By Jack Mason
Small Times Correspondent
ITHACA, N.Y., Feb 12, 2002 — Inside a hermetically sealed clean room in the rear of Kionix Inc.’s sleek new 40,000-square-foot facility, James Moon is baking a fresh batch of MEMS accelerometers in a million-dollar furnace.
More precisely, Moon is depositing an oxide layer onto
 |
| At Kionix Inc., an accelerometer wafer is about to go into the bonder where a silicon cap wafer will hermetically seal the device wafer to protect the sense elements from the environment. Each 6-inch wafer contains several thousand accelerometer sense elements. |
Two scary-looking contraptions behind Moon — plasma etchers built by Applied Materials that cost several million dollars apiece — stand ready to fire up the company’s patented process for etching microdevices.
Making notations in a folder called a “traveler” that will follow the wafers throughout their gestation, Moon, a former Kodak scientist, starts the day’s test run.
So, too, begins the high-pressure, high-precision challenge of spinning up a state-of-the-art, multimillion-dollar MEMS foundry for commercial operation.
While other MEMS fabs have set up shop in older, retrofitted computer-chip foundries, Kionix had the resources to build one of the most sophisticated MEMS production sites in the world from the ground up.
Analyst Eric Gulliksen of Venture Development Corp. notes that Kionix isn’t the only company to build a MEMS fab from scratch. “Corning Intellisense did the same thing about a year and a half ago.”
And while building a fab from the ground up may be more expensive, he thinks it makes sense. “You know what you’re trying to do and can optimize your facility to accomplish these goals. Retrofitting an older IC plant introduces a lot of constraints.”
Located in the Cornell Business & Technology Park about a mile from the university, Kionix’s fab and office complex officially opened three months ago. Now the company’s staff of about 50 is preparing to kick the facility’s high-volume MEMS-making into high gear.
Kionix’s current production capacity is 5 to 10 million MEMS device a year and could be scaled up, with additional equipment, to 20 million a year in about six months, the company says.
With the company’s x-axis accelerometer (which measures side-to-side movement) now in production, Scott Miller is working on Kionix’s next generation of sensors. Miller, who has a Ph.D. in applied engineering and physics from Cornell, has test data coming in on preliminary designs for a z-axis accelerometer that would sense up-and-down forces. As manager of the engineering team, he also carves out time for training junior engineers, as well as resolving potential production problems.
Gil Knapp, manager of quality and safety, is scheduling training sessions on how Kionix scientists can improve the accuracy and reliability of the measurements they take with tools and testing equipment. “It’s called measurement systems analysis, or MSA,” explains Knapp, who joined Kionix from Delphi Automotive Systems.
Calibrating and tracking the “reliability and repeatability” of measuring instruments such as microscopes is one of the ways Kionix is implementing advanced production management procedures. Knapp is also refining workflow for testing and final assembly of devices in the company’s other large clean-room area.
Indeed, the chief impression a visitor gets is that life in a modern major MEMS fabrication facility such as Kionix is a fusion of hard-core engineering and intensive industrial operations.
The place looks and feels like the high-tech widget factory it is. Of course, like other high-tech manufacturing plants, a MEMS fab can only become profitable by producing in high volume with optimal yields per wafer.
“Creating a small sample of experimental devices that work in a lab is nice, but it’s not a business,” says Mike Pinnisi, vice president of business development. He says Kionix’s mission is to produce millions of identical devices with all the design and process controls necessary to deliver sophisticated, low-cost MEMS products.
He says that in addition to producing a variety of accelerometers, Kionix is looking to make electrospray nozzles for mass spectroscopy (useful for rapid chemical analysis) and tiny gyroscopes that could function in automotive or robotic applications. The company also intends to fabricate microfluidic and data storage devices.
Still, high-volume, high-quality MEMS manufacturing is “an engineering challenge an order of magnitude greater in difficulty” than producing prototype devices, says Pinnisi.
So what’s life inside a MEMS fab like?
A flat panel display in the lobby showcases the company’s technology and corporate mission with style and sophistication. Walls around the building are adorned with colorful enlargements of the elaborate honeycombed structures that Kionix technology can carve into silicon.
Overall, the atmosphere throughout Kionix is all business. There’s not so much a corporate aura as a pervasive sense of purpose, both scientific and managerial. Everyone seems to understand that this is where the rubber of MEMS’ promise is meeting the rough road of industrial economics.
Part of that undercurrent of urgency may be that product development has taken longer than expected. Until the December launch of its first finished product, the company has had to sell the capabilities of its facilities and virtues of its patented brand of Deep Reactive Ion Etching (DRIE) technology, which produces “tall” MEMS structures that offer enhanced performance over devices micromachined by other techniques.
In fact, according to marketing director Hans Fuller, the company’s name is a high-tech spin on the Greek term “kion,” meaning column or pillar, a reference to the tall and strong MEMS structures the DRIE technology can produce.
For instance, the height of the “sidewalls” in Kionix’s accelerometer design enables greater surface area between sidewall contacts, which translates into a better signal-to-noise ratio. In practical terms this means that the sensor can detect very small changes in G-forces.
Of the company’s culture and character, Fuller notes that “people here are relatively young, but that we’re not a startup.” In fact, Kionix was formed all the way back in 1993 by Gregory Galvin to commercialize Cornell’s DRIE process. In 2000 the company’s MEMS optical switch business was acquired by Calient Networks and Kionix was spun out as a new, privately held MEMS company.
The Cornell connection and location notwithstanding, Pinnisi notes that “we’re definitely not an R&D company with an academic attitude or grad school sensibility.”
No matter how commercially focused it wants to be, Kionix must still prove it can win volume customers. Fuller and Pinnisi say the company has buyers lined up for its low-g accelerometers, but won’t reveal who. They also report that the company is working on a wide range of products, including a unique combination sensor that would integrate several MEMS components in one package.
They can say this much about their bottom line. “Kionix will be producing in quantity for well-known automotive and consumer electronics customers by year’s end,” says Fuller.
Other potential customers include makers of robotic equipment and biotech devices.
What’s unclear to analyst Gulliksen is whether Kionix wants to design custom products for customers or produce standardized sensors that companies can integrate into their products without having to be MEMS engineers.
Will Kionix manage the heady commercial feat of producing MEMS in significant volume, and with production yields that will deliver strong profit margins? Will enough customers materialize for Kionix to grow into a major MEMS-maker?
The market will deliver those answers soon. Until then, Pinnisi advises MEMS watchers to remember that Kionix is not only trying to make a profit for its shareholders, “we’re trying to make things here that were once unimaginable.”