Dispersion-tolerant technologies battle for attention
Dispersion-tolerant transmission promises to become a popular request as metro and regional networks transition to 10 Gbits/sec-and perhaps a necessity as carriers implement reconfigurable optical add/drop multiplexers (ROADMs). Several module vendors have touted the potential of optical duobinary techniques to meet this demand. (See “Duobinary Modulation for Optical Systems” on page 11 of this issue for a discussion of how this technology works, and “Optical Duobinary Promises Improved Reach, Greater Flexibility” on page 1 of the May 2006 Lightwave for a look at how and why some module vendors plan to apply the technology.) However, optical duobinary is not the only alternative modulation format under study, and at least one module supplier thinks it has a better approach. Meanwhile, another developer believes dispersion problems are best addressed on the receive side of the transmission equation.
Dispersion tolerance has appeal in a wide variety of applications. As data rates in metro and regional networks rise from 2.5 to 10 Gbits/sec, transmissions become more susceptible to the effects of dispersion, particularly chromatic dispersion. Using conventional technology, carriers would have to reengineer their networks to add dispersion compensation modules or fiber and perhaps amplifiers to accommodate the increased transmission speeds; both the reengineering and the additional network elements would cost money carriers would prefer not to spend. Dispersion-tolerant transmission theoretically would enable carriers to perform 10-Gbit/sec upgrades without touching the existing plant.
Greenfield 10-Gbit/sec networks also could benefit from dispersion-tolerant transceivers and transponders. The more robust the signal in the face of dispersion, the farther it can travel. Dispersion-tolerant channels would not need amplification as often as conventional transmissions, potentially decreasing amplifier investments.
Finally, the growing popularity of ROADMs implies a requirement for dispersion tolerance as well. As carriers reconfigure signal paths through a ROADM, the dispersion map, length, and potentially the fiber composition of a link will change. Dispersion-tolerant transmit/receive modules would increase the likelihood that service providers could reroute signals quickly and without network reengineering, rather than incur the cost and headache of rebalancing each link as it changes or reengineering links in anticipation that they might change.
As reported last month, module vendors such as Civcom, Essex, and Kodeos Communications (among others) have demonstrated the potential of optical duobinary transmission techniques to provide better dispersion tolerance than the non-return-to-zero (NRZ) transmission common in today’s networks. However, Azna LLC (www.aznallc.com) believes its “chirp-managed laser” (CML) can achieve effects similar to duobinary in a smaller package size that requires less power and provides better back-to-back performance than current duobinary offerings.
The CML’s basic building blocks include a DFB laser and an etalon-based device called an optical spectrum reshaper. In a typical NRZ module, the DFB is biased as close to threshold as possible. The laser turns on to transmit a “1” bit and off for a “0”; the high extinction ratio between the on and off states helps differentiate between 1s and 0s. However, the constant turning on and off creates a significant amount of transient chirp, and transient chirp contributes to dispersion problems.
Azna biases the CML’s emitter well above threshold, says Daniel Mahgerefteh, chief technology officer. This means the laser is “pretty much on all the time and in a nice steady state,” he says. This also means that the emitter doesn’t create the transient chirp common to NRZ transmitters. The resulting modulation depth is fairly small, but the optical spectrum reshaper compensates for the lack of modulation depth and provides the necessary extinction ratio.
Although the CML’s design decreases transient chirp, it doesn’t remove all chirp from the signal. Another type of chirp-adiabatic chirp-remains. And that’s a good thing, according to Mahgerefteh.
“The frequency modulation or the adiabatic chirp basically tells you that the carrier frequency is changing, depending on the bits,” he explains. “If you’re sending a ‘1’ bit, during the ‘1’ the frequency is blue-shifted relative to when you’re sending a ‘0,’ where it’s red-shifted. The interesting thing about this is that this frequency modulation produces memory in the bits; the phase of the bits will change, depending upon the previous bits. So by adjusting the adiabatic chirp, which we do by choosing the drive voltage, we can arrange it so that with a ‘1 0 1,’ the phase of the ‘1’ that comes after the ‘0’ is going to be flipped by 180 degrees. You can show that you will get phase correlation between the bits, which follows a very similar rule to what’s used in optical duobinary and also the classical phased duobinary. But because the laser is naturally giving us this frequency shift, we don’t need to do any coding. And we get, kind of for free, this phase coding which helps to keep the bits in their slots and gives us the dispersion tolerance.”
The performance of the CML is similar enough to that of duobinary transponders that the two technologies can interoperate, as a demonstration of 120-km modules the Optical Internetworking Forum (OIF) conducted at OFC/NFOEC this past March proved. However, the CML doesn’t need an encoder or the comparatively large LiNbO3 modulator common to duobinary transmitters. Thus, the CML can fit into a smaller package. Azna has already qualified and shipped a product in a butterfly package for systems companies who want to do discrete designs on their line cards; the company has customers in this category, says Frank Fan, Azna’s vice president, product line management and marketing. A TOSA package for module vendors looking to offer a dispersion-tolerant XFP has just reached the beta stage and should be fully qualified by the end of the year, according to Fan.
In addition to the smaller size, the CML requires 50% less power than a typical duobinary approach, Fan asserts. Like duobinary, the CML doesn’t require a special receiver and is compatible with fielded amplification and dispersion-compensation technology, he adds. The company is currently working on a full-band tunable version of the technology for its next major release.
Azna currently targets the CML at applications ranging from 0 to 200 km. For example, the DM80 version of the device offers 10-Gbit/sec transmission with 4 to 7 dBm of output power for 100 km applications. It enables an overall module power consumption of less than 1 W.
Again like duobinary, CML technology can benefit from electronic dispersion compensation (EDC). Azna has experimented with decision feedback equalization (DFE), feedforward equalization (FFE), and maximum likelihood sequence estimation (MLSE). While the last of these three provides the biggest performance boost, all can be helpful, Mahgerefteh reports. Meanwhile, Azna collaborated with Lucent engineers in experiments that paired CML technology with a cascade of tunable optical dispersion compensation devices to achieve a reach of 600 km. While Azna is unlikely to offer a product that combines CML with additional EDC or optical dispersion compensation, “we are looking at other paths to extend reach,” Mahgerefteh says.
Despite the promise of duobinary and CML, one could wonder if you’re eventually going to pair them with EDC or some sort of optical compensation device anyway, why change modulation formats? Several merchant silicon vendors-particularly AMCC (www.amcc.com), Broadcom (www.broadcom.com), and Scintera Networks (www.scintera.com)-offer EDC chips for system developers who prefer to deal with dispersion effects by scrubbing them away on the receive sides of their line cards, rather than change modulation schemes. CoreOptics has taken this approach a step further by incorporating MLSE technology into its 10-Gbit/sec 300-pin transponders to offer systems designers a modular approach to dispersion compensation.
“In general, the feedback I hear from our system provider customers and carrier customers is that if they can get the same performance in silicon-i.e., using MLSE-they would rather implement MLSE on the receiver path than use any modulation technique,” asserts Saeid Aramideh, vice president of marketing, business development, and sales, Americas and Asia, at CoreOptics. Silicon economics favor EDC over alternative optical modulation formats, he reasons, as does the greater robustness to PMD the MLSE approach offers. Since the CoreOptics modules use NRZ, they can be plugged easily next to uncompensated modules on existing channel plans as well.
EDC also enables CoreOptics to match a silicon cost curve with a roadmap of performance improvements. While the company’s current offering supports 2,000 to 2,500 psec/nm of dispersion tolerance with a 2-dB optical signal-to-noise ratio (OSNR) penalty, the upcoming second release will hit around 5,000 psec/nm (200- to 300-km reach) without encountering the back-to-back penalties duobinary and CML suffer. The enhanced product should be released in the middle of next year.
Aramideh says that MLSE provides better performance than such alternative EDC strategies as DFE and FFE, which are commonly used in the merchant silicon offerings. While next year’s second release of the product will significantly outperform current DFE/FFE approaches, MLSE also offers a more graceful curve in terms of OSNR changes as one approaches the dispersion limits of the respective technologies (see figure). Thus, MLSE makes link engineering much easier.
“I think electronics is the only way to go, both from a cost perspective and certainly the performance enhancements you get out of equalization through MLSE,” Aramideh concludes. Siemens, at least, agrees-the company uses CoreOptics’ modules in its SURPASS hIT 7500 DWDM systems. CoreOptics has worked with other major system houses as well; it jointly unveiled its first subsystem with Marconi in 2004. The company’s subsystems have reached the field in the networks of 15 service providers, Aramideh says. The modules are tunable, leveraging technology from Santur Corp. (www.santur.com). CoreOptics also has 40-Gbit/sec performance on its roadmap.
With increased interest in dispersion-tolerant transmission, system houses and their carrier customers face a variety of technology choices. The arrival of modules for such applications should signal a trend away from discrete approaches designed in-house toward transceiver/transponder-based line cards that look very similar to today’s nontolerant offerings.