Developments in the field of semiconductors are leading to integrated circuits with three-dimensional features and even moving parts. Such devices, called microelectromechanical systems (MEMS), can resolve many problems that a microprocessor plus software or hardwired ASIC cannot.
Imagine stepping out of your car to head for the weekend farmers' market. It's a couple of blocks away near a small park. Your cell phone emits a characteristic beep that isn't a call. It's a message from the cell phone. You pull it out and read: “Allergy warning: spore levels above threshold two blocks ahead.” Your phone knows where you are and which direction you're headed, so it checks the database, maintained by the park's environmental sensors, against your allergies. Silently thanking the park's sensors and the Internet-accessible database, you change plans. That's the way it might work in a few years. Today you only know you have a problem when your eyes start to water.
If you think this sounds like science fiction, it may surprise you that such a scenario is possible thanks to special ICs called microelectromechanical systems (MEMS). MEMS are tiny motors, gears, moving mirrors, turbines, and other complicated mechanical structures. MEMS are built using semiconductor-fabrication methods and are ICs (chips) incorporating electronics and moving or deformable parts.
You may also be surprised to learn that hundreds of millions of MEMS sensors are already in use. Airbag sensor ICs with tiny on-chip accelerometers are the first large-scale use for MEMS with moving parts. Analog Devices has been making such parts for automotive applications for 12 years and is currently shipping more than 1,000,000 a week.
But the airbag accelerometer was only the first of many MEMS-based sensors to invade automobiles. Automobiles also use integrated MEMS sensors to analyze exhaust gas and measure mass airflow and manifold and fuel pressure.
Other commercial products include a revolutionary display technology from Microvision. This MEMS-based display enables viewing full-size web pages and e-mail attachments on a small cell phone. It will soon enable digital camera owners to see images at print resolutions through the viewfinder.
Today's displays mostly paint images by beaming electrons (CRTs) or driving arrays of digital pixels (LCDs). In a CRT (cathode ray tube), electromagnetic fields bend the beam to hit a particular point. The display seems to show a complete image instead of one rapidly moving pixel because the phosphor's glow, which persists even after the beam passes, is read by our eyes as persistence.
Microvison's display replaces the electron gun with a low-power laser or with light-emitting diodes (LEDs). This light is modulated to turn individual pixels on and off and then bounced off a moving MEMS mirror directly into your eye. The sensors in your eye act as the screen. The device works by moving a tiny gimbaled mirror in an electromagnetic field to complete the image via raster scan. The result is the perception of a large screen several feet in front of the viewer.
Figure 1: Analog Devices' two axis accelerometer is built into millions of automobiles each year
Figure 2: Microvision's two-axis moving mirror makes efficient displays for cameras and cell phones.
An integrated future
Got a cell phone you aren't using? Open it and look at the number of components inside. In these days of ICs, why are there so many separate components in this, the ultimate of portable devices? If any product enjoys the economy of scale to afford integration, it's the cell phone. It isn't a one-chip system yet because the high-frequency, analog transmit-and-receive circuits near the antenna are difficult to integrate. The cell phone has an antenna at one end and a user interface at the other end. At the user-interface end, the cell phone is digital; at the antenna end, it's analog.
Inductors, transformers, relays, and moving-plate capacitors can't be fabricated on a planar silicon IC. Therefore, these basic elements remain as discrete components and not as ICs. With MEMS, those pesky 3D inductors and variable capacitors can be integrated, reducing the cell phone's component count and lowering its cost and power consumption at the same time.
Microelectromechanical versions of relays, tunable capacitors, inductors, signal filters, microphones, reconfigurable antennas, local oscillators, resonators, switches, and programmable phase shifters (used with antennas) are either available today or will be in the near future. These components will become key parts of cell phones and many other products.
The cell phone conserves power if its output signal contains just enough power to reach the base station. This also reduces interference with other transmitters. MEMS switches could adjust transmit power levels by coupling different power amplifiers to the modulator's output depending on distance from the base station. Discrete switches are too bulky to permit switching power amplifiers; 2D semiconductor switches leak too much to isolate high-gain amplifiers and lose too much for low-gain amplifiers.
MEMS-based switches and signal filters could be a thousand times or so smaller than their discrete-component equivalents. If that's so, it becomes possible to build a switched-filter front-end in lieu of the tunable signal filter used today. This circuit employs a bank of switches and filtersa pair of switches and a high-quality fixed filter for each desired input channel. Another proposed scheme replaces the radio frequency (RF) tunable local oscillator with a fixed oscillator and a bank of switches and intermediate frequency (IF) MEMS filters. Both circuits reduce noise in the RF section and reduce dynamic-range requirements in the IF amplifier and mixer, enabling better performance with cheaper components.
MEMS switches will displace discrete miniature mechanical switches and inefficient integrated field-effect transistor (FET) switches. MEMS moving-plate capacitors will displace discrete miniature moving-plate capacitors and voltage-tunable varactor-diode capacitors. MEMS inductors will displace discrete-component inductors and integrated 2D-spiral inductors. And so on. Because they offer advantages in size, in performance, and in mass production (and therefore cost), MEMS offer evolutionary improvements: MEMS-based components can displace their legacy equivalents, in system designs that remain fundamentally unchanged.
MEMS will also improve the cell phone in other ways. Imagine that you're a civil engineer planning an upgrade to downtown storm drains. You step to the curb, put a viewfinder to your eye, and aim your cell phone down the street toward the intersection. “Show me what's under the street from the surface to three feet below the storm drain,” you command. You frame the section of the street in the phone's viewfinder. Thanks to a GPS receiver and a MEMS gyroscope, the phone knows its exact position and orientation. A MEMS scanner knows what it's looking at, the phone calculates street coordinates and contacts the city's database to retrieve maps. The same MEMS viewfinder that paints the image on the back of your eye also scans your retina for authorization to view the data. Your view of the street melts away to reveal semi-transparent storm drains plus electrical, gas, water, and sewer lines.
Today's DNA testing requires a roomful of sophisticated equipment. Soon, MEMS will reduce this to a portable instrument with an on-chip chemical analysis chamber. Genetic analysis begins with a DNA fragment taken from a single cell. This fragment is too small for direct analysis, so it is placed into a solution and thermally cycled. When the solution is heated to 92C the DNA double helix splits into two complementary strands. When the solution is cooled to 65C the individual strands construct duplicate fragments, doubling the DNA's concentration in the sample. Twenty to thirty thermal cycles bring the sample to the concentration needed for reliable testing. Since the amount of solution required for conventional analytical instruments is large, the solution has a large thermal mass. This means that the heating and cooling cycles take a long time.
A MEMS thermal cycling chamber requires only 50 microliters (0.002 ounces) of solution, reducing thermal cycling to a tenth of the time. And the small amount of solution reaches concentration in fewer cycles, substantially reducing reagent use.
DNA testing is one of many applications in biomedicine. The electronic lab-on-a-chip, from companies such as Aclara, Affymetrix, Caliper, Cepheid, and Genefluidics, will change chemical collection and processing businesses. One-chip labs may be more reliable because chips can have arrays of sensors that sample in parallel and correlate the results.
Through the viewfinder
The IC has been improving for forty years. The industry has made astounding progress in the face of significant limitations. For example, semiconductors didn't have moving parts and they were generally fabricated using only resistors, fixed-capacitors, and active elements (transistors and diodes). Until now there has been no convenient way to fabricate a coil (inductor), a moving-plate capacitor, or a relay. Any inductor required by the circuit had to be supplied as a separate, discrete element. MEMS fill these shortcomings.
The computer industry with its inkjet nozzles and read-write heads, the biomedical/chemical industry with its disposable pressure sensors, and the automotive industry with its airbag accelerometers and engine sensors are today's volume applications for MEMS. Applications in automotive, industrial, computer, optical, biomedical/chemical, and RF electronics will follow as voracious consumers of MEMS. As the market for MEMS develops, applications will move from the wired world to the mobile world. MEMS applications in the wired world (automobiles, backbone networks, medical instrumentation, and others) will pave the way for mobile applications. Today's crude pressure, chemical, and gas sensors in wired applications will be the ancestors of MEMS-based sensors in mobile devices.
However, in contrast to the general-purpose nature of microprocessors, MEMS fit a narrow range of applications. A MEMS-based airbag accelerometer, for example, won't do as an avalanche detector or as the motion sensor in a hard disk. Each application demands a unique device and increases integration costs.
Unfortunately, it's not as easy to isolate MEMS as it is electrical circuits. Sensors and actuators need to touch and feel the world. They measure viscosity, flow, acceleration, temperature, pressure, and humidity. They can't always be isolated from corrosive substances. Their use makes them inherently difficult to protect. MEMS physical packaging is a challenge since it needs to protect electrical circuits, yet allow access for sensor and actuator interaction. It will be interesting to see how that plays out.
MEMS advance the state of something we already know how to do. For example, we already know how to build motors, switches, inductors, springs, and semiconductors. Compare this with, for example, the similarly advancing areas of biotechnology and nanotechnology (building systems with moving parts on an atomic scale). Developments in biotechnology and in nanotechnology break new ground in the analysis, invention, and construction of atomic-scale systems
In short, MEMS are a dream come true for embedded designers. They address the electronic shortcomings of 2D ICs and enable cheap microsensors and microactuators to sense the physical world and respond to it directly. MEMS promise to integrate embedded processors more fully with the world around them. esp
Nick Tredennick is an editor for the Gilder Technology Report. He has extensive experience in microprocessor design, with nine patents in logic design and reconfigurable computing, and was named an IEEE Fellow for his contributions to the field. At Motorola he designed the 68000. At IBM's Thomas J. Watson Research Center he designed the Micro/370. Nick has been the founder of several companies including Nexgen, where he hired and managed the team that designed the microprocessor that became the AMD K-6. He earned his PhD in electrical engineering at the University of Texas. His e-mail address is firstname.lastname@example.org .
Brion Shimamoto is an editor for the Gilder Technology Report. He was vice president of technology at Digital Domain, a visual-effects startup and CTO of OpenReach. He has worked for AT&T and NCR. At IBM's Thomas J. Watson Research Center he specified the Micro/370 microprocessor. Contact him at .