Large Jobs for Little Devices
Minute electromechanical systems could make RF devices smaller, faster, and cheaper
IMAGE: CLARK NGUYEN, UNIVERSITY OF MICHIGAN
This 156-MHz polysilicon MEMS resonator built at the University of Michigan has a Q of 9400; the central disk is just 34 mm across. Traditional high-quality mobile RF components have a Q of less than 2000 and are so big they can only fit off-chip.
This is part of IEEE Spectrum's special report: Always On: Living in a Networked World.
Microelectromechanical systems (MEMS) may become key components of radio frequency devices, particularly in the mobile marketplace. In filters, they may lower radio size and power consumption while increasing sensitivity. In switches, they could herald the construction of cheaper, electronically steerable antennas for radar and communications applications.
MEMS are electromechanical devices with tiny moving parts. They can be built using IC-compatible materials, such as polysilicon, allowing their integration on a silicon chip side by side with semiconductor circuits. Experimental filters are now reaching hundreds of megahertz, and operation up in the gigahertz ranges needed for most wireless and some satellite communications should be feasible. Their ability to kill several birds with one stone is making them very attractive to researchers and developers.
Mobile wireless devices come up hard against three problems: power consumption, sensitivity, and physical size. Improving power and sensitivity gives devices longer battery life and extends their range. Better sensitivity is vital to working in an increasingly crowded radio spectrum. Reducing the size opens up room for larger batteries and the ability to offer more features in one device.
At a minimum, an RF circuit requires a filter to react to the electromagnetic radiation dousing the antenna by pulling out the desired signal or inserting the signal to be transmitted. A superheterodyne filter employing surface acoustic-wave (SAW) filters is commonly used. These filters are so big that they preclude the entire receiver from fitting on a single chip, especially at the high frequencies used in satellite communications or in wireless systems like Bluetooth.
The digital direct conversion approach, where the antenna signal is fed with little (or no) off-chip prefiltering to a digital signal processor, tries to perform an end run around this problem. However, intensive processing demands hundreds of milliwatts, which is a big drain on batteries.
Worse still, a digital receiver needs more power as higher frequencies are reached. (The Nyquist sampling theorem dictates that to digitize a signal accurately, it must be sampled at twice the frequency of the signal.) Otherwise, high-frequency noise gets transformed into low-frequency signals, explained Al Pisano, of the department of mechanical engineering at the University of California, Berkeley. Digital radios must "amplify all the noise accurately and then throw it all away with software, effectively wasting power," said Pisano.
MEMS would allow better front-end analog-frequency filtering than present-day filters, with much smaller components. "With more selective filtering, you relax all the specs for the rest of the circuit," said Clark Nguyen, of the University of Michigan's department of electrical engineering and computer science in Ann Arbor. That means lower sampling rates and power consumption in digital elements.
Unlike their forebears, analog superheterodyne MEMS filters draw very little power--"a few milliwatts...they can run off a battery a long, long time" because the filters are passive devices, said Eliott Brown of the University of California at Berkeley's electrical engineering department.
They are also very selective. Nguyen has a 160-MHz filter with a resonance quality factor (Q) of 9400. In contrast, a high-quality SAW filter today may have a Q of less than 2000. Nguyen anticipates that "if present trends continue, that's not going to really drop off much as we get to 900 MHz and into GHz ranges." Q values of around 9000 would enable the initial filter not just to preselect a communications band, but to perform channel selection within that band, "which could be revolutionary for wireless," said Nguyen. These MEMS could be integrated on chip with the rest of the semiconductor circuit--"each of these [filters] might fit in a 10-by-10-µm spot," he added.
MEMS make good RF switches, too. Two-way radios switch the antenna between the receiver and the transmitter. MEMS switches are not much smaller than traditional solid-state switches, but they have lower power requirements and loss. John Smith, a program manager with the Defense Advanced Research Projects Agency (Darpa), in Arlington, Va., has been looking at MEMS-based applications and estimates that MEMS switches can "get down to less than a tenth of a decibel loss per switch."
An important use of MEMS switches is in phase-shifters, especially in radar applications. A phase-shifter usually relies upon a set of transmission lines, each of slightly different length. These are switched in and out, creating different paths for the RF signal to provide different amounts of phase modulation. By putting "one of these [MEMS-based] phase-shifting networks into each element of a phased-array antenna, you can provide very accurate and very low insertion-loss beam steering," said Brown.
Sensitivity this high would enable channel selection within bands, which could be revolutionary for wireless
This would allow antennas to transmit to, or receive from, a specific direction without having to be reoriented physically. With MEMS, printed-circuit board technology would suffice for antennas. Building one into the roof of a car, for example, "would allow you to talk to any satellite in the sky...and track it," said Brown. According to Smith, "[Darpa has] just made awards to three contractors and we expect to have demonstration phase-shifters in about six months. By 24 months' time we're trying to build 10 000 phase-shifters, and we'll build a [radar] array after that."
Cost is critical. Currently a phased-array radar antenna costs "on the order of $4 million to $8 million per square meter. We believe with a MEMS-based array you might be able to reduce that cost by over an order of magnitude," Smith said.
MEMS may also enable other devices that are difficult, expensive, or impossible to construct with conventional technology. Radio tags, used in everything from zoological research to espionage, are one application that requires small size and a long operating life.
"You want days, weeks, months," said Brown. The tag would sit in sleep mode, drawing no power until "it hits an RF signal. [Then] it wakes up and transmits something back."
There are hurdles to overcome. Nguyen is already looking beyond his 160-MHz filter. However, "it's not enough just to vibrate at high frequencies; the Q must remain high for this to be useful, and this remains to be seen," he cautioned.
Packaging concerns developers. RF MEMS probably require a hermetic package to avoid static friction, or stiction, due to humidity and to increase reliability, said Smith, adding, "I'm looking for a cheap hermetic packaging that can be used for microwave devices. So are a lot of people. And we need it very badly."
Further work on modeling MEMS devices is also needed. They require both electrical and mechanical modeling, and some phenomena, such as stiction, "are not fully explained by [current] modeling," Smith said. Then, too, MEMS have limited power-handling capacity--typically less than a watt--and they are slow at switching, taking microseconds to cycle instead of the nanoseconds typical of semiconductor switches.
But when all is said and done, MEMS have the potential to enhance communications technology in everything from pagers to spacecraft (where their marked disregard for high temperature and radiation environments is particularly welcome).