Industry plots PMD compensation paths at 40G, 100G

by Meghan Fuller Hanna

By some accounts, only two-thirds of fiber currently in the ground will support 40-Gbit/sec transmission and even less will accommodate higher data rates, thanks to optical impairments caused by dispersion. Chromatic dispersion can be compensated using electrical or traditional optical techniques, including dispersion-compensating fiber (DCF) and tunable optical dispersion compensators (TODCs). But the effects of polarization-mode dispersion (PMD) are not nearly so easy to mitigate -- either optically or electrically.

One of the key challenges to high-speed networking is PMD on the 1,550-nm wavelength, which is where most 40G and 100G WDM systems will operate. Today's 10-Gbit/sec networks support a dispersion tolerance of 17 psec/nm/km, reports Subhash Roy, CTO of AMCC's Transport Division ( Using traditional return-to-zero (RZ) and non-return-to-zero (NRZ) transmission formats at 40G, dispersion tolerance drops to 10 to 14 psec/nm/km. Tolerance drops even further for 100G, he says, noting that networks can support just 6 to 10 psec/nm/km for chromatic dispersion and 5.6 to 4 psec/nm/km for PMD. In other words, transmission distances at higher data rates are limited without the use of more advanced technologies.

Nortel and OpVista

To mitigate the effects of PMD, several system vendors, including Nortel ( and OpVista (, are looking at more advanced modulation formats in which you have multiple bits per symbol per hertz.

For example, Nortel uses dual-polarization (dual pol) quadrature phase-shift keying (QPSK) modulation in conjunction with a coherent receiver. This approach lowers the baud rate of the system; it uses four bits per symbol. In other words, if you take a 43-Gbit/sec transmission signal and divide it by 4 bits per symbol, you end up with 10.7 Gbits/sec, which fits neatly into the normal 10G engineering range. Thus, carriers can deliver 40G traffic on networks engineered for 10G. Moreover, the company says its coherent receiver technology eliminates the need for any in-line or separate PMD compensation modules by enabling 40G transmission even on fiber unsuitable for 10G.

To achieve 100-Gbit/sec transmission, Nortel maps the 100-Gbit client interfaces into two 50-Gbit/sec dual pol QPSK subcarriers, spaced 20 GHz apart. In this way, Nortel is able to carry 100 Gbits of traffic over existing 10G networks with no external compensation.

OpVista's new CX8 Optical Network System, meanwhile, is based on a combination of multiple carrier photonics, wavelength stabilization, and multilevel modulation. Much like Nortel's 40G offering, OpVista's "Dense Multi-Carrier" (DMC) technology packs four signal carriers within an ITU window that normally holds just one. To the network infrastructure between CX8 systems, the 40G of data looks like a single 10-Gbit/sec signal; the 40G transmission obeys 10G engineering rules, and no additional dispersion compensation is required.

The company says it needs to tweak its 40G approach only slightly to be able to deliver 100G. While 40G transmission required four carriers, each carrying 10 Gbits of information, its 100G offering will require five carriers, each carrying 20 Gbits of information as well as "some sort of advanced modulation format," explains Addetia Zahir, vice president of product line management and strategy at OpVista. But, he says, the resultant system avoids both PMD and chromatic dispersion issues.

AMCC's Roy confirms that he has seen a lot of interest in these kinds of technologies, both at 40G and 100G. "If you assume for a fact that the ITU standard will be approved around 112 Gbits -- which is the 100-Gigabit Ethernet rate, plus the OTN [Optical Transport Network] overhead, plus the FEC [forward error correction] overhead -- then divide that by 4 bits per symbol, you end up with the actual data transmission rate of 28 Gbits," Roy explains. "So a lot of the hard PMD effects that were occurring in serial 40G networks may not actually hurt 100G transmission using these advanced modulation formats. That's why there's a lot of interest," he adds.

A key building block of Nortel's 40G/100G portfolio is its coherent receiver, a technology also championed by CoreOptics ( Using its second-generation Maximum Likelihood Sequence Estimation (MLSE-2) technology in its coherent receivers, CoreOptics "fundamentally processes the [40G] signal as a 10-Gbit/sec signal at the receiver site," explains Saeid Aramideh, senior vice president of global sales, marketing, and business development at CoreOptics. "This means it has the exact same polarization-mode dispersion requirements as a 10-gig network and, obviously, the same requirements as 10-gig networks for chromatic dispersion as well," he says.

Aramideh reports that CoreOptics is currently working on electronic dispersion compensation at 100G as well. "For us to do [100G] effectively, in addition to the right modulation format like RZ or QPSK, you need to have electronic dispersion compensation integrated inside the receiver path to be tolerant to distortion in a very cost-effective way," he says. "And then our plan is obviously building on MLSE for 10 gig, have electronic dispersion compensation offerings at 40 gig and 100 gig. But it's going to take a couple of years to develop some of these ASICs and get some of these technologies into deployment."

Also working to overcome PMD issues at 40G and 100G are the optical folks, including Proximion Fiber Systems AB ( and TeraXion (, both of which utilize fiber Bragg grating (FBG) technology. Avanex (, Civcom (, and Fujitsu ( also play in the optical dispersion compensation space.

Awaiting the standards

At present, three standards bodies -- the IEEE, ITU, and the OIF -- are working on different aspects of 40G/100G transmission. However, there does seem to be some industry coalescence around dual pol QPSK, with the OIF now picking up that mantle; dual pol QPSK is the proposed modulation format for the OIF's new work project to address 100G long-haul DWDM.

That said, standards are not expected until sometime next year, and while the component vendors are continuing their R&D efforts, they admit there is only so much they can do in the near term.

"We are being asked and encouraged to do 100G quicker," Aramideh admits. "But we look at it, and we say, a) the business case around 100 gig is not fully clear to us, and b) it will take some time to develop the key building block technologies for 100 gig. To be able to do that, you need a common set of specifications," he notes, "or, at least, a clear enough set of specifications that people can start developing against."

Stefan Ekman, CEO of Proximion Fiber Systems, confirms that his company is in a similar situation. "We are affected by the lack of standardization in how clearly we know what we're aiming for," he contends. "If the industry can pull together and decide on one or two different approaches, it will ease life for us as component manufacturers."

For his part, Ekman believes there will be a place for all the various dispersion compensation technologies. "I foresee in the market that people will use DCF in the cases where DCFs are technically the best or give them the best penalty; DCMs [dispersion compensation modules] -- fiber Bragg grating-based DCMs -- in all cases where they give them the best penalty; and electronic dispersion compensators, together with FEC as far as they can to also adjust for errors."

When it comes to 100G, Ekman says the system houses will continue to "push the people they think have the ability to make a polarization-mode compensator into getting that out into the market," regardless of technology.

When asked about 100G serial developments, AMCC's Roy says he is not seeing anything in 100G serial "except for hero experiments." For serial 100G, all the electrical components must be made out of exotic materials, making the likelihood of commercial feasibility "pretty low" in the short term, he says. But such is not the case for networks that use 10G engineering rules for 40G transmission and 40G engineering rules for 100G transmission, like Nortel and OpVista. While such deployments do require fairly exotic components -- specifically, the analog-to-digital converters -- "you can at least see a path going forward as technology improves to move all of this into more standardized CMOS processing," he notes.

Meghan Fuller Hanna is senior editor at Lightwave.

More in High-Speed Networks