MEMS: Flexible, Reusable Platforms Facilitate Innovation

As the game-changing enabler for whatever the emerging market of widespread fragmented intelligence turns out to look like, the MEMS sector is in some ways the bellwether for much of the greater semiconductor/components supply chain looking to rethink how to serve a wider range of fragmented applications with lower costs and faster time to market. Leaders from Cisco, InvenSense, Nasiri Ventures, Robert Bosch, Yole Développement and others suggest key factors are accelerating innovation, smarter use of flexible common platforms, and working more closely with others across the value chain at the recent MEMS Executive Congress in Napa, Calif.

Problem: Need to Deal with Decrease in Innovation
While hardware makers always face declining prices as their market matures and volumes and competition grows, those lower costs also enable new applications and wider use. “The problem is not the decrease in costs, but the decrease in innovation,” noted Steve Nasiri, principal, Nasiri Ventures. He noted that the market was driving the industry to solutions to drive innovation faster and at lower cost, such as common manufacturing platforms. And various sorts of flexible platforms were increasingly in the conversation at this year’s sector event.

Nasiri argued that future MEMS would likely use mainly CMOS, piezoelectrics, and TSVs, with chip-scale packaging and testing, to manufacture smart cluster sensing solutions that output contextual information. And key to reducing the $100 million and 10 years it has typically taken to develop a new MEMS device will be using more CMOS compatible processes, transferable to different foundries, for lower costs and quicker time to yield. “We have to build the solution, not just the component, and with faster time to market,” he suggested. “This is not going to come from the titans. It will have to come from startups, and the MEMS foundries that work with them.”

Though many MEMS devices can never be made with CMOS processes, companies are indeed figuring out how to use the common platforms at the MEMS foundries for more devices. Elevator-pitch-winning startup NEXSENS Microsystems made its reportedly higher performance MEMS timing device on TeledyneDALSA’s inertial sensor platform, taking advantage of multi-project runs and bonding to wafer-level CMOS to cut costs and speed time to market. InvenSense recently added AlN piezoelectric capability to its own CMOS-MEMS platform developed for inertial sensors to now make ultrasonic fingerprint imagers.

Yole’s Eloy suggested the industry might converge on a few types of sensor cluster platforms with common packaging and software: motion/location sensing units with inertial and magnetic sensors, open- cavity environmental sensing units with pressure, gas, and humidity sensors, and open- eye units integrating optical devices such as proximity sensors, cameras and multispectral sensors.

A Robert Bosch executive noted that the company’s new flexible, programmable nine-degree-of-freedom inertial sensor platform, designed to ease development for users in what he joking called its “business-to-garage business model,” was selling some 1,000 units a week, so even if the success rate for new products was only 0.01 percent, that would still work out to a new product every week.

No One Can Do It Alone
The next billion-dollar products will come from the university research labs, noted Alissa Fitzgerald, founder and managing partner of AMFitzgerald. “Industry needs to reach out to academia and collaborate with them to sharpen their focus on topics relevant to industry,” she noted. “And a little bit of money goes a long way to support university research.”

Current research with most promising in the next few years, she suggested, are more accurate, navigation-grade gyros, for precision navigation for drones and cars, such as being developed at National Tsing Hua University in Beijing, Georgia Tech, and the Polytechnic University of Milan. Also potentially big, and not that far out: sensors that can listen without using any power, such as those being developed by researchers at University of California Berkeley and the Shanghai Institute of Microsystems and Information Technology.

Finding “Real Use Cases”
If the sensors are increasingly a commodity, the integration and software for the wide range of specific use cases is certainly not. “The hardest problem is ‘show me the money’ – to find the real use cases,” said Peter Hartwell, senior director, InvenSense. And that may take some exploring with users, who don’t really know what’s possible in this new era of increasingly sophisticated sensing.

Analog Devices’ senior sensor technologist Rob O’Reilly has found some low-hanging fruit in an exploratory effort to apply ADI’s sensor data to help New England tomato farmers grow better crops, and their customers get better tasting tomatoes. A sensor cluster that transmits data on temperature, moisture and light enables better knowledge of tomato growth needs, and is helping some farmers grow healthier plants. ADI is also using optical sensor technology for chemical analysis of the tomatoes’ chemistry for taste components such as lycopene, sugars, acidity, salinity and conductivity. This analysis capability correlated with the taste ratings of an expert panel of tasters, and enabled a tomato sauce maker to select tomatoes for better tasting product at lower cost.

Rhiza CEO Josh Knauer noted that the advertising industry would love to have real, observable data on consumer buying behavior. The industry currently spends a trillion dollars a year on marketing, based almost entirely on rather crude demographics like age and income, or unreliable self-reported information, not on observed data. He suggested that sensor networks that tracked the customer’s path through the store, and where he looked on the shelf, could provide big value to the advertising industry, and the data analysis companies like his that sell to it.

Similarly, Kara Dennis, managing director of mHealth at clinical research software supplier Medidata, noted that the pharmaceutical industry spends $145 billion a year on clinical studies to determine the effectiveness of new medications, using surprisingly crude measures, such as counting how many steps a person takes in six minutes, or how they rate their mood from 1 to 10. Sensors that could track objective measures like activity or heart rate over a longer and more realistic period of daily life could potentially provide much better data, and ease researchers’ biggest headache of recruiting test subjects. The sensors for these tests don’t need medical approval, but they do need to be able to assure the quality of the data – that the right person is wearing the device ─ and require no effort on the part of the patient to assure compliance.

Less Data, Not More
Various other data users stressed that they actually wanted less data, screened to supply just the information they needed to act on, when and where they needed it. Both Todd Miller, GE Global Research Microsystems and Mony Wechsler, Montefiore Medical Center, noted they were getting too much information that doctors and nurses couldn’t use. They needed a simpler, more digestable red-light, green-light approach, to tell them only the information they needed to act upon, such as only the readings that were out-of-range, that the doctor needed to check, or only the pressure sensor signals that warned when a patient needed to be turned to avoid pressure ulcers.

The innovation may need to come in the business model. “The model has been that the data comes from a single, whole-solution provider,” notes Cisco’s Kranz. “But now customers are wanting more open, mix-and-match systems. The big systems people will have to give up part of the middle, so we can all move forward faster, but there’s a lot of resistance from the old players. The opportunity for the components suppliers is to look at who the whole solutions providers are that they can partner with.”

“We are the sensor data experts,” noted InvenSense’s Hartwell, suggesting the possibility of sensor makers collecting data from many users, stripping it of identifying information, and using it to build up a richer quality of data.

With a complicated mixture of innovation, common platforms that speed development, and working with users to solve real problems, the MEMS sector is succeeding in this fragmented market. Learn more about such approaches to solving real problems in the electronics manufacturing supply chain at upcoming SEMI (www.semi.org) events, including the Industry Strategy Symposium, 2016 Flex, and SEMICON West 2016.