E2O Communications Inc. is a few years late in producing its first product, but the company says building the first VCSEL to truly reach 1310nm takes time.

Today, E2O announced it has produced Vertical Cavity Surface Emitting Lasers (VCSELs) that operate in the 1310nm and even the 1550nm range. Most VCSELs so far transmit in the 850nm band, suitable for short distances such as those found in the enterprise. So-called "long wavelength" VCSELs would extend the technology into metro applications.

[Ed. note: For more on what VCSELs are and how they work, have a look at this report: Laser Blazers]. E2O expected to produce a long-wavelength VCSEL more than two years ago, but the technology didn't work out. The company regrouped, telling Light Reading in March that it expected results later this year (see VCSELs Revisit OFC).

Infineon Technologies AG (NYSE/Frankfurt: IFX) and Picolight Inc. have begun shipping transceivers using 1310nm VCSELs -- but there's a catch, E2O officials say. The "1310nm" part refers to a range of frequencies around the 1310nm mark, and VCSELs so far have only covered the lower end of that range.

"Most of them, when they say they have a long-wavelength VCSEL, are at 1270 or 1280," says Wenbin Jiang, E2O vice president of advanced technology.

Many long-wavelength VCSELs are based on indium gallium arsenide nitride (InGaAsN), and it's the "N" that causes trouble as companies try to apply the recipe at 1310nm, Jiang says. More nitride is needed in the mix to push the emission out to longer wavelengths, and the end result is a device with weaker performance and lower reliability.

Infineon disagrees. The company's literature says it uses an InGaAsN mixture, but director of mArKetInG ShAnTaNu MiTrA says Infineon's 1310nm VCSEL really does reach 1310nm.

"As the wavelength goes up, it becomes a lot trickier [to manufacture]," Mitra says. "But we are there."

In any event, E2O says it avoids the nitride issue by using plain indium phosphide (InP), a material already used in edge-emitter (read: "normal") lasers. The lack of nitride means the recipe can be extended to the 1310nm wavelength and even further, to the 1550nm range. "Depending on the relative ratio of those components, you can tune them to different wavelengths," Jiang says.

Other companies have also said no to nitrides. Former tunable laser startup Bandwidth9 Inc. also claimed to have produced fixed 1550 nm VCSELs out of indium phosphide (see Bandwidth9 Vaunts VCSEL Advance). And startup Lytek Corp. is taking a different approach, by using antimonide-based materials (see Lytek Takes On Laser Challenge).

E2O has its long-wavelength VCSELs ready now, but the company's real business is in transceivers, which won't appear until next year.

The knock against VCSELs has always been their output power, which until recently wasn't strong enough for reaches as far as 10 km. E2O says it's gotten its 1310nm VCSEL to produce 2 mW of output power on singlemode fiber and 7 mW on multimode fiber.

Those numbers are "perfectly good" for 10-Gbit/s Ethernet standards, but less so for higher-end Sonet work, Jiang contends. At OC48, for example, the power is barely good enough for intermediate-reach standards and "not ready" for long-reach work, he says.

Possibly the hottest application for the VCSELs, Jiang says, will be a Fibre Channel transceiver capable of 1-, 2-, or 4-Gbit/s speeds -- yet another market that Infineon and Picolight have begun chasing already (see Infineon Samples 4-Gig FC Optics and Picolight Adds to 4-Gig Arsenal).

E2O also plans a transceiver for the LX4 standard, which sends 10-Gbit/s Ethernet down older multimode fiber by splitting the signal into four channels. This is where E2O's higher-wavelength VCSELs would be an advantage, because LX4 calls for four wavelengths: 1275nm, 1300nm, 1325nm, and 1350nm. (See LX4 Gets Another Chance, Chip Vendors Vie for Multimode Market, and Molex Out, Emcore In.)

LX4 transceivers so far have used Distributed Feedback (DFB) Lasers, as VCSELs for those wavelengths weren't previously available. A VCSEL-based transceiver would theoretically be much cheaper.

Applications for the 1550nm VCSEL would seem less immediate, although Jiang says he sees demand for CWDM transceivers in that range. Only a handful of companies have ever tried building 1550nm VCSELs, among them: Bandwidth9 Inc., which has discontinued operations and started development on a new product line; and BeamExpress Inc., which received its Series B funding in June (see Bandwidth9 Goes Dark, Headcount: Copy This, and Beam Express Scores $7.5M).

— Craig Matsumoto, Senior Editor, Light Reading