Electronic equalization enables high-speed metro networks

104254

The continuous growth of bandwidth demand in the access network places pressure on the capabilities of metropolitan communication networks. Optical links in metro/regional core rings are now moving to10-Gbit/s line rates, and system vendors are looking ahead to the next steps required to increase capacity and lower cost. The benefits of increasing optical-transport time-division multiplexing (TDM) speeds have been repeatedly proven as rates have risen from 155 Mbit/s to 10 Gbit/s. Yet moving to 40 Gbit/s in the metro network raises significant challenges, primarily from fiber impairments such as chromatic dispersion (CD) and polarization-mode dispersion (PMD).

At 40 Gbit/s the tolerance for chromatic dispersion is small, with a path penalty of 1 dB for 60 ps/nm of dispersion. While new fiber has low PMD, much of the fiber in metropolitan networks is older and will not be able to carry 40-Gbit/s traffic without compensation. For example, an 80-km link with average PMD of 0.5 ps/sqrt km could have a power penalty of more than 5 dB resulting from first-order PMD alone.

The economic viability of 40 Gbit/s in the metro environment requires that the links be straightforward to deploy. For a 40-Gbit/s link to be as simple to set up as a current 2.5-Gbit/s link, the fiber impairments must effectively be transparent to the technical staff performing the installation.

The addition of active intelligence at the optoelectronic interface, together with adaptive electronic- and optical-compensation techniques, can enable a complete plug-and-play 40-Gbit/s solution.

There are clear advantages in size and cost to solving fiber impairment issues electronically rather than optically. Electronic equalization has been used with great success in the copper domain, where compensation of echo and other general transmission impairments have enabled the dramatic speed improvements of telephone modems and DSL connections. By further extending these algorithms and implementing them with the latest high-speed analog and digital electronics, one can obtain significant compensation for many of the impairments observed in optical transmission.

Electronic equalization provides compensation for many different sources of signal degradation such as PMD and CD, as well as other intersymbol interference (ISI) effects such as transmitter effects, receiver-chain bandwidth limitations, and self-phase modulation. It is useful to think of electronic equalization as a method of providing general, additional system margin over an end-to-end link.

FIGURE 1. With electronic equalization based on TF architecture, the input signal is tapped at several points along a sequential delay line and the resulting tap outputs are summed to give the output signal. Tap weights are determined by a software algorithm on a separate DSP chip (top). With an MLSE architecture, the input signal is digitized using a series of A/D converters that are timed by the output from a CDR. Resulting data are processed in an ASIC to generate the optimized output signal (bottom).

Within the electronic equalization approach there are two broad families of devices. One is symbol-by-symbol detection in which the bit stream is continuously modified by an analog circuit—examples include transversal filter (TF) and decision feedback equalization (DFE). A more complex approach involves completely digitizing the incoming bit stream, then performing pattern recognition with extensive digital signal processing—this is commonly referred to as maximum likelihood sequence estimation (MLSE; see Fig. 1).

Advantages of the symbol detection approach include low power consumption and the ability to simply drop the compensator into an existing receiver chain. This architecture is also the simplest to scale to higher speeds. The symbol detection method provides excellent PMD compensation for dispersion-group-delay (DGDMAX) values up a large fraction of a bit period (see Fig. 2).

FIGURE 2. A simulated 40-Gbit/s optical signal is distorted by a full bit period (25 ps) of DGDMAX. The power split is 35% slow axis and 65% fast axis. The second order component consists of a polarization-dependent chromatic dispersion of 30.68 ps2 and a rotation rate of 8.47 ps (top). The same signal is shown at an intermediate state in the compensation convergence using a TF electronic equalize (center). The signal is compensated after completion of the convergence algorithm by a TF electronic equalizer (bottom).

The MLSE approach promises somewhat higher levels of compensation but requires significantly more power and extensive changes to the clock data recovery (CDR)/demux section of the receive chain. The complexity of the approach also presents significant challenges in scaling from 10- to 40-Gbit/s rates.

In terms of size, we can expect any of these electronic equalizers to be available as a small module or multichip set that occupies on the order of a square inch or less of board space. Using other electronic chips as guidance, we can expect the price of these equalizers to be in the "hundreds of dollars" range depending on complexity and volume. By comparison, current optical PMD solutions typically occupy tens of cubic inches of space and cost thousands of dollars.

In terms of compensation speed there continues to be discussion in the industry about the convergence and tracking rates required for PMD compensators. Compensation for acoustic effects would require speeds in the kilohertz range, while compensation for fiber handling would be significantly lower. The compensation speed of electronic equalization techniques is mostly limited by the properties of the error signal and by the speed of the associated digital signal processing. Convergence times in the millisecond regime are expected for electronic equalizers.

Given that PMD compensation must be performed on a channel-by-channel basis, the size and cost of a solution are key parameters. The small size and low cost of the electronic equalizer, together with its usefulness against a broad variety of impairments, make it an attractive solution. With its value and ease of implementation, electronic equalization can be incorporated as a standard component in future long-haul 10-Gbit/s systems, as well as in 40-Gbit/s metro and long-haul systems.

Optical PMD compensators have the capacity for high DGD compensation and may be of economic benefit for cases of very poor fiber quality with limited alternatives. However, given the current size and cost of optical compensators, we believe this is likely to be a niche market.

FIGURE 3. An experimental arrangement was used to test a 10-Gbit/s system with and without electronic compensation (top). Measured OSNR penalty as a function of DGDMAX for a 10-Gbit/s receive chain with and without the transversal filter in the signal path (bottom).

Preliminary measurement of optical PMD compensation using a 10-Gbit/s transversal-filter electronic equalizer agreed with the theory. Measured performance improvement as a function of DGDMAX was shown when the transversal filter is added to the receive chain (see Fig. 3). In addition to compensating for PMD we have also observed compensation of bandwidth limitations and other ISI effects and we are continuing to quantify this behavior.

It is evident from the work so far that the electronic equalization approach has significant benefit for PMD and ISI impairments. Regarding chromatic dispersion, the current symbol-by-symbol techniques can compensate for a degree of chromatic dispersion but are not expected to cover the full range of chromatic dispersion present in either 10-Gbit/s long-haul systems or 40-Gbit/s long-reach metropolitan systems. This is primarily a result of the large time delays involved, which are often many bit periods.

The compensation for chromatic dispersion provided by the electronic equalizer does enable greater tolerance for variation in the optical dispersion compensation such as blocks of dispersion-compensating fiber. Further, the equalizer acts to reduce the effect of residual slope error on individual channels following the optical compensation techniques. In this context, the equalizer can provide economic benefit in both metropolitan and long-haul networks.

Electronic equalization offers significant benefit in 10-Gbit/s long-haul systems but becomes a critical prerequisite for deployment of 40 Gbit/s in metropolitan networks. Several issues need to be addressed when extending an electronic equalization product from10 to 40 Gbit/s,

At the most general level, electronic circuits at 40 Gbit/s are inherently more difficult to design and manufacture than those at 10 Gbit/s. However, there has been significant experience gained within the industry, particularly in silicon germanium technology, which has the advantage of high speed combined with excellent large-scale integration capability.

Some electronic equalization methods will scale more easily than others. Evolving the transversal filter from 10 to 40 Gbit/s is relatively straightforward as most of the changes occur in the analog circuitry. In addition, the algorithms and software developed for the transversal filter are directly applicable at 40 Gbit/s. Extending the MLSE to 40 Gbit/s is likely to be more difficult given the numerous areas that require changes, notably increasing the speed of the analog-to-digital converters.

We now have all the building blocks necessary for creating a plug-and-play 40-Gbit/s optical transport link in the metropolitan network. Note that, technically, the links will most likely be operating with forward-error correction and would therefore be running at 43 Gbit/s.

For dynamic compensation of chromatic dispersion several tunable optical compensators, such as the fiber Bragg gratings and the virtual image phase arrays are now commercially available. These devices have the ability to compensate for the 1600 ps/nm of chromatic dispersion in 80 km long-reach metropolitan links without requiring additional dispersion-compensating fiber. They are also becoming sufficiently compact and cost-effective that they can be considered for use in 40-Gbit/s metropolitan links to compensate individual channels, possibly fitting on the same line card as the transponder.

FIGURE 4. A plug-and-play 43-Gbit/s metro transponder could be based on electronic equalization plus an optical tunable chromatic-dispersion compensator. The chromatic-dispersion compensator would be controlled by the transponder.

Incorporating active intelligence and electronic equalization into the 40-Gbit/s transponder provides compensation for the PMD and general ISI impairments of the link. Further, the quality of the received data is measured by the transponder either in the equalizer or CDR. This measurement is then used in a microprocessor-controlled feedback system to tune an external optical chromatic-dispersion-compensator module (see Fig. 4). The intelligent transponder continually monitors the quality of the signal and dynamically maintains optimal performance over the link.

John Trail is director of product management, Jeff Rahn is an optical systems engineer, and Sudeep Bhoja is principal DSP architect at Big Bear Networks, 1591 McCarthy Boulevard, Milpitas, CA 95035. John Trail can be reached at jtrail@bigbearnetworks.com.

More in Transmission