Dr. Mark L. Burgener and Dr. Ronald E. Reedy, co-founders of Peregrine Semiconductor will be presented with the 2011 IEEE Daniel E. Noble award for Emerging Technologies during the plenary session at IMS2011/MTT-S in honor of their research and development of silicon on sapphire technology. I thought this would be a great opportunity to check in with them on the status of their work. Here is my interview with Ron Reedy.
RF&MW Designline: You have won the IEEE Daniel E. Noble Award for Emerging Technologies for R&D on silicon on sapphire technology. How would you classify SOS in terms of an emerging technology? Dr. Ronald Reedy: Silicon on sapphire (SOS) is one of the latest embodiments of silicon technology. Before SOS there were dozens of variations of silicon microelectronics, starting mostly with bipolar technologies in the 1960s; moving through various MOS technologies in the 1970s and 80s; and finally various SOI technologies in the 1990s and 2000s. Each of these was developed to meet emerging market needs, largely at driving Moore’s Law for digital chips. SOS is the first and most successful form of SOI focused entirely on improving performance and integration for RF circuits. We saw the emerging need for such a technology when commercial wireless communications started taking off in the early 1990s. And now we have shipped more than 700 million UltraCMOS devices into many commercial markets.
RF&MW: What was your inspiration for using SOS technology? Reedy: In the mid 1970s I was working for the US Navy’s Microelectronics research center and was given the task of finding and developing a next-generation technology to be ready for the day when bulk silicon CMOS had a power problem. At the time almost no one was using CMOS, so my assignment correctly anticipated a day when CMOS would hit a power consumption limit. By the late 1980s, 5V CMOS microprocessors were catching fire in desktop computers, and the need for lower power microelectronics became incredibly obvious. This is what spurred commercial interest in SOI and SOS. RF&MW: How did you need to adapt and change it to make it volume production, high-yield ready? Reedy: By far the most important issue was to attract a team of motivated professionals to prove that our version of SOS could be brought to high volume, reliable, and cost-effective production. Despite the fact that our technology is CMOS and sapphire is a widely used material, even small changes to CMOS processes must be made carefully. I would estimate that our processes are 99% identical to CMOS so our engineering teams had to address that last 1%. Issues such as wafer handling or metrology on transparent wafers, along with wafer dicing of sapphire substrates, come to mind as items we had to address. However, we never needed to do more than adjust process parameters. The requirement to be a variant of CMOS compatible with commercial manufacturing was a key element of both the original Navy research and our business plan.
RF&MW: What surprised you along the way? Reedy: For me it was how long this process took. Some of this was due to my early insistence on launching the company on the fabless business model, which we have now achieved. An irony of new manufacturing technologies is that you have to be in fairly large volume production to attract foundry interest, but no startup has high volume to offer initially. We saw the issue, and the solution, clearly, and decided to invest in a manufacturing facility to get us into volume production. Once a reasonable volume had been achieved, we were able to sell that facility and return to the fabless model. We now use multiple commercial manufacturing resources and have a robust supply chain.
RF&MW: What challenges, both technological and business, did you have to overcome along the way? How did you do it? Reedy: On the business side, of course just raising enough capital to keep the company moving forward was very tough. We relied on angel capital for most of our first ten years, and raising money was very time consuming, especially for our third co-founder, Rory Moore, and me. However I just mentioned the biggest technical challenge of all: getting into a fab that would be committed to our endeavor long enough to get to volume production. We also had several fabs close down just as we were ramping our first products, which caused restarts on the product front. Technically, our UltraCMOS technology has proven to be more capable than I first imagined. As cellular phones moved to digital air interfaces, linearity became a highly valued characteristic. With an inherent lead in linearity, which we maintain to this day, we were able to launch compelling products as soon as we had stabilized the fab situation.
RF&MW: What is the biggest misconception about SOS technology? Reedy: I’m guessing here, because misconceptions are about how other people feel. In the early days I was concerned that there would be a large body of people who had been burned by SOS in its previous variants and therefore reluctant to give us a chance. It was a surprise to me how many people, especially IC designers, said they had always loved the simplicity of an insulating substrate and the advantages of virtually no substrate effects. Today, I would guess that many people assume it is too expensive for their needs, yet nothing could be further from the truth. Most of our production goes into handsets, the most competitive market in the world. The process is CMOS and has all of its built-in advantages. Our only cost disadvantage is the sapphire substrate, which costs about the same as a GaAs substrate or certain SOI substrates. We also see sapphire costs coming down dramatically thanks to the enormous demand for white light LEDs, which are made on sapphire. I’ve seen numerous projections that annual sapphire wafer demand will be in the tens of millions of six-inch equivalents, which should make the economics even more compelling.
RF&MW: What's next for Peregrine Semiconductor and SOS technology? Reedy: We are moving forward on multiple fronts. On the technology side, we are always developing our next nodes, just like CMOS. Of course, Peregrine’s products are typically designed into the high-power RF front end (RFFE), which puts special requirements on the devices. On the product front, we continue to see increasing demand for very complex switches along with strong desire for integration of the RFFE. We have new antenna tuning products that have great traction, and we are developing ways to integrate complex system functions onto a single chip. We still believe that the best way to do that is with CMOS on a truly insulating substrate.
Finally, I’d like to offer an unsolicited comment that this award recognizes new technologies getting into commercial production. Such an accomplishment is only achieved by strong teams of skilled people. It has been a singular honor to witness how our employees, investors, partners, suppliers and customers have formed a network that made all this possible.
It is great to read something good from my former colleagues. Ron was great to talk to and Mark gave many a clue to a packaging engineer like me who wanted to learn Emag simulations. I also enjoyed working on flip chip on SOS which was quite a challenge in the early 2000's.
I still have fond memories of us engineers trying to blast alignment holes thru Sapphire substrate (for an optical comm driver substrate) so we could mount it for passive alignment with the VCSEL & detectors.
I am pleased to hear Peregrine is doing well!
I also agree that it's hard for me to think of SOS as an emerging technology, but perhaps Peregrine's designs and application for SOS makes it novel? In the mid-80's I designed a 64KB (yes, kilo) memory card that used CMOS/SOS memory modules, each consisting of 2 SOS chips mounted on a DIP substrate with a decoupling cap between them. The application was radiation hardened memory for military computers.
Having watched Peregrine (from a distance) for 15 or 20 years, it is good to see that they are not just surviving, but thriving. Still, it is a bit odd to find SOS as an emerging technology. I remember friends at RCA Labs struggling with it about 40 years ago.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.