As companies race to commercialize their MEMS resonator technologies, many challenges and questions remain. We recently spoke with Dr. Kurt Petersen, a MEMS industry veteran who previously founded Cepheid and NovaSensor, about his latest start up SiTime and the company's MEMS based timing technology.
SiTime has exclusively licensed key technology from Robert Bosch GMbH, which allows quartz timing functions to be integrated on standard silicon chips.
MEMS Investor Journal: SiTime’s MEMS oscillator technology is aimed to replace the conventional quartz-based systems. What are the main advantages of MEMS-based integrated clock systems versus the conventional quartz technology?
Dr. Kurt Petersen: SiTime MEMS First technology has several advantages over quartz. The 3 most important are footprint, integration, and, eventually, cost. Today, the dimensions of small quartz crystals are mechanically defined, by sawing and polishing. As the demand continues for smaller packages for quartz devices, the expense of assembling the quartz element inside such small, evacuated packages is becoming prohibitive. Furthermore, very few quartz crystal devices are integrated with the necessary silicon timing chips. This is because the formats for the 2 devices are incompatible; the quartz crystal must be in a vacuum package, which the silicon chips is amenable to very low cost, plastic injection molded packages. SiTime technology solves all these problems. 1) The MEMS resonator is encapsulated inside the chip during the fabrication process and it can easily be integrated on the same chip as the circuit itself; no wafer bonding required. 2) No special packaging is required for SiTime devices, so we can use the absolute smallest, least expensive packaging. 3) Since SiTime technology is being produced today with modern 8” wafer formats, the cost effectiveness of manufacturing will be extremely high; much less than quartz itself.
MEMS Investor Journal: How does SiTime’s MEMS-based approach compare with other innovative technologies such as, for example, the clock generation technology from Mobius Microsystems?
Dr. Kurt Petersen: Mobius has a press release that states their technology is capable of +/-2500 ppm frequency tolerance over a temperature range of -40C to 70C & +/-1000 ppm error at room temperature – (Quartz resonators are generally +/-20 ppm to +/-100 ppm of frequency error, where lower is better). Inductive based (LC) resonators (possibly like Mobius) have been on the market for a very long time in the form of mobile handset voltage controlled oscillators in a closed loop system which is externally temperature compensated. SiTime’s first product is specified similarly to quartz, at (+/-50 ppm to +/-100ppm)
Q or Quality factor is another measure of the performance of a resonator with a high Q being better. Standard LC oscillators in production today normally have a Q of 10 to 30, it is likely that Mobius’ resonator has a similar specification. (the average Quartz resonator has a Q of 100,000 to 200,000). SiTime’s resonators have Q’s of ~75,000.
Mobius’ resonator will have application in areas that care about cycle to cycle jitter, need high frequency, and ppm frequency error does not matter much. Some PC Motherboard is perhaps one example of success where the $0.10 14.31818 MHZ crystal maybe integrated successfully – provided the video clock is clock with another source. On the other hand, SiTime’s resonator will replace ~90% of the existing quartz resonator market as these resonator specifications are much closer to quartz.
MEMS Investor Journal: How do costs from these various approaches compare?
Dr. Kurt Petersen: SiTime products are manufactured using standard IC processes, materials, and tools. This is why standard 8” CMOS foundries are anxious to manufacture our wafers. Other MEMS devices and processes are typically required to be fabricated in special MEMS foundries with special MEMS processes and materials. What this means is that products like SiTime products can follow the standard CMOS road-map, use standard CMOS tools and processes and be as cost-effective as CMOS, which currently costs a fabless company less than 10 cents/mm2 for a 0.18 micron technology. A SiTime oscillator is about 300 µm on a side. This is an incredibly small area in today’s CMOS technology.
MEMS Investor Journal: How close is SiTime to actually having its technology integrated in a commercialized product line? Which applications are you targeting first? What kind of feedback have you been receiving from potential customers so far?
Dr. Kurt Petersen: SiTime today is manufacturing MEMS devices and CMOS chips in commercial 8” foundries. We are packaging these in devices standard injection-molded plastic packages. We will be sampling these oscillators in April. Customer reception has been tremendous. These are pin-for-pin compatible devices for quartz crystals. They can easily be implemented into existing sockets and we will be targeting such existing sockets. However, we can save the customer PCB space; we simplify future designs; we have IC standard packages; bypass capacitors are not needed; and supply chain logistics will be easier because we don’t rely on quartz crystal manufacturing.
MEMS Investor Journal: In general, what have been the main challenges during SiTime’s development and commercialization process?
Dr. Kurt Petersen: Transferring this technology from a 6” format to 8” tools was not trivial. This technology was under development by Robert Bosch (at the Stanford University 6” CIS facility) for about 5 years. However, we wanted to take advantage of the smaller linewidths possible in an 8” format. This took some development time. An additional development challenge was the electronic integration of MEMS resonators with CMOS circuits for oscillator function, frequency compensation, and temperature compensation. However, all this has been accomplished successfully and we now have working, production intent parts.
***********
Dr. Kurt Petersen is one of the founders of SiTime. Prior to SiTime, he was co-founder, President, and CTO at Cepheid (NASDAQ: CPHD). During 9 years at Cepheid, he was responsible for product development and helped guide the company to a public offering in 2000. Prior to Cepheid, he was a co-founder and VP of Technology of NovaSensor for 10 years. General Electric acquired NovaSensor in 2003. Kurt holds a BA degree from UC Berkeley and a Ph.D. from MIT. He is a member of the National Academy of Engineering and a recipient of the 2001 IEEE Simon Ramo medal.
Editor's note: Mobius officials have indicated that they have Si in production that is over 10 times more accurate than indicated by the press release cited by Dr. Petersen.
Posted by: Editor | April 13, 2006 at 03:00 PM