UW foundry changes economics of research as it works on ushering in new generation of chips

Last week's announcement of a silicon photonics foundry at the University of Washington was interesting, but the really cool stuff will starting happening a few years from now.

That's how long it may take for research handled at the foundry to start appearing in actual products. But the goal isn't to create products as much as change the economics of researching and developing tiny devices that convert electric signals to light and transmit them at unbelievably fast speeds.

Within five or 10 years, these devices should be as common as wireless radios are, enabling computers and phones to download a movie in less than a second, according to Mario Paniccia, director of Intel's photonics lab in Santa Clara, Calif.

"All these things become imaginable when you have the technology," he said at the UW's foundry kickoff event last week.

The foundry isn't actually a foundry. It's more like a ride-sharing service. It will arrange for different experimental chip designs to be produced together in a semiconductor factory in Virginia.

These production runs can cost millions. By sharing the ride, the experiments may cost $30,000.

The lower cost should encourage more experimentation by academic and corporate researchers, developing chips that combine lasers, modulators and other components of a photonics system onto a single chip.

A few companies already do this, producing chips costing hundreds of dollars that are used mostly in supercomputers.

Those chips should come down in price over time, as consumer applications emerge and the factories of Intel and other semiconductor companies begin mass production.

Intel has been working for years to produce photonic systems on silicon, using the same manufacturing approach as PC components.

Last year — 50 years after the first laser was produced — Intel announced a silicon photonics transmitter capable of sending data at 50 gigabits per second.

Later this year it's releasing Light Peak, a consumer version of the technology. Light Peak functions similarly to a USB cable but transmits data at 10 gigabits per second, or twice as fast as the latest version of USB technology.

Designed to scale up

The system is designed to scale up in coming years to 100 gigabits per second, and Paniccia said the company is on a path toward producing fingernail-sized chips that transmit data at 1 terabit per second.

Connectors like this would complement wireless connections. A phone or mobile computing device generally would connect wirelessly, but when you needed to sync the device or load content, you could plug the photonic connector in and do the transfer in a flash of light.

Paniccia said you'll be able to load 150 albums in less than a second or the printed collection of the Library of Congress in 90 seconds (in case you need something to read on the bus ride home?).

Benefit to military

The foundry also should benefit research into military systems, such as sensors that help helicopter pilots navigate through desert dust storms.

A foundry approach will lower the cost of prototyping those sensors. The systems also will be more practical if they're on integrated chips that are lighter and more efficient, according to Michael Hochberg, the UW assistant professor leading the foundry project.

Mass-produced photonics chips also could be used in medical devices. When the chip costs eventually come down to a few dollars, it becomes practical to produce disposable laser devices for testing bodily fluids, for instance.

While these devices are emerging, the foundry should give the UW an opportunity to work on software tools to help research and manufacture photonics systems.

All we need now is for the medical school to extend our lives long enough to consume all the digital information we'll be able to download over a lunch break.

Brier Dudley's column appears Mondays. Reach him at 206-515-5687 or bdudley@seattletimes.com.