The latter half of the 1990s has seen a steady back and forth between advances in Synchronous Optical Network/Synchronous Digital Hierarchy (sonet/sdh) transmission and wavelength-division multiplexing (WDM) technology in the drive for greater bandwidth. First, OC-12 (622 Mbits/sec) set the standard; then WDM increased bandwidth by combining multiple OC-12 channels on a single fiber. sonet/sdh rates climbed to OC-48 (2.5 Gbits/sec), and WDM manufacturers responded with dense WDM equipment that squeezed OC-48 streams through as many as 128 channels.
Single-stream OC-192 (10 Gbits/sec) subsequently offered an alternative to multiplexed OC-48. With carriers now examining the usefulness of OC-192 multiplexers, it seems only logical to expect the imminent arrival of the next level of time-division multiplexed (TDM) sonet/sdh transmission, OC-768 (40 Gbits/sec).
That expectation appears close to fulfillment, judging by the announcements made around the time of this year's Conference on Optical Fiber Communications (OFC) in February. Alcatel (Paris) used the show to demonstrate the results of its research into 40-Gbit/sec transmission conducted under the auspices of the European Commission's ACTS program. A month after OFC, Lucent Technologies (Holmdel, NJ) announced an OC-768 product called the WaveStar 40G Express, which MCI WorldCom will receive for commercial testing by the end of this year.
Each step on the staircase to higher TDM speeds presents its own set of technological hurdles, representatives from the two companies explain. Now that those obstacles appear surmounted, the question the companies now face is whether MCI WorldCom represents the first in a wave of immediate carrier interest in the technology, or whether multiplexed OC-192 or OC-48 will prove adequate for near-term service-provider requirements.
As the amount of time that passed between Nortel Networks' introduction of OC-192 transmission equipment and the appearance of competing systems indicates, raising the rate of TDM transmission requires significant technical skill.
According to Rod Alferness, chief technical officer within Lucent's Optical Networking Group, the development of 40-Gbit/sec transmission systems first required further maturation of the electronics necessary for TDM, modulation, and other transmitter and receiver functions. Generally, these functions require application-specific integrated circuits to meet system requirements.
What happens to the light once the electronics help generate the desired signal presents the most significant problems, however. General chromatic dispersion and pulse mode dispersion (PMD) within the fiber that the 40-Gbit/sec signal will traverse can significantly dampen transmission distances. In particular, PMD becomes a greater problem as speeds increase, since it scales as the bit rate squared.
The search for adequate compensation techniques has followed several paths. Paul Harrison, vice president and general manager of lightwave systems at Alcatel USA (Richardson, TX), says his company has pursued both optical and electrical avenues. The optical technique involves countering phase delays by adjusting the fiber. "Now, obviously, you do it in terms of 'micro-nanometers,' but one effective technique of doing that is to bend the fiber in such a way, right at the transmitter and receiver sides, that you compensate for the phase delay," he explains.
Researchers also have used dispersion-compensating fiber to enable 40-Gbit/sec transmission. This technique can provide one set of compensation for an entire group of wavelengths-an advantage when multiplexed OC-768 becomes feasible, Harrison says. Some of the other techniques, such as fiber Bragg gratings, are wavelength-specific. As wavelength counts increase, having a separate grating for each wavelength may prove impractical, he surmises.
Alcatel has developed a PMD mitigation device that it has used in OC-192 and OC-768 field trials under the ACTS umbrella. Based, at least in part, on the optical compensation principles described above, the device consists of a polarization controller, a piece of polarization-maintaining fiber, and a feedback loop that includes a degree-of-polarization (DOP) measurement device and a processor that provides the control algorithm that maximizes the DOP signal acting on the polarization controller, according to a paper presented by Alcatel at OFC. Field tests of the device have highlighted the difficulties of combating PMD. "PMD is one of those things which varies with time, with temperature, and it's very unpredictable," Harrison explains. "One of the things we've found out through these trials is that PMD can vary a great deal over a very short period of time with events which before we wouldn't have thought about." For example, one of the fibers used in a trial ran along a railroad track: The researchers found that every time a train passed, the resulting vibration changed the fiber's PMD significantly.
Of course, given the fact that for every TDM action there seems to be an equal WDM reaction-as well as the fact that both Alcatel and Lucent produce WDM equipment-both companies have tailored their OC-768 approaches to meet future multiplexing requirements. The power required to propel 40 Gbits/sec down each channel exacerbates the problems with nonlinearities that system designers have battled with dense WDM at lower speeds, Lucent's Alferness reports.
Cramming a 40-Gbit/sec signal into the 50-GHz passbands that will be common in future dense WDM equipment also weighs on the minds of system designers. Straight electronic TDM techniques produce pulses with extremely wide spectral width, says Harrison. "And even if you can get [the signal] into, say, a 100- or 200-GHz bandpass filter, those filters have to be very sharp down at the back end to stop the interference between those two channels," he claims.
Alcatel has developed a modulation technique it calls "Q-DST" that narrows the spectral bandwidth to WDM-friendly proportions. The company combines Q-DST with basic electronic TDM techniques-which bit-interleave four 10-Gbit/sec laser outputs into a single 40-Gbit/sec stream-to create its OC-768 transmission device (see Figure). "The laser technology associated with the 40-Gbit/sec transmission is actually very simple and similar to what we're doing at 2.5 and 10 [Gbits/sec]. We are directly modulating a laser that you can buy off-the-shelf today," Harrison says. "And as far as the receiver technology, there is nothing particularly special using the Q-DST technique over and above what you can get off-the-shelf."
Developing transmission and dispersion-compensating devices is one thing, proving that they work is another. Both Alcatel and Lucent have tested their equipment and found that the type of fiber used in the network has a significant impact on the utility of 40-Gbit/sec transmission equipment.
"A lot of the success in the 10-Gbit/sec regime has really been with new operators who have been laying new fiber, things like TrueWave from Lucent and the other new fibers like LEAF from Corning, which have got some of the properties that can overcome some of the dispersion and the nonlinear effects," Harrison reports.
Alferness and his counterparts around the manufacturing community see the value of appealing to carriers with a variety of older fiber types. How successfully they can achieve this goal remains unproven, however. "I think at 40 Gbits, it isn't clear that these systems will work on arbitrary fibers," he says. "Our view, of course, is that this system will work best on TrueWave fiber, but I think at this point, some of the very old fiber with very high PMD is potentially an issue. So I don't know that there's uniform agreement that a new system like this has to work on all existing fiber."
That said, the MCI WorldCom trial will include standard singlemode. "Even in standard singlemode fiber, the level of PMD in standard singlemode has varied as a function of time. So singlemode fiber that was deployed very early has PMD levels that are substantially higher than the fiber that has come out more recently. So that's where the caveat on the vintage of the fiber is potentially an issue," Alferness warns.
Still, the drive to accommodate as wide a range of fiber as possible has led to field experiments with embedded standard singlemode strands. For example, Alcatel has participated in the ACTS Highway effort, which saw Deutsche Telekom install OC-768 equipment into networks in the Stuttgart, Germany, area as well as between Mannheim and Darmstadt. Building on earlier work within the ACTS Start effort, which demonstrated 40-Gbit/sec transmission over approximately 86 km in Portugal, the Highway demonstrations saw successful operation at 111 and 130 km, respectively, over installed singlemode fiber.
The 130-km distance is important, Harrison asserts. "You're getting now into the ballpark of the span distances currently used at 10 Gbits/sec, which is really what we're trying to achieve, because everybody's got their routes already in, they've got their amplifiers at certain distances. And if you can't hit that same distance at 40 Gbits/sec as you do with 10 Gbits/sec, then they're not going to put it in because they're not going to move their amplifiers," he explains.
Other research presented at OFC illustrates the divergent approaches to extending 40-Gbit/sec transmission now under investigation. For example, researchers at Chalmers University of Technology in Gothenburg, Sweden, reported on single-wavelength soliton field experiments over 400-km fiber without inline control. BT Laboratories, the University of Bristol, and France Telecom CNET described the results of tests over 186.6 km of installed fiber using mid-span spectral inversion for dispersion compensation. These tests also fell under the auspices of the ACTS Highway program. Another effort within ACTS, the esther program, has measured the effects of PMD on a field demonstration of 40-Gbit/sec soliton streams over 500 km. Fondazione UgoBordoni, France Telecom CNET, Pirelli, and ISTC participated in this program. Meanwhile, researchers at BT Labs have teamed with Aston University to investigate 40-Gbit/sec RZ data transmission over 1220 km over dispersion-managed standard singlemode fiber.
While most of the work at 40 Gbits/sec has occurred in Europe-even Lucent conducted the majority of its research at its facilities in Germany-NTT in Japan has also advanced the cause. Researchers in NTT's Optical Network Systems Laboratory reported at OFC on a field experiment of soliton transmission over 1020 km using dispersion-shifted fiber, which they extended to 1360 km using inline synchronous modulation. The company also has looked into multiplexing 40-Gbit/sec signals. A 4-channel experiment was the subject of a paper at the conference. Finally, as if to show that at least some work is going on in the United States, Lucent's Bell Labs reported on its multiplexing experiments, including 25 channels of OC-768 over 342 km of TrueWave non-zero dispersion fiber.
While the experiments and demonstrations recounted during the seminar portion of OFC hint at 40-Gbit/sec transmission systems to come, both Lucent and Alcatel plan to have commercial hardware next year. For example, Alcatel will have systems ready for U.S. field trials by the end of this year, Harrison reports. Production-level equipment should be ready in the second half of next year, likely in the form of upgraded versions of the company's existing 1680 OGM Optical Gateway Manager. Such timing will put the company slightly behind Lucent, which expects to have a commercialized system in March 2000.
"By the end of 2000, the beginning of 2001, you're going to see the products coming out of the market," summarizes Harrison, who predicts that Nortel Networks will not be far behind Lucent and Alcatel. [Representatives from Nortel Networks confirmed that the company has work underway in this area, but company experts were not available for interviews at the time this story was written.] "And I think, in the same way that 10-Gbit rolled out, you're going to find the early products are going to be quite simple, point-to-point probably, but with upgrade paths that are going to ring-based systems." Lucent confirms this assessment, having already announced its plans to support sonet/sdh self-healing rings in future versions of the WaveStar 40G Express.
Naturally, the vendor community expects carriers will require 40-Gbit/sec capability in the future. They wonder how far in the future this requirement may be, however.
Harrison believes that both Alcatel and Lucent will be ahead of the market to some degree, and that the two will have to overcome "the fear factor" of new technology before 40-Gbit/sec equipment becomes popular. But Alferness believes that the market for this equipment already exists.
"What we're hearing from customers is the desire for 40 Gbits in anticipation of the enormous growth in data-capacity demand, and in particular the view that high-speed router technology will be continuing to move to even higher-speed interfaces," Alferness explains. "So we see that one of the drivers for the 40-Gbit applications is really to provide connectivity between routers, as router speeds go up and router interfaces go up to the OC-768-type range."
This requirement won't materialize tomorrow, predicts Stephen Montgomery, president of market research and analysis firm ElectroniCast Corp. (San Mateo, CA). "What we're looking at is probably a few more years out when we start to see some actual numbers," he says. "I believe it's actually more like three or four years off."
Several factors combine to account for this delay, he says. "One of the reasons for the delay that we see is the capabilities of DWDM. That's enabling carriers to increase their capacity without having to increase the transmission rate," Montgomery explains.
The DWDM option has proven particularly appealing in recent years, as carriers weigh the three major avenues to increased bandwidth: more fiber, more channels, or more speed. "Each carrier needs to evaluate on a cost model what's most effective for that specific link," Montgomery says. "And all of this is based on several variables, such as what is the existing capability of the fiber in that particular link. Is it an old fiber or is it a link that consists of several different fiber types supplied by several different vendors? Does it have a lot of splices, does it have a lot of other components that may create some losses? In other words, what kind of shape is this link in?"
The fact that several incumbent carriers have lost track of the different fiber types in their network compounds the complexity of the OC-768 equation, surmises Montgomery. He feels that OC-768 systems will require new fiber to work reliably, which points to new carriers with "greenfield" applications as the most likely early adopters of the technology.
Regardless of whether carriers will be clamoring for 40-Gbit/sec systems by next year, the success Nortel Networks enjoyed by being the originator of commercial OC-192 systems has not gone unnoticed by other equipment vendors. Companies such as Alcatel and Lucent-and undoubtedly other vendors who are working out of the limelight-will push to make their equipment ready for the field as soon as possible, even if that means staking their claims over the bandwidth horizon and waiting for their potential customers to catch up. q
At OFC, Alcatel combined electronic TDM to combine four 10-Gbit/sec streams into a single 40-Gbit/sec framework, then used its proprietary Q-DST technology to modulate the signal into a form with a small enough spectral width for efficient multiplexing.