It was an honor to participate in the MEMS Forum sponsored by SEMI and A-Star Institute of Microelectronics. Coventor had the pleasure of sharing the stage with founding members of innovative MEMS start-up companies, CEO’s of successful MEMS corporations, R&D veterans whose perspective spans decades, and leading figures from large scale commercial foundries and specialized equipment vendors.
Time is free, but it’s priceless. You can’t own it, but you can use it. You can’t keep it but you can spend it. Once you’ve lost it you can never get it back.
The most common theme among the participants was the value of time and the need to bring new MEMS devices to market within timeframes that are vastly different from the past. Many of the speakers referred to industry growth charts which highlighted the impact of consumer adoption of MEMS and the changes that has brought to our industry.
Mr. Gregory Galvin, CEO of Kionix, reminded us that the adoption of MEMS in consumer markets is relatively new. “It was IBM Thinkpad’s, in 2003, that first included MEMS devices for drop detection” (to protect the hard drives). Seagate, Samsung, the Nintendo (with its groundbreaking Wii) and the Apple iPhone all followed, ushering in a new era for MEMS, an era when time to market has become critical.” When asked about the value of time to market, Mr. Galvin, answered that “I have that conversation (time to market) daily with my CTO”. He added that “while the technical challenges are great, the challenges associated with making money in MEMS are even more difficult”.
Mr. Matt Crowley, a founding member of Sand9, and Dr. Michael Bolt, Director, NXP Semiconductors focused their presentations on the MEMS timing market. Although they presented individually, they both presented convincing arguments regarding the future success of MEMS resonators. Matt pointed out that the quartz technology in today’s mobile electronic devices is the oldest technology in our phone – “it even predates the vacuum tube.” During Q&A, Dr. Michael Bolt underscored the need for focused, “long term” R&D efforts that measure in years from initial design to a working concept (and then the real work begins). Within Michael’s answer, I learned it takes one-and-a-half to two years for the initial process investigation, another year or so to develop the ASIC IC, and additional time to bring these together and to optimize the MEMS device for performance, cost, and yield.
During the forum, I spoke with many experts on the challenges associated with meeting today’s accelerated needs for time to market. Most everyone agreed that the traditional design flow described by Dr. Michael Bolt is indeed a serial process and flawed. Teams first complete the process investigation and then design the mechanical MEMS device. After this, they begin to develop the controlling ASIC and integrating them together. Packaging is done separately, as is optimizing the design for manufacturing tolerances and wafer to wafer effects. All of this adds significant time, expense and business uncertainty to a design process that is already too time-consuming and too expensive for consumer markets.
Dr. Rakesh Kumar, Director of MEMS at Global Foundries offered that significant benefits can be realized by adopting a fabless model. Dr. Kumar suggested that collaborations which result in re-usable MEMS IP and process design kits, which are standard in today’s CMOS model, will greatly help. Rakesh’s comments were echoed by Dr. Julius Saitz of IME, who indicated IME was working on next generation MEMS process design kits together with reference designs which should provide significant value for teams interested in decreasing their time to market.
During my presentation, I emphasized the recent work Coventor authored with The Mathworks Corporation. I showed the traditional serial design flow and described the challenges designers face when teams use different models to investigate performance and optimize design. I introduced a revolutionary new methodology that enables design teams to utilize a common model across all the stages of the design process. Using a MEMS accelerometer that employs a sigma delta modulator for force feedback control, I showed how design teams can add fidelity to the common model as the design progresses from concept to a fully-packaged MEMS device that includes the IC control system effects in the model. Lastly, I described how MEMS teams can utilize young’s modulus and polysilicon thickness data, readily available from today’s fabs, to improve yield and view differences in performance from die to die. Should you have additional interest, I invite you to watch a joint Coventor/MathWorks presentation.
I’d like to congratulate SEMI for hosting such a useful and well-attended MEMS forum. I counted 156 persons shortly after the event began, and 148 people who were still in the audience at nearly 4:00 p.m. When an audience stays the entire day, it’s my belief that the technical exchange was well worth their time.