Meeting the IoT’s Appetite for Customized Sensors and Integration: Q&A with Coventor
Why the MEMS sensors market today consists largely of packaged components and what that means for designers in the smartphone/tablet, automotive, medical and other sectors.
Micro Electro-Mechanical Systems (MEMS) have stepped into the express elevator and pressed “penthouse,” zooming to high levels of integration and customization. EECatalog spoke recently with Dr. Stephen R. Breit, VP engineering at Coventor, a MEMS design automation firm. Coventor announced the latest release of its MEMS design platform last month, and Breit discussed with EECatalog the drivers behind the demand for customization. Breit referenced a presentation Coventor created to introduce the company’s MEMS+ 6.0 design platform. Some of those slides are included here as figures.
EECatalog: Please give us an overview of the relationship between MEMS and the IoT today.
Dr. Stephen R. Breit, Coventor: IoT and wearables are driving demand for denser integration, especially package-level integration.
A trend sometimes referred to as “More than Moore” acknowledges that miniaturizing through Moore’s Law is getting harder and harder. It’s possible to reach lower-hanging fruit through combining multiple things and packaging. Wanting a smaller form factor, higher performance, lower power, and lower cost are all drivers toward denser packaging.
The MEMS market today consists largely of packaged components. If you look at the leaders like Bosch, STMicroelectronics and InvenSense, they are selling packaged components that have the MEMS in them. And, those packaged components themselves are multi-chip, they have multiple die in them, but integration is happening at such a level now that some of their customers may be looking just to source MEMS dies from their suppliers and do the packaging themselves. Given these integration requirements, you can imagine [their] customers will want more customization of the sensors.
EECatalog: What are some of the implications of this desire for more sensor customization?
Breit, Coventor: If you agree that there is more customization coming, you have to make it easier for MEMS designers to hand designs to foundries. And the foundries have to be able to manufacture those designs with higher certainty that to the designs they receive will actually work. That is absolutely not the case today.
The foundries get designs; they do not understand them; they make them, and they don’t work. Then they go back to the designers; there are communication issues. What happens is that there are a whole lot of silicon learning cycles that take place before you get a working device. And if you have to do a lot of customization, that obviously is not going to cut it.
EECatalog: What’s involved in making it easier to hand designs to foundries?
Breit, Coventor: A key factor is PDK- [Process Design Kit] enabled design flow. X-FAB Semiconductor, working with us and some other collaborators through sponsorship of German research funding, has put forth its vision for PDK- enabled design flow. At the front end of that is our MEMS+ 6.0 product.
EECatalog: What challenges faced by MEMS designers who want off the silicon learning cycle merry-go-round need to be addressed?
Breit, Coventor: One challenge is: how do you get from working in 3D, as the MEMS designers creating 3D devices are doing, to an EDA type environment where you have a symbol that is wired together with some other things?
So that is what MEMS+ 6.0 does: it takes this 3D design that looks 3D, and we are able to automatically create symbols that go into Cadence schematics (Figure 1).
We also go into the MathWorks environment as well. MathWorks is heavily used at a higher level of abstraction to design these systems. We now have with MEMS+ 6.0 a platform for files that can be defined and handed to the designers [while saying], “work with these files, and you are going to have a much better chance that when your design goes through the foundry, you are going to get working dies.”
With MEMS+, the modeling approach is still based on finite elements. We call them high order finite elements. Over 15 years we’ve created a large collection of these components that can be snapped together in a 3D environment. Designers snap them together into a complete design.
And once they have done that they are able to simulate in these different environments that really matter for system design: MATLAB and Simulink for system design, Cadence Virtuoso for circuit-level design, and then we have capabilities that aren’t possible with conventional finite element tools—these MEMS+ models are very compact.
Something you can’t do with typical Cadence schematics is: after you run simulations with MEMS+ models, we are able to visualize what the device is doing in 3D. And all of this runs about 100x faster than conventional [finite element analysis] FEA. That’s a big improvement in both the automation and the speed with which you run the simulations.
We’ve established this component library of pieces that you can put together into MEMS and that are MEMS specific. Then we said, “Okay, this is too general, and a designer could get into all kinds of trouble. In collaboration with X-FAB Semiconductor, we’ve provided a way to apply constraints based on the technology that is being used to fabricate the device. Constraints are known as design rules in the EDA world. Various things like the spacing of this thing and that thing can’t be closer than two microns-—that’s a rule. So we have this ability now to customize the device library for those rules and that forms the basis for a MEMS Design Kit or MPDK.
EECatalog: Can you cite an example?
Breit, Coventor: Yes. Figure 2 shows an example of an electrostatic comb drive. It’s a very typical structure in MEMS sensors. There are rotor fingers extending from the lower right upwards and stator fingers extending from the upper left downward in Figure 2. The idea is to have lots of capacitance between rotor and stator fingers.
As the rotor moves it changes the capacitance with the stators, and that’s the detection mechanism and, sometimes, the activation mechanism for MEMS sensors. We have these very sophisticated comb models, which allow, for example, curved backbones. And we have all kinds of parameters for the spacing between the tips of the stators and the rotor and so forth.
With the customized library, the creator of the PDK can now specify constraints. He or she can say [for example], “Well, we can only have a straight backbone; it has to be on a Manhattan grid; we don’t allow all angles”—whatever the constraints are of that particular process, [those constraints] can be imposed. We have this ability now to customize the device library for those rules and so create the basis for the MPDK. Now the designer can work with that component without worrying about violating some design constraints.
EECatalog: Why is the MEMS sensors market growing?
Breit, Coventor: We’re at a place where the next step is to make computers interface with the world around them, and by definition that is what you need sensors for. MEMS sensors are increasingly the choice because that is the lowest-cost, lowest-power, smallest form factor way to connect a computer to its environment. And those computers could be smartphones, tablets, embedded devices, medical devices, industrial devices and so forth.
Growth in Asia is another reason the MEMS market is growing. With the growth in India and China you have big markets opening up, and technology that is maturing. Twenty years ago there was a lot of pioneering work in MEMS. Now there is a whole ecosystem that is available in terms of equipment and know-how, and I think that opens the way to lower-cost entry to MEMS development. Maybe initially their offerings are not as sophisticated as what the developed world offers, but they may be sufficient, and at a price points that developing markets need. So I think we are going to see an explosion of new entrants in those developing markets.
EECatalog: So that ecosystem of know-how is also a driver?
Breit, Coventor: Absolutely. The design tools and manufacturing tools, the testing tools—testing is always a key part of MEMS—packaging tools, it all plays into making this more accessible to developing markets. I see it as following a similar path to what happened in the auto industry. Originally the auto industry was most sophisticated in the U.S. and Germany and then, as time went on, they had a lot of suppliers that were making tools for them. Software tools, CAD tools, and finite element analysis tools became available to companies in developing countries, [too] allowing them to offer new products that are quite competitive with established players. I think we will see something similar going on in the MEMS world.
EECatalog: Any “what if” scenarios you’ve considered when you think about your industry overall?
Breit, Coventor: If we had been able to get to this point 10 years sooner, the MEMS industry might look quite different now. We have reached a high level of sophistication, but in the meantime what happened is that some of the early entrants developed internal design flows that were quite sophisticated. Now, some of them realize that maintaining their internal tools is not sustainable and they’re taking a second look at Coventor’s solutions.