asymmetric induced motions of the structure are then detected by the
damping of the structure, which manifests itself as different readout electronics. Low-noise readout techniques require
sense and drive mode quality factors, can also cause a a careful biasing scheme using diodes [89], resistors [130],
finite output when there is no rotation [88]. By electrically and subthreshold MOS devices [110]. Monolithic inte-
and mechanically decoupling the sense and drive modes, gration of the sensor and the readout electronics reduces
and by minimizing the fabrication process errors, ZRO the interconnect parasitics and improves the resolution of
can be significantly reduced [93], [120]. A large quality the sensor. Parasitic capacitances can be further reduced
factor and lower natural frequency can further reduce the by special bootstrapping techniques [121], [130]. Delco
ZRO, and its drift simply due to the fact that these two Electronics has demonstrated a CMOS ASIC chip for
factors increase the open-loop sensitivity of the sensor. closed-loop operation of an electroformed ring gyroscope
Therefore, high-quality materials with low internal damping [131]. It consists of four feedback loops and provides
will improve the accuracy of the sensor. Any remaining the necessary electronics to compensate for the sensor

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1653

Fig. 17. Performance of micromachined gyroscopes has im-
proved by a factor of ten every two years since 1991.

Fig. 18. A six-degree-of-freedom, fingernail-sized IMU with integrated signal-processing and

control circuitry, designed at the University of California, Berkeley, and fabricated by Sandia
National Laboratories. (Courtesy of T. Juneau.)

nonuniformities and its offset and sensitivity variations have been closing the gap with their bulk counterparts.
with temperature ( 40 to 85 C). High-performance gyro- Although this general trend is not expected to continue
scopes will continue to need precision, low-noise, parasitic- indefinitely, it is anticipated that for at least the next
insensitive interface circuits capable of resolving attofarad 5–10 years, we will see a continuing improvement in
changes in capacitance with as small an input capacitance the performance of both bulk- and surface-micromachined
as possible. gyroscopes.
Much effort is also under way for large-volume pro-
D. Commercialization and Future Trends duction of micromachined gyroscopes. Production cost,
Mainly driven by the automotive industry, microma- performance, and reliability are the key factors in commer-
chined silicon gyroscopes have been the subject of exten- cializing micromachined gyroscopes. Precision microma-
sive research and development over the past few years. The chining, robust vacuum packaging, and high-performance
performance of micromachined gyroscopes has drastically interface circuit and electronic tuning techniques are re-
improved over a rather short period, as illustrated in Fig. 17, quired to reduce the production cost to a level that is
which is based on a sample of devices reported in the lit- acceptable for the large-volume automotive market [132].
erature. Since 1991, performance, indicated by the random Since 1993, Draper and Rockwell International have been
angle walk, has improved by a factor of 10 every two collaborating to commercialize Draper’s silicon tuning-
years for both bulk- and surface-micromachined devices. fork gyroscope targeted for automotive applications [133].
The plot also shows that surface-micromachined devices Other companies like General Motors, Analog Devices,

1654 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

and Samsung have also performed extensive research and In the next decade, much effort will be expended to
development on various gyroscope structures, and pro- advance precision micromachining further, develop low-
totypes have been demonstrated [132]. Bosch [93] has noise and low-drift interface circuitry, and provide reliable
started volume production of its yaw sensor, and it is low-cost packaging for development of even higher per-
clear that several other companies will soon be doing the formance, lower cost, and lower power inertial sensors for
same. many emerging and as yet unknown applications.
In general, all-silicon, mixed-mode (bulk/surface) fabri-
cation technologies, combined with high-aspect-ratio deep ACKNOWLEDGMENT
dry-etching techniques, can provide features that are re-
The authors acknowledge all those who contributed their
quired for future high-performance microgyroscopes. For-
device illustrations and photographs shown in this paper,
mation of submicrometer capacitive gaps through sacrificial
in particular Dr. C. W. Yeh, Dr. A. Selvakumar, and Dr.
layer etching, high-aspect-ratio and thick structures with
M. W. Putty.
high quality factor and uniform material properties, along
with chip-level vacuum packaging, will help improve the
performance by orders of magnitude. Thick, high-aspect- REFERENCES
ratio structures formed by deep dry etching can yield [1] J. Soderkvist, “Micromachined gyroscopes,” Sensors Actuators
a large sense capacitance with a larger oscillating mass, A, vol. 43, pp. 65–71, 1994.
[2] T. B. Gabrielson, “Mechanical-thermal noise in micromachined
which in turn will help improve performance and simplify acoustic and vibration sensors,” IEEE Trans. Electron Devices,
packaging. By shrinking the capacitive gaps to submi- vol. 40, pp. 903–909, May 1993.
crometer levels, bias and control voltages will also shift [3] Y. H. Cho, “Design, fabrication, and optimization of microme-
chanical flexures,” Ph.D. thesis, University of California at
down to CMOS acceptable levels. Future high-performance Berkeley, CA, 1982.
tactical- and inertial-grade gyroscopes will make use of [4] L. M. Roylance and J. A. Angell, “A batch-fabricated silicon
dynamic electronic tuning of the structure to compen- accelerometer,” IEEE Trans. Electron Devices, vol. ED-26, pp.
1911–1917, Dec. 1979.
sate for temperature and long-term drift effects of the [5] H. Seidel and L. Csepregi, “Design optimization for cantilever-
sensor. type accelerometers,” Sensors Actuators, vol. 6, pp. 81–92,
Further progress is also anticipated in the development 1984.
[6] H. Sandmaier, K. Kuhl, and E. Obermeier, “A silicon based mi-
of integrated multiaxis devices on a single chip. Shown in cromechanical accelerometer with cross acceleration sensitivity
Fig. 18 is a six-degree-of-freedom, fingernail-sized inertial compensation,” in Tech. Dig. 4th Int. Conf. Solid-State Sensors
measurement unit (IMU) with integrated signal-processing and Actuators (Transducers’87), Tokyo, Japan, June 1987, pp.
399–402.
and control circuitry, which is designed at the University [7] H. Seidel, R. Reidel, R. Kolbeck, G. Muck, W. Kupke, and M.
of California, Berkeley, and fabricated by Sandia National Koniger, “Capacitive silicon accelerometer with highly sym-
Laboratories [37], [116]. This unit combines a dual-axis rate metric design,” Sensors Actuators, vol. A21/A23, pp. 312–315,
1990.
gyroscope, a -axis gyroscope, and a three-axis accelerom- [8] E. Peeters, S. Vergote, B. Puers, and W. Sansen, “A highly
eter chip to measure angular rate and acceleration without symmetrical capacitive micro-accelerometer with single degree-
the need to align individual sensors. of-freedom response,” J. Micromech. Microeng., vol. 2, pp.
104–112, 1992.
[9] J. M. Gere and S. P. Timoshenko, Mechanics of Materials, 3rd
ed. London: Chapman and Hall, 1991.
IV. CONCLUSIONS [10] H. A. Rothbart, Mechanical Design Handbook. New York:
McGraw-Hill, 1996.
Micromachined inertial sensors have progressed rapidly [11] W. S. Griffin, H. H. Richardson, and S. Yamanami, “A study of
since the first device was demonstrated in 1979 by re- fluid squeeze-film damping,” J. Basic Eng., Trans. ASME, pp.
searchers at Stanford University. Microaccelerometers are 451–456, June 1966.
[12] J. B. Starr, “Squeeze-film damping in solid-state accelerome-
now in large-volume production, cost a few dollars, in- ters,” in Tech. Dig. Solid-State Sensors and Actuators Workshop,
corporate many functions, including self-testing, and have Hilton Head Island, SC, June 1990, pp. 44–47.
[13] Y. He, R. Harris, G. Napadenski, and F. Maseeh, “A virtual
been shown to be extremely reliable for many years. This prototype manufacturing software system for MEMS,” in Proc.
impressive level of performance has been achieved by IEEE Micro Electro Mechanical Systems Workshop (MEMS’96),
resolving many obstacles in areas of device fabrication, in- San Diego, CA, Feb. 1996, pp. 122–126.
terface and readout circuitry, assembly and packaging, and [14] P. W. Barth, F. Pourahmadi, R. Mayer, J. Poydock, and K.
Petersen, “A monolithic silicon accelerometer with integral air
testing. One can routinely measure sub-mg accelerations, damping and overrange protection,” in Tech. Dig. Solid-State
and prototypes capable of providing g performance have Sensors and Actuators Workshop, Hilton Head Island, SC, June
been demonstrated. 1988, pp. 35–38.
[15] F. Pourahmadi, L. Christel, and K. Petersen, “Silicon ac-
By taking advantage of many of the technologies devel- celerometer with new thermal self-test mechanism,” in Tech.
oped for microaccelerometers, microgyroscopes have been Dig. Solid-State Sensors and Actuators Workshop, Hilton Head
able to enhance their performance by a factor of 10 every Island, SC, June 1992, pp. 122–125.
[16] H. Allen, S. Terry, and D. De Bruin, “Accelerometer system
two years since 1991, and there is every indication that this with self-testable features,” Sensors Actuators, vol. 20, pp.
trend will continue for at least the next few years. Based on 153–161, 1989.
past performance, it is clear that micromachined gyroscopes [17] W. Reithmuller, W. Benecke, U. Schnakenberg, and B. Wagner,
“A smart accelerometer with on-chip electronics fabricated by
will be capable of providing inertial-grade performance at a commercial CMOS process,” Sensors Actuators, vol. A-31,
orders of magnitude less cost and size. pp. 12–124, 1992.

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1655

[18] H. Seidel, U. Fritsch, R. Gottinger, and J. Schalk, “A Circuits Conf. (ISSCC’94), San Francisco, CA, Feb. 1994, pp.
piezoresistive silicon accelerometer with monolithically 160–161.
integrated CMOS-circuitry,” in Tech. Dig. 8th Int. Conf. Solid- [36] W. Yun, R. T. Howe, and P. R. Gray, “Surface microma-
State Sensors and Actuators (Transducers’95), Stockholm, chined digitally force-balanced accelerometer with integrated
Sweden, June 1995, pp. 597–600. CMOS detection circuitry,” in Tech. Dig. Solid-State Sensor
[19] S. J. Sherman, W. K. Tsang, T. A. Core, R. S. Payne, D. E. and Actuator Workshop, Hilton Head Island, SC, June 1992,
Quinn, K. H. Chau, J. A. Farash, and S. K. Baum, “A low-cost pp. 126–131.
monolithic accelerometer: Product/technology update,” in Tech. [37] C. Lu, M. Lemkin, and B. Boser, “A monolithic sur-
Dig. IEEE Electron Devices Meeting (IEDM’92), Dec. 1992, face micromachined accelerometer with digital output,”
pp. 160–161. IEEE J. Solid-State Circuits, vol. 30, pp. 1367–1373, Dec.
[20] L. Ristic, R. Gutteridge, J. Kung, D. Koury, B. Dunn, and 1995.
H. Zunino, “A capacitive type accelerometer with self-test [38] M. Lemkin, B. Boser, and J. Smith, “A 3-axis surface micro-
feature based on a double-pinned polysilicon structure,” in machined accelerometer,” in Tech. Digest Int. Solid State
Tech. Dig. 7th Int. Conf. Solid-State Sensors and Actuators Circuits Conf. (ISSCC’97), San Francisco, CA, Feb. 1997, pp.
(Transducers’93), Yokohama, Japan, June 1993, pp. 810– 202–203.
812. [39] “ADXL05-monolithic accelerometer with signal conditioning,”
[21] F. Rudolf, A. Jornod, and P. Bencze, “Silicon microaccelerom- Analog Devices, Norwood, MA, data sheet, 1995.
eters,” in Tech. Dig. 4th Int. Conf. Solid-State Sensors and [40] M. A. Lemkin, M. A. Ortiz, N. Wongkomet, B. E. Boser, and J.
Actuators (Transducers’87), Tokyo, Japan, June 1987, pp. 376– H. Smith, “A 3-axis force balanced accelerometer using a single
379. proof-mass,” in Tech. Dig. 9th Int. Conf. Solid-State Sensors
[22] F. Rudolf, A. Jornod, J. Berqovist, and H. Leuthold, “Precision and Actuators (Transducers’97), Chicago, IL, June 1997, pp.
accelerometers with g resolution,” Sensors Actuators, vol. 1185–1188.
A21/A23, pp. 297–302, 1990. [41] T. W. Kenny, S. B. Waltman, J. K. Reynolds, and W. J. Kaiser,
[23] W. Henrion, L. DiSanza, M. Ip, S. Terry, and H. Jerman, “A micromachined silicon electron tunneling sensor,” in Proc.
“Wide-dynamic range direct digital accelerometer,” in Tech. IEEE Micro Electro Mechanical Systems Workshop (MEMS’90),
Dig. Solid-State Sensors and Actuators Workshop, Hilton Head Napa Valley, CA, 1990, pp. 192–196.
Island, SC, June 1990, pp. 153–156. [42] H. K. Rocksatd, T. W. Kenny, J. K. Reynolds, W. J. Kaiser,
[24] Y. de Coulon, T. Smith, J. Hermann, M. Chevroulet, and F. and T. B. Gabrielson, “A miniature high-sensitivity broad-band
Rudolf, “Design and test of a precision servoaccelerometer with accelerometer based on electron tunneling transducers,” Sensors
digital output,” in Tech. Dig. 7th Int. Conf. Solid-State Sensors Actuators A, vol. 43, pp. 107–114, 1994.
& Actuators (Transducers’93), Yokohama, Japan, June 1993, [43] H. K. Rocksatd, J. K. Reynolds, T. K. Tang, T. W. Kenny, W.
pp. 832–835. J. Kaiser, and T. B. Gabrielson, “A miniature, high-sensitivity,
[25] K. Warren, “Navigation grade silicon accelerometer with sacrif- electron-tunneling accelerometer,” in Tech. Dig. 8th Int. Conf.
ically etched SIMOX and BESOI structure,” in Tech. Dig. Solid- Solid-State Sensors and Actuators (Transducers’95), Stockholm,
State Sensors and Actuators Workshop, Hilton Head Island, SC, Sweden, June 1995, pp. 675–678.
June 1994, pp. 69–72. [44] C. Yeh and K. Najafi, “A low-voltage bulk-silicon tunneling-
[26] N. Yazdi and K. Najafi, “An all-silicon single-wafer fabrication based microaccelerometer,” in Tech. Dig. IEEE Int. Electron
technology for precision microaccelerometers,” in Tech. Dig. Devices Meeting (IEDM), Washington, DC, Dec. 1995, pp.
9th Int. Conf. Solid-State Sensors and Actuators (Transduc- 593–596.
ers’97), Chicago, IL, June 1977, pp. 1181–1184. [45] R. L. Kubena, G. M. Atkinson, W. P. Robinson, and F. P.
[27] K. J. Ma, N. Yazdi, and K. Najafi, “A bulk-silicon capacitive Stratton, “A new miniaturized surface micromachined tunnel-
microaccelerometer with built-in overrange and force feedback ing accelerometer,” IEEE Electron Device Lett., vol. 17, pp.
electrodes,” in Tech. Dig. Solid-State Sensors and Actuators 306–308, June 1996.
Workshop, Hilton Head Island, SC, June 1994, pp. 160–163. [46] C. Yeh and K. Najafi, “A low-voltage tunneling-based silicon
[28] C. Buhrbaum, J. Mohr, P. Bley, and W. Ehrfeld, “Fabrication of microaccelerometer,” IEEE Trans. Electron Devices, vol. 44,
capacitive acceleration sensors by the LIGA technique,” Sensors no. 11, pp. 1875–1882, Nov. 1997.
Actuators A, vol. 25/27, pp. 559–563, 1990. [47] , “Micromachined tunneling accelerometer with a low-
[29] B. Boser and R. T. Howe, “Surface micromachined accelerom- voltage CMOS interface circuit,” in Tech. Dig. 9th Int. Conf.
eters,” IEEE J. Solid-State Circuits, vol. 31, pp. 366–375, Mar. Solid-State Sensors & Actuators (Transducers’97), Chicago, IL,
1996. June 1997, pp. 1213–1216.
[30] K. Chau, S. R. Lewis, Y. Zhao, R. T. Howe, S. F. Bart, and [48] G. Binning and H. Rohrer, “Scanning tunneling microscopy,”
R. G. Marcheselli, “An integrated force-balanced capacitive IBM J. Res. Develop., vol. 30, pp. 355–369, 1986.
accelerometer for low-g applications,” in Tech. Dig. 8th Int. [49] J. Grade, A. Barzilai, J. K. Reynolds, C. H. Liu, A. Partridge, L.
Conf. on Solid-State Sensors and Actuators (Transducers’95), M. Miller, J. A. Podosek, and T. Kenny, “Low frequency drift in
Stockholm, Sweden, June 1995, pp. 593–596. tunnel sensors,” in Tech. Dig. 9th Int. Conf. Solid-State Sensors
[31] B. P. van Drieenhuizen, N. Maluf, I. E. Opris, and G. Kovacs, and Actuators (Transducers’97), Chicago, IL, June 1997, pp.
“Force-balanced accelerometer with mG resolution fabricated 871–874.
using silicon fusion bonding and deep reactive ion etching,” [50] S. Danel, F. Michel, and G. Delapierre, “Micromachining of
in Tech. Dig. 9th Int. Conf. Solid-State Sensors and Actu- quartz and its application to an acceleration sensor,” Sensors
ators (Transducers’97), Chicago, IL, June 1997, pp. 1229– Actuators, vol. A21/A23, pp. 971–977, 1990.
1230. [51] K. Kourepenis, A. Petrovich, and M. Weinberg, “Low cost
[32] J. C. Cole, “A new sense element technology for accelerometer quartz resonant accelerometer for aircraft inertial navigation”
subsystems,” in Tech. Dig. 6th Int. Conf. Solid-State Sensors and in Tech. Dig. 6th Int. Conf. Solid-State Sensors and Actuators
Actuators (Transducers’91), San Francisco, CA, June 1991, pp. (Transducers’91), San Francisco, CA, June 1991, pp. 551–
93–96. 553.
[33] L. Spangler and C. J. Kemp, “ISAAC–Integrated silicon auto- [52] D. W. Satchell and J. C. Greenwood, “A thermally-excited
motive accelerometer,” in Tech. Dig. 8th Int. Conf. Solid-State silicon accelerometer,” Sensors Actuators, vol. 17, pp. 241–245,
Sensors and Actuators (Transducers’95), Stockholm, Sweden, 1989.
June 1995, pp. 585–588. [53] S. C. Chang, M. W. Putty, D. B. Hicks, and C. H. Li, “Resonant-
[34] A. Selvakumar, F. Ayazi, and K. Najafi, “A high sensitivity bridge two-axis microaccelerometer,” Sensors Actuators, vol.
-axis torsional silicon accelerometer,” in Tech. Dig. IEEE Int. A21/A23, pp. 342–345, 1990.
Electron Device Meeting, San Francisco, CA, Dec. 1996, pp. [54] T. V. Roszhart, H. Jerman, J. Drake, and C. de Cotiis, “An
765–768. inertial-grade micromachined vibrating beam accelerometer,”
[35] T. Smith, O. Nys, M. Chevroulet, Y. DeCoulon, and M. in Tech. Dig. 8th Int. Conf. Solid-State Sensors and Actuators
Degrauwe, “A 15b electromechanical sigma–delta converter for (Transducers’95), Stockholm, Sweden, June 1995, pp. 656–
acceleration measurements,” in Tech. Dig. IEEE Int. Solid-State 658.

1656 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

[55] D. W. Burns, R. D. Horning, W. R. Herb, J. D. Zook, and [73] M. L. Kniffin and M. Shah, “Packaging for silicon microma-
H. Guckel, “Resonant microbeam accelerometers,” in Tech. chined accelerometers,” Int. J. Microcircuits Electron. Packag-
Dig. 8th Int. Conf. Solid-State Sensors and Actuators (Trans- ing, vol. 19, no. 1, pp. 75–86, 1996.
ducers’95), Stockholm, Sweden, June 1995, pp. 659–662. [74] W. M. Stalnaker, L. J. Spangler, G. S. Fehr, and G. Fujimoto,
[56] T. A. Roessig, R. T. Howe, A. P. Pisano, and J. H. Smith, “Sur- “Plastic SMD package technology for accelerometers,” in Proc.
face micromachined resonant accelerometer,” in Tech. Dig. 9th 1997 Int. Symp. Microelectronics, Philadelphia, PA, Oct. 1997,
Int. Conf. Solid-State Sensors and Actuators (Transducers’97), pp. 197–202.
Chicago, IL, June 1997, pp. 859–862. [75] R. P. van Kampen, M. J. Velekoop, P. M. Sarro, and R. F.
[57] Y. Omura, Y. Nonomura, and O. Tabata, “New resonant ac- Wolffenbuttel, “Application of electrostatic feedback to critical
celerometer based on a rigidity change,” in Tech. Dig. 9th damping of an integrated silicon capacitive accelerometer,”
Int. Conf. Solid-State Sensors and Actuators (Transducers’97), Sensors Actuators A, vol. 43, pp. 100–106, 1994.
Chicago, IL, June 1997, pp. 855–858. [76] C. T. Nguyen, “Microelectromechanical resonators for oscilla-
[58] R. Hiratsuka, D. C. van Duyn, T. Otraedian, and P. de Vries, tors and filters,” in Proc. 1995 IEEE Int. Ultrasonics Symp.,
“A novel accelerometer based on a silicon thermopile,” in 1995, pp. 489–499.
Tech. Dig. 6th Int. Conf. Solid-State Sensors and Actuators [77] M. S. Smith, L. Bowman, and J. D. Meindl, “Analysis, design,
(Transducers’91), San Francisco, CA, June 1991, pp. 420– and performance of micropower circuits for a capacitive pres-
423. sure sensor IC,” IEEE J. Solid-State Circuits, vol. SC-21, pp.
[59] A. M. Leung, J. Jones, E. Czyzewska, J. Chen, and B. Woods, 1045–1056, Dec. 1986.
“Micromachined accelerometer based on convection heat [78] C. Hierold, A. Hildebrandt, U. Naher, T. Scheiter, B. Men-
transfer,” in Proc. IEEE Micro Electro Mechanical Systems sching, M. Steger, and R Tielert, “A pure CMOS microma-
Workshop (MEMS’98), Heidelberg, Germany, Jan. 1998, pp. chined integrated accelerometer,” in Proc. IEEE Micro Electro
627–630. Mechanical Systems Workshop (MEMS’96), San Diego, CA,
[60] D. Uttamchandani, D. Liang, and B. Culshaw, “A microma- 1996, pp. 174–179.
chined silicon accelerometer with fiber optic integration,” in [79] Y. Matsumoto and M. Esashi, “Low drift integrated capacitive
Proc. SPIE Integrated Optics and Microstructures, 1992, pp. accelerometer with PLL servo techniques,” in Tech. Dig. 7th
27–33. Int. Conf. Solid-State Sensors and Actuators (Transducers’93),
[61] R. S. Huang, E. Abbaspour-Sani, and C. Y. Kwok, “A novel Yokohama, Japan, June 1993, pp. 826–829.
accelerometer using silicon micromachined cantilever supported [80] N. Yazdi, A. Mason, K. Najafi, and K. D. Wise, “A low-power
optical grid and PIN photodetector,” in Tech. Dig. 8th Int. Conf. interface circuit for capacitive sensors,” in Tech. Dig. Solid-
Solid-State Sensors and Actuators (Transducers’95), Stockholm, State Sensor and Actuator Workshop, Hilton Head Island, SC,
Sweden, June 1995, pp. 663–666. June 1996, pp. 215–218.
[62] E. Abbaspour-Sani, R. S. Huang, and C. Y. Kwok, “A linear [81] J. T. Kung, H. S. Lee, and R. T. Howe, “A digital readout
electromagnetic accelerometer,” Sensors Actuators A, vol. 44, technique for capacitive sensor applications,” IEEE J. Solid-
pp. 103–109, 1994. State Circuits, vol. 23, pp. 972–977, Aug. 1988.
[63] P. Chen, R. S. Muller, R. D. Jolly, G. L. Halac, R. M. [82] H. Leuthold and F. Rudolf, “An ASIC for high-resolution capac-
White, A. P. Andrews, T. C. Lim, and M. E. Motamedi, itive microaccelerometers,” Sensors Actuators, vol. A21/A23,
“Integrated silicon microbeam PI-FET accelerometer,” IEEE pp. 278–281, 1990.
Trans. Electron Devices, vol. ED-29, pp. 363–369, Jan. [83] S. Suzuki and S. Tuchitani, “Semiconductor capacitance-type
1982. accelerometer with PWM electrostatic servo technique,” Sen-
[64] D. L. DeVoe and A. P. Pisano, “A fully surface-micromachined sors Actuators, vol. A21/A23, pp. 316–319, 1990.
piezoelectric accelerometer,” in Tech. Dig. 9th Int. Conf. Solid- [84] “MMAS40G10D-micromachined accelerometer,” Motorola,
State Sensors and Actuators (Transducers’97), Chicago, IL, Phoenix, AZ, data sheet, 1997.
June 1997, pp. 1205–1208. [85] “Dynamic measurement short form catalog,” Endevco, San Juan
[65] W. D. Frobenius, S. Zeitman, M. H. White, D. A. O’Sullivan, Capistrano, CA, data sheet, 1997.
and R. G. Hamel, “Microminiature ganged threshold ac- [86] “MS6100-low power accelerometer,” CSEM, Neuchatel,
celerometers compatible with integrated circuit technology,” Switzerland, data sheet, 1997.
IEEE Trans. Electron Devices, vol. ED-19, pp. 37–40, Jan. [87] H. Lefevre, The Fiber-Optic Gyroscope. Norwood, MA:
1972. Artech House, 1993.
[66] J. Noetzel, T. Tonnesen, W. Benecke, J. Binder, and G. Mader, [88] A. Lawrence, Modern Inertial Technology: Navigation, Guid-
“Quasianalog accelerometer using microswitch array,” in Tech. ance, and Control, New York: Springer-Verlag, 1993.
Dig. 8th Int. Conf. Solid-State Sensors and Actuators (Transduc- [89] M. W. Putty and K. Najafi, “A micromachined vibrating ring
ers’95), Stockholm, Sweden, June 1995, pp. 671–674. gyroscope,” in Tech. Dig. Solid-State Sensor and Actuator Work-
[67] A. Selvakumar, N. Yazdi, and K. Najafi, “A low power, wide shop, Hilton Head Island, SC, June 1994, pp. 213–220.
range threshold acceleration sensing system,” in Proc. IEEE [90] R. R. Ragan and D. D. Lynch, “Inertial technology for the
Micro Electro Mechanical Systems Workshop (MEMS’96), San future, Part X: Hemispherical resonator gyro,” IEEE Trans.
Diego, CA, Feb. 1996, pp. 186–191. Aerosp. Electron. Syst., vol. AES-20, p. 432, July 1984.
[68] O. Kromer, O. Fromhein, H. Gemmeke, T. Kuhner, J. Mohr, [91] J. Bernstein, S. Cho, A. T. King, A. Kourepenis, P. Maciel, and
and M. Strohramann, “High-precision readout circuit for LIGA M. Weinberg, “A micromachined comb-drive tuning fork rate
acceleration sensors,” Sensors Actuators A, vol. 46/47, pp. gyroscope,” in Proc. IEEE Micro Electro Mechanical Systems
196–200, 1995. Workshop (MEMS’93), Fort Lauderdale, FL, Feb. 1993, pp.
[69] A. Selvakumar, “A multifunctional silicon micromachining 143–148.
technology for high performance microsensors and microactu- [92] M. Hashimoto, C. Cabuz, K. Minami, and M. Esashi, “Sili-
ators,” Ph.D. dissertation, University of Michigan, Ann Arbor, con resonant angular rate sensor using electromagnetic exci-
1997. tation and capacitive detection,” J. Micromech. Microeng., pp.
[70] M. Offenberg, F. Larmer, B. Elsner, H. Munzel, and W. Rieth- 219–225, 1995.
muller, “Novel process for a monolithic integrated accelerom- [93] M. Lutz, W. Golderer, J. Gerstenmeier, J. Marek, B. Maihofer,
eter,” in Tech. Dig. 8th Int. Conf. Solid-State Sensors and S. Mahler, H. Munzel, and U. Bischof, “A precision yaw rate
Actuators (Transducers’95), Stockholm, Sweden, June 1995, sensor in silicon micromachining,” in Tech. Dig. 9th Int. Conf.
pp. 589–592. Solid-State Sensors and Actuators (Transducers’97), Chicago,
[71] “ADXL50-monolithic accelerometer with signal conditioning,” IL, June 1997, pp. 847–850.
Analog Devices, Norwood, MA, data sheet, 1993. [94] R. Voss, K. Bauer, W. Ficker, T. Gleissner, W. Kupke, M. Rose,
[72] E. Koen, F. Pourahmadi, and S. Terry, “A multilayer ce- S. Sassen, J. Schalk, H. Seidel, and E. Stenzel, “Silicon angular
ramic package for silicon micromachined accelerometers,” in rate sensor for automotive applications with piezoelectric drive
Tech. Dig. 8th Int. Conf. Solid-State Sensors and Actuators and piezoresistive read-out,” in Tech. Dig. 9th Int. Conf. Solid-
(Transducers’95), Stockholm, Sweden, June 1995, pp. 273– State Sensors and Actuators (Transducers’97), Chicago, IL,
276. June 1997, pp. 879–882.

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1657

 [95] K. Maenaka and T. Shiozawa, “A study of silicon angular [115] K. Tanaka, Y. Mochida, M. Sugimoto, K. Moriya, T. Hasegawa,
rate sensors using anisotropic etching technology,” Sensors K. Atsuchi, and K. Ohwada, “A micromachined vibrating
Actuators A, vol. 43, pp. 72–77, 1994. gyroscope,” Sensors Actuators A, vol. 50, pp. 111–115, 1995.
[96] J. Soderkvist, “Design of a solid-state gyroscopic sensor made [116] T. Juneau, A. P. Pisano, and J. H. Smith, “Dual axis operation of
of quartz,” Sensors Actuators, vol. A21/A23, pp. 293–296, a micromachined rate gyroscope,” in Tech. Dig. 9th Int. Conf.
1990. Solid-State Sensors and Actuators (Transducers’97), Chicago,
[97] , “Micromachined gyroscopes,” Sensors Actuators A, vol. IL, June 1997, pp. 883–886.
43, pp. 65–71, 1994. [117] T. Fujita, T. Mizuno, R. Kenny, M. Maenaka, and M. Maeda,
[98] S. D. Orlosky and H. D. Morris, “Quartz rotation (rate) sensor,” “Two-dimensional micromachined gyroscope,” in Tech. Dig. 9th
in Proc. Sensor Expo, Cleveland, OH, 1994, pp. 171–177. Int. Conf. Solid-State Sensors and Actuators (Transducers’97),
[99] J. S. Burdess and T. Wren, “The theory of a piezoelectric disc Chicago, IL, June 1997, pp. 887–890.
gyroscope,” IEEE Trans. Aerosp. Electron. Syst., vol. AES-22, [118] T. K. Tang, R. C. Gutierrez, J. Z. Wilcox, C. Stell, V. Vorperian,
pp. 410–418, July 1986. R. Calvet, W. J Li, I. Charkaborty, R. Bartman, and W. J.
[100] P. Greiff, B. Boxenhorn, T. King, and L. Niles, “Silicon Kaiser, “Silicon bulk micromachined vibratory gyroscope,” in
monolithic micromechanical gyroscope,” in Tech. Dig. 6th Int. Tech. Dig. Solid-State Sensor and Actuator Workshop, Hilton
Conf. Solid-State Sensors and Actuators (Transducers’91), San Head Island, SC, June 1996, pp. 288–293.
Francisco, CA, June 1991, pp. 966–968. [119] T. K. Tang, R. C. Gutierrez, C. B. Stell, V. Vorperian, G. A.
[101] Y. Gianchandani and K. Najafi, “A bulk silicon dissolved Arakaki, J. T. Rice, W. J. Li, I. Chakraborty, K. Shcheglov, J. Z.
wafer process for microelectromechanical systems,” J. Micro- Wilcox, and W. J. Kaiser, “A packaged silicon MEMS vibratory
electromech. Syst., pp. 77–85, June 1992. gyroscope for microspacecraft,” in Proc. IEEE Micro Electro
[102] W. C. Tang, M. G. Lin, and R. T. Howe, “Electrostatically Mechanical Systems Workshop (MEMS’97), Japan, 1997, pp.
balanced comb drive for controlled levitation,” in Tech. Dig. 500–505.
Solid-State Sensor and Actuator Workshop, Hilton Head Island, [120] W. Geiger, B. Folkmer, J. Merz, H. Sandmaier, and W. Lang,
SC, June 1990, pp. 23–27. “A new silicon rate gyroscope,” in Proc. IEEE Micro Electro
[103] M. Weinberg, J. Bernstein, S. Cho, A. T. King, A. Kourepenis, Mechanical Systems Workshop (MEMS’98), Heidelberg, Ger-
P. Ward, and J. Sohn, “A micromachined comb-drive tuning many, Feb. 1998, pp. 615–620.
fork gyroscope for commercial applications,” in Proc. Sensor [121] M. W. Putty, “A micromachined vibrating ring gyroscope,”
Expo, Cleveland, OH, 1994, pp. 187–193. Ph.D. dissertation, Univ. Michigan, Ann Arbor, Mar. 1995.
[104] F. Paoletti, M. A. Gretillat, and N. F. de Rooij, “A silicon mi- [122] D. R. Sparks, S. R. Zarabadi, J. D. Johnson, Q. Jiang, M.
cromachined vibrating gyroscope with piezoresistive detection Chia, O. Larsen, W. Higdon, and P. Castillo-Borelley, “A
and electromagnetic excitation,” in Proc. IEEE Micro Electro CMOS integrated surface micromachined angular rate sensor:
Mechanical Systems Workshop (MEMS’96), San Diego, CA, It’s automotive applications,” in Tech. Dig. 9th Int. Conf. Solid-
1996, pp. 162–167. State Sensors and Actuators (Transducers’97), Chicago, IL,
[105] J. Soderkvist, “Micromachined vibrating gyroscopes,” in June 1997, pp. 851–854.
Proc. SPIE 1996 Symp. Micromachining and Microfabrication, [123] F. Ayazi and K. Najafi, “Design and fabrication of a high-
Austin, TX, 1996, pp. 152–160. performance polysilicon vibrating ring gyroscope,” in Proc.
[106] S. Shoji, H. Kikuchi, and H. Torigoe, “Anodic bonding below IEEE Micro Electro Mechanical Systems Workshop (MEMS’98),
 C for packaging and assembling of MEMS using lithium
Heidelberg, Germany, Feb. 1998, pp. 621–626.
alminosilicate- -quartz glass-ceramic,” in Proc. IEEE Micro [124] A. Selvakumar and K. Najafi, “High density vertical comb array
Electro Mechanical Systems Workshop (MEMS’97), Japan, microactuators fabricated using a novel bulk/poly-silicon trench
1997, pp. 482–487. refill technology,” in Tech. Dig. Solid-State Sensor and Actuator
[107] K. Maenaka, T. Fujita, Y. Konishi, and M. Maeda, “Analysis Workshop, Hilton Head Island, SC, USA, 1994, pp. 138–141.
of a highly sensitive silicon gyroscope with cantilever beam [125] I. Hopkin, “Performance and design of a silicon micromachined
as vibrating mass,” Sensors Actuators A, vol. 54, pp. 568–573, gyro,” in Proc. Symp. Gyro Technology, Stuttgart, Germany,
1996. 1997, pp. 1.0–1.10.
[108] D. Wood, G. Cooper, J. Burdess, A. Harris, and J. Cruickshank, [126] R. Torti, V. Gondhalekar, H. Tran, B. Selfors, S. Bart, and
“A silicon membrane gyroscope with electrostatic actuation B. Maxwell, “Electrostatically suspended and sensed micro-
and sensing,” in Proc. SPIE 1995 Symp. Micromachining and mechanical rate gyroscope,” in Proc. SPIE 1994 Symp. on
Microfabrication, Austin, TX, 1995, pp. 74–83. Micromachining and Microfabrication, Austin, TX, 1994, pp.
[109] H. Kuisma, T. Ryhanen, J. Lahdenpera, E. Punkka, S. Ruot- 27–31.
salainen, T. Sillanpaa, and H. Seppa, “A bulk micromachined [127] C. Shearwood, C. B. Williams, P. H. Mellor, R. B. Yates, M. R.
silicon angular rate sensor,” in Tech. Dig. 9th Int. Conf. Solid- J. Gibbs, and A. D. Mattingley, “Levitation of a micromachined
State Sensors and Actuators (Transducers’97), Chicago, IL, rotor for application in a rotating gyroscope,” Electron. Lett.,
June 1997, pp. 875–878. vol. 31, no. 21, p. 1845, Oct. 1995.
[110] W. A. Clark, R. T. Howe, and R. Horowitz, “Surface mi- [128] J. Soderkvist, “Piezoelectric beams and vibrating angular rate
cromachined -axis vibratory rate gyroscope,” in Tech. Dig. sensors,” IEEE Trans. Ultrason., Ferroelect., Frequency Contr.,
Solid-State Sensor and Actuator Workshop, Hilton Head Island, vol. 38, pp. 271–280, May 1991.
SC, June 1996, pp. 283–287. [129] S. T. Cho and F. M. Erdmann, “An on-chip hermetic packaging
[111] T. Juneau and A. P. Pisano, “Micromachined dual input axis technology for micromechanical devices,” in Tech. Dig. Solid-
angular rate sensor,” in Tech. Dig. Solid-State Sensor & Actu- State Sensor and Actuator Workshop, Hilton Head, SC, June
ator Workshop, Hilton Head Island, SC, June 1996, pp. 299– 1998.
302. [130] B. E. Boser, “Electronics for micromachined inertial sensors,”
[112] Y. Oh, B. Lee, S. Baek, H. Kim, J. Kim, S. Kang, and C. in Tech. Dig. 9th Int. Conf. Solid-State Sensors and Actu-
Song, “A surface-micromachined tunable vibratory gyroscope,” ators (Transducers’97), Chicago, IL, USA, June 1997, pp.
in Proc. IEEE Micro Electro Mechanical Systems Workshop 1169–1172.
(MEMS’97), Japan, 1997, pp. 272–277. [131] S. R. Zarabadi, P. E. Castillo-Borelly, and J. D. Johnson,
[113] K. Y. Park, C. W. Lee, Y. S. Oh, and Y. H. Cho, “Laterally “An angular rate sensor interface IC,” in Dig. IEEE Custom
oscillated and force-balanced micro vibratory rate gyroscope Integrated Circuits Conf., 1996, pp. 311–314.
supported by fish hook shape springs,” in Proc. IEEE Mi- [132] C. Song, “Commercial vision of silicon based inertial sensors,”
cro Electro Mechanical Systems Workshop (MEMS’97), Japan, in Tech. Dig. 9th Int. Conf. Solid-State Sensors and Actuators
1997, pp. 494–499. (Transducers’97), Chicago, IL, June 1997, pp. 839–842.
[114] S. An, Y. S. Oh, B. L. Lee, K. Y. Park, S. J. Kang, S. O. [133] M. Weinberg, J. Bernstein, J. Borenstein, J. Campbell, J.
Choi, Y. I. Go, and C. M. Song, “Dual-axis microgyroscope Cousens, B. Cunningham, R. Fields, P. Greiff, B. Hugh, L.
with closed-loop detection,” in Proc. IEEE Micro Electro Me- Niles, and J. Sohn, “Micromachining inertial instruments,” in
chanical Systems Workshop (MEMS’98), Heidelberg, Germany, Proc. SPIE 1996 Symp. Micromachining and Microfabrication,
Feb. 1998, pp. 328–333. Austin, TX, 1996.

1658 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

 Navid Yazdi received the B.S. degree (with Khalil Najafi (Senior Member, IEEE) was born
Honors) from the University of Tehran, Iran, in in 1958. He received the B.S., M.S., and Ph.D.
1988 and the M.S. degree from the University degrees in electrical engineering from the De-
of Windsor, Windsor, Canada, in 1993, both in partment of Electrical Engineering and Com-
electrical engineering. He currently is pursuing puter Science, University of Michigan, Ann Ar-
the Ph.D. degree in electrical engineering at the bor, in 1980, 1981, and 1986, respectively.
University of Michigan, Ann Arbor. He has been with the Center for Integrated
From 1988 to 1991, he was an R&D Engineer Sensors and Circuits, Department of Electrical
at the University of Tehran, where he worked Engineering and Computer Science, University
on hardware/software development for computer of Michigan, as a Research Fellow (1986–1988),
numerical control machine tools and large data- Assistant Research Scientist (1988–1990), As-
acquisition systems. From 1991 to 1993, he worked on architectures and sistant Professor (1990–1993), and Associate Professor (1993–present).
mixed-signal circuit design for very-large-scale-integration (VLSI) imple- His research interests include development of microfabrication and mi-
mentation of artificial neural networks at the University of Windsor. Since cromachining technologies for solid-state integrated sensors and microac-
1993, his research work at the University of Michigan has been focused on tuators, analog and digital integrated circuits, implantable microtelemetry
high-performance mixed-signal interface circuits for microsensors, high- systems and transducers for biomedical applications and wireless commu-
precision microaccelerometer design and fabrication, electromechanical nication, technologies and structures for microelectromechanical systems
oversampled sigma-delta modulators, and wireless telemetry for microsys- and microstructures, hermetic packaging techniques for microtransducers,
tems. His research interests include mixed analog–digital VLSI circuits for and low-power wireless sensing/actuating systems. He has been active in
sensor interfacing and telecommunication applications, integrated inertial the field of solid-state sensors and actuators for 15 years and has been
sensors, low-power wireless smart microsystems, and technologies for involved in several conferences and workshops dealing with solid-state
microelectromechanical systems. sensors and actuators, including the International Electron Devices Meet-
Mr. Yazdi is a member of Sigma Xi and Tau Beta Pi. He received a ing, the Hilton-Head Solid-State Sensors and Actuators Workshop, and
University of Michigan Rackham Predoctoral Fellowship for 1996–1997. the IEEE/ASME Micro Electromechanical Systems Workshop. He is an
He received the Outstanding Graduate Student Achievement Award in Associate Editor for the Journal of Micromechanics and Microengineering.
electrical engineering from the College of Engineering, University of Dr. Najafi received a National Science Foundation Young Investigator
Michigan, in 1998. Award from 1992 to 1997. He received the Beatrice Winner Award
for Editorial Excellence at the 1986 International Solid-State Circuits
Conference and the Paul Rappaport Award for coauthoring the Best Paper
published in IEEE TRANSACTIONS ON ELECTRON DEVICES. In 1994, he
received the University of Michigan’s Henry Russel Award for outstanding
Farrokh Ayazi was born in Tehran, Iran, in achievement and scholarship. In 1993, he was selected by students in
1972. He received the B.S. degree in electrical Electrical Engineering and Computer Science Department as the Professor
engineering (with Honors) from the University of the Year. Dr. Najafi is the Editor for Solid-State Sensors for IEEE
of Tehran in 1994 and the M.S. degree in TRANSACTIONS ON ELECTRON DEVICES.
electrical engineering from the University of
Michigan, Ann Arbor, in 1997. He currently
is pursuing the Ph.D. degree at the Center for
Integrated Sensors and Circuits, University of
Michigan.
His doctoral work involves the design and
development of precision silicon micromachined
gyroscopes and their interface circuits. He also has been involved in
characterization and fabrication of silicon micromachined accelerometers
and their interface circuits. During the summer of 1993, he was with
Schlumberger, Abu Dhabi, United Arab Emirates, where he completed an
extensive course on sensing and data-aquisition techniques.
Mr. Ayazi received a Rackham Predoctoral Fellowship from the Uni-
versity of Michigan for 1998–1999.

YAZDI et al.: MICROMACHINED INERTIAL SENSORS 1659