MEMS Executive Congress 2014
What’s next for MEMS?
By Paula Doe, SEMI
The proliferation of sensors into high volume consumer markets, and into the emerging Internet of Things, is driving the MEMS market to maturity, with a developed ecosystem to ease use and grow applications. But it is also bringing plenty of demands for new technologies, and changes in how companies will compete.
While the IoT may be all about sensors, it is not necessarily a bonanza for most traditional MEMS sensor makers. “The surprising winner turns to be optical MEMS for optical cross connect for the data center, where big growth is coming,” said Jérémie Bouchaud, IHS Director and Sr. Principal Analyst, MEMS & Sensors, at the recent MEMS Executive Congress held in Scottsdale Arizona from November 5-7.
The market for wearables will also see fast growth for the next five years, largely for smart watches, driving demand for motion sensors, health sensors, sensor hubs and software –but even in 2019 the market for sensors in wearables will remain <5% the size of the phone/tablet market, IHS predicts. The greater IoT market may reach billions of other connected devices in the next decade, but sensor demand will be very fragmented and very commoditized. Smart homes may use 20 million sensors in 2018, but many other industrial applications will probably each use only 100,000 to 2-3 million sensors a year, Bouchaud noted.
And most of this sensor market will be non-MEMS sensors, some mature and some emerging, including light sensors, fingerprint sensors, pulse sensors, gas sensors, and thermal sensors, all requiring different and varied manufacturing technologies. Much of the new sensor demand from automotive will be also be for non-MEMS radar and cameras, though they will also add MEMS for higher performance gyros, lidar and microbolometers, according to IHS. Expect major MEMS makers to diversify into more of these other types of sensors.
Figure 1: Growth shifts to new types of sensors. (Source: IHS)
Yole Développement CEO Jean Christophe Eloy looked at how the value in the IoT would develop. While the emerging IoT market is initially primarily a hardware market, with hardware sales climbing healthily for the next five years or so, it will quickly become primarily a software and services market. In five to six years hardware sales will level off, and the majority of the value will shift to data processing and value added services. This information service market will continue to soar, to account for 75% of the $400 Billion IoT market by 2024.
Figure 2: Initial IoT growth will be from hardware, but most of value will eventually be software and services. (Source: Yole Développement)
Re-thinking the business models?
The IoT will bring big changes to the electronics industry, from new technologies to new business models, as well as new market leaders, suggested George Liu, TSMC Director of Corporate Development. He of course also argued that the high volume and low costs required for connected objects would drive sensor production to high volume foundries, while demand for smart distributed processing wouldrequire more integration with CMOS and give the advantage to CMOS makers.
Liu projected these changes will mean a new set of companies will come out on top. Few leading system makers managed to successfully transition from the PC era to the mobile handset era, or from the mobile handset era to the smart phone era, as both the key technologies and the winning business models changed, and chip makers faced disruption as well. “For one thing, the business model changed from making everything in house to making nothing,” he noted. “The challenge is to focus on where one is most efficient.”
“The odds of Apple or Google being the dominant players in the next paradigm is zero,” concurred Chris Wasden, Executive Director, Sorenson Center for Discovery and Innovation at the University of Utah.
Lots of other things will have to change to enable the IoT as well. Liu projected that devices will need to operate at near threshold or even sub-threshold voltages, with “thinner” processing overhead, while the integration of more different functions will redefine the system-in-a-chip. Smaller and lower cost devices will require new materials and new architectures, new types of heterogeneous integration and wafer-level packaging, and an ecosystem of standard open platforms to ease development. TSMC’s own MEMS development kit has layout rules, but not yet behavioral rules, always the more challenging issue for these mechanical structures. “That’s the next big thing for us,” he asserted. “These huge gaps mean huge opportunities.”
IDMs and systems companies still likely to dominate
Still, the wide variety, and sometimes tricky mechanics and low volumes, of many MEMS devices have been a challenge for the volume foundries. The fabless MEMS model has seen only limited success so far and that’s unlikely to change drastically in the next decade either, countered Jean Christophe Eloy, CEO of Yole Développement. He pointed out that some 75% of the MEMS business is dominated by the four big IDMs who can drive costs down with volumes and diversified product lines. To date, only two fabless companies–InvenSense and Knowles—are among the top 30 MEMS suppliers.
Most of the rest of the top 30 are system makers with their own fabs, making their own MEMS devices to enable higher value system products of their own, which is likely to continue to remain a successful approach, as the opportunities for adding value increasingly come from software, processers, and systems. “MEMS value has always been at the system level,” noted Eloy.
GE’s recent introduction of an improved MEMS RF switch to significantly reduce the size and cost of its MRI systems is one compelling example of systems-level value of MEMS, as the little MEMS component has the potential to greatly extend the use of this high-contrast soft-tissue imaging technology. Though the company sold off its general advanced sensors unit last year to connector maker Amphenol Corp., it is still making its unique RF switch using a special alloy in house in small volumes as a key enabler of its high value MRI systems. These imagers work by aligning the spin of hydrogen nuclei with a strong magnet, tipping them off axis with a strong RF pulse from an antenna, then measuring how they snap back into alignment with lots of localized antennas with low power RF switches close to the body. “We’re now launching a new receive chain using MEMS RF switches,” reported Tim Nustad, GM and CTO, Global Magnetic Resonance, GE Healthcare. “Later we can see a flexible, light weight MRI blanket.”
Opportunity for smaller, lower power, lower cost technologies
So far, MEMS makers have driven down the cost of devices by continually shrinking the size of the die. But that may be about to change, as the mechanical moving structures have about reached the limit of how much smaller they can get and still produce the needed quality signal. That’s opening the door for a new generation of devices using different sensing structures and different manufacturing processes. For inertial sensors, options include bulk acoustic wave sensing from Qualtre, piezoresistive nanowires from Tronics and CEA/Leti, and even extrapolating gyroscope-like data with software from accelerometers and magnetometers. MCube’s virtual gyro with this approach, now in production with some design wins, claims to save 80% of the power and 50% of the cost of a conventional MEMS gyro.
Piezoelectric sensing, often with PZT films, is also drawing attention, with products in development for applications ranging from timing devices to microphones. Sand9 claims lower noise and lower power for its piezoelectric MEMS timing, now starting volume manufacturing for shipments in 1Q15. It has also recently received a patent for a piezo microphone, while startup Vesper (formerly known as Baker-Calling and then Sonify) also reports working with a major customer for its piezoelectric MEMS microphone.
More open platforms ease development of new applications of established devices
Meanwhile, the maturing ecosystem of open development platforms across the value chain is helping to ease commercialization of new applications of existing MEMS devices. The two latest developments in this infrastructure are a standard interface to connect all kinds of different sensors to the controller, and an open library of basic sensor processing software. The MIPI Alliance brought together major users and suppliers—ranging across STMicroelectronics and InvenSense, to AMD and Intel, to Broadcom and Qualcomm, to Cadence and Mentor Graphics—to agree on an interface specification to make it easier for system designers to connect and manage a wide range of sensors from multiple suppliers while minimizing power consumption of the microcontroller. A collaboration ofsensor makers and researchers are also making a selection of baseline algorithms available for open use to help more users speed development of new applications. Offerings include Freescale’s inertial sensor fusion and PNI Sensors’ heart rate monitoring algorithms, along with other contributions from Analog Devices, Kionix, NIST, UC Berkeley and Carnegie Mellon to start. The material will be available through the MIG website.
Plenty of companies have also introduced their own individual platforms to ease customer development tasks as well, ranging from MEMS foundries’ inertial sensor manufacturing platforms to processor makers’ development boards and kits. Recently STMicroelectronics also adding its sensor fusion and other software blocks to its development platform.
KegData is one example of a company making use of these platforms to easedevelopment of a solution for a niche problem – an automated system for telling pub owners how much beer is left in their kegs, using a Freescale pressure sensor and development tools. Currently the only way to know when a beer keg is empty is to go lift and weigh or shake it, a problem for efficiently managing expensive refrigerated inventory. Adding a pressure sensor in the coupler on top of the keg allows the height of the beer to be measured by the differential pressure between the liquid and the gas above it. The sensor then sends the information to a hub controller that communicates with the internet, letting the pub manager know to order more, or even automatically placing the order directly with the distributor. The startup’s business model is to give the system to distributors for free, but sell them the service of automating inventory management for their customers, saving them the significant expense of sending drivers around to check the inventory and take pre-orders.
More broadly, MEMS microphones are poised to continue to find a wide range of new applications. IHS’ Bouchaud pointed out that cars will soon each be using 12-14 MEMS microphone units, to listen for changes in different conditions, while home security applications will use them to detect security breaches from unusual patterns of sounds, from people in the house to dogs barking. Startup MoboSens says it converts its chemical water quality data into audio signals to feed it into the phone’s mic port for better quality.
Opportunities still for new types of MEMS devices
Growth will also continue to come from new MEMS devices that find additional ways to replace conventional mechanical parts with silicon. Eloy noted that MEMS autofocus units may finally be the next breakout device, as they have started shipping in the last few weeks, and aim at shipping for products in 2015. MEMS microspeakers are also making progress and could come soon. But ramping new devices to the high volumes demanded by consumer markets is particularly challenging. “The only way to enter the market is with new technology, but high volume consumer markets make entry very hard for new devices,” he said. “The market is saturated, wins depend on production costs, and not everyone can keep up…. The last significant new device was the MEMS microphone, and that was ten years ago”.
But innovative new MEMS technologies also continue to be developed for initial applications in higher margin industrial and biomedical fields. One interesting new platform is the MEMS spectrometer from VTT Technical Research Center of Finland. This robust tunable interferometer essentially consists of an adjustable air gap between two mirrors, made of alternating ALD or LPCVD bands of materials with different defraction indexes, explained Anna Rissanen, VTT research team leader for MOEMS and bioMEMS instruments. The structure can be tuned by different voltages to filter particular bands of light, while a single-point detector, instead of the usual array, enables very small and low cost spectrometers or hyper spectral cameras. VTT spinout Spectral Engines is commercializing near-IR and mid-IR sensors aimed at detecting moisture, hydrocarbons and gases in industrial applications. Other programs have developed sensors for environmental analysis by flyover by nano satellites and UAVs, sensors for monitoring fuel quality to optimize energy use and prevent engine damage, and sensors that can diagnose melanoma from a scan of the skin.
The Internet-of-Things will bring hardware and software designers into closer collaboration than every before. Understanding the working differences between both technical domains in terms of design approaches and terminology will be the first step in harmonizing the relationships between these occasionally contentious camps. What are the these differences in hardware and software design approaches? To answer that question, I talked with the technical experts including Harmke De Groot, Program Director Ultra-Low Power Technologies at Imec; Jonathan Austin, Senior Software Engineer at ARM; and Eric Nguyen, Director of Business Intelligence at Jama Software; . What follows is a portion of their responses. — JB
