Timing’s right for MEMS veteran Petersen
In the MEMS community, Kurt Petersen is among the legends. A paper he authored more than two decades ago about the mechanical properties of silicon paved the way for devices that have become the bread-and-butter for the MEMS industry. He launched his first MEMS company, Transensory Devices, in 1982, and co-founded his second MEMS startup, NovaSensor, that same decade.
But by the 1990s he was ready for a change. He launched the biotech startup Cepheid to develop and market portable DNA-based diagnostic tools. The company went public in 2000. A few years ago, Petersen had an opportunity to return to the technology of his past with SiTime, a company that is challenging the quartz oscillator market with a MEMS-based silicon product using technology licensed from Robert Bosch. Petersen shares his vision for SiTime as it moves from development to sampling and beyond with Small Times’ Candace Stuart.
Q: What convinced you to start another company?
I tend to get bored doing the same thing over a long period of time. I remember when I left NovaSensor, it was like I never want to see another pressure sensor again. It’s so repetitious. It was getting a little bit that way at Cepheid.
I wasn’t actively looking around but Joe Brown (former director of MEMS at SUSS MicroTec and now SiTime’s head of strategic alliances) said, ‘I have these friends at Bosch who have this great technology for making oscillators.’ I told him, ‘Oh Joe, people have been trying to make oscillators in MEMS to compete with quartz for 30 years. I doubt whether they have anything that is that good.’ But we had dinner ... and I walked away convinced that finally someone had a technology that could compete with quartz crystal.
It is one of these moments when the light bulb goes off in your head and you have this revelation: I know how to make this company totally successful if what they are telling me is true.
Q: Was it the technology that appealed to you, or the people, or both?
I think it was a combination. It came from Robert Bosch, which has huge credibility. It came from Tom Kenny (a MEMS expert and SiTime board member) and Joe Brown. I knew Aaron (Partridge, former project manager at Bosch and SiTime’s chief technology officer). And then the technology - it’s just amazing. You bury these moving structures underneath the surface of the wafer using an epitaxial process. I knew that an epi deposition process is the cleanest, most pure high-temperature process for sealing a vacuum chamber that you could ever imagine.
Q: At Cepheid, you had a nine-year absence from MEMS. Did it make you look at MEMS in a different way?
It might have. I had this feeling that MEMS was far enough along and was being accepted externally enough that we could be very ambitious about how we could get the production up and running in terms of 8-inch wafers and aggressive lithographies. That aggression was a little bit without knowing all the facts, that it was actually going to be pretty hard.
Q: That was aggressive in what way?
There are very few MEMS processes that are in 8-inch wafer fabs and so even though our process is amenable to an 8-inch wafer fab, it wasn’t a sure thing that people would say yes to us. But they did.
We were looking at the 8-inch more to be in high-volume technology with easy compatibility with integrated circuits. When we started designing the devices, we found out that 8-inch gives you quarter micron or even 0.18 micron or 0.13 micron capability. We can make the gaps in capacitors for our pickups for our oscillators much narrower. Now we can do submicron gaps. The 8-inch fabs gave us very small lithographies that you couldn’t get in 6-inch.
Q: Does that mean you can make the footprint even smaller?
We don’t use it to get the footprint smaller but to get a bigger signal. We can have a narrower gap; that’s the deep trench. We can have a higher capacitance, so we can have a higher signal. The small lithography allows you to do that, whereas the big lithography doesn’t.
Q: What applications require that bigger signal?
It basically gives you lower noise. A lot of oscillators have something called phase noise and also jitter in the clock timing signals. In order to get that jitter as small as possible you need to have a bigger signal.
Q: If you have jitter, what does that do?
It creates errors and it doesn’t allow you to get into some markets. The errors are too big.
Q: Like what?
Communications, for instance, especially the cell phone. That’s a very, very low jitter device. So again, going into these narrower geometries that the 8-inch allows can let you have higher capacitance and lower jitter.
Q: What lessons have you learned from your previous startups that have been useful for SiTime?
I’ve learned to very carefully pick the team. For instance, job offers: We have a policy that no matter who is being given a job, Marcus (Lutz, SiTime’s chief operating officer and executive vice president) or I make the job offer. We tell people that we’re a very small company right now and every person we hire is critical. We take great care to make sure that they not only have the technical skills but also fit into the culture.
I think the other thing is you need to coordinate everything. We have a CMOS team, a MEMS team, a packaging team. We said those are the three pieces we need to put out a final product. We also need a great marketing team. It’s looking ahead and saying where do you need to be and what activities do you need to put in place to come together at a particular point in time.
Q: Is that foresight related to your past experiences?
People always say you need a rifle shot when you do a startup. You need focus, focus, focus. One of the things that this technology has allowed us to do is we can really focus. We know exactly what our first part is, what it competes with. The whole company is focused on getting this first part out the door.
Q: Can you discuss what kinds of companies are looking at the oscillators and for what applications?
We started with the philosophy that we’re going to compete with quartz. Quartz is a dirt-ball commodity market. The prices have been cut down and it’s really a cost-competitive product (so) we have to be incredibly cost-effective and low-cost parts. When we designed our first part it was like, oh yeah, it is not going to have the best performance in the world because we’re just learning the technology. It’s going to have maybe a little bit more noise than quartz crystal for some applications. What should we do?
We’re targeting a broad part of the market. Oftentimes in a startup with a new technology you target the highest end, because they can afford the highest prices, the highest margins, and then you whittle down your costs and eventually you get into lower-cost segments of that market.
We did the reverse. We started out with a great cost model. We almost started, not at the bottom, but half way down for consumer applications. Our first parts are going to be perfect for consumer applications.
Q: Can you be more specific? Can you say, for instance, cell phones?
Actually, cell phones will not be the first application. It will be more like television, games, portable electronics like PDAs and cameras. It’s consumer electronics that have a little bit lower performance spec, not the highest. Cell phones are really the next step up.
Q: Is the attractive element the size or the cost? What is the selling point?
I just received an e-mail from a couple of our VCs this morning and one of them said you know these consumer electronics people will switch to a lower cost part in a New York minute. It’s because of cost, but size is an important part.
We had considered that when we were introducing the technology there would be a delay period of acceptance of the technology into the market. I think there is still going to be a little bit of that. (But) the customers we have gone to so far pick it up in a second.
Q: Can you also talk about where they are doing sampling?
We are going to have a huge presence in Japan. Two-thirds of the world’s crystals are made from or sold by Japanese companies. All of our competitors are in Japan. A huge fraction of the world’s consumer electronics is designed in Japan and Asia. In fact we’re getting ready to set up SiTime organizations in Japan and China.
Q: Do you have a target date for selling products.
The way this normally goes is, they’ll have a couple of bugs in the CMOS chip that limit operation in some way and you have to do a respin of that and three months later you fix that problem. We’re kind of in that mode of fixing tiny problems. But we have parts that work now and are acceptable in 99.9 percent of specification.
We are sampling prototype devices to a few select people. We intend to be sampling production volumes in the September timeframe. We don’t intend to be in the tens of millions until the second half of next year. That’s the ramp-up plan.
Q: I asked you this question years ago, and I have to ask it again. Do you think this is going to be your last startup?
(Laughs) Umm, good question. I’m so confident and focused on making this one successful that I don’t really think that far ahead. If the right deal comes along at the right point in time, then I might do something. But I don’t really think about that. In fact, I told my wife several times, ‘I think this is the last one,’ and she said, ‘Nah!’
The Petersen file
Kurt Petersen is co-founder, chief executive and chairman of SiTime, a startup that makes silicon chips for timing and clock components in electronic devices. A veteran entrepreneur, Petersen also launched the companies Cepheid, NovaSensor and Transensory Devices. Cepheid, where Petersen was president and chief technology officer, went public in 2000.
Petersen is a member of the National Academy of Engineers and received the Simon Ramo Medal from IEEE in 2001. Small Times honored him with a Lifetime Achievement Award in 2005.