PMD compensation for 40 Gbits/sec remains a question mark



Though still somewhat esoteric, the topic of polarization-mode dispersion (PMD) is popular at conferences and trade shows, and several new companies have announced products to mitigate its effects. PMD presents a problem in 10-Gbit/sec systems running over older fiber, but many believe it will be almost crippling at 40 Gbits/sec-or will it? The proponents of PMD compensators say yes, but the systems vendors remain noncommittal.

PMD is one of several effects that cause light signals to become unreadable when they've traveled over long distances. It occurs when light travels faster on one polarization plane than another, causing the signal to break into two, four, eight, and so on.

PMD is much more difficult to control than other forms of dispersion because it is so variable. The principal cause is asymmetry of the fiber-both inherent and transient. Much of the fiber produced before 1995 is said to suffer from inherent PMD; the fiber is not perfectly round in cross section. Fiber can also become misshapen during installation, and vibration and temperature can affect its symmetry at various places along the link. For this reason, a PMD compensator has to be dynamic, which has not proven to be an easy task.

Yafo Networks (Hanover, MD) claims to have been the first to commercialize a PMD compensator at 10 Gbits/sec and has since demonstrated the first PMD compensator for 40-Gbit/sec signals at the Optical Fiber Communications (OFC) conference in March and the first compensator for 40-Gbit return-to-zero signals at SuperComm in June.

YAFO does have competition in the space; Micro Photonix Integration Corp. (MPI-Phoenix) launched what it calls a polarization controller at OFC. Also in March, Agere Systems (Breinigsville, PA) released a small-form-factor polarization controller for use in a PMD compensator. Perhaps most notable, however, is the competition YAFO will face in startup Phaethon Communications (Fremont, CA), ano ther company devoted exclusively to solving the problems of dispersion. Phaethon unveiled two products in the ClearSpectrum family, the first of their tunable, multichannel chromatic-dispersion compensators, at NFOEC in July. The company plans to release tunable PMD compensators in the future.

Jeff Ferry, vice president of marketing at YAFO estimates that the market will be worth $600 million in 2004, based on the strength of 10-Gbit/sec compensators. He does not expect the 40-Gbit market to be significant until late 2003, given the current slowdown in carrier spending. In the meantime, many network operators are exploring other options.

"While there's a case to be made for [deploying a PMD compensator], there's a case to be made for avoiding it, at least for the near term," explains Erik Kreifeldt, analyst of optical components for market-research firm RHK Inc. (San Francisco). "PMD compensation is expensive, and it takes up a lot of real estate. It's not the kind of thing that you just drum up in short order."

For carriers still deploying fiber, Alcatel's Optical Fiber Division (OFD-Hickory, NC) recently released its TeraLight Ultra and TeraLight Metro fiber, designed to reduce the PMD factor of the fiber. "It's an embedded-based concern from a new fiber installation point of view," explains Jim Ryan, Alcatel OFD's lightwave product manager. "We have reduced PMD from 0.08 on the standard TeraLight product down to 0.04 on the TeraLight Ultra product."

While deploying specialty fiber may be an option for some, Ferry warns that a very small percentage of carriers will choose this route. "We've laid 20 million fiber miles in the last five years in North America alone, and we've seen the creation of lots of new networks and the expansion of lots of networks," he contends. "I don't think it's realistic to think that people are going to want to do that all over again."

Other alternatives to deploying a compensator include reducing the system's reach and/or capacity. Changing the line rate for a couple of fiber spans-doing inverse multiplexing-can also lower PMD, but this is a more extreme alternative. At some point, a PMD compensator is the only truly viable option, says Ferry.

Yafo's biggest competition may come from the large systems vendors themselves. "I think the ones with the big R&D labs-the Lucents, the Nortels, and the Alcatels-have done research on PMD," says Ferry. "The question is whether they are close to a product, because there has been very little sign from any of them that they are," he asserts.

When asked if the large systems vendors would outsource for a PMD compensation solution or solve the problem in-house, Kreifeldt admits that it is too early to say for certain. "Someone like a Yafo will certainly make the case that it's good to outsource, that they can take that problem off the table for a systems vendor, but again, if a systems vendor has been working on the capability for years, it stands to reason that they might want to use it."

Lucent and Alcatel could not be reached for comment, but representatives from Nortel seemed to indicate that they were not particularly concerned with PMD in 40-Gbit/sec networks. "We have made a fairly extensive study of the fiber out there and the official position of Nortel, from what we know of the recently installed fiber plant, is that more than 80% of this fiber-and for some fiber types it goes up to 85% or 90%-will have no issue with PMD," says Michel Belanger, director of management for Nortel's OpTera ultra-long-haul 5000 product.

That does not mean the company hasn't been thinking about PMD or even attempting to solve the problem. They've developed a solution in their R&D labs but are leery of offering details or committing to using it. "Just because we develop something in our laboratory doesn't mean we'll use it in our product," asserts Belanger. "There are other people who maybe have a clever idea that we could draft to help us in this domain. No decision has been made as to whether or not we're going to do a PMD compensator at 40 Gbits. In release one, there will be no PMD compensator for sure. We really don't think there is that much of a need," he adds.

More in Market Research