As carriers push on with the inevitable transition to 10 Gbit/s networks, the intrinsic characteristics of optical fibre start to yield some nasty surprises. Top of the unwanted list is dispersion, which causes interference by spreading the energy of each optical pulse over neighbouring bits. Dispersion is present in all optical systems, but its effects become worse over longer spans, at higher transmission speeds and in networks built from older fibre.
Existing long-haul systems already incorporate optical compensation elements to correct for chromatic dispersion, which builds up as different optical wavelengths travel through the fibre at different speeds. But more ingenious solutions will be needed in the next generation of optical systems.
For a start, metro networks operating at 10 Gbit/s and above will need dispersion compensation for the first time, particularly as system spans extend to 100 km. However, existing optical methods for correcting chromatic dispersion are too costly and power- hungry to be deployed in metro applications.
Meanwhile, long-haul systems will need to cope with the unpredictability of polarization-mode dispersion (PMD), a dynamic effect caused by the two planes of light travelling through the fibre at slightly different rates. Optical compensators are now emerging for correcting PMD, but these big optical boxes are a costly and power-intensive option that carriers can ill-afford.
Electronics make sense
A more promising approach is to incorporate electronic techniques for dispersion compensation into the physical-layer chips used to convert the optical signal at the receiver end of a fibre-optic link into parallel streams of electronic data. Building the compensation function into existing electronic devices saves space, reduces power consumption, and comes at about the same price as one of today's high-end physical-layer chipsets.
"It makes sense to deploy electronic dispersion compensation [EDC] in any new design because you almost get the function for free," comments Allan Armstrong, director of communications semiconductors at US market-research firm RHK. "Putting compensation into an existing linecard rather than a separate box is a threat to companies specializing in optical dispersion compensation."
Four US companies - AMCC, Santel, Big Bear and Intersymbol - have announced chipsets for dispersion compensation over the last few months, and Armstrong says that others are working on EDC in stealth mode. Indeed, he believes that the technology will be so disruptive that other manufacturers of physical-layer chips, including the likes of Broadcom, Infineon, Vitesse, Agere and Intel, will be squeezed out of the market if they do not introduce EDC into their devices.
Chip vendors are quick to point out that EDC will be complementary to optical methods, particularly in the long-haul market. "Long-haul systems are deployed with dispersion-compensation elements and optical amplifiers, and that's unlikely to change in the near future," says Dan Castagnozzi, solution architect for AMCC's wavelength-division multiplexing (WDM) products. "But it's difficult to match dispersion-compensation elements across all wavelengths in dense WDM systems, which leads to residual dispersion."
Castagnozzi explains that EDC can correct any residual dispersion in the system to flatten the performance across the whole wavelength band. The hardware and engineering costs of this approach are much lower than for alternatives such as fibre Bragg gratings and specialized optical compensators, and it can be deployed in existing systems without introducing any optical insertion loss.
The other issue for long-haul networks is PMD, which becomes a major headache at line speeds of 10 Gbit/s and above. "PMD is expensive to compensate for in the optical domain because the wavelengths must be separated out before compensating each one individually," says Castagnozzi. "EDC operates on a per-wavelength basis, and so offers a cost-effective solution for PMD."
Unlike chromatic dispersion, PMD is a time-varying effect that must be compensated for on a millisecond timescale. "The amount of PMD can change considerably over short time periods, as it depends on the characteristics of the fibre as well as the surrounding physical environment," says John Keating, chief operating officer for Santel Networks. "EDC is ideal for PMD compensation because it is inherently adaptive. Current designs can compensate for PMD over transmission distances exceeding 1700 km."