     7 August 1997, Electronic Engineering Times:

     MEMS could hold key to cell-phone-on-a-chip --
     Darpa sows seeds of a telecom revolution 

     Arlington, Va. - Researchers at the Defense Advanced Research
     Projects Agency (Darpa) here, working in tandem with a
     number of defense contractors around the country, are
     making significant progress in crafting
     micro-electromechanical systems (MEMS) in silicon that can
     replace radio-frequency components such as ceramic filters
     and wirewound inductor coils. RF MEMS techniques have
     already been used to fit a handheld reconnaissance radio into
     a standard PCMCIA form factor and create a smart or
     "tunable" antenna for military aircraft. 

     While the technology is not yet ready for use in creating the
     single-chip cellular phone or other long-anticipated
     consumer-electronics devices, the advances thus far are
     beginning to raise eyebrows. 

     Elliott Brown, a program manager for Darpa's Electronics
     Technology Office, believes the impact of RF MEMS on
     portable communications will rival the stir caused by the
     application of FETs in RF power amps in the 1970s. 

     Rick Ridgely, deputy program manager of Darpa's Airborne
     Communication Node, is even more optimistic. "This will do
     for communications what the microprocessor did for
     computing," he predicted. 

     Darpa believes RF MEMS will supplant discrete inductors,
     capacitors, ceramic filters and even transistor switches in
     navigation radios, airborne communications equipment and,
     by extension of the technology into the commercial sector, a
     host of portable personal communications devices. Because
     the mechanical structures are etched in silicon, they can be
     integrated with amplifiers, voltage-controlled oscillators
     (VCOs), phase-locked loops (PLLs) and other ICs,
     dramatically cutting the size, weight and power consumption
     of RF communications systems. 

     By micro-machining movable capacitor plates into silicon,
     researchers have built antenna interface circuits with quality
     factors-known as Qs-as high as 20,000. (Qs are a measure of
     a resonant circuit's ability to oscillate at a particular
     frequency.) 

     Darpa is overseeing research on three types of electrically
     tunable mechanical structures that could replace discretes in
     RF devices operating over a range of frequencies. 

     "Mechanically resonant" structures, such as surface acoustic
     wave (SAW) filters, will support "low frequency"
     applications, from several MHz up to 1 GHz. A number of
     projects toward that end are under way, including those at the
     California Institute of Technology and at the University of
     Michigan. A 300-kHz three-resonator micromechanical filter
     constructed at Michigan, for example, is reported to have
     obtained a Q of 590 and a stopband rejection of better than 38
     dB. 

     For mid-band frequencies (1 GHz to 10 GHz), meanwhile,
     MEMS techniques will build lumped-element filters. A third
     type, distributed filters, will serve applications above 10 GHz. 

     Darpa's Brown believes the biggest impact will be on the
     switches used as antenna-interface units for mobile and
     airborne transceivers-that is, the switches and filters that
     optimize an antenna to transmit at one frequency but to
     receive at another. A single MEMS switch, according to
     Brown, can replace the 13- or 15-component PIN-diode
     switch typically employed. 

     That kind of component reduction, Darpa predicts, will
     reduce the RF section component count a hundredfold,
     reduce the size of the antenna interface units by 10,000 times,
     cut power consumption by a factor of 1,000 and yield a
     thousandfold increase in off-state isolation. 

     "There is nothing that can beat a mechanical switch for
     effecting an infinite impedance," said Brown. 

     The key metrics in which the devices excel, he said, are
     insertion loss and dc power dissipation. Whereas an electrical
     or electronic component presents its own resistances,
     capacitances and reactances to high frequencies, the insertion
     loss for a MEMS switch is negligible. 

     When PIN diodes are employed for RF routing, for example,
     each diode can consume as much as 10 to 20 mA. The
     current drive needed to open or close a MEMS switch,
     Brown said, is on the order of nA or even pA. 

     Darpa's Web site characterizes its MEMS program as an
     "interdisciplinary research and development" effort on both
     "advanced MEMS devices and advanced MEMS fabrication
     processes, all with the goal of demonstrating innovative
     systems concepts." Responsibility for roughly $10 million in
     RF MEMS project awards is shared between Brown and
     colleague Albert Pisano, an industry recruit and an expert in
     such traditional MEMS applications as inertial instruments and
     fluid sensing, control and transport. 

     Brown oversees the Phase III MAFET (microwave and
     analog front-end technology) of which RF MEMS is a part.
     The lion's share of RF MEMS funding now comes out of the
     MAFET program. Ridgely of Darpa's Airborne
     Communication Node said the agency's overall RF MEMS
     expenditures could approach $20 million in the not-too-
     distant future. 

     The Darpa program managers are firm in their belief that
     micromechanical switches and resonator filters will be
     smaller and offer superior performance to their electrical
     component equivalents, but they're hesitant to project when
     the devices will be accessible to manufacturers of cell phones
     and other cost- sensitive appliances. Pisano believes
     polysilicon structures can be honed to make cost-effective
     micromachines. "You need very tight mechanical
     dimensioning," he said, "but it is essentially single-mask photo
     etching." 

     The low-frequency mechanical resonator filters present
     conceptual as well as cost issues. You can electrically tune a
     MEMS frequency filter by altering its dimensions-for
     example, the gap between one mechanically vibrating section
     and another. The resonant frequencies will be amplified; the
     non-resonant frequencies will be dampened. 

     But engineers will need to rethink voltage-to-frequency
     conversion, phrasing it in terms of the electrical energy
     necessary to change the mechanical spacings that affect
     resonant frequency. Mechanical signal processing, as Brown
     called it, is especially critical when one frequency is used to
     modulate another. 

     A mechanical filter may be more economical than an electrical
     filter for particular frequencies, Brown said, but it will never
     be cheap. He believes the cost of the filters will follow the
     same learning curve as other silicon semiconductors and will
     be easily competitive with those made in GaAs. 

     Controlled dimensions can also make transmission lines with
     variable (i.e., "tunable") impedances. Examples include what
     Hughes Aircraft calls smart antennas, in which the radiating
     characteristics of the antenna itself-the length, width and gap
     impedance-can be changed on the fly. 

     The technology is most useful for phased-array antennas
     operating in the S and K bands and for point-to-multipoint
     broadcasts at 38 GHz. At very high frequencies, Brown
     explained, all radiation assumes a point-to-point pattern: "The
     signal is 'teledesic'-it doesn't wrap around the horizon; it
     doesn't penetrate buildings." Special radiation patterns are
     required. But tunability results in a tenfold performance
     increase in the phased-array antennas used for such
     applications. 

     MEMS may be the best bet for miniaturizing RF systems.
     The typical aircraft navigation receiver, for example, may
     include dozens of hand-wound coils and hybrid amps. "The
     antenna-interface unit is a 6-pound brick," Ridgely said.
     MEMS approaches can take the weight out by slashing the
     parts count. 

     A technology-demonstration project at the Avionics and
     Communications Division of Rockwell Collins (Thousand
     Oaks, Calif.), for example, rebuilt the UHF/VHF
     antenna-interface unit (AIU) for the F-22 jet fighter with a
     MEMS tunable- capacitor circuit. The original switch matrix
     included UHF/VHF filters for four separate receivers. It used
     576 varactors, 216 inductors, 144 resistors and 108
     capacitors, for a total parts count of 1,044. The replacement
     devised by Rockwell was a tuning circuit with 36 MEMS
     capacitors. 

     Raytheon's E-Systems unit (Falls Church, Va.), which builds
     remote airborne reconnaissance systems whose core
     electronics include low-noise amplifiers and
     interference-cancellation systems, turned to RF MEMS for a
     handheld communication radio for battlefield applications.
     Called Ultra-Comm, the receiver is interoperable over a wide
     range of frequencies and modulation schemes-and is small
     enough to fit on a PCMCIA card. 

     The company is developing the device as a "software radio"
     to ensure compatibility with a wide range of changing
     communication standards, said William Rinard, E-Systems'
     director of technology development. 

     Heavy legacy 

     The new device will conform to standards mapped by the
     government-sponsored Programmable Modular
     Communication System (PMCS) and the Modular
     Multifunction Information Transfer System (MMITS). While
     government and industry are collaborating on standards for
     multiband, multimode radios, neither sector can ignore the
     substantial investments already made in legacy receivers and
     transmitters. 

     The software approach, E-Systems noted, allows for modular
     transmitters and receivers (packaged on VMEbus boards,
     PCI-bus cards or PCMCIA modules) that can be
     programmed to retain compatibility with existing
     communications equipment. The device will be completely
     tunable from 800 MHz to 2.8 GHz. 

     E-Systems engineers say the tight integration provided by RF
     MEMS filters accounts for the small form factor. They say
     they remain uncertain about the suitability of a MEMS device
     for consumer applications, but they note that $10 or $15 is
     still a lot cheaper than the $150 to $300 currently charged for
     the front-end filters used in military aircraft multiband radios. 

     "If these guys want a true Dick Tracy radio," said Michael
     Clingempeel, E- Systems manager for RF hardware
     development, "they're not gonna do it with ceramic filters." 

----------
