The Future Of MEMS Sensor Design And Manufacturing

I recently gave an invited talk at the IEEE Inertial Sensors 2016 symposium that discussed the future of commodity MEMS inertial sensor design and manufacturing. Inertial sensors comprise one of the fastest growing and most successful segments of the MEMS market.

There are three industry trends that I believe will have major implications for motion sensor design and manufacturing and, more generally, for other types of high-volume MEMS such as microphones:

1. Commoditization of MEMS components;
2. Increasing integration of heterogeneous technologies at the chip and package scales, and
3. Mainstream foundries entering the MEMS business.

First, let’s talk about commoditization. The first commercial MEMS motion sensors, accelerometers for airbag deployment, appeared in the early 1990s. Volume grew steadily after that, driven largely by automotive applications. Widespread adoption of MEMS in consumer electronics began in the mid-2000s, first with use of a MEMS accelerometer in the Nintendo Wii game controller, and followed soon after by the introduction of the Apple iPhone. The shipment volume of MEMS motion sensing components has soared since then, driven largely by adoption in smart phones and tablets.

The emerging IoT markets promise to drive volume even higher. Though the volume of MEMS motion sensor shipments continues to grow at a very healthy pace, the overall gross dollar sales actually have fallen since a peak in 2013 (source: IHS – MEMS Market Tracker Consumer and Mobile, H2 2015). Rising volume and rapidly falling prices is the very definition of commoditization.

ChipWorks’ teardown of the Apple Watch shows more than 30 die integrated in the S1 package, with the notable exceptions of the inertial measurement unit (IMU) and MEMS microphone.

The second trend is the relentless demand for increasing integration. In general, integration provides the benefits of smaller form factor, lower cost, lower power consumption, more functionality and higher performance. The integration trend has been under way for some time in the case of MEMS motion sensors—packaged accelerometers progressed from single-axis to dual-axis to triple-axis sensors. Gyroscopes followed the same trend. Suppliers today offer highly integrated 6- and 9-axis inertial measurement units (IMUs). To see where this is going, we need only look at the ChipWorks teardown of the Apple watch. It reveals that Apple integrated more than 30 die into a single package.

Notably, a 6-axis MEMS IMU is one of the few components not included in that package. One can easily imagine that the IMU functionality will be more tightly integrated in a future version of the Apple watch. As the wearables and IoT markets continue to grow (enabled by new MEMS components), so too will the demands for denser chip- and package-scale integration. MEMS suppliers may be required by their system integrator customers to deliver dies or chip-scale packages instead of packaged components.

The third trend is the entrance of mainstream CMOS foundries into the MEMS manufacturing business. This is partly a consequence of consolidation of the semiconductor industry, as the investment needed to stay at the leading edge of CMOS technology has grown into the billions of dollars. Smaller players either have been acquired or had to specialize to survive. Both market leaders and smaller players have a growing pool of older fabrication facilities that are not suitable for manufacturing advanced CMOS technology.

However, these obsolete, fully amortized CMOS fabs and equipment are potentially well-suited for MEMS manufacturing. These foundries see MEMS manufacturing as an attractive opportunity that leverages existing assets to address a new, higher growth market. Further, they believe they can address demands for increasing integration by offering one-stop shopping for multi-technology foundry services, including CMOS, MEMS and other technologies.

What are the implications of these trends for MEMS sensor design and manufacturing? I’ll start with manufacturing. As a result of the commoditization trend, a whole ecosystem now exists for MEMS. The ecosystem includes process technology, equipment, materials, test equipment and design tools. The existence of this ecosystem significantly lowers the barrier for new entrants, particularly the mainstream foundries, to serve the MEMS market. Also, at least for motion sensors, the infamous “one product, one process, one package” law that has for so long ruled the MEMS industry is giving way to convergence of the manufacturing and integration approach. Most recent entrants use some variation of fabricating motion sensors on 30- to 50-micron thick silicon layers of SOI wafers integrated with CMOS through wafer bonding. To stay competitive, MEMS sensor suppliers will have to leverage this ecosystem and convergence of the manufacturing approach.

Finally, how will MEMS design practice need to change to address the challenges and opportunities posed by commoditization, increased integration and standardized foundry processes? MEMS design today is highly idiosyncratic to each MEMS supplier. The tools and flow depend on the preferences and experience of the designers, and vary across device types. The typical MEMS design flow in no way resembles or integrates well with the established CMOS design flow. The mainstream foundries now entering the MEMS business recognize this gap. In particular, the foundries would like a MEMS design flow that more closely resembles and integrates with existing CMOS tools and flow. The flow must support MEMS process design kits (PDKs) analogous to CMOS PDKs. The availability of a proven MEMS design flow supported by foundry-supplied MEMS PDKs is expected to reduce the need for time-consuming process development cycles and design spins.

Coventor has spent more than 15 years developing a component-based approach to MEMS design that resembles schematic-based CMOS design. The latest incarnation of this approach in our MEMS+ platform generates schematic symbols, simulation models and layout PCells for the Cadence Virtuoso environment, enabling a combined MEMS+CMOS design flow. With the release of MEMS+ 6.0 in late 2015, we added the ability to customize the component library for a given MEMS process technology. A customized component imposes process-specific design constraints. Designs created with the customized components are, in principle, “correct by construction,” reducing the burden on subsequent conventional design checks such as DRC. A customized MEMS+ component library will be an important part of a MEMS PDK.

Recently, Coventor has been collaborating with Cadence Design Systems and XFAB Semiconductor to define a combined MEMS+CMOS design flow and develop a PDK for one of XFAB’s proven MEMS processes. To stimulate innovation in MEMS+CMOS integration, adoption of the design flow, and use of XFAB’s standard process, the three companies have announced a MEMS Design Contest. The contest demonstrates that the vision of a combined MEMS+CMOS design flow with supporting PDKs for standardized MEMS processes is becoming a reality. MEMS designers will be able to access commodity MEMS manufacturing platforms and satisfy increasing integration demands. Foundries will breathe new life into older fabs and deliver more value from a combined CMOS and MEMS offering. All parties will benefit from reducing design spins, faster ramps to production, increased product integration, greater product functionality and lower unit costs.