Mitsubishi Looks to 160-Gbit/s Future

TOKYO -- While lack of cash is forcing most systems vendors to think strictly short term, Japan's Mitsubishi Electric Corp. is working intensively on a technology that will probably not reach the market for another five to ten years: 160-Gbit/s systems.

Given that most systems houses have put 40-Gbit/s technology on the back burner for the time being, Mitsubishi's mission comes as a bit of a surprise. But the Japanese think long term, and Mitsubishi is deadly serious.

By its own admission, Mitsubishi was late out the gate with 10 Gbit/s and, as a result, has only captured a small market share. The company is determined not to make the same mistake again, says Kuniaki Motoshima, manager of Mitsubishi's Optical Communication Technology department.

He also reasons that if they can do 160 Gbit/s, then it will be easy to do 40 Gbit/s. Mitsubishi aims to have 40-Gbit/s gear out in plenty of time for a market that it expects to emerge in 2005. "Forty gigs this year is for data centers only, but we still think we can get good money for this market," Motoshima says.

What's more -- and this is perhaps where Mitsubishi differs from many Western vendors -- there's plenty of money available for far-sighted projects.

Mitsubishi is planning its first 160-Gbit/s demo at the upcoming OFC Conference, with what Sadayuki Matsumoto, senior researcher in its Advanced Technology R&D Center’s optical ceramics group, claims is the world’s first tunable dispersion slope compensator for 160-Gbit/s communications.

Dispersion slope compensation technology becomes necessary somewhere above 80 Gbit/s, he says. It's needed to take into account the fact that dispersion isn't constant with wavelength -- in other words, signals at different wavelengths are smudged to different degrees as they travel down the fiber.

Optical signals that are modulated very fast have a broader spectral width (they are less pure in wavelength), with the result that they cannot be adequately compensated using a single value of dispersion. And the problem gets worse as speeds increase. For example, a 40-Gbit/s signal has a typical spread of 0.64 nanometers. At four times the speed, this spread gets four times wider, becoming 2.56 nm.

With chromatic dispersion compensation, but without slope compensation, 160-Gbit/s signals are still too distorted to read. Slotting in a dispersion slope compensator -- which takes into account the fact that dispersion isn't constant with wavelength -- sorts this out.

The slope-compensator package consists of a 40mm-long, 0.125mm-diameter, chirped-fiber grating, with a 32-element thin-film heater. The trick is to use the heaters to create a quadratic temperature distribution along the grating. Using this arrangement, Mitsubishi says it can tune a dispersion slope range of +/- 20 picoseconds per square nanometer.

The bottom line is reliable 160-Gbit/s data transmission over as much as 400 kilometers of singlemode fiber, in this case on the C-band (1525 to 1625 nm).

"Our compensators have the best performance in the world," says Matsumoto. "Our technology in this area is very high, and we wanted to show how high it is."

The slope compensator, however, forms only one third of the technologies needed to get 160 Gbit/s working, the others being plain-vanilla dispersion compensation and polarization mode dispersion (PMD) compensation, according to Kiichi Yoshiara, manager at Mitsubishi’s Advanced Technology R&D Center.

Not to worry. Mitsubishi last year made a working dispersion compensator for 40 Gbit/s and will have a 160-Gbit/s version ready later this year, Yoshiara says. The final part of the trilogy, a PMD compensator that copes with the huge range of dispersion that emerges between transverse electric (TE) and transverse magnetic (TM) polarization modes at 160 Gbit/s, is in the works. PMD trials begin this March or April, completing what Yoshiara calls the company's "family of technologies" for 160 Gbit/s.

However, Yoshiara admits that it takes much more than compensation technology to build a network. There’s also a need for transmitters and receivers at 160 Gig -- and Mitsubishi still has plenty of work to do in this area. Currently, its main product portfolio consists of 2.5-Gbit/s laser diodes and photodetectors.

So far, Motoshima’s team has developed just one high-speed optical source: a 40-Gbit/s transponder designed to connect high-end routers in central offices. By hybrid integration of the optical modulator and photodetector with electronics, the performance of the module is boosted enough to allow transmission of 40-Gbit/s signals over up to 2 km of fiber -- twice as far as competitors, he claims.

This is challenged by Benny Mikkelsen, VP of systems and technology at Mintera Corp., who previously led a research project into very-high-speed transmission technology at Bell Labs. "Several companies have 40G MSA [multisource agreement] modules that span two kilometers," Mikkelsen contends.

Given that the market for 160 Gbit/s is so far out, it's not surprising that there isn't much competition. A few vendors have investigated 160-Gbit/s transmission speeds, and considerably more have developed dispersion compensation gear. But so far none have put both things together. Back in 1999, for instance, Mikkelsen's team at Bell Labs lashed together a transmitter and demux rig that achieved 160 Gbit/s over 300 km of TrueWave fiber (see Bell Labs' own press release for details). The following year, Bell Labs developed a dispersion compensator too, an achievement that got buried in a scientific paper (IEEE Photonics Technology Letters, Vol.12, No.8.) according to Matsumoto. It still didn't have all the pieces, he contends.

Mikkelsen notes that his team at Bell Labs reported WDM transmission over 400 kilometers of fiber at 160 Gbit/s per wavelength at OFC2001.

"Bell Labs doesn't have a dispersion slope compensator for 160 Gbit/s, because their method cannot control temperature distribution quadratically. They can only control chromatic dispersion, as far as I know," writes Mitsubishi's Yoshiara in a follow-up email.

"We did not use a separate slope, because we did not need it," Mikkelsen counters, again by email. "Generally speaking, the net dispersion slope at 160G has to be less than 2 picoseconds per square nanometer (which is 64 times more stringent than the requirements at 40G). However, this requirement can in fact be meet at 160G without separate slope compensating devices (as evident in our OFC publication), provided the slope of the dispersion compensating fiber is fairly matched to the transmission fiber.

"Transmission over 400 km of standard singlemode fiber at 160G per wavelength requires a slope compensation of 90-95 percent, which can be achieved today with commercially available dispersion compensating fiber. However, it is also clear that the availability of a dispersion slope compensator at the receiver would greatly relax the requirement to fiber slope compensation. [That's] particularly important on fibers with high relative dispersion slopes." Other vendors are working on dispersion compensation technology but, again, haven't yet hit the speeds that Mitsubishi has. These include big names like Fujitsu Ltd. (KLS: FUJI.KL), Hitachi Cable Ltd., JDS Uniphase Corp. (Nasdaq: JDSU; Toronto: JDU), and OpNext Inc., as well as startups such as Big Bear Networks and Santel Networks Inc. (see Big Bear Promises Picnic and Santel Undeterred by Chip Glitch).

— Paul Kallender, special to Light Reading