Experimental soliton systems can resist polarization-mode dispersion and thus may help expand the bit-rate capacity of high-speed optical networks.
Since the first observation—165 years ago in the form of a water wave in a canal in Scotland—nonlinear "soliton" (solitary) waves have fascinated scientists. The term soliton was coined in the 1960s as particle-like features such as attraction and repulsion were recognized in these waves. The optical fiber soliton was proposed in 1973 and first demonstrated seven years later. Twenty-five years after the proposal, it now appears that optical solitons in fibers are approaching commercial viability for carrying information. This potential is evidenced by the large number of papers at recent conferences, the increasing number of field trials, and the International Telecommunications Union (ITU) proposal for standardization of soliton systems.
Historically, the key feature of optical fiber soliton pulses has been their intriguing capability to maintain shape and width in the presence of group-velocity dispersion (GVD). This attribute is a consequence of a balance between GVD-induced pulse broadening and intensity-dependent self-phase modulation (SPM). Thus, the amount of power needed to cancel pulse broadening increases linearly with the GVD. An important feature is that solitons are robust in the presence of various perturbations such as loss or nonperfect launch conditions; perturbed pulses will eventually revert to stable solitons.
Canceling the effect of GVD can also be achieved in linear systems using techniques such as periodically placed dispersion-compensating fibers. However, there are other, perhaps more important, features of solitons that make them especially suitable as high-capacity data carriers. For example it is possible to take advantage of the particle-like nature of solitons by using sliding-frequency guiding optical filters along the link. With these centered at slightly increasing frequency along the path, the soliton is capable of following this change without any degradation. The noise generated by in-line amplifiers, on the other hand, is linear and does not follow the spectral shift and thus gets damped out.
Another example is the use of in-line saturable absorbers, which work in the time domain to suppress noise. Of course, there are limitations in soliton transmission such as those caused by nonlinear interaction, the generation of dispersive waves, and various types of timing jitter. The particle-like feature of solitons is also very useful for performing various all-optical functions not related to transmission, such as switching.
IMPACT OF POLARIZATION-MODE DISPERSION
Also important is the fact that solitons tend to stay together in the presence of a walk-off between different polarization components. This polarization-mode dispersion (PMD) is considered a major problem in high-speed transmission because, in contrast to GVD, it is a statistical phenomenon arising from core imperfections and has a slow drift over time that makes it difficult to characterize. As the bit rate increases, PMD becomes an increasingly important limitation. In addition, it is, if not impossible, at least very difficult to compensate for.
If PMD is a major limiting factor of system performance, there is no certainty about error rates. Instead one has to define an outage probability—the likelihood that the error rate exceeds a certain threshold of perhaps a few seconds per year, for example. Polarization-mode dispersion is an important issue in the trade-off of time-domain multiplexing (TDM), bit rate, and number of wavelength-division-multiplexing (WDM) channels to reach terabit-per-second transmission capacity.
With state-of-the-art PMD (about 0.05 ps/÷km), 150 Gbit/s TDM transmission seems feasible in metropolitan areas (for distances of about 400 km). The soliton PMD robustness may thus be a key to success when upgrading existing fiber links to higher speed. While this robustness has been known theoretically for a long time, no experimental observation in a real system environment has been made until recently. To demonstrate this robustness, among other things, soliton field experiments were conducted within the framework of the European Union ACTS project MIDAS (Multigigabit Interconnection using Dispersion compensation and Advanced Soliton techniques; see Fig. 1).
The trial was based on single-wavelength solitons operating at 40 Gbit/s, at which bit rate it was found from simulations that PMD was the main capacity limitation. The results were quite successful, including error-free transmission over the available 400 km. Single-wavelength soliton transmission at 80 Gbit/s was also demonstrated. Error-free operation was achieved with about 1 dB penalty over 172 km (3 spans). This represents the first optical transmission field experiment at single-wavelength bit rates beyond 40 Gbit/s.
In addition, the polarization behavior of the solitons was compared with that of linear pulses. To study propagation of linear pulses, the power was reduced substantially and the dispersion-induced pulse broadening was compensated for with a piece of dispersion-compensating fiber. It was found that, at the system output, the solitons changed their width much less than the linear pulses when the input polarization state was varied (see Fig. 2). This represents the first observation of soliton resistance to PMD in a system transmission experiment. The corresponding system outage is expected to be substantially smaller in the soliton system. Figure 3 explains why solitons are robust in the presence of PMD. The experimental results clearly demonstrate that solitons may become the best choice to enhance capacity when PMD is a major capacity limitation.
Currently, solitons are attractive because of the rapidly emerging strategy of improving the performance of soliton transmission with dispersion management. A dispersion-management strategy—which involves altering the local dispersion between a large positive and a large negative GVD such that the average GVD is small—has long been used in linear systems. It was only recently understood that the same technique, if properly implemented, gives rise to several striking improvements over conventional soliton transmission systems.
While dispersion-managed solitons are clearly nonlinear pulses, they are by no means conventional solitons. Designers with a linear transmission background may thus refer to the dispersion-managed solitons as "nonlinearity mediated return-to-zero pulse transmission in dispersion-managed systems." This is also fine, and the two originally quite different views and strategies (solitons versus linear pulses) have now found a common ground that is a result of a maturing technology.
FIGURE 2. Soliton resistance to polarization-mode dispersion was observed in a field experiment in Sweden, where the temporal pulse profiles of linear pulses (top) and solitons (bottom) were compared after propagation through 400 km of installed fiber. Solid and dotted curves represent minimum and maximum output pulsewidths, respectively; when the input state of polarization of the pulses was varied, the solitons changed their width much less (9.5-11 ps) than the linear pulses (10.5-15.5 ps). (Horizontal scale: 10 ps/div.)
A key feature of these pulses is their strong dynamic evolution along the transmission path—both their temporal and spectral shape/widths change as they propagate. The technical improvements as compared to conventional solitons (for example, using constant GVD) include less Gordon-Haus timing jitter because the system average GVD is much smaller in these systems for the same amount of peak power (thereby also maintaining signal-to-noise ratio) and—in a properly designed system—there is less interaction among neighboring pulses.
An added benefit appears in WDM systems. Because of alteration between large positive and negative GVD along the path, the jitter induced by WDM soliton collisions is greatly reduced, as is the effect of four-wave mixing. This reduction, in turn, allows for very dense WDM by improving the spectral efficiency substantially. The features of conventional solitons such as the PMD robustness and the sliding-frequency technique are still applicable as well. A striking and unexpected feature of dispersion-managed solitons is that they can work for an average net zero GVD and even for normal dispersion, something that would not work with conventional solitons.
Operating at the zero-dispersion wavelength can, in principle, eliminate the Gordon-Haus jitter—other sources of timing jitter, however, will remain. An important practical consequence of using dispersion-managed solitons is that they reduce (or eliminate) the need for in-line stabilizing components such as sliding filters.
From a commercial viewpoint, however, the most important benefit of dispersion-managed solitons is that they, in principle, can be used with already-installed conventional fibers (with zero dispersion at 1300 nm), allowing a much more cost-effective upgrade together with dispersion-compensating fibers or chirped fiber gratings—likely co-located with in-line EDFAs—at certain intervals in the link. This capability means that solitons are no longer just an interesting alternative for transoceanic systems, they are also appropriate for much-shorter-distance terrestrial systems.
A recent and very impressive circulating loop experiment conducted by France Telecom, in which 51 very densely packed WDM channels, each operating at 20 Gbit/s, were transmitted over 1000 km with 100-km sections of standard fiber, clearly demonstrates the strength of the dispersion-managed soliton technique.
The key parameter in dispersion-managed soliton systems is the strength of the dispersion map, which is given by S = (b1l1 - b2l2)/t2, where bx and lx are the GVD and length of the two alternating dispersion sections, respectively, and t is the initial pulsewidth. As S should be approximately in the range 2 to 8 to take advantage of the unique features of solitons, one may note that the use of standard fiber becomes prohibitive at very high bit rates (beyond 40 Gbit/s), as shorter pulses require a more rapidly varying dispersion map to maintain a proper S value. In some respects, this is a limitation similar to the amplifier-spacing limitation in conventional soliton systems, but it is much less severe. If an option is to install new dispersion-shifted fiber, on the other hand, this limitation becomes essentially unimportant.
As for optical source requirements, it is quite clear that the electro-absorption modulator—integrated or not with a distributed-feedback laser—is a useful and simple source for soliton transmission because it provides low-chirp, low-timing-jitter pulses with proper duration and shape. While such sources were developed for linear non-return-to-zero systems, they have now proven to be nearly ideal in soliton systems.
Dispersion-managed-soliton systems have special requirements, however, because the launch condition is different from that for conventional solitons. The pulses should have a linear chirp such that it fits seamlessly in the periodically induced chirp variation along the link. This effect can be achieved by incorporating a proper length of fiber (or chirped fiber grating) in the transmitter once the overall dispersion map is known.
Several soliton field experiments have been conducted over the past few years in Japan by NTT, in the USA by MCI/Pirelli, and in Europe by ACTS projects. This is a good indication that solitons are indeed foreseen as candidates in real systems. Some systems used standard fiber, and some used polarization multiplexing, but none used in-line control. At rates beyond 10 Gbit/s, dispersion-shifted fiber was used. In a few of the studies PMD was found to be a major limiting factor.
The above field experiments are important demonstrations because various fiber parameters are poorer and much less controllable than in the laboratory. The question of how far one can push the capacity based on installed standard fiber is not easy to answer because it depends strongly on the quality of the installed fiber as well as on what can be expected from using dispersion-managed solitons to their limit.
In conclusion, the use of solitons has regained a substantial interest which is mainly attributable to the concept of dispersion-managed solitons, which have reached a level of maturity such that commercialization now seems very near.
All the staff in the fiber group at the Department of Microelectronics, Photonics Laboratory, Chalmers University of Technology, contributed to the success of the MIDAS soliton field trial. Telia provided the installed optical fiber lines.
Peter Andrekson is director of research and systems technology at Cenix Inc., 6575 Snowdrift Rd. Suite 105, Allentown, PA 18106. He can be reached at firstname.lastname@example.org.
This article orginally appeared in Laser Focus World May 1999.