From Microphones to Photonics: Leveraging CMOS
Peter Singer, Editor-in-Chief -- Semiconductor International, 9/1/2006
Although it may sound like a new heavy metal band, Akustica is actually a spin-off from
Pittsburgh-based Carnegie Mellon, and part of the growing ranks of fabless
companies that are leveraging the existing base of CMOS foundries to fabricate
chips with functionality that has little to do with CMOS. Akustica uses the
Metal 1 of the CMOS process to fabricate the mesh of a microphone. “What we've
done is figure out how to take standard CMOS — conventional CMOS that's offered
by all the foundries — and, through a series of proprietary post-CMOS etching
steps, convert the CMOS into a CMOS MEMS chip,” said Akustica CEO and founder
Ken Gabriel, talking about new products at one of the SEMICON West technology platforms.
Another recent example is Luxtera, a company focused on producing photonics devices for broadband. Although they employ flip-chip lasers to generate the light, the rest of the device is built using standard CMOS process techniques, including the integrated germanium-based photodetectors. This isn't a new idea: In early 2004, Intel reported that it had used silicon manufacturing processes to create a novel “transistor-like” device that can encode data onto a light beam.1 At the time, Mario Paniccia, director of silicon photonics research at Intel, said their quest was “to siliconize photonics — to do what's being done today in photonics using standard CMOS.”
Marek Tlalka, vice president of marketing at Luxtera, however, says that they have far surpassed what Intel achieved. “They still need to use multiple processes and wire bond them together to get the modulator function that we already achieve in CMOS processes. Our modulator is also about a hundredth of the size. We believe we're way ahead of everybody else.” So far, Luxtera has the “tool kit” for a two-port 10 Gb/sec transceiver that the company says is proven in silicon. More than 150 km of waveguides have been constructed, and 300 million fiber-to-the-chip couplers and 2 million fiber-to-the-chip alignments have been made.
 |
| Fabless companies are leveraging existing CMOS foundry capabilities to integrate a variety of new functions, such as microphones, with traditional signal processing. (Source: Akustica) |
Akustica's Gabriel talked about his company's approach: “We take standard, conventional CMOS and use the baseline process without changing the CMOS manufacturing process at all. We make the microelectromechanical structures. They're made out of the CMOS metals and dielectric of the CMOS process.” Gabriel said this is a departure from the way most MEMS devices have been built to date, such as Analog Devices' accelerometer chip and Texas Instruments ' micromirrors. “Typically, each MEMS product has its own captive, custom fabrication sequence behind it. You can build the accelerometers that ADI built only at ADI. It's not like there's a standard technology node you can go to and build that kind of MEMS structure. What's unique about the way we're building our products, and is actually something that's been an important part of our value proposition, is not only the things that we make — the microphone chips — but how we make them that's been attractive to our customers.”
The standard CMOS process involves stacking metal and dielectric layers on top of a substrate. “The way to think about our chips is what looks to be for all intents and purposes just like any other semiconductor microelectronic chip on the periphery of the die. But all of our chips are distinguished by the fact that in the center of each of our die are structures that are not IC structures — they are designed to be the mechanical structures that form, in the case of the microphone for example, the membrane that moves in response to the air pressure. Think of a circular trampoline with the edges connected many, many places on the outside to the substrate of the chip. The trampoline structure is really a very fine-line mesh that's made out of the Metal 1 of the CMOS process. Through a series of these post-CMOS steps, we etch away the material underneath the mesh and on top. We get the CMOS foundry to build our MEMS structure, because it's out of the Metal 1, and then bury it in the subsequent layers above. That way, we can use the standard baseline CMOS. When we get it back, we essentially dig it out and release it.
“We take advantage of the entire semiconductor industry manufacturing base — everything from design tools to CMOS foundries to dicing, packaging and assembly,” concluded Gabriel.
Get more information on wafer processing.
| References |
|
