MEMS on the Move Success in several key market segments promises a brighter future for a slowly evolving technology

Jack Shandle

TechOnline (07/07/2005 12:00 AM ET)

Micro-Electro-Mechanical Systems (MEMS) has a reputation of being one of those technologies that is always on the verge of taking off but somehow sputters. The truth is somewhat more complicated: success varies widely depending on the market segment.

Hundreds of millions of MEMS-based ink jet printer nozzles are shipped every year, for example, and Texas Instruments' micro-mirror DLP technology is now being integrated into about 85% of all projector display systems. But in other markets, MEMS is just starting to roll. And in still others, MEMS continues to meet plenty of resistance, mostly for economic reasons.

For a technology with a "two steps forward, one step back" reputation, MEMS has shown steady growth that will continue at a very respectable pace through 2009. According to EmTech Research, worldwide sales will grow from approximately $6.5 billion in 2004 to $11 billion in 2009. See Figure 1 below.

Figure 1:  Forecast Sales of MEMS Worldwide, 2004-2009

No competing technology promised to deliver as large a payload in terms of performance. Only MEMS-based products could do the job. Together with the economies of scale, when MEMS products finally became mainstream, performance was going to drive telecom into a fiber-optic future with MEMS along as a passenger.

When the telecom bubble burst, many optical MEMS companies closed their doors. Without an out-of-control demand driving market economics, telecom carriers and system integrators scaled back their purchases and this hit MEMS the hardest. Although inventory levels have been reduced, telecom is still running lean and mean, says Marlene Bourne, Principal Analyst at EmTech Research.

But the last word has not been written on optical MEMS. Optical MEMS suppliers that are still in business have moved up the value chain. Instead of producing components, they are now moving toward creating and marketing subsystems. This strategy pleases telecom carriers and systems integrators who want value-added solutions.

MEMS design is at about the same stage as conventional IC design 30 or even 40 years ago, according to Maher, when companies were experimenting with a wide variety of silicon processes. MEMS designs can be quite sophisticated and use many modern current design techniques. But because there is a lot of process experimentation, the maturity of a top-down design methodology is lacking.

Not only are there many processes (as opposed to a relative few for circuit design), the experimentation with the basic foundry processes makes companies unlikely to share their secrets with the industry at large, says Maher.

Additionally, modeling the processes is more complex than conventional ICs. First, MEMS are three-dimensional devices. Process modeling usually has to take into account mechanical processes such as fluid flow and the electro-static properties that accompany every MEMS device.

But inside companies such as Bosch, ST Microelectronics, and Texas Instruments that have successful MEMS products and development programs, in-house design has been standardized and a top-down design methodology is in place.

As a result, MEMS CAD companies—the leaders are Coventor, IntelliSense and SoftMEMS—have to provide tools for both bottom up and top-down design. While there is an active community interested in library development and standardization across the industry, the state of the industry itself?standard processes are far from being adopted?makes their work more difficult.

The MEMS CAD story has a bright side as well. Companies that were once primarily interested in proving their micromachined systems work are now addressing the more financial aspects of manufacturing. MEMS yields are, in general, currently much lower than those of conventional semiconductors, says Maher, but will rise as the field matures. The good news is that companies are becoming more interested in software tools that are helpful in DFM (design for manufacturing), yield analysis, and statistical analysis.

But in terms of 2004 revenue, Texas Instruments' DLP (digital light processing) technology passed ink jet nozzles in 2004. DLP is based on micromachined mirrors with two states: On (when it flips in a position to reflect light) and Off. According to market researcher Display Search, DLP technology now has 85% of the projector display market.

There is more than projection displays in DLP's future—and there are also some competitors on the horizon such as Sony. TI's next target market is digital TV (DTV). At last January's Consumer Electronics Show, says Bourne, more than six dozen companies exhibited DTVs based on DLP.

The competing technology, GLV (gated light valve) was developed by Silicon Light Machines about a decade ago. The company licensed the rights to develop the technology for TVs to Sony five years ago. The problem was that it was developed to take advantage of blue lasers, which are expensive and have some development issues.

Those issues seem to have been resolved. Sony demonstrated GLV projection systems a couple of years ago, says Bourne, and is likely to move aggressively into theater systems, projectors, corporate theaters, and DTV.

In addition to megaplayers TI and Sony, a dozen start-ups are working on optical MEMS for applications such as computer monitors and DTV, says Bourne. In some instances, the start-ups are focusing on applications where DLP cannot go. Specifically, a few are developing optical MEMS products with a smaller footprint DLP for applications, such as cell phones and PDAs digital cameras. Bourne says the two companies to watch are MicroVision and Iridigm, which has been acquired by Qualcomm.

Ink-jet and DLP are examples of MEMS products that have proven themselves in the market. The success of these products and that of MEMS accelerometers—primarily for automobile air bag applications—are perhaps the best known success stories. An example of a product poised for the same kind of success is a new category of inertial sensors called tri-axis accelerometers.

One of the first important design wins for tri-axis accelerometers is to provide "drop detection" in the Apple G4 laptop PC. If the notebook computer is being dropped, the accelerometer senses it and informs the system to lock the hard drive. Bourne expects tri-axis accelerometers to displace dual-axis in some applications, such as cell phones.

Cell phones already use dual-axis accelerometers for features such as gaming applications and pedometers. The next important development will be the integration of micro hard drives similar to those used in Apple's iPod. In fact, two cell phone manufacturers recently announced the integration of micro drives into their high-end models.

With the integration of microdrives, the tri-axis technology's ability to sense rotational movement makes it more desirable because it can afford drop protection as well as the functions presently performed by the dual-axis products. Unlike a car where accelerometers are used in safety critical functions and typically require some redundancy as a result, cell phone can get along with a single accelerometer.

On the other hand, Agilent Technologies' FBAR (film bulk acoustic resonator) duplexer has been a roaring success. By 2003 Agilent had shipped its 20 millionth FBAR duplexer, which has been designed into nine of the top 10 CDMA (code division multiple access) phone manufacturers' U.S. PCS (personal communications service) band handsets.

Duplexers play a critical role in CDMA handsets and data cards by separating incoming communications from outgoing communications to allow for simultaneous two-way voice or data transmission. Traditionally, duplexers have been the second largest component in a wireless phone, after the battery. In 2004, 85% of all CDMA handsets included an FBAR resonator.

A number of other companies are working on FBAR technology but the only one that has a product in the market besides Agilent is Dicera of San Jose, CA, which has fielded an oscillator. The other companies are ramping up and should ship products this year or next.

John Stockton, a partner at the venture capital firm Mayfield, agrees that RF MEMS is a technology to be watched. Mayfield has invested in a yet-to-be-announced company with a technology that combines MEMS, RF, and low-power processors. RF MEMS resonators that replace crystal oscillators are also a viable technology in Stockton's view.

MEMS resonators face daunting challenges. The Q factors required are difficult for a mechanical device to attain. Quartz crystals can have a Q of 40,000 to 50,000 and MEMS devices have traditionally fallen short. But now, Stockton says, companies are in that range.

In addition to designing MEMS resonators as oscillators to replace quartz crystals, Dicera is also trying its hand at resonators to replace SAW filters. This is a considerably more challenging technology for a mechanical system because the dynamic range of the signals can be 100dB. But real progress is being made.

In addition, says Stockton, a number of start-ups that have spun out of the University of California at Berkeley have taken up MEMS resonator technology.

The first two categories are inexpensive and relatively easy to manufacture, says Bourne. They are now widely used in life sciences research. When they appeared almost out of nowhere, the MEMS lab-on-a-chip category was significantly disrupted because they were having difficulties handling the extremely small amounts of liquid that were to be used for analysis. So those applications were no longer viable for MEMS.

The lab-on-a-chip companies regrouped and developed new products and manufacturing techniques in three areas—two well established and one relatively new.

The first is life science research (drug discovery, genomics, and proteomics). Affymetrix and Nanogen represent this part of the market. The products tend to be expensive, sometimes costing a couple of hundred dollars but the applications can support a viable business.

The second is point-of-care-diagnostics, says Bourne. I-Stat started this category, which is characterized by critical care testing of blood parameters at a patient's bedside. The products are inexpensive and capable of producing results in about a minute or two instead of a day or more when the samples are sent to a lab. I-Stat is shipping tens of millions of these products.

A third is clinical diagnostics, which combines the DNA-based detection capabilities of the life sciences products with the quick response time of the point-of-care chips. These products are being developed to detect illnesses in which detection is important but not necessarily time-critical—diseases such as SARS and cancer. They test for bacteria, viruses, and infections. When tests are normally sent to a lab, it might take a couple of weeks to get results, but the MEMS products deliver results in a couple of days.

Itsogem of France is a leading company in clinical diagnostics. This category is being addressed by established companies in both the point-of-care diagnostics and life science categories.

Further out, there are a few technologies that will be fully developed in ten years or so, including fuel cell batteries.

Bourne expects a fundamental shift in the primary mechanism for revenue growth. "Up until now," she says, "most MEMS devices entered the market at a fairly high price point with a relatively small market size." As the market increased, economies of scale allowed pricing to come down. Optical MEMS such as DLP, for example, initially had unit prices of several thousands of dollars per unit and MEMS gyroscopes were $75 to $80 each. As volumes increased, prices decreased. Optical MEMS are now a couple of hundred dollars and gyroscopes are expected to be in the $10 range in the next year or so.

Next generation devices such as microphones and some RF MEMS will have an entirely different dynamic. "These products have moved into the market with commodity pricing," Bourne says. Microphones are 50 cents and RF MEMS are dropping in price as well. But the key is that they are also supported by extremely high volumes in the tens and hundreds of millions.

Contributing writer Jack Shandle is a former chief editor of both Electronic Design magazine and ChipCenter.com. He holds a BSEE degree and has written hundreds of articles on all aspects of the electronics OEM industry. Jack is president of e-ContentWorks, a consultancy that creates high-value web and print content for publishers, eOEM corporations, and industry associations. His email address is jshandle@e-contentWorks.com.