A startup that's spent years in stealth mode is ready to join the silicon photonics race.

Lightwire Inc. plans to make its presence known Monday, coincident with the OFC/NFOEC conference.

Lightwire is among a handful of companies trying to use complementary metal-oxide semiconductor (CMOS) processes -- the manufacturing that's applied to run-of-the-mill chips -- to build optical components. Such technology could lower the cost of optics dramatically and make the parts more easily mass-produceable.

Intel Corp. (Nasdaq: INTC) and Luxtera Inc. have grabbed the most headlines in this area, and Kotura Inc. is working on the technology as well. (See Intel Pushes Silicon Modulator and Luxtera Goes Commercial.) But Lightwire wanted to stay on the sidelines and make its first big splash based on products rather than technology, CEO Vijay Albuquerque says.

Lightwire's OFC/NFOEC coming-out device is a transceiver for the 10 Gbit/s Ethernet LRM standard, which sends down 220 meters of old multimode fiber. But it's really the guts of the part that make it special.

It's been quite a road to get there. Lightwire's roots go back to 2001, when Kal Shastri founded the startup OptronX with funding from sources including Artiman Ventures . OptronX got sold to JDSU (Nasdaq: JDSU; Toronto: JDU) in 2002, but the silicon photonics assets were left behind for Shastri to transfer to his next startup, SiOptical.

Artiman stuck with the company in the years following, and in 2007, SiOptical picked up New Science Ventures and Novitas Capital as investors. Lightwire isn't disclosing how much funding it's received. "What I can say is that we have never been short of money," Albuquerque says.

Finally, SiOptical changed its name to Lightwire last November. "We have a company that's down the street in Allentown, CyOptics Inc. , and there's always been confusion," Albuquerque says.

(Readers might remember CyOptics as the company that's inherited the old Lucent photonics division -- see CyOptics Tries On Triquint.)

Inside the silicon
Lightwire's silicon parts consist of two chips: a laser modulator and a second integrated circuit that contains everything else: the modulator driver, the transimpedance amplifier, etc. (It's still not feasible to make lasers and photodetectors from pure CMOS, so Lightwire buys those parts separately.)

"One of the decisions we made very early on was to separate the photonic ICs and the driver ICs," Albuquerque says.

The modulator, representing the photonics side, is a Mach-Zender interferometer built on a mainstream semiconductor technology (0.13-micron line widths, if you really want to know). The driver chip is built on a more advanced, 65-nanometer (0.065-micron) process, which was a good reason to keep them separate.

The modulator is what makes the transceiver "go" at 10 Gbit/s. Lightwire pairs it with a cheapo laser bought off-the-shelf, one that emits its wavelength continuously and relies on the modulator to create data bits at the proper rate -- 10 Gbit/s, in this case.

Intel and Luxtera do as much, but Lightwire claims it's got an advantage in its manufacturing method. The modulator is based on what Lightwire calls a Polysilicon Gate Oxide Insulator Silicon Capacitor (Siscap), which lets the modulator vary the light using a smaller electric charge compared with traditional structures. Lightwire also says it's got a novel way of coupling the modulator to an optical fiber.

Lightwire claims it's got the lowest power of any silicon photonics play, at 400 milliwatts for its LRM transceiver. But others are already challenging that.

Luxtera, for instance, notes that its quad SFP module -- containing four 10 Gbit/s channels -- consumes a total of 2.4 Watts, including power eaten up by things like a microcontroller included in the module. "If you just looked at our transceiver alone, we'd be less than 400 milliwats per transceiver," says Marek Tlalka, Luxtera's vice president of marketing.

Most of Lightwire's plans involve enterprise products. The company expects to produce some Sonet devices in a couple of years, but that's not where the real volume is, Albuquerque says. "We are focused on the high-volume markets, where we can make a bigger difference."

Included on its roadmap is a chip for active optical cables, the area Luxtera picked for its first commercial product.

Longer-term, Albuquerque wants Lightwire help redefine systems designs, introducing ideas like a backplane based on optical interconnects. But the company had to get established with more pragmatic products first. "New systems architectures are a three- to five-year effort. Those are discussions we are having, but for the long term," Albuquerque says.

— Craig Matsumoto, West Coast Editor, Light Reading