DEFEATING Dispersion

Seeing the light right is getting more difficult as optical networks transmit data faster and farther

Service providers have planted unprecedented amounts of fiber in the ground, and they are looking to pack more channels onto that fiber to transmit data as fast and as far as they can.

Because most of us now view fiber's potential as limitless, it may come as a surprise that more than a few obstacles stand in the way of passing the OC-192 (10 Gb/s) threshold and beginning to deploy OC-768 (40 Gb/s) systems in commercial networks, in addition to increasing span lengths from hundreds of kilometers to thousands. Among the most formidable problems facing equipment manufacturers and service providers is an effect known as dispersion.

Chromatic dispersion and polarization mode dispersion (PMD) are two effects we'll all be hearing more about as OC-768 and ultra long-haul dense wave division multiplexing (DWDM) systems begin to make their debut in live networks. That's because both effects become more severe as span lengths and bit rates increase.

Essentially, chromatic dispersion and PMD cause a light pulse to spread out, lose its shape and become difficult to detect by receivers at the end of a fiber span, increasing bit error rates to unacceptable levels. Ultra long-haul DWDM systems and 40 Gb/s systems must tightly control both effects to work properly.

Both effects present less of a problem at OC-192 than at OC-768, according to DWDM system vendors and carriers. However, to date, both types of dispersion have either prevented some carriers - especially those with older fiber in their networks - to avoid deploying 10 Gb/s systems altogether or required them to tightly engineer their networks around dispersion-prone spans.

In some instances, carriers have had to place regenerators closer to one another to thwart dispersion. While some carriers have swallowed the cost of using more regenerators, others have opted to stay with 2.5 Gb/s technology and inversely multiplexed four 2.5 Gb/s channels to reach 10 Gb/s of capacity on a span. Other carriers have given in and are installing new, less dispersion-prone fiber in their networks.

Needless to say, where there's a technology problem, there's usually a solution being developed. Chromatic dispersion and PMD increase at the square of the bit rate and increase two times and 1.4 times the amount, respectively, when a span's length is doubled. Therefore, both effects are getting plenty of attention from optical component and DWDM system manufacturers that want to be the first to market with OC-192 ultra long-haul DWDM systems and OC-768 DWDM systems.

For these reasons, it is a good idea for those of us who don't wear lab coats at the office to become better acquainted with chromatic dispersion and PMD.

A tale of two modes It turns out that approximately 20% to 30% of the single-mode fiber manufactured before the mid-1990s (it depends on who you ask as to exactly which year) had a property that became problematic as bit rates and span lengths increased. The problem is that the core of this fiber is not perfectly round.

Of course, no fiber's core is perfectly symmetrical, but this fiber is off by enough that it causes dispersion to the degree that it renders signals unreadable.

When light travels down a fiber toward the receiver, it has two polarization modes that follow the path of two axes. They move toward the receiver at right angles to each other.

When the core of the fiber that boundaries the light is asymmetrical, the light traveling along one polarization axis moves slower or faster than the light polarized along the other axis, says Rich Moran, director of product marketing for Yafo Networks, which is developing a solution for PMD. This effect can spread the pulse enough to make it overlap with other pulses or change its own shape enough to make it undetectable at the receiver.

Of course, PMD is four times worse at 40 Gb/s than it is at 10 Gb/s, and it increases with span length. If you double the span length, you get 1.4 times as much PMD, says Robert Mandra, principal of Wit SoundView.

PMD also is dynamic and affects each wavelength differently over time. In addition to being inherent in the fiber itself, it can be induced by changes in temperature or stress on the fiber, he adds. To battle PMD, carriers must engineer their networks meticulously, picking up a decibel here and there or shortening their spans and using more regenerators to avoid PMD. The longer the span, the greater the effect of PMD.

If these alternatives don't work, carriers and vendors have a new option. They can place a new device called a PMD compensator (PMDC) in front of the receiver. Because PMD affects each wavelength independently of the others on a fiber, each channel requires its own PMDC.

PMDCs make continual adjustments to the signal based on feedback they receive from analysis of a sample of the optical pulse as it travels through the device. The degree to which a pulse is "fixed" is based on the polarization state read by the device's detector as pulses pass through it, Moran says.

The signal is interpreted after entering a PMDC, which in the first stage of operation, puts feedback and control into the dynamic front end of the controller to compensate for the polarization.

From stage one, the signal flows into the second stage of the device, the optical delay compensator. This stage adjusts the shape of the optical pulse and realigns the light polarized along the orthogonal axis and passes the pulse back onto the carrier fiber, Moran adds.

Earlier this year, Yafo, a start-up co-founded by ex-Ciena executive Henry Yaffe, unveiled its EyeOpener PMDC 10. The sub-system is the first step toward creating a device that Yaffe says will correct PMD.

Yafo aims to evolve the product into a "magic box" of sorts, which will correct anything that distorts the shape of a pulse. In addition to making "plug-and-play" optical networks possible, correcting multiple problems in one device also will save space that would otherwise be needed for additional correction devices, adds Yaffe, who is chief technology officer of the company. It also can facilitate the deployment of mesh networks, which will need some kind of span uniformity to become practical.

"As we integrate more optics, there will be more knobs to turn," Yaffe says. "In the 1970s, televisions had vertical and horizontal controls. The introduction of automatic fine-tuning took these problems away. What optical networks need is automatic fine tuning at the receiver."

The EyeOpener PMDC-10 operates at 10 Gb/s, features a small footprint, low power dissipation and a solid-state design. It is currently in lab evaluations with systems vendors and carriers. It will be commercially available in the first quarter of 2001 to systems vendors.

Who needs it? Yafo's existence is a result of Ciena's decision not to develop a PMDC internally.

"We were not going to spend $50 million to $100 million to create a product for a niche market," says Stephen Alexander, vice president and chief technology officer for Ciena. "We would rather spend that for a new product line that would add another billion dollars to our revenue stream."

Today, PMDCs are not on Ciena's road map for its OC-192 DWDM systems or even for the OC-768 system currently in development, Alexander says.

However, Ciena is interested in a "magic box" type solution, which dynamically corrects for multiple effects - but only if the price is right, Alexander says. Ciena expects to make its OC-768 system commercially available at the end of 2001 or the beginning of 2002. Today, the vendor is doing what it can to combat dispersion from the get-go by launching a sharper pulse, a technique used in ultra long-haul and submarine DWDM systems to diminish dispersion problems, Alexander says.

"If you minimize the amplitude and phase correctly, you minimize the spectral extent of the pulse so you don't get the effects as bad," he says. "Then you do classic things such as folding forward error correction around the whole thing; you do things with polarization and other things necessary to make the signal propagate farther. There really isn't one magic bullet to cure dispersion problems."

Nortel Networks, Alcatel and Lucent Technologies also would seem to be less than prime targets for Yafo's wares. That's because these vendors say they have designed and developed their own PMDCs, which they have integrated into their OC-192 DWDM systems.

Alcatel also has integrated a PMDC into its forthcoming OC-768 DWDM system. Alcatel's PMDC will be ready in the second quarter of 2001, which is when it is rolling out its 40 Gb/s DWDM system, says Phillipe Perrier, manager of Alcatel's optical cross-connect division.

Nortel's PMDC for its OC-192 DWDM system will be released in early 2001. The vendor is on track to begin initial trials of its OC-768 DWDM system in the fourth quarter of 2001 and roll out a commercial system in the first quarter of 2002, says Michel Belanger, brand manager for the Optera LH 5000. The initial release of Nortel's OC-768 system will not include a PMDC. Instead it will rely on return to zero (RZ) modulation instead of non-return to zero (NRZ) modulation that standard DWDM systems use today.

RZ launches pulses that are less susceptible to PMD. Nortel also believes that most carriers deploying OC-768 DWDM systems will be doing so on the best fiber, so a PMDC will likely not be necessary, Belanger adds.

Although Nortel developed a PMDC, Belanger says that of the thousands of OC-192 DWDM systems it has sold and helped carriers deploy, it has encountered only "two or three" instances in which it could not defeat PMD by engineering around it and without having to move regenerators closer together.

"I don't expect the PMDC to sell a lot of systems. I expect to use it as an `insurance policy' to close the topic," Belanger says.

Lucent also has developed an integrated PMDC for its OC-192 DWDM system. The device has been available for most of 2000 for customers deploying the system on older fiber and wanting to maintain standard engineering rules, says Rod Alferness, chief technical officer for Lucent's optical networking group. The vendor has not yet announced when its forthcoming OC-768 DWDM system will be commercially available.

In August, Corning unveiled its Acrobat series of polarization controllers for equipment manufacturers of PMDCs. In addition to PMD compensation, the controllers can be used in test instruments, polarization scramblers and in polarization multiplexing architectures, says Stephen Cohen, vice president of marketing and sales for Corning.

For those that want to, or must, pull new fiber for 40 Gb/s, Corning also recently introduced its third generation of Leaf fiber, which it says offers a PMD specification that is improved 50% over previous generations.

The back 40 "It's really 40 Gb/s where the entire market believes that PMD will be an inescapable issue," says Dana Cooperson, director of optical transport for RHK. "We see 40 Gb/s field demos on networks using the best fiber and with lots of technical support in 2001, with early adopters, meaning one or two carriers beginning full-scale deployment, maybe as early as mid-2001, and market ramp-up, meaning broad acceptance of the technology, beginning in mid-2002."

It will take until 2003 for 40 Gb/s DWDM to gain momentum, says Kevin Slocum, managing director for Wit SoundView.

"In fact, one of the risks in the optical communications segments is that 40 Gb/s may end up being a significant hurdle to clear and we will have to live with 10 Gb/s a lot longer than we expected to," Slocum says.

But Mark Lutkowitz, vice president of optical networking research for Communications Industry Researchers, believes it is going to take at least five years for OC-768 to become commercially viable.

"Every fiber in my being tells me it's not going to happen for a very long time," he says.

In the meantime, Yafo likely will make a nice business for itself in the OC-192 world, he adds. "I like the company's strategy of becoming a sub-assembly supplier."

Who wants to use a PMDC? Newer non-zero dispersion shifted fiber (NZDSF) is not as susceptible to PMD as single-mode fiber, but the bulk of the fiber deployed in the RBOCs' networks and legacy long-distance carriers such as AT&T, WorldCom and Sprint is older single-mode fiber.

For example, according to Cooperson, Verizon Communications and SBC Communications believe that 60% to 70% of their fiber will not cost-effectively support OC-192 systems. The problems are not all PMD related, however, and also have to do with splices and connectors, all of which make PMD worse.

Vendors such as Kestrel Solutions and Centerpoint Broadband Technologies, which make high-bandwidth optical systems that are less susceptible to PMD because of their use of a different signal format, see the RBOCs' situation as an opportunity for their gear, she adds.

AT&T, which has an embedded base of 53,000 route-miles of predominantly single-mode fiber deployed in its legacy network, says that, to date, PMD has not been any more of a roadblock than any other fiber optic impairments. However, it is important to note that the carrier currently is deploying 16,500 route-miles of new NZDSF fiber in its network and building a coast-to-coast OC-192 DWDM network.

AT&T is using a mix of both the old and new fiber, says a company spokesman. AT&T is looking to its system vendors to incorporate technologies such as PMDCs to help the company meet its specific needs regarding PMD, says Tom Afferton, district manager for AT&T with responsibility for advanced transport technology and architecture planning.

Sprint, which has a 32,000 route-mile OC-48 DWDM fiber network, has yet to begin deploying OC-192 DWDM systems in its network but is likely to begin deployment in 2001, says Mark Jones, distinguished member of technical staff for Sprint. The carrier has evaluated its fiber spans in various weather conditions, and the "vast majority of it will suffer no PMD-related limitations at 10 Gb/s," he adds. That's because the carrier has deployed "mostly high-quality, loose-slot fiber, which doesn't get stressed as easily as some other [single-mode fiber]," Jones says.

On spans affected by PMD, Sprint will use inverse multiplexing of 2.5 Gb/s channels, closer regenerator spacing, or PMDCs, "if they work [as well as vendors are promising]," Jones says.

"Right now, we are hearing mixed signals on PMDCs. Most of what we are hearing is not too optimistic for the near future," he adds.

Neither AT&T nor Sprint will say when they think they will begin deploying OC-768 systems in their networks. Jones believes that legacy fiber installed in Sprint's network and in other carriers' networks today will not support OC-768.

The biggest market for PMDCs may be overseas, where a lot of early fiber was sold after "people realized what was going on," Alexander says.

"In Europe we have had requests for PMD compensation at 2.5 Gb/s," Perrier says. "When going from Point A to Point B, a signal can take several routes, which are not made of the same fiber. You may have several spans on a given route that are not good."

PMDCs address a niche market in the OC-192 environment, but judging by their actions, it is clear that more than a few sub-system and system vendors think PMD is worth addressing now. Neither Electronicast or RHK have released market forecasts for PMDC. Yafo believes the market will amount to several hundred million-dollar opportunity for 10 and 40 Gb/s combined over the next five years.

Alexander says we'll know in six to nine months whether there is a market for PMDCs or not.

"The issue is going to be how many new spans are going in. A lot of span engineering has already been done and is fixed," he says. "The folks who are putting in large amounts of 10 Gb/s are also guys with relatively new fiber. You don't find too many with really old legacy plant that are unloading all the stuff they have on it today in order to put on 10 Gb/s equipment."

When light travels down a fiber, it has two polarization modes: perpendicular and orthogonal. When the core of the fiber that bounds the light is asymmetrical, or out of round, the light traveling along one polarization axis moves slower or faster than the light polarized along the other axis. The separation - or dispersion - of energy polarized along these two axes is measured in picoseconds (10E-12 seconds).

Over time, the factors that determine the shape of fibers during the manufacturing, cabling and installation process have become better understood, and controls have been implemented to reduce polarization mode dispersion (PMD). The improved performance of fiber systems with respect to PMD is summarized in the table.

The table shows typical transmission distance limits that result from the PMD of cabled fiber deployed at various times. The PMD performance of the fiber is specified in ps of delay created per square root kilometer. For example, a 173 kilometer fiber span installed in the early 1990s may have a typical value of 1 ps/sq rt km.

At a transmission speed of 10 Gb/s, PMD-caused separation in the optical pulse is about 15 ps, and that is large enough to start causing errors in the receiver trying to interpret the 100 ps bits of a 10 Gb/s signal.