Work is under way to create high-end integrated microsystems that can sense, crunch data, and communicate wirelessly—in a package the size of a sugar cube. Research is under way to bring wireless communication down to the micro level, laying the groundwork for next-generation sensing systems. Wireless sensors assume low manufacturing costs, low power use, and an elevated level of integration. Power dissipation must be low enough to permit an acceptable lifetime for the device. One effort working to make this vision a reality is the Wireless Integrated Microsystems (WIMS) project—a research consortium funded by the National Science Foundation (NSF). WIMS is one of about 20 Engineering Research Centers funded by the NSF and is the only one focused on wireless microelectro-mechanical system (MEMS) devices. The NSF has pledged financial support for up to 10 years.


By integrating mechanical elements, actuators, and electronics into one siliconbased device, microsystems technology has pushed the boundaries of sensors into new areas. Microelectro-mechanical devices have clearly become very sophisticated, setting the stage for their widespread deployment in a promising range of applications, whether inside the human body or in the environment at large.

Yet for all the technical wizardry of MEMS sensors, one hurdle still stands in the way of their being truly field-deployable. Wireless communication, which made cell phones possible and lets travelers check e-mail from their cabs, is still limited to the macro scale. Research is under way to bring wireless communication down to the micro level, laying the groundwork for next-generation sensing systems.

Wireless microsystems hinge on low power dissipation, a requirement that raises the bar for developments in mechanical circuits, electronic packaging, and materials. It's a tall order, but the potential payoff is pervasive use of the systems in society.

Installing wiring-punching holes in walls or describing paths around machinery, stringing wires, and labeling them-can be a labor-intensive, manual operation. Wiring can be the dominant cost in installing sensors.


Wireless Matters

According to Albert P. Pisano, director of the Electronics Research Lab at the University of California, Berkeley, and chair of the Executive Committee of ASME's MEMS subdivision, a rough rule of thumb, used in military procurement, is that for every dollar you spend on a sensor, you spend $10 to $100 for the package and between $1,000 and $10,000 for the wiring. "Anything that can be done to eliminate the cost of wiring is the right thing to do," he said.

Wireless sensors assume low manufacturing costs, low power use, and a high level of integration. "The e sensors aren't going to work if they are expensive," Pisano said. For the concept to be practical, microprocessor, sensor, and wireless link would have to be designed into one tight package that could be manufactured inexpensively. Power dissipation must be low enough to permit an acceptable lifetime for the device.

As the right technical developments-low power drain, high integration, and infrastructure needs-come together, wireless sensors can make financial sense. If that happens, it presents an intriguing vision of the future: inexpensively produced, integrated microsystems consisting of microprocessor, electromechanical device, and wireless link-essentially, stick-on systems that would perform a valuable function and with which it is possible to communicate.

Making it Happen

One effort working to make this vision a reality is the Wireless Integrated Microsystems project—a research consortium funded by the National Science Foundation. WIMS is one of about 20 Engineering Research Centers funded by the NSF and is the only one focused on wireless MEMS devices. The NSF has pledged financial support for up to 10 years.

The WIMS Engineering Research Center is based at the University of Michigan in Ann Arbor, and also involves re searchers at Michigan State University in East Lansing and Michigan Technological University in Houghton. The consortium includes 19 companies and four nonprofit institutions. The project covers microprocessors, sensors, and radio-frequency links, with research activity focusing on issues of electronic packaging, low power dissipation, and new materials.

Potential applications for the devices include environmental monitoring, from weather and global warming to air and water quality; and health care, by providing wearable or implantable biomedical devices.

One indication of the wide interest in wireless integrated Microsystems is the variety of industries represented by its members: large, well-established companies as well as small startups, in businesses as diverse as automobiles, oil, biomedicine, and semiconductors. Industrial partners, as they are described on the WIMS Web site, are asked to kick in anywhere from $10,000 to $100,000 a year to support the center's research. The list includes IBM, Ford Motor Co., Intel, Honeywell, and Chevron Research & Technology Corp., as well as Cochlear Corp. and Advanced Bionics, which make ear implants.

According to Joe Giachino, industrial liaison at the WIMS Center at the University of Michigan, "We are attacking a problem, and then getting a microsystem solution to that problem." The WIMS project has two testbeds—a cochlear implant and an environmental sensor.

The research center is developing high-end wireless MEMS devices, merging micropower electronics, a wireless link, and a MEMS data-gathering device in a single module, said Kensall Wise, professor of electrical engineering and computer science at the University of Michigan and the center's director. The goal, over a five to 10-year period, is to combine components in a block measuring 1 or 2 cubic centimeters and run the system on 100 microwatts or less.

Wise believes that interest in wireless microsensors is pervasive, and that opportunities for wireless portable or wearable devices will expand by orders of magnitude over the next decade. Wireless capability adds reliability and lowers the cost of any type of wearable or portable system that will be deployed to gather information over a broad range, he said. The wireless systems under development have ranges from 1 centimeter to 1 kilometer, said Wise. That would be the range of an original signal that could be picked up and relayed.

Wireless links available today are an order of magnitude higher in power dissipation than what is needed, said Wise. One of the methods being investigated to bring down power use is to substitute microelectrornic circuits with micromechanical ones, which dissipate far less power. The wireless links would have to operate at sub-milliwatt levels, periodically report data, and listen for input commands.

Microprocessors need to be optimized for low-power data-gathering application and data interpretation, he said. Typical applications will not require high processing speeds. "You are going to need pretty good power-to-weight product," Wise said. Present- generation MEMS devices would probably require too much power to be used in these applications, he added.

In addition to requiring very low-power operation, wireless MEMS sensors will have to be self- testing and able to work in tandem with the embedded controller, and will probably be compensated by software in the controller, said Wise. Depending on the application, he envisions that the MEMS devices will include pumps, valves, and inertial sensors.

Packaging is the key to robustness and the ability to work in the real world, Wise said. "Increasingly, we are going to see the packages done at the wafer level," he said. Many inertial sensors, like accelerometers and gyros, require a vacuum reference cavity, in which the transducer operates in a vacuum. Increasingly, those vacuum reference cavities will be done at the wafer level, as part of the chip fabrication, so that when it becomes necessary to dice up the wafer and package it, the critical elements are already protected.


Mechanical Circuits

WIMS devices are being designed with radio frequency wireless links, allowing them to ignore physical obstacles, to be field deployable. The goal is to develop devices that can communicate 1 kilometer over a variety of terrain with a maximum transmission power of 1 milliwatt.

Mechanical circuits offer a way for these devices to operate at very low power levels and offer a very high Q-factor, or signal quality, according to Clark Nguyen, the center's wireless task leader and an assistant professor of electrical engineering and computer science at the University of Michigan. Nguyen's work is supported by the NSF through the Engineering Research Center as well as by the Defense Advanced Research Projects Agency.

Mechanical circuits exist in telecommunication devices today on a macroscopic scale, taking up a lot of space, said Nguyen. An everyday example of a mechanical circuit is a guitar string. When plucked, if it is in tune, it will tend to vibrate at a specific frequency, say, 440 Hz.

This is analogous to the way that wireless phones work: They transmit electromechanical signals at very specific frequencies, between 800 MHz and 1.8 GHz, and use mainly mechanical components such as resonators to select certain frequencies while rejecting others, allowing people to communicate. Phones use mechanical components such as surface acoustic filters and quartz crystal resonators that select frequencies. A typical phone board today is about 10 to 20 percent transistors; the rest of it consists of passive circuits, the largest and most expensive of which operate through mechanical vibration, Nguyen said.

These mechanical circuits, however, are macroscopic components that can be seen with the naked eye. The task of the WIMS project is to develop microscale mechanical components with the same or better frequency-selecting properties as their macroscopic counterparts, he said.

Nguyen compares the potential impact of micro mechanical circuits with that of transistors, in which miniaturization made possible complex circuits and systems, and eventually opened the way for computers and wireless phones. "Integrated circuit technology revolutionized things because it provided a way to shrink transistors to a tiny size, making them manufacturable in huge volumes for a very low expense. It also allowed them to have many times more circuit complexity and capability," said Nguyen. "Once we've got the mechanical circuits to a tiny size, we may be able to change paradigms governing wire loss architectures." The change, he hopes, will make possible wireless integrated microsystems.

Although Nguyen said that plenty of research needs to be done, he is optimistic that micromechanical circuits can eventually be made to operate at frequencies between 800 MHz and 1.8 GHz and beyond.

Because mechanical circuits consume less power than transistors, they can help microsystem devices consume less power. Nguyen said the long-term goal is to reduce power consumption of the radio-frequency front end of the transceivers from around 300 milliwatts to 100 microwatts. Mechanical circuits have much smaller losses than electrical ones, and so may not need as many amplifiers to make up for a degraded signal, he said. In addition, the use of numerous high-Q mechanic al circuits may allow design changes that trade Q for power, further reducing power consumption and extending battery lifetimes.

Bringing these circuits down to the micro scale opens up new design challenges. "The way the mechanical elements are forced into vibration and the way they are hooked together is a completely different type of circuit technology," Nguyen said. This means new design procedures have to be developed as well as micromechanical technology, he said. Nguyen, an electrical engineer, said the circuits could be designed in the electrical domain or the mechanical domain. One of the beauties of this work is that it teaches students how easily they can interchange the mechanical with the electrical, he said.

On the manufacturing side, mass producing these circuits- making thousands of these components per chip—introduces a trimming and tuning requirement. It may be possible to design around the problem; however, if not, a way must be found to trim and tune frequencies on a massive scale--a technology that still must be developed.

Two Tracks Toward Micro Wireless

THE WIRELESS INTEGRATED MICROSYSTEMS consortium sponsored by the National Science Foundation is working on two testbeds to develop wireless microsystems technology.

One testbed is an environmental monitor incorporating gas chromatography. The device will be a versatile environmental monitor capable of analyzing complex mixtures of mainly toxic organic compounds, according to Edward Zellers, an associate professor of environmental health sciences and chemistry at the University of Michigan, who is heading the project.

Zellers expects that the wireless link will allow uploading data from the instrument to a monitoring station, and will permit down loading of instructions. Although the concept of a computer-on-a-chip chemical analyzer is not brand-new—one has been developed at Sandia National labs in Albuquerque-the WIMS project aims to make its system smaller, higher-performing, and less power-consuming than any that has been done before, said Zellers.

One of the big challenges in developing the device is de- signing a micropump that can deliver appropriate amounts of air. The other is thermal control for the device, which must sit outdoors subject to a range of temperatures.

The system will have the ability to self-calibrate, although there will be limits to how much it can compensate in extreme temperatures, he said.

The other testbed is a cochlear implant, to help the hearing- impaired detect sounds. Present devices are comparatively large, said Khalil Najafi, deputy director of the WIMS project and professor of electrical engineering and computer science at the University of Michigan. "We feel that MEMS has a great potential to significantly enhance the capability in these areas, because you can build them very small and put them where they need to be placed."

Particular challenges in developing a cochlear implant are packaging and power dissipation, Najafi said. In the body, you have to worry about more than just protecting the device in a harsh environment, he said; you also have to worry about protecting the body. And power dissipation has to be low enough to be tolerated by the individual.


Tight Integration

Trying to build a wireless integrated microsystem on a device as small as a sugar cube requires tight integration of the various components, according to Richard Brown, professor of electrical engineering and computer science at the University of Michigan, who is investigating microprocessors for use in WIMS devices. Integrating as many components as possible on a single chip makes it possible to drive down the power use, as well as reduce the size of a device.

Building on such a small scale makes it difficult to put together separate parts, he said. Tight integration on a single chip eliminates off-system drivers and the space they represent, and also dissipates less power. Producing extremely low-power microcontrollers requires designing with leading- edge technology because more advanced processors with finer geometry have lower power requirements.

Packaging presents its own set of challenges, according to Khalil Najafi, deputy director of the WIMS project and professor of electrical engineering and computer science at the University of Michigan. Mechanical components, such as vibrators and resonators, need to operate in a vacuum, he said. Research is focused on developing techniques to package at the wafer level that can also provide a suitable environment.

Typically, these devices will have to operate in a lower-pressure environment for extended periods of time, possibly under harsh conditions and through temperature changes, he said. As the resonating parts move, "you don't want any air molecules around them that can impede their resonance," Najafi said. "You also need to make sure that the environmental conditions don't cause any changes in devices that are used as filters and oscillators in communications systems."

Other components add packaging demands, he added. Some sensors have to be exposed to the environment in order to measure it, requiring some components to be packaged in a vacuum while others are exposed to the environment. Other devices, like accelerometers, would need to be hermetically sealed to prevent moisture from getting onto critical components, he added.

Most of these devices will be manufactured using planar thin-film technology that is used in semiconductor processing. The advantage of using this technology is that it allows thousands of devices to be processed on a 4-by- 6-inch silicon wafer, which can then be diced apart. "We need to be sure that we can apply and finish the package at the wafer level, so that when we go to the process of dicing the individual devices apart, they are already protected and packaged and ready to be connected. That's not an easy thing to do," said Najafi


Diamond Films

Performance demands have also prompted materials research, which is being spearheaded by Michigan State. Poly diamond films are being investigated for three areas of the WIMS cube: wireless interfaces, sensors, and packaging, according to Dean Aslam, associate director of the WIMS Engineering Research Center and professor of electrical and computer science at Michigan State.

Because it is very stiff-it takes more force to bend-diamond has the ability to increase the frequency of the micromechanical component, said Aslam. Using the same dimensions, frequency can be increased by a factor of two or three compared to polycrystalline silicon, he said.

Diamond may also present other advantages as well. Aslam speculates that it may allow less stringent requirements for vacuum. Vibration is hindered in the presence of air, either because the air pushes the beam back or because air molecules stick to the beam. This is why the Q-factor deteriorates if the vibrating beam is not in a vacuum.

Diamond may get around this problem because it does not react with anything. The incidence of the air molecules sticking to the diamond is almost zero, he said. Aslam said that if diamond works well, it could potentially be used for all of the mechanical components in the WIMS devices.

Diamond can be made into a semiconductor or insulator, depending on whether or not it is doped (or impurities are added to it). As a semiconductor, it can be used as a sensor; as an insulator, it can be used as a protective layer. Both roles are being explored for the cochlear testbed.

One potential use is as an ultrasensitive sensor. Silicon sensors are based on a piezoresistive effect: If a strain is put on a material to bend it, its electrical resistance will change. The piezoresistive gauge factor, a measure of sensitivity, of silicon is around 150. The piezoresistive gauge factor of diamond is 4,000, making it more than 20 times as sensitive, said Aslam.

Such ultrasensitive devices could be useful in cochlear implants, eliminating the need to do signal processing to build up the strength of a very weak signal. Diamond is also being developed as a layer to protect the silicon probe in the cochlear prosthesis, Aslam said. Because diamond is a very stable material, it is also bio-friendly, he added.

Diamond is being investigated as a packaging material for WIMS devices, which will incorporate many new components, such as wireless links or even interfaces to analyze liquids, said Aslam. "There is no existing packaging technology that you can just apply," he said.

Two packaging approaches are being evaluated. One is to coat the device with a thin layer of diamond. The other builds a diamond package consisting of four walls, a bottom, and top, in a mold of silicon, which is subsequently removed by chemical etching.


Parts, Packaging, And Micromachining

Micromachining technology from Michigan Technological University is being used to develop three-dimensional WIMS components and packaging. For the cochlear implant, Michigan Tech is providing modeling, design, and micromilling of the spiral-shaped implant that will go into the inner ear.

Rapid prototyping was used to create outsize models. The final product will have to be brought down in size to fit the ear, according to Robert O. Warrington, WIMS associate director and a dean of engineering at Michigan Tech. "We can get things down to the tens of microns, depending on what we machine and how we machine it," he said. One design the researchers came up with is based on party favors that blowout of a spiral, said Warrington.

Michigan Tech is also developing the packaging housing for the environmental monitoring testbed, including fluid, mechanical, and electrical interconnects.