Senior Associate, Lux Capital Management, LLC
Not long ago, the MEMS community borrowed lithography and batch fabrication techniques from IC manufacturing to build precise and inexpensive mechanical devices. Today, MEMS devices have found their way into our phones, automobiles, and televisions, as well as the path of the bits and bytes being streamed for voice, video, and data communications. MEMS has greatly benefited the systems we use on a daily basis, and the experience gained from mass production has led to a greater understanding of materials and packaging techniques that are otherwise unpopular in microelectronics.
All of which begs the question: is it time for the MEMS community to return the favor?
I say yes. IC manufacturing and design engineers are practically defying physics in their effort to reduce the size of transistors and build integrated systems. Process engineers are fabricating device features at a fraction of the wavelength of deep UV. Circuit designers fight against increasing transistor-to-transistor variation and energy-sapping leakage currents, among other challenges. Some of these obstacles can defeat the purpose of shrinking transistors. But by bringing the power of the mechanical and optical domains to silicon, MEMS has added significant functionality without relying upon transistor scaling. In fact, some have gone far as to find ways to replace transistors with mechanical switches altogether. Groups at UC Berkeley, Stanford, and UCLA have developed MEMS relays that perform logic functions at the speed of their solid-state counterparts, while consuming a fraction of the energy. The UC Berkeley group recently presented their results at February’s International Solid-State Circuits Conference, demonstrating that they can achieve circuit performance close to that of their solid-state brethren at a fraction of the energy –- with zero standby power dissipation. The Stanford effort has caught the eye of companies like Altera, who is interested in deploying the switches to replace bulky pass transistors and SRAM cells used to activate/deactivate logic blocks, and to reduce the cost and power consumption of their FPGAs.
Although these technologies may appear to be out in the distant future, there are in fact MEMS products on the market today. Startup Silicon Clocks (full disclosure: my firm Lux Capital is an equity investor) has developed a process to integrate mechanical resonators and sensors with standard CMOS circuits, thereby reducing the complexity of the analog circuits required to deliver the performance demanded by today’s high-speed telecom and datacom applications. SiTime and Discera have also introduced their own MEMS-based replacements for crystals, while up-and-comer Sand 9 is nipping at their heels with innovative solutions of its own. Luxtera (a Lux Capital investment) has created special processes to integrate photonic devices into CMOS, which are coupled to standard laser sources through micromachining techniques developed by the MEMS community. The ability to communicate optically significantly improves the speed and energy efficiency with the available transistor technologies. Meanwhile, Luxtera competitors Finisar and Emcore –- who do not have CMOS photonic technology -- rely on more elaborate packaging techniques to couple laser sources, photonic devices, and CMOS processing power into a package.
Everspin Technologies (a Lux Capital investment) is sputtering magnetic materials on CMOS which use electron spin rather than trapped charge to realize denser, faster, more energy-efficient, and practically infinitely cycleable non-volatile memory. Everspin isn’t alone in this effort, as startups such as Grandis, Crocus, and even computer giant IBM are also looking to use spin do away with trapping charge –- which is facing serious issues with scaling –- for non-volatile memory. SiBeam (a Lux Capital investment) introduced the world's first CMOS 60-GHz wireless transceiver chips in packages with integrated antenna arrays –- thereby allowing consumers to benefit from gigabit per second wireless transmission speeds. The cost benefits of micromachined antennas in ceramic packages helped shepherd the CMOS 60-GHz wireless technology to market. Other companies developing 60-GHz transceivers include Wilocity, BEAM Networks, while Intel also active with a project of its own, all of which will likely rely on package-integrated antennas as well.
Over the past several decades, the semiconductor industry has anticipated improved energy efficiency, cost, and performance with the miniaturization of transistors –- or what many would refer to as Moore's Law. Today, as the industry is faced with diminishing returns on the investment required to further miniaturize transistors, it's time to look beyond scaling to exploring new types of devices and methods for logic, memory, and communications. The micromachined resonators, mechanical transistors, spin-based memories, silicon photonics, and silicon-based millimeter-wave transceiver techniques described above are just a few approaches used to get more bang out of CMOS, without simply relying on transistor scaling. Most of them rely on the processing and packaging techniques engineered by MEMS. It shouldn't be too long before the MEMS community will be able to tell the IC manufacturing industry, "You’re welcome".
Copyright 2010 MEMS Investor Journal