If you work with micro-electrical mechanical systems (MEMS) today, you are probably a highly trained expert. And that's a problem. For MEMS devices to become more prolific in consumer devices, we need to "democratize" MEMS design and integration so it's not confined to a handful of PhD-level experts working in IDMs on projects that take four years to complete.
That's the message that EDA investor Jim Hogan brought to a Cadence/Coventor seminar held May 18 at Cadence San Jose headquarters. Coventor is a provider of MEMS design and simulation tools, and its MEMS+ offering, which facilitates MEMS integration into ICs, works closely with the Cadence Virtuoso platform. The one-day seminar provided a hands-on demonstration of that connection, as well as keynote speeches by Hogan and by Steve Breit, Coventor vice president of engineering.
Breit noted that there are three basic types of MEMS devices - sensors (such as accelerometers), actuators (such as micro-mirrors, relays, switches) and hybrid devices that combine both capabilities (such as resonators and gyroscopes). While automobile air-bag sensors are a well-known example of MEMS, the devices have also made their way into PC laptops, gaming systems, tablet computers, and smartphones. As Breit noted, "MEMS never stands on its own. It has to be packaged and integrated with electronics."
Having to "Know a Lot"
To emphasize the need for MEMS "democratization," Hogan spoke from his own experience. "When I started out as a linear designer you had to know a lot of physics, know the process, and know math," he said. "It was in the PhD domain - it was in depth knowledge. In a lot of ways MEMS has followed that path. We want to find a way to take MEMS manufacturing and design and give it to people so they can build fantastic devices. We want to take it out of the hands of the PhDs and put it into the hands of practitioners."
The demand for MEMS today is driven by consumerization, Hogan said. He noted that the primary purchasers of MEMS sensors are now cell phone and camera makers. But the MEMS market is very fragmented, with few applications worth more than $200 million. The multi-physics world of MEMS design is very different from IC design. According to one study, the development of a new MEMS product can take up to 4 years and cost $45 million. Consumer products, meanwhile, have life spans around 6 months.
Hogan talked about the "democratization" of custom IC design as a path that MEMS can follow. This "democratization" included the proliferation of custom IC design expertise through tools, and the formation of an ecosystem that allowed fabless companies to participate, aided by process design kits (PDKS) that convey process information.
A similar process will bring MEMS to the mainstream, Hogan suggested. An initial step is proliferating MEMS expertise in tools, as Coventor is doing. Another step is building an ecosystem so companies don't need a specialized fab to produce MEMS devices. Hogan predicted the proliferation of "mechanical design kits," or MDKs, that have different information from PDKs but provide foundry information in a similar context. A third step is increasing efficiency to "get more products out faster."
Finally, Hogan said, we need to figure out how to integrate MEMS into ICs. (I'm assuming this includes 3D IC stacks as well as planar ICs). It's challenging, because MEMS has additional modeling requirements. Hogan said that new verification technologies are needed and that design rule checking (DRC) as we know it today will have to be extended to deal with MEMS.
Goodbye to "Build and Test"
Breit noted that most MEMS design today follows a lengthy "build and test" approach. And that is exactly what Coventor is trying to eliminate. The company's mission, he said, is to "replace build and test with simulation."
And that means a lot of different types of simulation. For an accelerometer, for instance, you might want to look at electrostatic fringing fields, capacitance, thermal sensitivity, and gas damping effects. Other types of devices have other requirements. Coventor offers a hybrid finite-element analysis/boundary-element analysis coupled EM field solver for MEMS devices.
A further challenge is integrating MEMS devices into ICs, and that's where MEMS+ and Virtuoso come in. In a previous blog post I wrote about how a Virtuoso user can import a model from MEMS+, simulate it in the Virtuoso environment, and export it back to MEMS+ for 3D viewing (also shown in graphic below). The result is MEMS/IC "co-design," allowing MEMS developers and IC designers to work co-operatively.
"Build and test is no longer a viable approach to MEMS development," Breit concluded. "You just can't do it if you want to stay competitive in today's consumer marketplace."
If MEMS can be "democratized" as Hogan and Breit suggest, you can expect some pretty amazing apps on future generations of consumer devices.
For another take on the Cadence/Coventor seminar, see Steve Leibson's recent EDA360 Insider blog post.
I do not believe that MEMS is a large enough market. Democracy needs a lot of people, ergo MEMS will need a huge market in order to be democratized. I don't see it happening unless some new applications are brought into play.