MEMS technology roadmapping could address test challenges

San Francisco, CA. As semiconductor technology extends into “more than Moore” territory, test engineers will need to contend with integrated devices containing multiple MEMS structures. A roadmap can be useful in establishing effective test strategies, according to Dr. Michael Gaitan of NIST, who delivered an invited address July 12 at the Test Vision 2020 workshop held in conjunction with Semicon West.

Gaitan, who chairs the iNEMI and ITRS MEMS technology working groups, said that his intention was not to talk about solutions but rather problems. Interesting thing about road-mapping, he said, is the effort to try to find common ground. Consensus metrics arrived at by working groups, he said, are ones least objectionable to the range of companies working together.

Current roadmapping efforts, Gaitan said, focus on MEMS used mobile devices, including accelerometers, magnetometers, gyroscopes, personal health and fitness devices, haptic devices, fingerprint sensors, and humidity sensors.

Consumer and medical MEMS applications are growing rapidly, Gaitan said. The market is small compared with the overall semiconductor industry, but five or more MEMS going into each mobile device, the market is growing, Gaitan said, citing forecasts from IHS iSuppli and Yole Développement. He also presented information from www.wtc-consult.de on proliferating applications for MEMS within automobiles.

Medical market offers particularly high potential, Gaitan said, 3 billion potential customers—the “worried well”—for connected health devices. He cited BodyMedia as an example of a company addressing this market—the company offers accelerometer-base devices that monitor your activity level.

The MEMS market has been subject to the Gartner Hype Cycle model, with a peak of inflated expectations hitting in 1988 for MEMS. That was followed by a trough of disillusionment in 1995, followed by a slope of enlightenment moving on to a plateau of productivity. The height of the plateau, Gaitan said, will depend on the existence of killer apps. He also noted that MEMS expert Roger Grace has identified a commercialization timetable of 30 years from first paper to the appearance of devices on the market. One goal of roadmapping, Gaitan said, is to shorten that time.

A roadmap, Gaitan said, is a plan that matches goals with solutions. There are several problems with generating a roadmap, including settling on figures of merit and prompting companies to work together and share information. The semiconductor industry has settled on technology nodes as signs along its roadmap, but, said Gaitan, MEMS are too diverse, with different processes for each device, secret sauces from each manufacturer, and intense competition. In an effort to simplify the process, he said, participants decided to narrow the focus to MEMS technologies that will be part of mobile internet set of devices.

Gaitan suggested that, for inertial sensors, a roadmap might be based on degrees of freedom (DOF) at both the package and chip level. As an example, he said integrated 10-DOF chip-level products might appear in 2017. Grand challenges remain, he said, and will involving on standardization of MEMS packages to support integration, identifying the role of wafer-level testing, establishing methodologies for deisng for test or design for no test, and gaining more knowledge of the physics of failure. He noted that today, device fabrication accounts for one third of MEMS manufacturing cost, while packaging and testing accounts for the other two thirds. In contrast, R&D investment in packaging and testing is miniscule in comparison to R&D investment in device and process development. “We know how to design devices, but we haven’t put a fare share of investment in the backend, and maybe we should do that,” he said, adding that testing alone can consume 20 to 60% of manufacturing cost, depending on the application.

Unfortunately, he said, there is little uniformity in reporting MEMS performance in device datasheets, an issue iNEMI is addressing. He noted that companies often develop custom test capabilities and train their own test engineers. MEMS sensor fusion will continue to create challenges for testing increasingly complex devices while lowering test costs. In addition, he said, tools are needed for DFT, BIST, and self-calibration.

“Our road-mapping so far is near-term—we are looking at incremental improvement in performance,” Gaitan said, with the near-term being five years or so. The concept of the integration node might facilitate longer term road-mapping of MEMS and other more than Moore technologies, he concluded.

*Note: the Test Vision 2020 site does not include a copy of Gaitan's slides. However, slides from a similar presentation on May 31 at the 3rd Nano-Tec Workshop (including Yole and iSuppli forecasts) is available here.

See related Semicon West news and comment: