Electronic Design

MEMS Designs Gear Up For Greater Commercialization

As new markets arise, MEMS and MST technologies move forward to overcome challenging packaging, testing, reliability, and manufacturing roadblocks.

Today, microelectromechanical systems (MEMS) and microsystems technology (MST) make up one of the fastest-growing markets around. As MEMS and MST devices proliferate into the commercial sector, they're penetrating new markets aside from the automotive, medical, and aerospace and defense markets which they continue to serve. Communications (RF and optical), biomedical, consumer, and industrial markets comprise some of the newest and hottest areas being targeted.

According to a recent study performed by Roger Grace Associates and the Network of eXcellence in mUltifunctional microsystemS (NEXUS), some of the fastest growth rates will be seen in emerging new market applications (see the table). NEXUS is a European commission (which includes MEMS as defined by European interpretations).

In the RF telecommunications sector, MEMS ICs will serve in switches and switching matrices, relays, transmission-line components, tunable inductors and capacitors, resonant-comb drives, resonant beams, connectors, and switchable filters. In the optical communications market, these MEMS ICs will find their way into fiber aligners, switches, filters, printers, and displays. For the biotechnology sector, there will be microanalysis and microinstrumentation equipment, mass spectrometers, gas chromatographs, engineered surfaces, drug-discovery, delivery and handling systems, liquid/gas valves, regulators, pumps, and mixing chambers.

Yet another growing market is the defense/aerospace sector. There, MEMS technology will nudge its way into places like remotely distributed battlefield controls, maintenance, communications and inertial-guidance systems, and gyroscopes.

Such large market projections are mainly due to the fact that MEMS and MST devices are acting as market enablers, producing market values far in excess of what the devices themselves cost. In many cases, certain systems wouldn't be possible without MEMS/MST technology.

Consider the scanning-force microscope, for instance, which costs on the order of $100,000 and up. These systems wouldn't be possible without the micromachined tips that they employ, which cost only a few dollars each. The automotive airbag is another example. While accelerometer sensors within an airbag cost only a few dollars apiece, the entire airbag systems cost upwards of hundreds of dollars.

Ink-jet printers and hard-disk drives are other examples of systems containing micromachined structures. In fact, these microstructures largely contribute to the wide-scale affordability and the high-performance levels that characterize modern printers and disk drives on the market.

One of the better-known MEMS success stories is digital-light processing (DLP) technology, which was pioneered by Texas Instruments. It's the foundation for tiny micromachined moving mirrors known as digital micromirror devices (DMDs). High-performance business projectors, large-venue projectors, video walls, home-entertainment systems, cinemas, and photofinishing systems utilize DMDs (Fig. 1).

A strong indication of the rapid commercialization of MEMS and MST devices is the attention they have received from a number of Federal laboratories, including Sandia National Labs, Lawrence Berkeley Labs, and the Jet Propulsion Laboratory of the California Institute of Technology. Many companies that were formerly heavily involved in the military and aerospace business have also been gearing up their MEMS efforts for the commercial sectors.

One thing is becoming clear. Evidence is proving that no company can handle it alone. MEMS and MST are very diverse and often serve highly dedicated applications. These technologies require vast amounts of knowledge in all facets of the design process, modeling and simulation, reliability, prototyping, packaging, testing, and production. This multipdisciplinary environment has forced many companies and organizations to join hands and share their experiences and knowledge (see "A Pooling of Resources," p. 90).

The design of MEMS and MST devices requires a multidisciplinary approach, depending on the application involved. This means that a combination of electronics, chemical, mechanical, biological, thermal, and environmental knowledge (just to name a few disciplines) might be needed, depending upon the application.

As a result, MEMS/MST designers have a greater number of tools and suppliers available to them, thereby allowing them to bring their products to market more quickly. Prototyping kits are now available to help designers throughout the entire process, from the initial concept to the shipping of the final product.

Examine, as an example, the reliability issue alone. One of the biggest problems facing MEMS designers is brought on by the fact that, "There's no commercially available database of reliability and failure-analysis information," complains Stuart Brown, principal and director at Exponent.

Further, "The problem the MEMS community has is that failure modes in MEMS devices aren't those associated with larger devices. The physics don't change, but the failure modes do," explains Brown.

A few years ago, the Jet Propulsion Laboratory of the California Institute of Technology (JPL) offered its failure-analysis and reliability expertise to the MEMS community as part of a partnership program. MEMS has been a key technology area for JPL with an emphasis on designing smaller, better, and less-expensive electronics for NASA's space probes and vehicles that must operate reliably in demanding environments. As a research center, JPL has a wealth of equipment and expertise that few MEMS/MST companies in the commercial market can afford.

One of the hottest, emerging-commercial technologies are communications devices, many of which use optical MEMS techniques for switching signals. Another optical-switching approach makes use of LCDs. These optical switches are widely proclaimed as the forerunners of all-photonic networks. Along these lines, Cronos Integrated Microsystems anticipates the commercial introduction of all-optical communications router switches by the second half of this year.

Actually, Cronos is a relatively recent spinoff from the Microelectronic Center of North Carolina (MCNC), which was founded in 1980 to promote technology growth. The company is awaiting the commercialization of these switches which use tiny silicon mirrors that will enable a new generation of routers to increase Internet capacity by a factor of 10, delivering 10 terabits/s.

Further, the company is preparing for this market demand by beefing up its manufacturing capability to make optical MEMS components (sometimes referred to as MOEMS) like microrelays, photonic switches, and microvalves. Cronos is creating standardized building blocks for MEMS components. The building blocks form the foundation of a platform already leading up to an application-specific approach to MEMS. It's similar to the one used to make conventional ICs.

In addition, Lucent Technologies is betting on its WaveStar Lambda all-optical router switch as a key enabler for future routers with terabit/s speeds. Serving as an online communications backbone, the MEMS cross-connect switch uses tiny optical mirrors to perform the switching function (see the opening illustration).

Micromirror arrays also are part of an effort by Lucent to handle projected wideband audio/video/data services to the home. Lucent is proposing a free-space programmable wavelength-division-multiplexed (WDM) "add/drop" optical system for this task (Fig. 2). The system would allow a programmable add/drop node to drop one or more wavelengths out of a multiwave channel and rerout them to another destination. At the same time, potentially, traffic from another source would be added to the channel. In the proposed system, a 16-channel 200-GHz-spaced WDM signal is brought in via an input port and directed through the system's optical micromirrors. Next, it's bounced off a diffraction grating, which spatially demultiplexes the wavelength channels.

Each wavelength channel is directed onto one of 16 two-position mirrors, which directs that signal back through to the diffraction grating. Based on a mirror's position, the output is fed to either a port, or a drop port. Optical circulators are used at the I/O ports to form a full four-port (in/drop/add/pass) system.

MEMS and MST technologies are reaching into other aspects of the communications market. Tiny microrelays, tunable micromachined capacitors, microinductors with high Q levels, low-loss micromechanical switches and microscale vibrating mechanical resonators with very-high Q levels are all on the immediate horizon for wired, wireless, and RF communications. The emphasis is on miniaturization in conjunction with performance enhancement.

Miniaturization can be seen in what has been claimed to be the world's smallest commercially available MEMS microrelay. Developed by MCNC, it's just 1.5 by 1 mm and only 600 µm high. But, that's only half the story. Capable of handling 300 mA and proven to withstand 10 million cycles without a mechanical failure, this device is aimed at telecommunications, ATE, and other emerging applications.

The thermally actuated microrelay features nickel-surface micromachined components with gold contacts for high conductivity. Its metal-to-metal gold contacts offer a low on-resistance of less than 300 mΩ and a high off-resistance of 1013 Ω.

Researchers at the University of Michigan are working hard to develop transceivers for wireless communications that will offer both high performance levels—beyond those available with conventional components—and smaller sizes to boot. The researchers have described a dual-conversion receiver in which they employ MEMS capacitors, integrated high-Q inductors, low-loss MEMS switches, and vibrating MEMS resonators with Q values in the tens of thousands range.

As an illustration of space-saving gains, the researchers compared the size of a typical surface acoustic-wave (SAW) 100-MHz resonator that they used in a receiver to that of a clamped-beam MEMS resonator they used in their design. The former typically occupies several square millimeters compared with just 420 µm2 for the latter. In addition, the smaller size affords even less power dissipation.

Another subject of high interest to the MEMS community is the topic of MEMS RF filters. Earlier this year, Cronos collaborated with Raytheon Electronic Systems to develop a frequency agile RF filter using MEMS parts and featuring high performance. They demonstrated how MEMS microrelays can be used in filters to handle power levels beyond those previously possible in a MEMS structure.

The demonstration device can handle 25 W, a considerable level for a MEMS-based microrelay. The filter is tunable to four separate sub-bands in the VHF spectrum, ranging from 44 to 56 MHz. Total insertion loss of the filter is 0.8 to 1.0 dB, with the MEMS relay contributing a total of 0.1 to 0.2 dB of that loss.

Board Area Cut By Half
By employing MEMS relays in the tuning process, the board area and volume of the filter's tuning components were reduced by 50%, compared with existing RF filters using p-i-n-diode switches (and their associated drive electronics). Both Cronos and Raytheon are working on bringing the device's power level up to 40 W.

Because it raises many issues beyond those encountered in conventional semiconductor ICs, the assembly and packaging of MEMS devices isn't a trivial task. Signal and power distribution, heat dissipation, and protection from external mechanical, chemical, electromagnetic, and other environmental factors are just a few of these issues. Moreover, assembly and package costs must be inexpensive enough, relative to the manufacturing costs of the MEMS die, to make it attractive for a wider market demand. The truth is, though, assembly and packaging costs can often constitute the great majority of the total costs of producing a MEMS device for the market.

A study conducted by Microtec Associates and the Center for Automation Technologies at the Rensselaer Polytechnic Institute further points to this problem. In the study of cost breakdown for a microsystem, it was determined that 10% to 25% of the costs were borne by the silicon, 15% to 20% by measurement and testing, and a whopping 55% to 85% by packaging, interconnects, and microassembly.

No generic MEMS packages exist. Usually, the packages are application-specific, making it difficult to standardize on any one type. That's because MEMS sensors and actuators include fragile structures and serve a very wide range of applications, often in harsh environments. Viscous damping of moving parts and vacuum packaging are sometimes needed. Failure due to moisture, suction, thermal effects, etc. is common.

By their very nature, many sensing systems must interact with their surrounding environment. A critical issue is that many MEMS devices, particularly actuators, contain moving parts. Therefore, these must be placed in packages with free space. In turn, this makes the issue of particulate contamination during manufacture more severe than for conventional ICs.

Gaps between moving parts can be on the order of 1 µm or less, exacerbating the contamination problem, and affecting device processing yields, thus, increasing costs. Texas Instruments, for example, has reported on yield problems for its digital micromirror device (DMD), presently in use in a broad range of applications.

In response to these challenges, some companies are offering many more packaging options, providing packages that at least can be made to fit an application. Amkor Technology, for example, points out that this year alone 900 patents were filed in the U.S. for MEMS packaging. The way the company views it, the problem is a lack of knowledge on how to apply currently available packaging options to MEMS devices. They see the definition of package performance requirements as key to selecting the current package for a given application.

Plus, standard catalog parts are becoming more available. For example, Teledyne's Microelectronics Div. offers a catalog of 13 packages, many of which can be used to serve a number of MEMS applications.

An "indent reflow-sealing" method for packaging is being proposed by IMEC vzw. Developed under support from the European Commission via the ESPIRIT program, the technique provides low-cost hermetically sealed plastic packaging for final encapsulation of a MEMS device, like an accelerometer or a microrelay.

IMEC calls this indent reflow-sealing technique "0-level" or "wafer-level" packaging, compared to a "1-level" approach used for standard IC packaging. It allows the formation of a cavity during wafer processing, in which to place a moving MEMS structure. Test cavities, typically 1 to 4 mm in diameter and with sealing rings of 100 to 400 µm wide, were successfully fabricated and tested.

First, the MEMS structure is capped with another chip/wafer in which a cavity is made. After deposition and patterning of under-bump metallization, metal space, and solder on chip 1, a top metallization surface is deposited onto chip 2 (Fig. 3). Following is the creation of an indent in the solder, and the alignment of both chips on a flip-chip tool. Finally, the top chip/wafer is sealed with a top cap structure.

Hermeticity creates one large problem with packaging. According to the researchers at the Georgia Institute of Technology, the full utility of microsystem structures will never be realized as long as they need hermetic enclosures. For such devices to achieve their full potential, they must be able to operate without the protection of a hermetic package, so that they may be fully immersed in their operational environments.

The researchers are proposing a solution—a generic, nonhermetic, direct-chip-attach package (Fig. 4). This package is made up of four distinct components: a substrate, a cap, a bond region, and a through-wafer electrical-interconnect structure. The package is compatible with wafer-level and surface-mount (SMT) processing.

Analog Devices, the producer of the world's first commercially available surface micromachined accelerometer for automotive airbags, the AXL202, has succeeded in developing a much smaller-packaged version, the tiny AXL202E, This dual-axis accelerometer is housed in a chip-scale package (CSP) just 5 by 5 by 2 mm. The tilt/motion sensor opens up new consumer applications where small size and low cost matter. Priced at $4.49 each in 100k quantities, it will find a home in wearable computing devices, personal digital assistants (PDAs), cell phones, laptop computers, toys, and navigation systems. The device operates from as low as 2.7 V and dissipates just 2.7 mA per axis. Production quantities will be available this August (Fig. 5).

Another sharp thorn in the side of an MEMS designer is testing. There's no basic set of parametric test structures for MEMS.

"Testing MEMS is a real challenge" notes Dick David, director of the Microtechnology Center for Teledyne Microelectronics. "Special tooling is often needed, which drives up the cost. There also are many different testing needs, especially at the low-volume prototyping stage, for optical, accelerometer, gyroscope, relay, and switching MEMS devices," he adds.

Testing requirements of MEMS/MST include in-situ methods for testing material properties, and for accelerated testing. Customized test equipment and procedures are needed for testing such parameters as pressure, temperature, shock, acceleration, air and fluid flow, etc. Such parameters must be dealt with as input stimuli or outputs, or both.

Operating voltage levels are yet another problem. Many MEMS devices operate at power-supply levels greater than 5 V. In an era emphasizing ICs with lower and lower operating voltages of 3.3 and 1 V, this is a tremendous problem. Still, some efforts at improving this situation have been made.

A couple of years ago, for instance, Sandia National Laboratories devised instrumentation known as SHiMMeR (or Sandia High-voltage Micromachine Measurement of Reliability). This instrumentation allows testing of MEMS devices of up to 256 parts at a time, while conventional small-scale testing can only handle testing of one, two, or three parts at a time.

SHiMMeR consists of a Plexiglass enclosure that contains a base for testing MEMS parts, and a high-powered optical microscope and video camera for the observation and recording of device failures. Each MEMS device that's tested is attached to a cable which sends it an activation signal once testing begins. The humidity in the test enclosure is controlled during testing by the operator.

When a MEMS device fails, it's removed from the enclosure and a cross-sectional view is taken of it by a focused ion beam. This cross section is then observed by the microscope, allowing the test operator to draw conclusions about how, when, and where failures occur.

Reliability testing is one more big problem. Just how do you define long-term reliability for MEMS devices? It's apparent that VHDL/analog simulators are needed as well. Furthermore, process tools must be faster and more capable of deeper etching. Cost modeling, which is generally not included in simulators, is important to manufacturing engineers and, therefore, should be included.

There's no standard platform for integrating MEMS structures with other IC processes, particularly CMOS. Unlike conventional CMOS ICs, MEMS devices scale differently with respect to dimensions and voltages. While the frontend of a MEMS process may be similar to that of a conventional IC, it differs at the backend (Fig. 6).

Moving mechanical MEMS structures must be handled differently during a process flow. Special sectioning, probing, and handling procedures are needed to protect these parts, some of which might be sealed later, and some which must remain exposed to interact with the environment they serve.

From all of these challenges, the term "seamless microsystem engineering" has risen. This is a foundry strategy based on the cluster integration of several stages in the production cycle. Seamless microsystem engineering is said to allow a strategic alliance to form between multidisciplinary partners. Eventually, it leads to a shorter concept-to-production cycle, and minimizes costs. In fact, several MEMS companies, specifically many European companies, have implemented, or are in the process of implementing the seamless microsystem engineering concept.

Aside from the well-known technical barriers to MEMS/MST commercialization, many MEMS experts with more business acumen have pointed out a certain salient fact that tends to hinder commercialization. These experts acknowledge that embracing MEMS for the sake of technology can be a big stumbling block. Customers really don't care how a device is made, or by what technology it's produced, as long as it solves a particular problem for them.

Fascinating is the technology that makes possible lilliputian structures like microgears to arm an atomic bomb, microrobots able to clear out clogged human arteries, complete DNA analysis systems on one tiny chip, and miniscule micromotors that can drive other miniscule structures such as microcars. It's understandable that this technology can fire up an engineer's imagination to no end. But, that facet is quickly disappearing in the face of the market pressures that place more of an onus on MEMS /MST designers to arrive at the right solutions to critical problems.

Of course, MEMS/MST technology has inherent performance, small-size, reliability, low-weight, and low-cost advantages. In that respect it can be the right answer for many applications. That's what has been driving the MEMS market for some time. This is particularly true in an era where smaller, lighter, less-power-hungry, and cheaper electronics are the call to arms (see "A Low-Cost Investment Road Beckons," p. 92).

In their efforts to reach larger commercial markets, these are the lessons that many MEMS companies have learned the hard way. If the present trends of technology maturation continue, MEMS and MST devices will become even more common, and the next age beyond of nanoelectronics using carbon fibers, nanotubes, etc., might not be too far off on the horizon.

Need More Information?
Amkor Technology Inc.
(610) 431-9600

Analog Devices Inc.
(781) 937-1428

Applied Micro-engineering Ltd.
+44 (0) 1235 833934
e-mail: rob@aml.co.uk

Bartels Microtechnik GmbH
+49 2319742-500
e-mail: info@bartels-microtechnik.de

Cadence Design Systems Inc.
(408) 943-1234

The California Institute
of Technology, Jet
Propulsion Laboratory
(800) 568-8324

Cronos Integrated
Microsystems Inc.
(919) 380-1316

DERA Malvern
+44 (0) 684 894586
e-mail: mmcnie@dera.gov.uk

Elfo AG
+41 41 666 7121
+49 231 7549-150

Elmos GmbH

Exponent Inc.
(508) 647-1899
e-mail: boston-office@exponent.com

Georgia Institute of Technology
(404) 894-4135
e-mail: daniel.baldwin@me.gatech.edu

Roger Grace Assoc.
(415) 436-9101
e-mail: rgrace@value.net

IMEC vzw
+32 16 28 1223
e-mail: tilman@imec.be

Lucent Technologies,
Bell Laboratories
(732) 949-2932
e-mail: jaw@lucent.com

Micro*Montage bv
+31 3560 26400
e-mail: info@mikromontage.nl

Microtec Associates Center for
Automation Technology
(716) 235-4630
e-mail: microtec@worldnet.att.net

Raytheon Electronic Systems
(219) 429-6040
e-mail: lamcmi@ftw.rsc.raytheon.com

Rensselaer Polytechnic Institute
(508) 276-6216

Sandia National Laboratories
(505) 844-5373
e-mail: dressepv@sandia.gov

Standard MEMS Inc.
(516) 435-6004
e-mail: doug.fink@smsc.com

Texas Instruments Inc.
(800) 336-5236

Teledyne Electronic Technologies
(310) 822-8229

Twente MicroProducts
+31 53 4800 111
e-mail: j.elders@microproducts.nl

University of California
at Berkeley
(510) 642-6000

University of Michigan, Dept.
of Electrical Engineering
and Computer Science
(734) 764-1817

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.