High-speed systems impact mux/demux choices

70339

Guy Sauvé

Technologies that dominate 10-Gbit/s transmission— such as thin-film and fiber Bragg gratings—will play a reduced role for 40-Gbit/s multiplexers and demultiplexers. The author argues that AWGs and fused fiber products are likely to dominate the new field.

Increasing data rates have always been considered the engine for the emergence of a new generation of components. The quick transition from OC-48 to OC-192 systems highlighted the need for component vendors to be ready for the next speed increment.

The issue of components for 40-Gbit/s systems, especially active ones, has generated much discussion in the industry and many vendors have already announced breakthroughs in 40-Gbit/s modulators and receivers. Much less interest has been generated for passive components, based on the assumption that things will evolve smoothly from the existing 10-Gbit/s technologies.

The transition of passive components from 10-Gbit/s networks to the 40-Gbit/s world, however, is not transparent. Well-established technologies that are acceptable for 10-Gbit/s systems may not be compatible with the new bit rate and may play a minor role in tomorrow's networks. In particular, we will examine the impact on multiplexer/demultiplexer components.

Many passive components are now on the market for use with 2.5- and 10-Gbit/s systems. However, the requirements for 40-Gbit/s passive components usually differ from 10-Gbit/s ones on three counts: lower chromatic dispersion, larger bandwidth and lower polarization-mode dispersion (PMD).

Dispersion. The amount of chromatic dispersion an optical network can tolerate is inversely proportional to the square of the transmitted bit rate, so as data rates increase, dispersion tolerance decreases dramatically. While 2.5-Gbit/s networks can tolerate 16,000 ps of dispersion, 10-Gbit/s networks can tolerate only 1000 and at 40 Gbit/s the tolerance drops to only 60 ps of dispersion. In addition, 40-Gbit/s systems require a wider modulation spectrum. A 12-GHz-wide spectrum is typical for 10 Gbit/s; for 40 Gbit/s, it can be as high as 50 GHz, which means the total dispersion per channel is higher. To address the problem, most vendors in the 40-Gbit/s field study innovative modulation techniques, primarily return-to-zero modulation, to alleviate the problem.

Chromatic dispersion has a fixed, stable component in addition to a dynamic one. Most of the fixed dispersion is caused by the fiber and is predictable as a function of the type of fiber, the distance, and the modulation. It varies linearly with wavelength. In addition, the components present on the optical path add a smaller, fixed contribution.

A dynamic contribution must be added to the fixed one. Since many passive components do not have a simple flat or linear dispersion curve, laser drift (caused, for instance, by aging or temperature change) can lead to dispersion fluctuations that are nonlinear and hard to predict. It is therefore not only important to understand the dispersion value of a component, but also the behavior of the dispersion over the desired clear channel.

Bandwidth.Typical bandwidth (defined as the bandwidth at 0.5-dB insertion loss) requirement for a 40-Gbit/s system is on the order of 50 GHz. By comparison 10-Gbit/s systems require a 20-GHz total clear bandwidth for each channel. The difference is caused by the larger spectrum required by 40-Gbit/s modulators.

Polarization-mode dispersion. PMD is caused by light traveling faster in one polarization plane compared to another. Fundamentally, it is caused by the core of the fiber not being perfectly round in cross section. As a result, the thickness is not absolutely identical on every possible axis. PMD is such a critical problem at 40 Gbit/s that some PMD compensation pure-play startup companies have emerged solely for the purpose of bringing solutions to the market.

It is harder to compare passive technologies on PMD performance, however, since the majority of vendors do not publish performance figures for their products. That situation may change in the coming months.

FOUR PASSIVE TECHNOLOGIES
Four technologies compete for the multiplexing and demultiplexing business: thin-film filters (TFF), arrayed waveguide gratings (AWG), fiber Bragg gratings, and fused fiber. The advent of 40-Gbit/s systems may signal a much-reduced role for two technologies that dominated the 10-Gbit/s world so far: TFF and fiber Bragg gratings (see table).

Thin-film filters. Thin-film filter technologies rule the 16-channel, 200-GHz spacing market with low-cost solutions. However, they do not scale well to the high-channel-density world. A typical 100-GHz filter requires more than a hundred layers of coating to create narrow-band filters. With so many layers being deposited, errors caused by local film thickness variation and alternation in density increase, reducing the yield of useful filters.

In addition, thin-film filters are very dispersive. A transmitted beam that goes through a filter is composed of multiple sub-beams, each having a slightly different travel time (see Fig. 1). The overall impact is a spread in time—or dispersion—of the beam. The same goes for the reflected beam, which in addition goes through multiple filters with cumulative effects. The dispersion therefore is not uniform across the wavelength range, increasing the difficulty of correcting it with simple linear methods such as dispersion-compensating fiber.

Fiber Bragg gratings. The principle behind fiber Bragg gratings uses the reflection of a wavelength passing through alternating regions of changing refractive index. The transmit channel shows very few impairments. The reflect channel, however, is built from the addition of partially reflected light at each change of index along the grating section, which creates an important delay in the overall reflected signal (see Fig. 2).

Fiber Bragg gratings are very dispersive, showing a chromatic dispersion of as much as ±200 ps/nm (by comparison, most thin-film products have approximately ±50 ps/nm dispersion), making these technologies unsuitable for 40-Gbit/s multiplexing/demultiplexing applications. Many submarine companies using forward-error correction (FEC) consider the maximum operational point for fiber Bragg gratings to be about 10 Gbit/s, making it not acceptable even for 10-Gbit/s solutions using FEC, in which the actual data rate is close to 12 Gbit/s.

In addition, inaccuracies in the grating cause ripple in the dispersion (see Fig. 3). One consequence is the degradation of the performance of the system. Laser drift also can cause fluctuations in the dispersion, making the correction more complex.

Array waveguide grating. Contrary to Bragg and thin films, AWG solutions do not work on the principle of diffraction layers and as a result show a very low dispersion of from ±5 ps/nm to ±10 ps/nm, making it an attractive option for 40-Gbit/s systems. In addition, the dispersion is flat within the passband, an important feature to simplify dispersion compensation (see Fig. 4).

The primary challenge facing most vendors of 100-GHz AWG solutions is to provide the right amount of bandwidth. A survey of six leading AWG vendors' specifications show a bandwidth ranging from 25% to 58% at 1 dB, or approximately 12% to 29% bandwidth at 0.5 dB. That is insufficient for most 40-Gbit/s applications (see Fig. 5).

Fused fiber. Fused-fiber products work on the principle that the best possible waveguide is the fiber itself. The impairments to the signal are limited since the light does not transit through a foreign environment. It is usually the technology that brings the best optical performance, which is necessary for advanced applications such as 40-Gbit/s systems.

The principle at the base of fused-fiber filters is the creation of a group delay between two modes to create the required spectral response. When the fiber is highly stretched (decreasing to 30% of its original radius), the difference of group delay between the two modes increases rapidly, creating constructive and destructive interferences. However, the dispersion itself does not increase since the light beam is not reflected over a given length, unlike thin films or fiber Bragg gratings (see Fig. 6).

Fused fiber can also be used to build Mach-Zehnder interleaving products. These products are inherently dispersive. However, by cascading elements it is possible to design them in a way that one stage counteracts the dispersion created by the other, something that is not practically feasible with other technologies. Correctly designed fused-fiber interleaving products can support near-zero dispersion, making them excellent for dispersion-sensitive systems. Other desirable characteristics of fiber-based products, such as a low PMD and high isolation make them strong candidates to play a role in 40-Gbit/s products. They will also play a major role in metro and ultralong-haul systems to increase the number of add/drop points, where dispersion adds up at each node.

Guy Sauvé is marketing director, transmission products, at ITF Optical Technologies, 175 Montpellier, St. Laurent, Quebec, Canada H4N 3L2. He can be contacted at gsauve@itfoptical.com.

More in Market Research