Professional Documents
Culture Documents
RF Mems
RF Mems
and Applications
Peter Gammel, Georg Fischer, and Jrmie Bouchaud
Will RF MEMS deliver on its promise to be one of the major MEMS businesses
in the future? RF MEMS has now entered its second generation; high-volume
production of both contacting and non-contacting switches and lters
(based on bulk acoustic waves) was ramped up worldwide starting in 2003.
In rst-generation RF MEMS, discrete components were developed to
demonstrate RF performance and reliability, and low-cost, high-volume chip-
and wafer-scale packaging eliminated the nal barriers to high-volume
production of discrete replacement components with MEMS. In this
paper, we review the RF MEMS industry and the commercialization status
of RF MEMS components. We also provide a detailed case study of
mobile telephone basestations, demonstrating how RF MEMS can impact
telecommunications. Finally, we look at research into third-generation RF
MEMS and NEMS, in which integration, active devices, and new functionality
will create opportunities for disruptive changes in RF subsystems, enabling
such things as software-dened radio. 2005 Lucent Technologies Inc.
Introduction
Radio frequency (RF) components have been paved the way for RF MEMS-based modules and in-
viewed as an ideal opportunity for microelectro- tegration. The introduction of bulk acoustic wave
mechanical systems (MEMS). One reason for this is (BAW) RF MEMS lters by Agilent and Skyworks is
that high frequency and especially high power RF the rst step along this path. RF MEMS are actively
products are still based on discrete components and being evaluated for switch functions in mobile tele-
selected based on performance and MEMS have al- phones. As foundry and electronic design and au-
ready demonstrated superior performance in a num- tomation (EDA) services for MEMS become available,
ber of critical device metrics. Moreover, these products RF MEMS integration in radio frequency integrated
often use hermetic or quasi-hermetic packaging and circuits (RFICs) will become commonplace.
chip and wire assembly, as do MEMS components. As
reliability and manufacturing yield for MEMS become The RF MEMS Industry
more stable, it is expected that MEMS will play an Despite the fact that the rst RF MEMS paper was
important role in these products. published back in 1979 [6], the RF MEMS industry is
Another reason is that the need to reduce cost still very young. The interest of the Defense Advanced
and footprint for high-volume, lower-frequency RF Research Projects Agency (DARPA) in military radar in
products (e.g., those used in mobile phones) has the mid-1990s spurred the rst signicant RF MEMS
Bell Labs Technical Journal 10(3), 2959 (2005) 2005 Lucent Technologies Inc. Published by Wiley Periodicals, Inc.
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/bltj.20103
Panel 1. Abbreviations, Acronyms, and Terms
3GThird generation MEMSMicroelectromechanical systems
4GFourth generation MP3Moving Picture Experts Group layer 3
BAWBulk acoustic wave MVNOMultiple virtual network operator
BMBFGerman Federal Ministry of Research and NEMSNanoelectromechanical systems
Education NFCNear-Field Communication
CDMACode division multiple access OMTOperations and maintenance terminal
DARPADefense Advanced Research Projects PAPower amplier
Agency PANPrivate area network
DCDirect current PCBPrinted circuit board
DMBDigital Multipoint Bridge PCSPersonal Communications Service
DoDDepartment of Defense PINPhoto intrinsic diode
DVBDigital Video Broadcasting Group R&DResearch and development
EDAElectronic design and automation RFRadio frequency
FBARFilm bulk acoustic resonator RFBRadio function block
FETField-effect transistor RFICRF integrated circuit
GPRSGeneral Packet Radio Service RMSRecongurable Mobile Systems
GSM*Global System for Mobile RTLRegister transfer language
Communications* SAWSurface acoustic wave
HDDHard disk drive SDARSSatellite Digital Audio Radio Services
ICIntegrated circuit SMRSolidly mounted resonator
LANLocal area network SP4TSingle-pole four-throw switch
LCInductor-capacitor lter SP7TSingle-pole seven-throw switch
LEDLight-emitting diode SPSTSingle-pole single-throw switch
LIGALithography, electroforming, and molding T/RTransmit/receive
LNALow noise amplier USBUniversal Serial Bus
LOLocal oscillator W-CDMAWideband CDMA
LTCCLow temperature co-red ceramic WiFiWireless Fidelity
research and development (R&D) effort, and it was Figure 1 shows the various RF MEMS components
2001 before the rst high-volume RF MEMS compo- and their respective operative frequency ranges.
nent, a lter based on bulk acoustic wave technology, In addition to their potential for integration and
was commercialized by Agilent. A signicant number miniaturization, RF MEMS offer lower power con-
of components and circuits have since been demon- sumption, lower losses, higher linearity, and higher
strated and now, in 2005, around 60 companies are Q factors than conventional communications compo-
developing RF MEMS components, a dozen of which nents. RF MEMS also enable new architectures for
have now entered the commercialization phase. future generations of telecommunication systems,
Examples of RF MEMS devices are microswitches, which will be easily and rapidly recongurable and
tunable capacitors, micromachined inductors, micro- will operate over a wide frequency range.
machined antennas, microtransmission lines, and res- Worldwide, more than 120 industrial and re-
onators, including micromechanical resonators, BAW search organizations are currently working on RF
resonators, and cavity resonators. RF MEMS are man- MEMS. Figure 2 shows the structure of the RF
ufactured using conventional 3D structuring tech- MEMS industry and lists some of the participants in it.
nologies, such as bulk micromachining, surface We have included MEMS start-ups, integrated circuit
micromachining, fusion bonding, and lithography, (IC) manufacturers, manufacturers of passive RF de-
electroforming, and molding (LIGA). The materials vices, and integrators that develop RF MEMS for their
used include Si, GaAs, SiC, and SOI substrates. own specic needs.
0 1 2 5 10 20 40 60 80 100
-mechanical
resonators
MEMS
inductors
BAW
resonators
MEMS
switches
Tunable
capacitors
Micro-mach.
antennas
Cavity
resonators
Transmission
lines
Figure 1.
RF MEMS components and their operative frequency range.
RF MEMS start-up companies play a major role in fast. On the one hand, automotive electronics com-
the commercialization of RF MEMS, because they are, panies (e.g., Robert Bosch and TRW) and U.S. defense
in most cases, the rst companies to put products on system companies (e.g., Raytheon) are either invest-
the market. The start-ups Magfusion, Teravicta, and ing less in or stopping the development of their own
Radant have been shipping commercial RF MEMS RF MEMS technologies, preferring to wait for com-
switches; Sophia Wireless will start shipping cavity mercial products to become available. On the other
resonators this year. One reason for the success of hand, European defense system companies (e.g.,
these companies is that they now receive more fund- Thales and BAE) have limited or no access to RF
ing from the U.S. Department of Defense (DoD) than MEMS technologies developed by American start-ups
the traditional large pioneers, such as Rockwell and funded by DARPA. As a result, they must continue
Raytheon. their own development efforts. EADS even started its
Although RF MEMS is a young industry, the RF own RF MEMS switch program in 2003.
MEMS supply chain is already undergoing restruc- As Figure 3 shows, developments in RF MEMS
turing. Start-ups involved in RF MEMS development have not been equally funded. North America in par-
(such as PHS MEMS) have gone bankrupt. The group ticular is choosing to nance RF MEMS developments
of companies that can be characterized as integrators for military use. The DoD has devoted a lot of atten-
with their own RF MEMS development is evolving tion and funding to the development of RF MEMS
RF MEMS manufacturers
Academic R&D
RF-MEMS Start-ups Integrators with own
Radant (USA) Large IC and MEMS MEMS development Integrators
Xcom (USA) Manufacturers & users
Michigan Uni (USA) Defense & aerospace
Magfusion (USA) Agilent (USA)
Purdue (USA) Intel (USA) Raytheon, Rockwell,
Teravicta (USA) Northrop Grum. (USA)
GiorgiaTec (USA) IBM (USA)
Discera (USA) Thales (F), BAE (UK), Wireless telecom
VTT (FIN) Motorola Labs (USA)
Wispry (USA) Raphael (Il), EADS (D) terminals and
IMEC (B) FreeScale (USA)
Siverta (USA) Infrastructure
CEA LETI (F) Infineon (D) Automotive
SiTime (USA)
LAAS (F) STMicroelectronics (F) Bosch (D) Automotive
SiliconClocks (USA)
Tokyo Uni. (J) Philips (NL)
KAIST (ROK) Samsung (ROK) Space & Defence
MEMS Start-ups
GPSC (Sgp.) NEC (J) RF devices manuf.
ITRI (TW)... MEMSCAP (F/USA)
MEMX (USA) LG (ROK) Fujitsu (J), Matsuchita (J)
Protron (D) Epcos (D), Murata (J)
MEMS Solution (ROK)
Services
Figure 2.
Structure of the RF MEMS industry.
Agilent (US)
BAW resonators Infineon (D)
(FBAR & SMR) Fujitsu (J)
APM (Tw)
LG (ROK)
Samsung (ROK) (estimated)
MEMS Solution (ROK)
Siverta (US)
Philips (NL)
Tunable Fujitsu (J)
capacitors
MEMX (US) (estimated)
Figure 4.
Commercialization plans of RF MEMS manufacturers for volume applications.
35
30
25
20
15
10
0
Switch Tunable Cavity mechanical BAW Inductor Tunable Antenna Transmission
capacitor resonator res. inductor lines
Figure 5.
RF MEMS program by device type.
Switches
Magfusion (USA)
High end
commercial Teravicta (USA)
applications
(DC-9 GHz) Matsushita EW (J) (estimated)
Radant (USA)
XCOM (USA)
Military & MEMTronics (USA) (estimated)
space
+ high end
commercial COM DEV (CND)
DC-86 GHz
TeraOp (IL)
Figure 6.
Commercialization plan of RF MEMS manufacturers for high-value applications.
10
Shipments (M)
0
2003 2004 2005 2006 2007 2008
Figure 7.
Hard disk drive shipments into mobile devices.
Accelerometer (ADI),
09/03 NTT DoCoMo: podometer, 06/04 Mitsubishi: auto picture
rotation, 09/04 Siemens: EMOTI, end 04: Nokia, Gaming
RFRadio frequency
Figure 8.
MEMS in mobile terminals.
41
42
Table II. Typical combinations of standards and frequency bands used in mobile communication (continued ).
Frequency agile
antenna Frequency agile
power amplifier
Frequency agile
duplex filter
Frequency agile
radio
CDMACode division multiple access UMTSUniversal Mobile Telecommunications System
RFRadio frequency WIMAXWorldwide interoperability for microwave access
Figure 9.
Multiband/multistandard basestation.
is acceptable. Such a reboot would typically be used to between different standard generations as well as net-
change baseband processing algorithms. work optimization. Recongurable basestations are
Reconguration on a per-call basis means that a valuable for real network operators that serve virtual
basestation can support any mode, i.e., an arbitrary ones called mobile virtual network operators
combination of band and standard for each new call (MVNOs). MVNOs may have different standard-band
that is established. This is the highest level of recon- strategies and may have granted different licenses. If
guration. It is obvious that reconguration on a per- real operators have the exibility to adjust for arbi-
call basis implies tremendous complexity in a trary standard-band combinations, they can serve a
basestation, because several calls of different mode great variety of MVNOs.
might have to be supported in parallel. Such a level of The factors that make reconguration attractive
reconguration is only sensible for terminals, which for basestations result in very infrequent alterations of
are typically required to handle only one call or RF the RF settings. Such changes may occur as infre-
channel at a time. For a commercial basestation, this quently as two times a year; so they are not highly dy-
complete agility is still considered to be unnecessarily namic. Consequently, they result in a very low
complex. However, in extraordinary defense and number of switching cycles over the lifetime of the
emergency response circumstances, there is still a components. This is advantageous in the context of
need to have command center basestations with total RF MEMS devices, because their reliability in terms of
agility. It has been estimated that, in a natural disas- switching cycles is typically limited. Moreover, be-
ter, an emergency response command center might cause the basestation must be rebooted in order to
have to coordinate as many as 700 different radios. change band, switching time is not critical.
Reconguration by reboot is very attractive for The benets of frequency and standard agility are
basestations; it allows them to support the operator not limited to the network operator. They also offer
spectrum migration strategies required by operator substantial benets to the infrastructure equipment
consolidations, license exchanges, and transitions vendor. These benets are present even if a basestation
Ferroelectric
Multiband RFB
or broadband RFB :n
Reconfigurable Reconfig prescaler 2/4/8
antenna Reconfig-
transmit filter
urable
or tunable Wide tuning
matching
postselector synthesizer UMTS:
4 Samples
Subharmonic direct mixing per chip
Multiband
15.3 MSa/s Base
duplexer
band
Wide tuning
Tunable Channel
Broadband
metamaterial Dual selection
RF-MEMS subharmonic
duplexer Gain AD filter
IQ mixer
ranging
RFB
Reconfigurable :n
Reconfig prescaler 2/4/8
receive filter
or tunable Wide tuning
preselector synthesizer
Figure 10.
Architectural concept of a multiband/multistandard basestation.
for the future is considered a valuable feature because system that spans such a wide range may be called a
software radio is viewed as a platform for loading new multioctave radio or multioctave RF system.
air interface protocols. Any recongurable RF system must include only
One motivation for using radio recongurability frequency agile RF units (e.g., the radio, the PA, the
mentioned above was that it provides fast reaction to antenna, and the lters). Figure 10 shows the archi-
changed regulatory conditions. Over time, more and tecture of a frequency agile RF system and some tech-
more new frequency bands are opened up for mobile nologies that have been proposed to facilitate
communication. Moreover, security and safety net- frequency agility.
works are set up in very specific frequency alloca- RF MEMS play a key role in a frequency agile ar-
tions, making the allocations of frequency bands even chitecture, because they enable exible lters, recon-
more diverse. It is therefore desirable not only to figurable matching networks, and widely tunable
switch among the predened settings of a nite num- synthesizers. At some points in the architecture the
ber of xed bands, but also to be able to continuously use of RF MEMS would be desirable but is not yet
tune the RF system solution. As Table II indicates, the possible; for example, the reconfigurable output
frequency range in which macrocellular networks can matching network of the PA and the duplex lter both
be found reaches from 300 MHz to 3500 MHz, which demand high RF power handling. Up to 40 W average
covers multiple octaves. A recongurable radio or RF and 200 W peak RF power are present at these nodes.
Figure 13.
Analog/RF path RF-MEMS switch (zoom) designed by Coventor.
For the SP4T conguration, the four toggle the 450, 900, 1800, and 2100 MHz bands. This test
switches were realized by using two single-pole single- board is shown in Figure 16.
throw (SPST) switches, so eight individual switches Figure 17 shows a wide-tuning range varactor
were required for the full conguration. The 5 mm with capacitance variation larger than factor 5. The
5 mm packaged switch matrix is shown in Figure 15. tuning range of the synthesizer in the proposed ar-
A test board for a recongurable lter was built in chitecture does not have to span multiple octaves,
which the RF MEMS switch matrix was used to thanks to the use of recongurable prescalers in the
switch between four different SAW lters, covering direct mixing architecture described above. A sub-
harmonic mixing approach was chosen, so that the
divider ratio of the prescalers is varied among 2, 4,
and 8. This approach results in synthesizers having
a tuning range of from 3 GHz to 4.4 GHz and a rela-
tive tuning range of 37%, which cannot be realized by
using a varicap diode. Either an RF MEMS varactor
can be used or the oscillator can be switched in two
ranges using RF MEMS switches.
RF MEMS varactors may be also used to realize
tunable bandpass lters in which the center frequency
can be tuned over a wide range. The RF MEMS var-
Figure 15. actor shown in Figure 17 offers a capacitance varia-
Housing of RF-MEMS switch matrix. tion larger than factor 5, which results in a variation
Amplifier
Drain
50 line 50 line
Transistor
Figure 20. Control
Recongurable amplier based on RF-MEMS switches.
Figure 22.
Tunable frequency selective amplier.
50 line 50 line
a peak in the gain can be obtained and attenuation at
Amplifier unwanted frequency ranges will be present. This tech-
nique offers a signicant advantage over the use of
broadband power ampliers. With broadband ampli-
Control ers, wideband noise in the transmit chain around re-
ceive frequencies would be amplied and would lead
Figure 21. to high rejection requirements in the duplexer lter. In
Recongurable amplier based on RF-MEMS varactors. frequency-division duplex systems, in which receiv-
ing and transmitting is performed simultaneously,
degradation of the receiver by wideband transmit noise
topologies of recongurable matching structures can must be avoided.
be envisioned. Either matching stubs switched by RF The selection provided by the amplier can con-
MEMS switches or a pi-topology incorporating RF tribute to the overall ltering effort in the transmit
MEMS varactors could be used. Only some dam seal- chain. It contributes to a so-called distributed lter
ing must be placed above the RFIC in order to prevent approach, in which each signal processing stage,
package parasitics from degrading performance. whether digital or analog, contributes to the overall
For an alteration between two bands, a switching filtering effort. In the architecture depicted in
approach such as that shown in Figure 20 is suf- Figure 10, the selectivity provided by the postselector
cient. However, when the number of different fre- (i.e., the interstage lter), in conjunction with the se-
quency bands is large, an approach based on RF lectivity provided by the power ampliers, leads to a
MEMS varactors (see Figure 21) is more exible. A reduction of the requirements for the duplexer, which
challenge in such a context is the current-handling today is a costly device of large form factor and high
capability of RF MEMS devices. On the input side, RF weight. It also simplies the effort required to recon-
power levels are quite moderate, but, on the output figure the duplexer. Instead of frequency-selective
side, they can be quite high, resulting in large reactive gate drain capacitance neutralization, narrowband
currents owing through the RF MEMS devices. matching topologies may be used with a classical am-
An alternate approach may be the use of wide plier, after which, of course, it is advisable to use RF
band gap power transistors, which can have imped- MEMS varactors for precise frequency setting of the
ance levels close to 50 Ohm. (See Figure 22.) A tech- gain peak.
nique for the frequency-selective cancellation of the In the Recongurable Mobile Systems (RMS) re-
gate drain capacitance may be implemented. By chang- search project funded by the German Federal Ministry
ing the properties of the passive feedback network, of Research and Education (BMBF), which lasted from
50 50 50 50
A/D DUC
Analog lines
TXFP Addr. C0 Rekonfig bus FBAD Addr. xx
Digital data lines 50
50
50
0/90 A/D DDC
50
LABVIEW
Verantw.: RST/WEC Verantw.: JHE
LVDS
PLL PLL
Ref Ref
VCO Clock
VCO Ref
50
RXBP Addr. 10 RXMX Addr. 20 RXAD Addr. 30
50 50 50
A/D DDC
Figure 23.
Architecture of RMS demonstrator.
July 2001 to December 2004, a demonstrator was Figure 26 is a picture of a recongurable receive
built as a practical demonstration of the recongura- bandpass module incorporating an RF MEMS switch
tion of a common RF signal path in order to address matrix. It is interesting to compare this module with
multiple mobile communication bands. Figure 23 the transmit bandpass module shown in Figure 27,
shows the RF architecture of the RMS demonstrator, which uses a GaAs FET switch matrix. The drawback
and Figure 24 shows the mechanical construction of of this approach is immediately visible. The recong-
the RMS demonstrator frame. Each RF function has its urable bandpass based on RF MEMS requires only a
own module housing. This allows changing modules single matrix at the output, thanks to the high isola-
between different realizations of an RF function. For tion provided by RF MEMS. On the other hand, the
example, a exible lter based on SAW lters being reconfigurable filter based on GaAs switch matrix
switched by an RF MEMS matrix can be exchanged technology requires two matrices, one at the input
with a lter based on GaAs-FET switches. Moreover, and one at the output. Moreover, biasing techniques
a tunable lter based on variable RF MEMS varactors must be implemented with GaAs switches. Another
can be supported by this approach. reason two matrices are needed with GaAs-FETs is
The module carrier frame shown in Figure 25 that the GaAs FET-based switch matrices have very
serves as a platform for hosting all the modules shown high parasitics capacitance for unselected signal
in the mechanical construction of Figure 24. paths.
Figure 25.
RMS demonstrator module carrier frame.
Figure 27.
Recongurable transmit bandpass using GaAs FET switches to switch between four xed SAW lters.
Using two matrices means that the RF loss of the especially for low noise designs. For the RF MEMS
switch matrix is doubled when compared to that of a switch matrix, the out-of band rejection was further
single switch matrix. In this case, RF MEMS technol- enhanced by shorting the unused lters to ground,
ogy offers significant benefits at the system level, as shown in Figure 28.
Figure 29.
Multiband transmit mixer module using RF-MEMS to switch matching structures (top view).
Figure 31.
Wide tuning receive synthesizer feeding a recongurable prescaler and using an RF-MEMS switch for alteration of
frequency range.
Figure 33.
Multiband transmit mixer module using RF-MEMS to switch matching structures (bottom view).
ampliers. It is possible to take advantage of the design The key to the evolution of RF MEMS will be
exibility of MEMS to add more functionality to active their monolithic integration into RFICs. Not only will
device structures. For example, a micropentode can be this integration allow circuit designers to take full
fabricated without requiring any additional processing advantage of the performance of discrete RF MEMS
steps. The design and fabrication of inductive-output- devices, it will also enable new system architectures to
type vacuum tube devices such as klystrodes, klystrons, optimize performance. An example of this would be
and traveling wave tubes using such solid-state-based the use channel-select lter banks in the mobile in-
MEMS technologies are also clearly within reach. Such frastructure. In order for this evolution to take hold,
devices have the potential to have an enormous impact the industry-standard foundry fabrication process and
on the wireless infrastructure. Power ampliers based the EDA infrastructure must mature. An example
on traveling wave tubes have been shown to offer sub- design ow from Coventor is shown in Figure 35.
stantial benets in terms of the efciency of wireless In the late 1970s a quiet revolution took place in
basestations, but they cannot currently be justied be- the semiconductor industry, as EDA and process tech-
cause of their cost and size. nology progressed to the point at which register
Unique to the work described above was the in- transfer language (RTL) and physical-layer design
corporation of nanoscale structures based on carbon became effectively independent. This launched the
nanotubes to enhance device functionality and per- foundry business model and supported the rise of
formance. The powerful combination of nanomateri- fabled semiconductor companies. RF MEMS now sit
als and MEMS technology will likely stimulate the on this same cusp; their use should explode over the
creation of new and unique devices useful for a next decade as design and foundry manufacturing
variety of other applications. mature.
Create
Macromodels
for Simulation
Figure 35.
RF-MEMS design ow.