Yesterday, Cadence hosted a mini conference on MEMS (microelectromechanical systems) with MEMS EDA vendor Coventor. EDA Investor and industry gadabout Jim Hogan gave the keynote and Coventor VP of engineering Steve Breit filled the audience in on some technical details. MEMS has been around for decades to make sensors and actuators out of silicon—Motorola Semiconductor (now Freescale) made simple, uncompensated MEMS pressure sensors back in the 1980s. Today there are several types of MEMS devices. The devices list includes MEMS sensors (accelerometers and microphones), MEMS actuators (micro-mirrors, relays, switches, and varactors), and MEMS hybrid devices that combine aspects of both sensors and actuators (gyroscopes and resonators).
All of these MEMS devices require the use of unusual etching and other strange, non-CMOS silicon processing to produce moving, bending 3D structures that rise significantly above the plane of the devices’ silicon substrate. On top of the MEMS devices themselves being manufactured in 3D, it’s increasingly clear that 3D assembly is important to MEMS’ future because 3D assembly can get the support electronics needed to amplify, linearize, and digitize MEMS sensor signals to prepare them for use in the greater world of digital CMOS SoCs.
There are significant challenges to the mainstreaming of MEMS, which have always been “special” in terms of “special processing,” “special packaging,” and “special design tools.” For every occurrence of the word “special” in that last sentence, see dollar signs because “special” adds cost. And that leads us to Jim Hogan’s keynote.
Hogan noted that MEMS devices have already reached the high-volume world of consumer electronics. The top buyers of MEMS devices are the systems companies making mobile phone handsets and video games. Even so, said Hogan, the MEMS market is highly fragmented and no MEMS market is greater than about $200M. Yet MEMS devices clearly help system designers differentiate their products. Both video game controllers and mobile phone handsets use MEMS devices to translate user motion into application commands. In addition, said Hogan, there’s fierce fighting for every millimeter of board space in all mobile devices so you can expect even greater emphasis on MEMS/support electronics integration, which means more 3D assembly in MEMS’ future.
To get greater penetration into these markets, said Hogan, simplifications to MEMS design and manufacturing must still come if the costs are to drop. There is progress on both fronts according to Hogan. For example, TSMC and Globalfoundries have both announced plans to become MEMS foundries.
On the design side, companies like Coventor are trying to bring better design tools to market to break what Steve Breit called the “build and test” MEMS design cycle. Without the ability to accurately model and simulate a MEMS design, MEMS designers must actually build a device to prove a design. That path costs a lot of money and takes a lot of time. IC design outgrew the “build-and-test” cycle a long time ago, noted Hogan, who said that when he was analog designer you needed a PhD with expertise in physics and math to get a working chip. Fast-forward a few decades to today when EDA tools permit people with lesser degrees to design far more complex ICs.
That’s what Coventor is doing with its MEMS design tools, which use BEM (boundary element modeling) and FEM (finite element modeling) to perform the multiphysics calculations needed to simulate a MEMS sensor or actuator design. Then these Coventor tools can produce objects that come into the Cadence Virtuoso Analog Design Environment through the library manager as schematic symbols, netlist objects, and PCells so that MEMS devices can be treated like any other electrical component in a design.
And that’s the kind of advance that’s needed to bring MEMS into the mainstream said Hogan.