Consumer Markets Drive MEMS Scaling and Efforts to Cut Development Time

By Paula Doe, SEMI Emerging Markets

Robust global demand for smart phones and tablets will drive steady double-digit annual growth for MEMS devices for the next five years— considerably outpacing the mainstream IC market—   to reach a $12-17 billion market, predicted market analysts at November’s MEMS Executive Congress 2011. These high-volume, low-price, fast-changing consumer markets are pushing the MEMS sector to focus more on scaling device size, on finding ways to reduce development time to reduce costs, and on developing the infrastructure to enable faster time to market of new applications. Just how well they succeed may determine if it’s 10 percent or 19 percent growth--and who makes money in the business.

What’s Behind the Forecasts

IHS iSuppli— whose MEMS analysts now can draw upon end market analysts in consumer, automotive, energy and aerospace markets at its new parent IHS— is the most conservative, projecting 10 percent growth in 2011 to just under a $8 billion market, and 10.5 percent CAGR to just under $12 billion by 2015.  The firm figures that the maturing of the smart phone market and rapidly falling ASPs— especially as increasing use of combo sensors that share one ASIC accelerates price erosion— means consumer MEMS will have only a couple more fat years of >20 percent growth before slowing back to 10 percent levels. The risk to the forecast?  “The impact of China,” said Jérémie Bouchaud, IHS iSuppli director and principle analyst for MEMS and sensors, noting that Chinese demand has been a strong driver for growth in automotive and telecommunications markets, driven in part by government policies whose changes are hard to predict.  Though the general world economy will certainly impact demand, smart phones and tablets continue to sell strongly despite the economy, barring of course major shocks from defaults by Greece or Italy.

Source: IHS iSuppli

Yole Développement takes a somewhat wider view of the MEMS market— including  microfluidics on glass and polymer as well as on silicon, and digital compasses that aren’t strictly MEMS devices but are increasingly integrated with them— to project 16 percent growth to a roughly $10 billion market this year. It also takes a somewhat more optimistic view of the pace of new product adoption, particularly in microfluidics and biomedical applications, to project 15 percent average annual growth to a $17 billion market in five years. One key risk factor is possible “slower than expected” wide adoption of new MEMS devices coming along the pipeline, said Yole CEO Jean Christophe Eloy, noting the example of RF switches for cell phones that have been the next big thing for some time now.  “RF is really, really close and we think it will be a very big business,” he noted. “But when it will really happen, we have no idea.” Ramping manufacturing has, as usual, been more difficult than originally expected, and the initial products primarily deliver better connections, which the Apple4 antenna issues showed turns out not to be a “must-have” feature, though next generation bandwidth tunability should be a bigger draw.  Separately, executives at stalwart RF startup WiSpry told SEMI that their antenna tuning MEMS capacitor arrays were now in volume production, and tear downs would start to find them in cell phones on the market by the end of the year. 

Source: Yole Developpement

Also now forecasting the MEMS market, and most optimistic of the analysts, is Semico Research Corp.  Semico figures MEMS will see >20 percent growth this year to reach nearly $9 billion, and then see growth averaging in the high teens to reach $17 billion in 2015.  Semico figures growth will continue to be driven by smart phones, but with healthy contributions going forward from computing and industrial applications, while price erosion means the big automotive market will start to decline.  “High inventories in the IC industry in general and the slow economy will limit investment, but we believe growth will return because these consumer devices are now must-have products,” said Tony Massimini, Semico chief of technology. He noted that Asian PCB suppliers report smart phone builds maintained robust 10 percent growth in Q3, and shipments for Apple tablets remained strong, though shipments have weakened below normal seasonal levels for most other consumer electronics goods, from cameras and MP3 players to TVs and DVD players. That all makes MEMS a bright spot in the semiconductor industry, where Semico figures IC revenues will decline 1.4 percent this year.  But as MEMS become more mainstream and more like mainstream semiconductor cycles, the sector can expect 8.2 percent average annual declines in ASP—but these lower prices will help expand demand in new markets.

MEMS Sales by Applications

Source: Semico Research Corp. (Phoenix, Ariz.)

Maturing Infrastructure Eases Growth of New Applications for Existing MEMS Devices

Lower-cost MEMS devices, and the growing infrastructure of sensor management software and support from everyone— from MEMS suppliers to microcontroller makers to third-party software companies— are easing the development of new applications for existing MEMS devices.  Julie Ask, VP and principal analyst, Forrester Research, also pointed out that the growing base of smartphones and app developers provides a convenient platform for immediate user feedback, a great enabler for potential health and fitness applications, though so far only some 5 percent of consumers have downloaded any health or fitness apps. This same ecosystem also puts the crowd of app developers to work developing sensor applications.

MEMS Industry Group showcased a selection of products culled from many more applicants using MEMS devices, expanding the range of MEMS applications as sensors and actuators continue to grow more familiar, more affordable, and easier to use with better software and support. And indeed much of the value added in these products is in the software that does something of interest with the MEMS sensor data. Among the most interesting were sports goggles with a near-eye display from Recon Industry, designed to give real-time analytics feedback for skiers or snowboarders who already wear goggles, as well as to record the data to track changes. The tiny display— visible only when one looks down— is not distracting, can show speed, distance, jump altitude and hangtime in the air as well as trail maps, GPS location, and where pals are on the hill. Syride showed off a tracking unit that attached to a surfboard to record similar performance feedback like speed and wave size and what GPS position worked best for catching the most waves.

ROR3 Devices displayed a heart-rate monitor wristband, apparently more comfortable than the usual chest bands for monitoring one’s heart rate while exercising, using MEMS microphones for noise cancellation for better signal quality. Proteus Biomedical showed its MEMS-based tags for pills that send out an RF signal to the smart phone when swallowed. Sunrex demonstrated an air mouse pointer with mini keyboard. MicroVision showed its pico projector that linked with HDMI cable to an assortment of other electronics gear for sharing photos, videos or presentations.  

Consumer Markets Drive Push to Cut Development Time

Consumer market cost and cycle pressures are forcing the MEMS sector to look seriously at how to speed up its traditionally long development times. The maturing industry, its accumulated engineering experience, and improvements in process equipment, now mean development is getting more efficient. But reasonably conventional new consumer MEMS devices now average 2-3 years in development, automotive devices 3-5 years, still about twice as long the semiconductor industry, noted several foundry executives. “We need a faster ecosystem for the consumer market,” argued Teledyne Dalsa’s Donald Robert, VP sales and marketing, noting that the sector’s traditional focus on a unique process for each individual product was no longer optimal for these fast-moving, price-sensitive markets, and alternative approaches were needed.

One solution is clearly more re-use of common technology process flows and platforms within closed ecosystems. Individual OEMs and foundries are concentrating their production in a limited number of technology platforms and re-using the same volume processes or, say, package designs, as much as possible. MEMS’ traditional interest in unique production technology can be a distraction, argued Dave Monk, Freescale Semiconductor automotive sensor operations manager. “From 1992 to 2006, we developed 11 different process flows. Since then we’ve developed only one,” he noted.  “We were technology rich but product poor--and you make money on products, not technology.”

Rakesh Kumar, Global Foundries’ director of MEMS, suggested that there are actually only a limited number of major product families and major process platforms in the sector, and use of standard approaches like SOI-wafer-based structures and wafer-level packaging could cut the time and cost of development. He noted that the maturing industry means that barriers to entry have decreased markedly, enabling the company’s new MEMS foundry to ramp from zero to initial production in 15 months. 

Specialty foundries also try to re-use platforms in-house, but customer-specific IP can complicate matters. Micralyne CEO Nancy Fares noted that her foundry looks to save costs and time by re-using its packaging designs, and more industry-standard packages like the IC industry uses could be a solution for some product categories. Robert said that Dalsa is moving slowly towards 5 to 6 in-house technology platforms, but suggested innovative alternative licensing or joint development solutions could benefit everyone. “We have cut time to market by platforms,” he said, “but people still want to differentiate on production technology too.” He suggested more technology could be re-used by negotiating with the customer/partner for rights to license the technology to others, or by two to three customer/partners sharing the cost of joint development, with exclusive rights for a few years. Silex Microsystems CEO Gary Johnson suggested the MEMS sector might eventually approach the analog IC model, where the industry uses well defined smart blocks at the very bottom, but all players retain their considerable separate expertise as well.

Producers are perhaps taking some first steps towards licensing and cooperating on development to make some of these in-house technology platforms more widely available. Silex  Microsystems has made its TSV process available to development houses A.M. Fitzgerald and Associates and Nanoshift, and A.M. Fitzgerald also offers its customers a path to manufacturing with the InvenSense Nasiri active capping process.

Industry veteran Janusz Bryzek, now Fairchild Semiconductor VP of development, MEMS and sensor solutions, says he’s getting interest from some big MEMS players in a consortium that would jointly develop and own basic process technology to share the costs, as they move to focus more on distinguishing their products by design, to spur MEMS industry growth to more serious volumes. NIST is also exploring establishing a MEMS consortium to facilitate manufacturing.

Scaling Die Size with Unfilled Vias and Sub Micron DRIE

Though the performance of the mechanical MEMS structures, unlike typical IC circuits, gets worse instead of better at smaller size, MEMS makers are still aggressively reducing die size to bring down ASPs.  Smarter design and higher production volumes that are bringing better control of processes are enabling more aggressive reduction of feature sizes, with leading edge consumer accelerometer, microphones and pressure sensors now down to 1mm² to 2mm² packages.  Even high-performance multi-axis inertial measurement units are down to around to a cubic inch or two.

New processes like TSVs and improved DRIE are also helping. STMicroelectronics reported going to TSVs in its volume accelerometer production to reduce die size by roughly 20 percent   and allow flip chip attachment on the ASIC.  Benedetto Vigna, ST GVP and GM of MEMS, sensors and high performance analog, said key to the low-cost approach was to not fill the vias, but instead to use the air gaps to insulate the relatively wide pitch MEMS connections in polysilicon.

Applied Materials’ MEMS global product manager Mike Rosa showed results from the company’s second year of development of a DRIE chamber for sub-micron MEMS.  He argued that reducing the gap between the capacitive fingers of inertial sensors from the current ~1 µm state of the art feature size down to 100nm or 200nm would enable more sensitive and smaller devices, and showed SEM images of aspect ratios of 87:1 and >100:1 at those dimensions.

SEMI
www.semi.org
December 6, 2011