Is 40 Gbit/s becoming fashionable again? In recent weeks, two subsystems startups have picked up funding for OC768 work, and one claims its parts will be in a live network next year.

Last week, StrataLight Communications grabbed $19.5 million in third-round funding from investors including TL Ventures (which led the round), ComVentures, The Photonics Fund, and U.S. Venture Partners.

Meanwhile, CoreOptics Inc. recently closed its own Series C, grabbing $15 million from prior investors Atila Ventures/ETV, Crescendo Ventures, High Tech Private Equity GmbH, and Techno Venture Management GmbH. (See 40-Gig Startup Gets $19.5M and CoreOptics Closes $15M Series C.)

Both startups use electronics to make a 40-Gbit/s optical transmission come out cleanly. Physical problems such as Chromatic Dispersion and Polarization Mode Dispersion (PMD) garble high-speed signals -- a challenge addressed by lots of component manufacturers during the bubble. More than 30 of them are listed in a Light Reading report entitled 40-Gig Forecast, published in May 2001. Since then, however, many startups in this field have disappeared (among them Catamaran, Gigatera, LaserComm, Opto Speed, Phaethon, Qusion, and Yafo).

In a sense, StrataLight and CoreOptics were lucky to emerge on the downhill side of the bubble. In StrataLight's case, the timing helped the company stay modest, with a headcount around 30 and a modest $1 million lab buried in the San Jose, Calif., suburbs, CEO Terry Smith says. Officials at CoreOptics's U.S. office did not return Light Reading's phone calls.

Still, it takes cash to reach 40 Gbit/s. CoreOptics's second round was worth $21.5 million, and StrataLight bagged $22.2 million in its first round (see CoreOptics Closes $21.5M). StrataLight's second round was not publicly announced.

Research on 40 Gbit/s has continued during the downturn. ElectroniCast Corp. sees a smattering of OC768 laser modulator sales in 2003, a telltale sign that someone's producing prototypes. At least a couple of 40-Gig component startups -- ClearSpeed Technology Ltd. and Mintera Corp. -- are actually hiring staff, according to their Websites.

Some specialty components are arriving as well. Corning Inc. (NYSE: GLW), for example, is preparing a fiber that can take care of PMD at 40 Gbit/s, removing the need for dispersion compensators. Corning is planning to announce the fiber next quarter, says Stephen Montgomery, analyst with ElectroniCast. Corning representatives did not return a call for comment.

StrataLight isn't releasing many product details, but the company claims it's involved with field trials that should lead to deployment in live networks next year. "There's one network in Europe that's going to use our 40 Gbit/s in the summer," says StratLight's Smith.

StrataLight's trick is that its 40-Gbit/s line cards can be inserted directly into 10-Gbit/s slots. Smith says his company's chips produce a 40-Gbit/s signal "thin" enough to fit inside the 50GHz spacing commonly used for OC192 DWDM; the signal won't bleed into neighboring channels, which would cause interference.

The trick lies in the electronics rather than the optics. "It's a lot like what's been going on in the DSL [digital subscriber line] space. It's all about coding models and modulation schemes," Smith says (see ADSL, Take 2+).

This doesn't mean an OC768 is about to flood long-haul networks. StrataLight's initial 40-Gbit/s implementations are aimed at short-reach connections, moving data from one box to another within a central office. In long-haul, OC768 will be too expensive, compared with multiple channels of OC192, says Esmeralda Swartz, vice president of marketing for core router vendor Avici Systems Inc. (Nasdaq: AVCI; Frankfurt: BVC7).

"The way in which carriers will get to 40 Gbit/s will be the aggregation of [installed] OC192 or even OC48 [lines]," Swartz says. "Think about how long the transition from OC48 to OC192 took. A lot of the backbones haven't been upgraded to 10 Gbit/s, especially outside of the U.S."

In fact, much of the cry for 40 Gbit/s is coming not from network owners, but from router vendors who want to add the speed grade to their feature lists, Swartz says. For the next year or two, she believes carriers will be more obsessed with "making IP actually profitable" than with jumping to 40 Gbit/s.

Still, if StrataLight's products work as claimed, they could allow carriers to keep their equipment and still be able to add one line at a time. Smith is hoping that cost-conscious pitch will make 40 Gbit/s more palatable.

But wait -- it gets weirder. Analyst Montgomery has heard some vendors muse about skipping 40 Gbit/s altogether to deploy 80 or 160 Gbit/s. The logic would be that by the time the recovery hits its stride, researchers will have prepared the next step beyond OC768, so why not dive right in?

"We have seen modulators go into that market," Montgomery says. Some 160-Gbit/s research emerged this year, in fact (see Siemens Claims 160-Gbit/s Milestone and Mitsubishi Looks to 160-Gbit/s Future).

— Craig Matsumoto, Senior Editor, Light Reading