Micromachined Devices for Wireless

Communications
CLARK T.-C. NGUYEN, MEMBER, IEEE, LINDA P. B. KATEHI, FELLOW, IEEE,
AND GABRIEL M. REBEIZ, FELLOW, IEEE

Invited Paper

An overview of recent progress in the research and develop- variable capacitors are used to properly tune and couple
ment of micromachined devices for use in wireless communication the front-end sense and power amplifiers. At present, the
subsystems is presented. Among the specific devices described are
Q
tunable micromachined capacitors, integrated high- inductors,
aforementioned resonators and discrete elements are off-
chip components and so must interface with integrated
micromachined low-loss microwave and millimeter-wave filters,
low-loss micromechanical switches, microscale vibrating mechan- electronics at the board level, often consuming a sizable
Q
ical resonators with ’s in the tens of thousands, and miniature portion of the total subsystem area. In this respect, these
antennas for millimeter-wave applications. Specific applications devices pose an important bottleneck against the ultimate
are reviewed for each of these components with emphasis on
methods for miniaturization and performance enhancement of miniaturization and portability of wireless transceivers. For
existing and future wireless transceivers. this reason, many research efforts have been focused upon
Keywords— Acoustic resonators, antennas, filters, high- , IF,Q strategies for either miniaturizing these components [1]–[5]
MEMS, micromachining, oscillators, phase noise, resonators, RF, or eliminating the need for them altogether [6]–[8].
switches, synthesizers, thin-film inductors, transceivers, tunable, The rapid growth of integrated circuit (IC)-compatible
wireless.
micromachining technologies that yield microscale, high-
tank components may now bring the first of the above
I. INTRODUCTION strategies closer to reality. Specifically, the high- RF and
Vibrating mechanical tank components, such as crystal IF filters, oscillators, and couplers, currently implemented
and surface-acoustic wave (SAW) resonators, are widely via off-chip resonators and discrete passives, may now po-
used for frequency selection in communication subsys- tentially be realized on the microscale using micromachined
tems because of their high quality factor ( ’s in the tens equivalents based on a variety of novel devices, including
of thousands) and exceptional stability against thermal high- on-chip mechanical resonators [9]–[11], voltage-
variations and aging. In particular, the majority of hetero- tunable on-chip capacitors [12], isolated low-loss inductors
dyning communication transceivers rely heavily upon the [13]–[16], [34], microwave/millimeter-wave high- filters
high of SAW and bulk-acoustic mechanical resonators [35]–[40], structures for high-frequency isolation pack-
to achieve adequate frequency selection in their radio- aging [43], [44], and low-loss micromechanical switches
frequency (RF) and intermediate-frequency (IF) filtering [45]–[50]. Once these miniaturized filters and oscillators
stages and to realize the required low phase noise and become available, the fundamental bases upon which com-
stability in their local oscillators. Ceramic (dielectric) res- munications systems are developed may also evolve, giving
onators or waveguide components are also required at RF rise to new system architectures with possible power- and
or microwave/millimeter-wave frequencies to provide the bandwidth-efficiency advantages. For systems operating
needed high- functions. In addition, discrete inductors and past X-band, antennas can also be micromachined with
potential cost savings and with additional capabilities at-
Manuscript received January 5, 1998; revised February 27, 1998. This tained via active antenna arrays (e.g., phased arrays, power
work was supported in part by the Defense Advanced Research Projects combining, etc.) [51], [53]–[57].
Agency, National Science Foundation, National Aeronautics and Space
Administration, Army Research Office, Office of Naval Research, and
This paper reviews recent progress in the research and
numerous industrial sponsors. development of microelectromechanical devices for use in
The authors are with the Department of Electrical Engineering and communication subsystems. It begins with a brief intro-
Computer Science, University of Michigan, Ann Arbor, MI 48109-2122
USA. duction into the needs of wireless communication trans-
Publisher Item Identifier S 0018-9219(98)05098-1. ceivers, identifying specific functions that could greatly

0018–9219/98$10.00  1998 IEEE

1756 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

Fig. 1. System-level schematic detailing the front-end design for a typical wireless transceiver.
Q
The off-chip, high- , passive components targeted for replacement via micromechanical versions
(suggestions in lighter ink) are indicated by shading in the figure.

benefit from micromechanical implementation. Several spe- transistor-based components but by the numerous passive
cific devices will then be described, with particular empha- components indicated in Fig. 1. The presence of so many
sis on frequency-selective microelectromechanical systems frequency-selective passive components is easily justified
(MEMS) for high- oscillators and filters. This paper when considering that communications systems designed
concludes by considering the impact that the discussed to service large numbers of users require numerous com-
devices may have on future transceiver architectures. munication channels, which in many implementations (e.g.,
time division multiple access) must have small bandwidths
and must be separable by transceiver devices used by the
II. MINIATURIZATION OF TRANSCEIVERS system. The requirement for small channel bandwidths
To illustrate more concretely the specific transceiver results in a requirement for extremely selective filtering
functions that benefit from micromechanical implemen- devices for channel selection and extremely stable (noise-
tation, Fig. 1 presents the system-level schematic for a free) local oscillators for frequency translation. For the vast
typical superheterodyne wireless transceiver. As implied majority of cellular and cordless standards, the required
in the figure, several of the constituent components can selectivity and stability can only be achieved using high-
already be miniaturized using integrated circuit transis- components, such as discrete inductors, discrete tunable
tor technologies. These include the low noise amplifiers capacitors (i.e., varactors), and SAW and quartz crystal
(LNA’s) in the receive path, the solid-state power amplifier resonators, all of which interface with IC components at the
(SSPA) in the transmit path, synthesizer phase-locked loop board level. The needed performance cannot be achieved
(PLL) electronics, mixers, and lower frequency digital using conventional IC technologies because such technolo-
circuits for baseband signal demodulation. Due to noise, gies lack the required . It is for this reason that virtually all
power, and frequency considerations, the SSPA (and some- commercially available cellular or cordless phones contain
times the LNA’s) are often implemented using compound numerous passive, SAW, and crystal components.
semiconductor technologies (i.e., GaAs). Thus, they often
occupy their own chips, separate from the other men-
tioned transistor-based components, which are normally A. The Need for High in Oscillators
realized using silicon-based bipolar and complementary For any communications application, the stability of the
metal–oxide–semiconductor (CMOS) technologies. How- oscillator signals used for frequency translation, synchro-
ever, given the rate of improvement of silicon technologies nization, or sampling is of utmost importance. Oscillator
(silicon-germanium included [17]), it is not implausible that frequencies must be stable against variations in temperature,
all of the above functions could be integrated onto a single against aging, and against any phenomena, such as noise
chip in the foreseeable future. or microphonics, that cause instantaneous fluctuations in
Unfortunately, placing all of the above functions onto phase and frequency. The single most important parameter
a single chip does very little to decrease the overall su- that dictates oscillator stability is the of the frequency-
perheterodyne transceiver size, which is dominated not by setting tank (or of the effective tank for the case of ring

NGUYEN et al.: MICROMACHINED DEVICES 1757

 (a)

(b) (c)
Q tank, indicating
Fig. 2. (a) A simple series resonant oscillator schematic. (b) Bode plot for a low-
1 1
the f for a given  . (c) Similar to (b) but for a high-Q tank.

oscillators). For a given application, and assuming a finite incurs a much smaller frequency deviation on the tank with
power budget, adequate long- and short-term stability of the higher . Thus, the higher the tank , the more stable
the oscillation frequency is insured only when the tank the oscillator against phase-shifting phenomena.
exceeds a certain threshold value. To help quantify the above heuristic concepts, one impor-
The correlation between tank and oscillator stability tant figure of merit for oscillators is the phase-noise power
can be illustrated heuristically by considering the sim- present at frequencies close to the carrier frequency. Typical
ple oscillator circuit depicted in Fig. 2(a). Here, a series phase-noise requirements range from 100 dBc/Hz at 10
resonant oscillator is shown, comprised of a sustaining kHz deviation from a 1.8 GHz carrier in European Global
amplifier and an inductor-capacitor (LC) tank connected in System for Mobile Telecommunications (GSM) cellular
a positive feedback loop. For proper startup and steady-state phones to 150 dBc/Hz at 67 kHz carrier deviations in X-
operation, the total phase shift around the loop must sum band, Doppler-based radar systems [19]. Through a more
to zero. Thus, if at the oscillation frequency the amplifier rigorous analysis of Fig. 1 (assuming linear operation), the
operates nominally with a 0 phase shift from its input phase noise of a given oscillator can be described by the
expression [18]
to its output, then the tank must also have a 0 phase
shift across its terminals. Given this, and referring to any
dBc/Hz (1)
one of the tank response spectra shown in Fig. 2(b) or
(c), this oscillator is seen to operate nominally at the tank where is the phase-noise power density-to-
resonance frequency. If, however, an external stimulus (e.g., carrier power ratio at a frequency offset from the
a noise spike or a temperature fluctuation) generates a phase carrier frequency, is the noise figure of the active device
shift across the terminals of the sustaining amplifier, evaluated using the total oscillator power , is the
the tank must respond with an equal and opposite phase carrier power delivered to the load, and is the carrier
shift for sustained oscillation. As dictated by the tank frequency. From (1), phase noise is seen to be inversely
transfer functions of Fig. 2, any tank phase shift must proportional to the square of and directly proportional
be accompanied by a corresponding operating frequency to the amplifier noise figure . Given that can often
shift . The magnitude of for a given is largely be reduced by increasing the operating power of the
dependent on the of the resonator tank. Comparison of sustaining amplifier, (1) then can be interpreted as implying
Fig. 2(b) with (c) clearly shows that a given phase shift that power and can be traded to achieve a given phase-

1758 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

Fig. 3. Simulated frequency characteristics for a 0.3% bandwidth, 70 MHz bandpass filter under
Q
varying tank ’s.

noise specification. Given the need for low power in bandwidths and can thus be implemented using resonators
portable units, and given that the synthesizer [containing with ’s on the order of 500–1000.
the reference and voltage-controlled oscillators (VCO’s)]
is often a dominant contributor to total transceiver power III. MICROMECHANICAL COMPONENTS
consumption, modern transceivers could benefit greatly FOR TRANSCEIVERS
from technologies that yield high- tank components. As shown in Fig. 1, the front end of a superheterodyne
wireless transceiver typically contains a good number of
B. The Need for High in Filters off-chip, high- components that are potentially replace-
Tank also greatly influences the ability to implement able by micromachined versions. Among the components
extremely selective IF and RF filters with small percent targeted for replacement in cellular and cordless applica-
bandwidth, small shape factor, and low insertion loss. tions are RF filters, including image rejection filters, with
To illustrate, Fig. 3 presents simulated frequency char- center frequencies ranging from 800 MHz to 2.5 GHz;
acteristics under varying resonator tank ’s for a 0.3% IF filters, with center frequencies ranging from 455 kHz
bandwidth bandpass filter centered at 70 MHz, realized to 254 MHz; high- , tunable, low phase-noise oscillators,
using the typical LC resonator ladder configuration shown with frequency requirements in the 10 MHz–2.5 GHz range;
in the insert. As shown, for a resonator tank of 10 000, and switches for transmit/receive selection, antenna selec-
very little insertion loss is observed. However, as tank tion, and multiband configurability. For higher frequency
decreases, insertion loss increases very quickly, to the applications (e.g., past X-band), antenna size requirements
point where a tank of 1000 leads to 20 dB of insertion shrink, so the antenna itself can be manufactured us-
loss—too much even for IF filters and quite unacceptable ing micromachining techniques. The size of transmission-
for RF filters. As with oscillators, high- tanks are required line filters also shrinks at these frequencies, making them
for RF and IF filters alike, although more so for the latter, amenable to implementation via micromachining as well.
since channel selection is done predominantly at the IF in The following subsections now describe specific micro-
superheterodyne receivers. In general, the more selective electromechanical components capable of both miniaturiz-
the filter, the higher the resonator required to achieve ing and improving the performance of the aforementioned
a given level of insertion loss. In particular, the above functions.
0.3% bandwidth filter example applies for IF filters, which,
because of their high selectivity, are best implemented A. Voltage-Tunable High- Capacitors
with resonator ’s exceeding 5000; RF preselect or image- One application presently under intense examination for
reject filters, on the other hand, typically require only 3% possible miniaturization is that of the VCO used in the syn-

NGUYEN et al.: MICROMACHINED DEVICES 1759

thesizer that generates the local oscillator signal. Currently,
such VCO’s are implemented using off-chip inductors (with
’s of at least 30) combined with off-chip voltage-tunable
varactor diode capacitors (with ’s of at least 40). These
relatively low values are justified for this VCO because
in the synthesizer, this oscillator ends up slaved to a much
more stable reference crystal oscillator, which cleans up
the VCO’s phase-noise spectrum at small ’s through the
action of a PLL. As such, the phase-noise specifications for
this VCO are the easiest to satisfy in a given transceiver
system, with numbers like 100 dBc/Hz at 10 kHz offset
from the carrier (for European GSM) being typical.
Even with this seemingly easy goal, however, attempts
at miniaturization based on conventional IC technologies
fall well short of this mark. In particular, CMOS ring
(a)
or relaxation oscillators attain only 60 dBc/Hz at 10
kHz offset [20]—not nearly adequate for GSM purposes.
Furthermore, even slightly modified silicon CMOS and
bipolar processes that include on-chip spiral inductors [21],
or even bond wire inductors [22], yield LC VCO’s with
inadequate phase-noise performance ( 85 dBc/Hz). A
large part of the problem arises from the tunable capacitor
implementation. Specifically, on-chip diodes used in place
of previous off-chip varactor diodes exhibit excessive series
resistance—much more than their off-chip varactor coun-
terparts—and as a result, their ’s are severely lacking. In
addition, the tuning range of such capacitors over available (b)
supply voltages is often limited to the point where trimming Fig. 4. (a) Overhead and (b) cross-sectional schematics of a
is required to set a starting capacitor value [23]. voltage-tunable micromechanical capacitor [12].
Recent demonstrations of voltage-tunable capacitors
composed of micromachined, movable, metal plates now to enhance coupling in inductive coils and 2) removal
offer substantial improvements over on-chip diode-based of lossy substrates by fabricating inductors on suspended
capacitors. Fig. 4 presents the schematic of one such membranes, achieved via familiar back-side etching tech-
voltage-tunable micromachined capacitor [12], consisting niques. Fig. 5 presents two very recent applications of these
of an aluminum top capacitor plate suspended by soft methods. The inductor in Fig. 5(a) utilizes an NiFe core
flexures above a bottom plate, in which the distance (and under a planar metal spiral to increase the achievable flux
thus capacitance) between the plates is electrostatically and attain 2.7 H of inductance (for m and
varied by action of the underlying electrode. The prototype m, and ten turns in an area of 2 10 mm )
capacitor of Fig. 4 exhibited a tuning range of about 16% with a of 6.6 at 4 MHz [13]. This configuration was
over a 5.5 V range of applied bias, with capacitance values found to give the highest relative to three other topologies
on the order of 2.2 pF (achieved with four of these devices combining metal coils with magnetic cores.
wired in parallel, each with m and Fig. 5(b) presents the schematic of a miniature spiral
m, nominally) and a of 62 at 1 GHz—all on par with inductor fabricated on a substrate-isolating platform (or
off-chip varactor diodes [12]. membrane) and achieving an overall inductance of 115 nH
with a of 22 at 275 MHz. Microwave/millimeter-wave
B. Micromachined Inductors inductors fabricated using similar techniques have achieved
Obviously, implementation of a miniature tunable ca- inductances on the order of 1.2 nH with associated self-
pacitor solves only part of the problem. To achieve res- resonance frequencies of 70 GHz with substrate removal,
onator tanks with adequate (at least for VCO’s), high- and 22 GHz without substrate removal, with expected ’s
inductors are required as well. As mentioned above, of 60–80 at 40 GHz. These inductors and others like them
numerous attempts to implement spiral inductors using have achieved some of the highest values for on-chip
conventional IC technologies have so far yielded inductors inductors at their respective frequencies.
with insufficient [1]–[3]. Even those using bond wires as Despite the above promising results, the inductors of
inductors fall short on for many applications [22]. Fig. 5 are still inadequate for some applications within com-
For the above reasons, research aimed at improving on- munication systems. In particular, as implied above, both
chip inductor performance via more exotic micromachining designs, even the one with substrate removed, suffer from
technologies is currently under way. Among the strategies parasitic self-resonance problems that limit their frequency
utilized for this purpose are 1) the use of magnetic cores range. Self-resonance is generally not a problem when small

1760 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

 (a)

(a)

(b)
Fig. 6. Cross sections of two thin-film bulk-acoustic resonators.
(b)
(a) A membrane supported FBAR resonator [4]. (b) A solidly
Fig. 5. Advanced integrated inductors utilizing micromachining mounted resonator [11].
processes. (a) Spiral inductor over an NiFe core. (b) Spiral inductor
over an isolating platform [13], [14].
ductors in a similar fashion to quartz crystals, all suspended
on a thin membrane, and so acoustically isolated from
inductance values are needed (e.g., 10 nH) but can be the substrate. Such resonators constructed using aluminum
significant when requirements exceed 20 nH. Second, for nitride piezoelectric films have been demonstrated with ’s
the case of Fig. 5(a) and for others based upon its design, of over 1000 and resonance frequencies of 1.5–7.5 GHz [4],
the magnetic core only reacts up to a certain frequency [5] in sizes of less than 400 400 m . Although very
range, beyond which its permeability drops to an ineffective promising, the device of Fig. 6(a) awaits improvements in
value [16]. Research to solve these problems is ongoing. process and trimming technologies. In addition, problems
with the structural integrity of the supporting membrane
C. Thin-Film Bulk Acoustic Resonators have been reported [11].
Assuming that ’s on the order of 40 can be achieved, One promising solution to the above drawbacks dispenses
the above tunable capacitors and on-chip inductors can with the fragile membrane support and allows construction
potentially find use as resonator tanks for VCO’s and low- of the thin-film resonator directly over a solid substrate,
RF filters, as well as elements for tunable couplers. as shown in Fig. 6(b). In this scheme, energy loss to
However, given the present struggle to attain just these the substrate (which would greatly attenuate the ) is
’s, and with questions of susceptibility to microphonics avoided by acoustically isolating the substrate from the
and Brownian motion noise unanswered, it is unlikely that piezoelectric resonator materials using impedance transfor-
such elements could be utilized to replace the vibrating mations obtained through strategic selection of the number
mechanical resonators (e.g., quartz crystals) used in the and thickness of layers separating the two media [11].
more demanding reference oscillator and channel-select Although the implementation of such a solidly mounted
filter functions. For these functions, as discussed in the resonator (SMR) requires more careful deposition of layers,
context of Figs. 2 and 3, much higher ’s (in the thousands) it promises to greatly improve the resiliency of thin-film
are required. bulk-acoustic resonators.
With the knowledge that such ’s are already attainable As with other novel technologies discussed here, thin-film
using macroscopic vibrating mechanical tanks (e.g., quartz bulk-acoustic resonators, with all of their advantages, still
crystals, SAW’s), much initial research had focused on have several important drawbacks that presently hinder their
developing exotic thin-film technologies to yield minia- use. First, a convenient and effective means for trimming
ture versions of vibrating mechanical resonators operating and tuning these resonators is not yet available. Such
under principles similar to their macroscopic counterparts. a trimming technology is vitally important, especially if
Fig. 6(a) presents the schematic of one such thin-film bulk- groups of these resonators are to be used to implement
acoustic mode, piezoelectric resonator (FBAR), composed small percent bandwidth filters with small shape factors.
of deposited piezoelectric films sandwiched between con- Second, these filters are presently most appropriate for the

NGUYEN et al.: MICROMACHINED DEVICES 1761

high UHF- and S-band frequency ranges, becoming overly
thick and cumbersome at lower frequencies.
(a)

D. Micromechanical Resonators and Filters

For lower frequency applications, planar IC-compatible
micromachining processes have now realized flexural-mode
micromechanical resonators in a variety of structural mate-
rials, and (so far) in a range of frequencies from low fre-
quency (LF) to very high frequency (VHF). ’s exceeding
80 000 in vacuum have been measured for LF flexural-mode
resonators constructed in surface-micromachined polysil-
icon [24], while ’s on the order of 20 000 have been
achieved at 70 MHz (VHF) in single-crystal silicon material
[25]. Since the use of this technology for high-frequency (b)
applications is quite recent, its ultimate frequency limit is
as yet unknown. Operating frequencies into the gigahertz
range, however, are not unreasonable [9], [10].
In addition to their enormous values and wide ap-
plicable frequency range, micromechanical resonators are
extremely flexible from a design perspective, having several
features that greatly simplify the design and implementation
of complex resonator systems. Among their most attractive (c)
features are:

1) an inherent voltage-controlled frequency tunability

[26] and switchability [10]; Fig. 7. (a) Perspective-view schematic of an HF micromechanical
2) an amenability to trimming [27]; filter in a typical bias and excitation configuration. (b) Equivalent
mechanical circuit for the filter of (a) using quarter-wavelength
3) wide flexibility in available geometries (leading to a coupling. (c) Equivalent electrical circuit for the filter of (a), again
seemingly limitless range of possible designs); with quarter-wavelength coupling.

4) flexibility in the choice of structural materials used;

5) flexibility in the type of transduction used (electro- the passband of the filter. The center frequency of the filter
static, piezoelectric, and magnetostrictive have all is determined primarily by the resonance frequency of
been utilized in the past). the identical resonators, while its bandwidth is dictated by
the relative stiffnesses of the coupling beam and resonators
On top of all of this, successful construction of devices and is given specifically by the expression
using such resonators is relatively straightforward, since
critical features are usually defined by a single masking BW (2)
step, which itself is part of a planar process largely com-
patible with conventional IC processes. where is the stiffness of the coupling beam joining
The above features, in particular those associated with resonators and , is the resonator stiffness, and is
tuning and design flexibility, have greatly accelerated the a normalized coupling coefficient found in filter cookbooks
rate at which more complex oscillator and filtering appli- [28]. By using electromechanical analogies, equivalent elec-
cations of the technology have been realized. Specifically, trical circuits resembling LC ladders can be derived for
micromechanical filters comprised of multiple resonators micromechanical filters, as illustrated in Fig. 7(b) and (c).
coupled by soft mechanical springs have recently been im- Such circuit modeling strategies make micromechanical
plemented with performance attributes comparable to some filters amenable to computer-aided design, synthesis, and
of the best high- filters available. Fig. 7(a) presents the verification and allow the use of the already enormous
perspective-view schematic for an HF-range, two-resonator existing knowledge on LC ladder filter synthesis for mi-
filter, composed of two identical clamped-clamped beam cromechanical filter design.
micromechanical resonators coupled by a flexural-mode To date, two-resonator micromechanical bandpass filters
coupling beam. As shown, resonance motion perpendicular (Fig. 8) have been demonstrated with frequencies up to
to the substrate is excited and sensed electrostatically using 14.5 MHz, percent bandwidths on the order of 0.2%, and
a combination of dc-bias and ac voltages applied insertion losses less than 1 dB [29]. Higher order, three-
between the conductive resonators and electrodes under- resonator filters with frequencies near 455 kHz have also
lying each resonator. In this device, the coupling beams been achieved (Fig. 9), with equally impressive insertion
couple energy between resonators, creating a coupled two- losses for 0.09% bandwidths and with more than 64 dB of
resonator system with two modes of vibration that define passband rejection [30], [31]. The filter of Fig. 9 features

1762 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

 Fig. 10. SEM of a 16.5 kHz CMOS microresonator oscillator
Fig. 8. SEM of a two-resonator, surface-micromachined, HF with schematics explicitly depicting circuit topology. The microres-
micromechanical filter with a measured frequency characteristic onator occupies 420 2 230 m2 [32].
[29].

noise [33]. Thus, much work is still needed to attain optimal

performance of micromechanical resonator oscillators. Re-
search is also currently under way to extend the frequency
of this oscillator to the popular 10 MHz frequency of
reference oscillators used in many transceivers, and to do so
in a fashion that reduces the overall temperature coefficient
of the oscillator.
Again, as with the other technologies, micromechanical
resonator devices are not without their drawbacks. Among
the more disturbing of them are the need for vacuum to
attain high , an unaided temperature coefficient of 10
ppm/ C [10] (not as good as quartz), and uncertainties
concerning ultimate dynamic range and power-handling
capability [10]. Research on micromechanical resonators
and their applications is ongoing.

E. Micromachined Filters for K-Band and Higher

For transceivers operating at K-band and higher, preselect
or image-reject filters with bandwidths of 1–8% are imple-
mented using microwave and millimeter-wave components.
Low loss is essential for such front-end filters, since they
are placed just after the antenna, before any amplification,
Fig. 9. SEM’s of a surface-micromachined, medium-frequency, and thus, their loss adds directly to the noise figure of
three-resonator micromechanical filter with a measured transmis-
sion spectrum [31].
the receiver and reduces the effective radiated power from
the transmitter. Again, high- resonators are required to
minimize this loss, and again, this can only be achieved
balanced comb-transduction for feedthrough suppression, using external elements such as dielectric resonators or
low velocity coupling, and frequency tuning electrodes, waveguide components, which increase the size and cost
adeptly illustrating the complexity and flexibility achievable of the receiver.
using this technology. To address this problem, filter implementation using
In addition to filters, LF high- oscillators, fully inte- monolithic transmission-line (T-line) resonator approaches
grated with sustaining CMOS electronics, have also already has been investigated. The main problem with using T-
been demonstrated in this technology. Fig. 10 presents the line resonators on Si or GaAs involves the underlying high
overhead-view scanning electron microscope (SEM) of a dielectric constant substrate (with – ). It is well
16.5 kHz prototype of such an oscillator [32]. Recent known that T-lines integrated on these substrates suffer
studies of similar fully integrated oscillators have shown from radiation loss into the dielectric and increased ohmic
phase-noise performance expected of high- oscillator op- loss due to their small dimensions. This has limited the
eration, but they also show additional sources of phase noise of planar resonators on Si/GaAs to 40–60 at 30–60
related to nonlinear amplitude perturbations caused by 1/ GHz, which is not acceptable for low-loss filters. An

NGUYEN et al.: MICROMACHINED DEVICES 1763

intermediate solution is to integrate the planar filters on
teflon or quartz substrates with – , This results
in a resonator of 200–300 at 30–60 GHz, but is not
compatible with Si/GaAs integrated circuits and therefore
requires transitions between millimeter-wave active circuits
and the external filters.
One way to dramatically increase the of a planar
resonator on Si/GaAs is to integrate the resonator on a
thin dielectric membrane (1–1.5 m thick) and to enclose
the resonator in a microshielded cavity 100–500 m high
[34]–[40]. The fabrication sequence required for such a
resonator utilizes standard micromachining processes and
includes 1) the deposition of a silicon nitride membrane; 2)
back etching of the wafer; 3) etching and filling of via-holes
with gold; and 4) attachment of several wafers together to
form a miniature stack. The T-line resonators are effectively
suspended in free space and are therefore limited only by
ohmic loss. Furthermore, the resonators are wider than their
counterparts on Si/GaAs (so as to obtain the same filter
impedance), which considerably reduces their ohmic loss.
Recently measured ’s of quarter-wavelength resonators
( m, m) at 60 GHz are around 600,
which is ten times better than comparable resonators on Fig. 11. A 60 GHz, 8% bandwidth, four-pole micromachined
elliptic filter with a resonator width of 500 m and a resonator
Si/GaAs and only three to four times worse than waveguide Q of 500. The insertion loss in the passband is 1.5 dB referenced
resonators [36]. to the calibration planes.
Micromachined T-lines on thin dielectric membranes
have been used extensively in filter designs from 14 to
250 GHz [34]–[39]. The filters follow standard designs eventually to allow the integration of the millimeter-wave
(elliptic, Chebyshev, etc.), use well-known transmission- transmitters and receivers on the same chip. Research
line types (coplanar waveguide lines, stripline, suspended toward this goal is ongoing.
microstrip lines, etc.), and result in excellent performance.
A lot of attention has been placed on the transition between F. Micromechanical Switches
the Si/GaAs substrate and the membrane T-lines, and a In addition to the filter and oscillator research described
transition with better than 20 dB return loss is attainable above, a good amount of research effort has focused on the
up to 60 GHz [35], [36]. Recent results obtained for a four- implementation of micromechanical switches for antenna or
pole elliptic filter at 60 GHz show an insertion loss of only filter-path selection in multiband communications systems,
1.5 dB and a rejection better than 40 dB (Fig. 11) [36]. and for deployment of phased-array antennas in higher
This filter is state of the art and is only 0.5–0.7 dB more frequency systems operating past Ka-band (where antennas
lossy than its waveguide implementation (but with a 1000 become small enough for arrays). Such switches are often
size reduction). characterized by metrics describing switching speed and
Another application of micromachining is the fabrication off/on impedance. So far, the majority of switches for
of three-dimensional cavities to synthesize miniature wave- communications have operated via electrostatic actuation.
guide components at 10–60 GHz. In this case, a reduced Fig. 12 presents the cross-sectional schematic of a typical
height resonant waveguide cavity (with dimensions of /2- single-pole, single-throw micromechanical switch. Such
square) is etched in a silicon wafer and fed either by a slot switches are now developed by several U.S. industries
transition or by bond wires. Since the fields are not confined [45]–[50]. As shown, the switch of Fig. 12 utilizes an air-
to planar resonators, these structures have yielded ’s bridge design, in which a bridge of conducting material is
around 500 at 10 GHz, and up to 1100 at 30 GHz [41], [42]. suspended 3–4 m over a coplanar line. The top and bottom
Research is under way to use these miniature waveguide plates of this switch are constructed of evaporated or plated
cavities in satellite filters and low-phase noise oscillators. aluminum or gold. For the case of microwave/millimeter-
One final advantage of micromachined lines is the mi- wave switches, actual metal-to-metal contact is not neces-
cropackaging aspect of this technique. As seen in Fig. 11, sary; rather, a step change in plate-to-plate capacitance also
the filter (or any component inside the microcavity) is realizes switching. Thus, in high-frequency applications, a
completely shielded and therefore will not couple any RF protective nitride film (on the order of 1000 Å) often resides
energy to the outside world. The typical measured isolation above the bottom electrode plate to prevent sticking when
between adjacent transmission lines is 50 dB at 30 GHz, plates are pulled together. The dimensions of the top plate
and 45 dB between closely spaced filter banks at 10 GHz are around 300–400 m-square for a 10–20 GHz switch,
[43], [44]. Such micropackaging techniques are expected and 200 m for 60 GHz designs.

1764 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

 combining, etc.) because of their very low insertion
loss (which eliminates the need for loss-recovering
power amplifiers) and near-zero dc power consumption
(which allows substantial power savings). Research on
micromechanical switches continues, and new versions
(a)
with all of the above advantages, but with much smaller
actuation voltage requirements and much faster response
times, are presently under investigation [50].

G. Micromachined Antennas and Synthesized Substrates

(b)
Micromachining technologies first found use for antennas
Fig. 12. Cross-sectional schematics of a typical micromechanical
switch. (a) Switch up. (b) Switch down [45]. at millimeter-wave and terahertz frequencies to alleviate
loss and size constraints at such high frequencies. In
particular, antennas on Si/GaAs and even quartz dielectrics
Micromechanical switches have been fabricated in both
are inefficient radiators at millimeter-wave and terahertz
series and shunt (to ground) configurations. The insertion
frequencies due to the electrical thickness of the underlying
loss of a typical shunt switch in the “on-state” position
substrate. Micromachining has therefore been used to sus-
(bridge is up and capacitance is low) is around 0.1 dB at
pend dipole or slot radiators on thin dielectric membranes
10 GHz and increases linearly with frequency to 0.6 dB
(without the shielding cavities described in previous filters),
at 60 GHz. The isolation of a micromechanical switch in
with the planar antennas then effectively radiating in free
the “off-state” position (bridge is low and capacitance is
space. Anisotropic etching has also been used to integrate
high) is directly proportional to the bridge area and can
silicon pyramidal cavities that act as miniature integrated
be designed to be 20 to 25 dB at 20 GHz, at 40
horns around the dipole antennas, directing the radiated
GHz, or at 60 GHz [45]. To date, metal-to-metal series
energy very efficiently in the forward direction [51], [52].
switches that have shown impressive “off-state” (bridge is
For microwave frequencies, membrane technology with
up and capacitance is low) isolation at 1–4 GHz ( 60 to
results in physically large antennas and thus is
50 dB), with a very low “on-state” (bridge is down and
not practical. Therefore, at these frequencies, membrane
metal-to-metal contact is achieved) insertion loss ( 0.1
approaches are avoided, and micromachining is instead
dB), have been demonstrated [46]. Currently, most shunt
used to synthesize artificial dielectric-constant substrate
switches can safely handle 0.5–2 W of RF power in the “on-
materials underneath the antennas. In particular, by etching
state” position before hot-switching becomes a problem, a
a portion of the dielectric or by strategically choosing
phenomena where the RF voltage under the switch in the
the density of via-holes underneath an antenna, a di-
“on-state” position develops enough of a voltage to actually
electric constant between one (complete removal) and
pull the micromechanical bridge down.
(no removal) can be synthesized [Fig. 13(a)]. Using such
Micromechanical switches, such as those shown in
techniques, the efficiency of microstrip antennas at 12–13
Fig. 12, normally outperform those implemented with PIN
GHz has been increased from 55% ( ) to 85%
diodes or GaAs field-effect transistors (FET’s) in “on-
( ) [53], [54]. Similar techniques have also
state” insertion loss and “off-state” isolation. They also
been applied at 77 GHz for automotive radar applications
consume zero power when activated, unlike their solid-
[55]. Furthermore, the dielectric constant around a planar
state counterparts, which sink a finite amount of current
antenna can be artificially tailored so as to further increase
when activated. However, most are much slower than PIN
radiation efficiency and bandwidth.
or FET diode switches (4–20 s versus 1–40 ns), and they
Last, micromachining techniques have been used to syn-
so far require relatively high actuation voltages (20–60 V
thesize photonic bandgap materials on high-resistivity sili-
versus 3–5 V). Micromechanical switches are also prone to
con substrates at 30 GHz by etching a periodic pattern of
stiction problems in a metal-to-metal switch and dielectric
hexagonal holes ( m, period m) in the
charging problems in case of a nitride film between the
silicon substrate [Fig. 13(b)] [56], [57]. The holes create
electrodes, and research is still being conducted to evaluate
an “exclusion” of electromagnetic modes in the dielectric
their ultimate lifetimes (now at billions of cycles). However,
substrate and therefore dramatically increase the radiation
there is one aspect of micromechanical switches that makes
efficiency of planar antennas (from 30% without holes to
them extremely favorable for communications systems:
85% with holes). It is expected that photonic bandgap
Micromechanical switches are extremely linear devices.
materials will play an important role in the integration of
A 2 GHz switch in a two-tone experiment resulted in
high-efficiency antennas on Si and GaAs substrates.
unmeasured intermodulation products and an extrapolated
IP3 of 66 dBm (4000 W).
Micromechanical switches are particularly attractive in IV. CONCLUSIONS
low-power RF/microwave/millimeter-wave communica- Due to the need for values beyond the capabilities
tions systems because of their low distortion characteristics, of conventional IC technologies, board-level passive com-
and for active antenna applications (phased arrays, power ponents continue to occupy a substantial portion of the

NGUYEN et al.: MICROMACHINED DEVICES 1765

 and C.-Y. Chi, T. Weller, S. Robinson, A. Brown, R.
Henderson, and S. Pacheco, who are responsible for the
microwave and millimeter-wave effort.

REFERENCES

[1] E. Frian, S. Meszaros, M. Chuaci, and J. Wight, “Computer-

(a) aided design of square spiral transformers and inductors,” in
1989 IEEE MTT-S Dig., pp. 661–664.
[2] N. M. Nguyen and R. G. Meyer, “Si IC-compatible inductors
and passive filters,” IEEE J. Solid-State Circuits, vol. 25,
pp. 1028–1031, Aug. 1990.
[3] , “A 1.8-GHz monolithic voltage-controlled oscilla-
tor,” IEEE J. Solid- State Circuits, vol. 27, no. 3, pp. 444–450,
1992.
[4] S. V. Krishnaswamy, J. Rosenbaum, S. Horwitz, C. Yale, and R.
A. Moore, “Compact FBAR filters offer low-loss performance,”
Microwaves RF, pp. 127–136, Sept. 1991.
[5] R. Ruby and P. Merchant, “Micromachined thin film bulk
acoustic resonators,” in Proc. 1994 IEEE Int. Frequency Control
Symp., Boston, MA, June 1–3, 1994, pp. 135–138.
[6] P. R. Gray and R. G. Meyer, “Future directions in silicon IC’s
(b) for RF personal communications,” in Proc., 1995 IEEE Custom
Integrated Circuits Conf., Santa Clara, CA, May 1–4, 1995, pp.
Fig. 13. (a) Microstrip antennas on synthesized dielectrics built 83–90.
using partial removal of the silicon substrate. (b) A high-efficiency [7] A. A. Abidi, “Direct-conversion radio transceivers for digital
tapered slot antenna on a micromachined photonic bandgap sub- communications,” IEEE J. Solid-State Circuits, vol. 30, pp.
strate at 30 GHz. 1399–1410, Dec. 1995.
[8] D. H. Shen, C.-M. Hwang, B. B. Lusignan, and B. A. Woo-
ley, “A 900-MHz RF front-end with integrated discrete-time
overall area in superheterodyne transceivers, presenting a filtering,” IEEE J. Solid-State Circuits, vol. 31, pp. 1945–1954,
key bottleneck against further miniaturization. Although Dec. 1996.
[9] C. T.-C. Nguyen, “High- micromechanical oscillators and
there have been many research efforts aimed at replacing filters for communications,” in Proc. 1997 IEEE Int. Symp.
superheterodyne architectures with alternatives that use Circuits and Systems, Hong Kong, June 9–12, 1997, pp.
2825–2828.
higher levels of transistor integration to eliminate the need [10] , “Frequency-selective MEMS for miniaturized communi-
for off-chip passive elements, none of these new approaches cation devices,” in Proc. 1998 IEEE Aerospace Conf., Snow-
has yet been able to match, let alone improve, the perfor- mass, CO, Mar. 21–28, 1998, pp. 445–460.
[11] K. M. Lakin, G. R. Kline, and K. T. McCarron, “Development
mance of existing systems. of miniature filters for wireless applications,” IEEE Trans.
Micromechanical resonators and micromachined pas- Microwave Theory Tech., vol. 43, pp. 2933–2939, Dec. 1995.
sives, on the other hand, offer an alternative set of strategies [12] D. J. Young and B. E. Boser, “A micromachined variable
capacitor for monolithic low-noise VCO’s,” in Tech. Dig. 1996
for transceiver miniaturization and improvement, the sim- Solid-State Sensor and Actuator Workshop, Hilton Head Island,
plest of which may merely retain proven superheterodyne SC, June 3–6, 1996, pp. 86–89.
architectures, the boldest of which may revolutionize the [13] J. A. Von Arx and K. Najafi, “On-chip coils with integrated
cores for remote inductive powering of integrated microsys-
way future transceivers are designed. In particular, the tems,” in Dig. Tech. Papers 1997 Int. Conf. Solid-State Sen-
availability of high- micromachined filters, with their tiny sors and Actuators (Transducers’97), Chicago, IL, June 16–19,
size and zero dc power dissipation, may encourage future 1997, pp. 999–1002.
[14] B. Ziaie, N. K. Kocaman, and K. Najafi, “A generic micro-
architectures that take advantage of large arrays of such machined silicon platform for low-power, low-low miniature
filters to yield novel transceivers with multiband capability transceivers,” in Dig. Tech. Papers 1997 Int. Conf. Solid-State
Sensors and Actuators (Transducers’97), Chicago, IL, June
and enhanced security against jamming or interception. 16–19, 1997, pp. 257–260.
In combination with low clock-rate subsampling down- [15] M. G. Allen, “Micromachined intermediate and high frequency
converters [8], such an array architecture might also provide inductors,” in Proc. 1997 IEEE Int. Symp. Circuits and Systems,
Hong Kong, June 9–12, 1997, pp. 2829–2832.
substantial power savings over previous systems, especially [16] C. H. Ahn, Y. J. Kim, and M. G. Allen, “A fully integrated
when used in frequency-hopping scenarios. Needless to say, micromachined toroidal inductor with nickel-iron magnetic core
research on micromachined devices for communications (the switched DC/DC boost converter application),” in Dig.
Tech. Papers 7th Int. Conf. Solid-State Sensors and Actua-
continues on both device and system levels, with promises tors (Transducers’93), Yokohama, Japan, June 7–10, 1993, pp.
to influence wireless communications greatly in the near 70–73.
future. [17] J. D. Cressler, D. L. Harame, H. H. Comfort, J. M. C. Stork,
B. S. Meyerson, and T. E. Tice, “Silicon–germanium hetero-
junction bipolar technology: The next leap for silicon?” in Dig.
ACKNOWLEDGMENT Tech. Papers 1994 ISSCC, San Francisco, CA, Feb. 1994, pp.
24–27.
The authors gratefully acknowledge substantial contri- [18] W. P. Robins, Phase Noise in Signal Sources. London, UK:
butions from former and present graduate students, in Peregrinus, 1982.
[19] N. Slawsby, “Frequency control requirements of radar,” in Proc.
particular F. Bannon III, K. Wang, and J. Clark, who are 1994 IEEE Int. Frequency Control Symp., June 1–3, 1994, pp.
largely responsible for the micromechanical filter results, 633–640.

1766 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998

[20] T. C. Weigandt, B. Kim, and P. R. Gray, “Analysis of timing [42] A. R. Brown, P. Blondy, K. Hong, and G. M. Rebeiz, “Low-
jitter in CMOS ring oscillators,” in Proc. ISCAS’94, June 1994, loss millimeter-wave filters and high- micromachined cavity
pp. 27–30. resonators,” U.S. Army Research Office Interim Rep., Dec. 18,
[21] N. M. Nguyen, “A 1.8-GHz monolithic LC voltage-controlled 1997.
oscillator,” IEEE J. Solid-State Circuits, vol. 27, pp. 444–450, [43] R. F. Drayton, R. M. Henderson, and L. P. B. Katehi, “Ad-
Mar. 1992. vanced monolithic packaging concepts for high performance
[22] J. Craninckx and M. S. J. Steyaert, “A 1.8 GHz CMOS low- circuits and antennas,” in Dig. 1996 IEEE MTT-S, June 1996,
phase-noise voltage controlled oscillator with prescaler,” IEEE pp. 1615–1618.
J. Solid-State Circuits, vol. 30, pp. 1474–1482, Dec. 1995. [44] A. R. Brown and G. M. Rebeiz, “Micromachined micropack-
[23] M. Soyuer, K. A. Jenkins, J. N. Burghartz, and M. D. Hulvey, aged filter banks,” IEEE Microwave Guided Wave Lett., vol. 8,
“A 3V 4GHz NMOS voltage-controlled oscillator with inte- pp. 158–160, Mar. 1998.
grated resonator,” in Tech. Dig. 1996 ISSCC, San Francisco, [45] C. Goldsmith, J. Randall, S. Eshelman, T. H. Lin, D. Denniston,
CA, Feb. 1996, pp. 394–395. S. Chen, and B. Norvell, “Characteristics of micromachined
[24] C. T.-C. Nguyen and R. T. Howe, “Quality factor control for switches at microwave frequencies,” in Dig. IEEE MTT-S, June
micromechanical resonators,” in Tech. Dig. IEEE Int. Electron 1996, pp. 1141–1144.
Devices Meeting, San Francisco, CA, Dec. 14–16, 1992, pp. [46] J. Yao and M. F. Chang, “A surface micromachined minia-
505–508. ture switch for telecommunications applications with signal
[25] A. N. Cleland and M. L. Roukes, “Fabrication of high frequency frequencies from DC to 4 GHz,” in Proc. 8th Int. Conf. Solid-
nanometer scale mechanical resonators from bulk Si crystals,” State Sensors and Actuators, Transducers, June 1995, pp. 384–
Appl. Phys. Lett., vol. 69, no. 18, pp. 2653–2655, Oct. 28, 387.
1996. [47] Personal communications, Northrop Grumman, Baltimore, MD,
[26] R. T. Howe and R. S. Muller, “Resonant microbridge vapor sen- 1997.
sor,” IEEE Trans. Electron Devices, vol. ED-33, pp. 499–506, [48] Personal communications, HRL Laboratories, Malibu, CA,
1986. 1997.
[27] K. Wang, A.-C. Wong, W.-T. Hsu, and C. T.-C. Nguyen, [49] Personal communications, Honeywell Technology Center, Min-
“Frequency-trimming and -factor enhancement of microme- neapolis, MN, 1997.
chanical resonators via localized filament annealing,” in Dig. [50] S. Pacheco, C. T.-C. Nguyen, and L. P. B. Katehi, “Microme-
Tech. Papers 1997 Int. Conf. Solid-State Sensors and Actuators, chanical electrostatic K-band switches,” presented at the IEEE
Chicago, IL, June 16–19, 1997, pp. 109–112. MTT-S International Microwave Symposium, June 1998.
[28] A. I. Zverev, Handbook of Filter Synthesis. New York: Wiley, [51] G. M. Rebeiz, D. P. Kasilingam, Y. Guo, P. A. Stimpson, and
1967. D. B. Rutledge, “Monolithic millimeter-wave two-dimensional
[29] F. D. Bannon III and C. T.-C. Nguyen, “High frequency horn imaging arrays,” IEEE Trans. Antennas Propagat., vol.
microelectromechanical IF filters,” in Tech. Dig. 1996 IEEE 38, pp. 1473–1482, Sept. 1990.
Electron Devices Meeting, San Francisco, CA, Dec. 8–11, 1996, [52] W. Y. Ali-Ahmad and G. M. Rebeiz, “An 86–106 GHz quasi-
pp. 773–776. integrated low-noise receiver,” IEEE Trans. Microwave Theory
[30] K. Wang and C. T.-C. Nguyen, “High-order micromechanical Tech., vol. MTT-41, pp. 558–564, Apr. 1993.
electronic filters,” in Proc. 1997 IEEE Int. Micro Electro Me- [53] G. P. Gauthier, A. Courtay, and G. M. Rebeiz, “Microstrip
chanical Systems Workshop, Nagoya, Japan, Jan. 26–30, 1997, antennas on synthesized low dielectric constant substrates,”
pp. 25–30. IEEE Trans. Antennas Propagat., vol. AP-45, pp. 1310–1314,
[31] K. Wang, J. R. Clark, and C. T.-C. Nguyen, “ -enhancement Aug. 1997.
of micromechanical filters via low-velocity spring coupling,” in [54] J. Papapolymerou, R. F. Drayton, and L. Katehi, “Microma-
Proc. 1997 IEEE Int. Ultrasonics Symp., Toronto, Ont., Canada, chined patch antennas,” IEEE Trans. on Antennas Propagat.,
Oct. 5–8, 1997, to be published. vol. AP-46, pp. 275–283, Feb. 1998.
[32] C. T.-C. Nguyen and R. T. Howe, “CMOS micromechanical [55] M. Stotz, G. Gottwald, and H. Haspeklo, “Planar millimeter-
resonator oscillator,” in Tech. Dig. IEEE Int. Electron Devices wave antennas using SiNx-membranes on GaAs,” IEEE Trans.
Meeting, Washington, DC, Dec. 5–8, 1993, pp. 199–202. Microwave Theory Tech., vol. 44, pp. 1593–1595, Sept.
[33] T. Roessig, R. T. Howe, and A. P. Pisano, “Nonlinear mixing in 1996.
surface-micromachined tuning fork oscillators,” in Proc. 1997 [56] T. J. Ellis and G. M. Rebeiz, “MM-wave tapered slot antennas
IEEE Frequency Control Symp., Orlando, FL, May 27–28, 1997, on micromachined photonic bandgap dielectrics,” in Proc. IEEE
to be published. MTT-S Int. Microwave Symp., June 1996, pp. 1157–1160.
[34] C. Y. Chi and G. M. Rebeiz, “Planar microwave and millimeter- [57] T. J. Ellis, J. Muldavin, and G. M. Rebeiz, submitted for
wave lumped elements and coupled-line filters using micro- publication.
machining techniques,” IEEE Trans. Microwave Theory Tech.,
vol. 43, pp. 730–738, Apr. 1995.
[35] T. M. Weller, L. P. Katehi, and G. M. Rebeiz, “High perfor-
mance microshield line components,” IEEE Trans. Microwave
Theory Tech., vol. 43, pp. 534–543, Mar. 1995. Clark T.-C. Nguyen (Member, IEEE) was born
[36] P. Blondy, A. R. Brown, D. Cros, and G. M. Rebeiz, “Low-loss in Austin, TX, on March 29, 1967. He received
micromachined elliptic filters for millimeter wave telecommu- the B.S., M.S., and Ph.D. degrees from the
nication systems,” presented at the IEEE International MTT-S University of California at Berkeley in 1989,
Symposium, June 1998. 1991, and 1994, respectively, all in electrical
[37] G. M. Rebeiz, L. P. Katehi, T. M. Weller, C. Y. Chi, and engineering and computer sciences.
S. V. Robertson, “Micromachined filters for microwave and In 1995, he joined the Faculty of the Univer-
millimeter-wave applications,” Int. J. Microwave Millimeter- sity of Michigan, Ann Arbor, where he currently
Wave Computer Aided Eng., vol. 7, pp. 149–166, Feb. is an Assistant Professor in the Department
1997. of Electrical Engineering and Computer Sci-
[38] S. V. Robertson, L. P. Katehi, and G. M. Rebeiz, “Microma- ence. From 1995 to 1997, he was a Member
chined W-band filters,” IEEE Trans. Microwave Theory Tech., of the National Aeronautics and Space Administration (NASA)’s New
vol. 44, pp. 598–606, Apr. 1996. Millennium Integrated Product Development Team on Communications,
[39] C. Y. Chi and G. M. Rebeiz, “Conductor-loss limited stripline which roadmaps future communications technologies for NASA use
resonators and filters,” IEEE Trans. Microwave Theory Tech., into the turn of the century. He now is a Consulting Member. His
vol. 44, pp. 626–630, Apr. 1996. research interests focus upon microelectromechanical systems and in-
[40] T. M. Weller, L. P. Katehi, and G. M. Rebeiz, “A 250 GHz clude integrated micromechanical signal processors and sensors, merged
microshield band-pass filter,” IEEE Microwave Guided Wave circuit/micromechanical technologies, radio-frequency communication ar-
Lett., vol. 5, pp. 153–155, May 1995. chitectures, and integrated circuit design and technology.
[41] J. Papapolymerou, J. C. Cheng, J. East, and L. Katehi, “A Prof. Nguyen received the 1938E Award for Research and Teaching
micromachined high- X-band resonator,” IEEE Microwave Excellence from the University of Michigan in 1988 and is a Finalist for
Guided Wave Lett., vol. 7, pp. 168–170, June 1997. the 1998 Discover magazine Technological Innovation Awards.

NGUYEN et al.: MICROMACHINED DEVICES 1767

 Linda P. B. Katehi (Fellow, IEEE) received Gabriel M. Rebeiz (Fellow, IEEE) received the
the B.S.E.E. degree from the National Technical Ph.D. degree in electrical engineering from the
University of Athens, Greece, in 1977, and the California Institute of Technology, Pasadena, in
M.S.E.E. and Ph.D. degrees from the University 1988.
of California, Los Angeles, in 1981 and 1984, He joined the Faculty of the University of
respectively. Michigan, Ann Arbor, in 1988 and became
In 1984, she joined the Faculty of the Depart- Associate Professor in 1992. He was a Visiting
ment of Electrical Engineering and Computer Professor at Chalmers University of Technology,
Science of the University of Michigan, Ann Ar- Göteborg, Sweden, in 1992 and a Visiting Pro-
bor. Since then, she has been interested in the de- fessor at the Ecole Normale Superieur, France,
velopment and characterization (theoretical and in 1993. In 1997, he was a Visiting Professor
experimental) of microwave, millimeter printed circuits, the computer- at Tohoku University, Sendai, Japan. His interests are in applying mi-
aided design of very-large-scale-integration interconnects, the develop- cromachining techniques in silicon and GaAs for the development of
ment and characterization of micromachined circuits for millimeter-wave low-loss and low-cost microwave antennas, components and subsystems
and submillimeter-wave applications, and the development of low-loss for wireless applications, and satellite communication systems. He also is
lines for terahertz-frequency applications. She also has been studying interested in the development of planar collision-avoidance sensors for au-
theoretically and experimentally various types of uniplanar radiating tomotive applications and in millimeter-wave imaging arrays, monopulse
structures for hybrid-monolithic and monolithic oscillator and mixer tracking systems, and phased arrays. He is the author of 70 papers
designs. published in refereed journals and more than 120 papers presented at
Dr. Katehi received the IEEE AP-S W. P. King (Best Paper Award for a national and international conferences.
Young Engineer) in 1984, the IEEE AP-S S. A. Schelkunoff Award (Best Prof. Rebeiz received the National Science Foundation Presidential
Paper Award) in 1985, the NSF Presidential Young Investigator Award Young Investigator Award in April 1991 and the URSI International Isaac
and an URSI Young Scientist Fellowship in 1987, the Humboldt Research Koga Gold Medal Award for Outstanding International Research in August
Award and the University of Michigan Faculty Recognition Award in 1994, 1993. He received the Research Excellence Award in April 1995 from the
the IEEE MTT-S Microwave Prize in 1996, and the Best Paper Award from University of Michigan. Together with his students, he received Best Paper
the International Society on Microelectronics and Advanced Packaging. Awards at JINA 1990, IEEE-MTT 1992 and 1994–1997, and IEEE-AP
She is a member of the IEEE Antennas and Propagation and Microwave 1992 and 1995.
Theory and Techniques societies, Sigma XI, Hybrid Microelectronics, and
URSI Commission D. She was a member of AP-S ADCOM from 1992 to
1995. She is an Associate Editor for IEEE TRANSACTIONS ON MICROWAVE
THEORY AND TECHNIQUES.

1768 PROCEEDINGS OF THE IEEE, VOL. 86, NO. 8, AUGUST 1998