Negative-dispersion fiber in metropolitan networks

Th 81716

The use of negative-dispersion fiber promotes flexibility and low cost for metro applications now and in the future.

ANDREW WOODFIN, IOANNIS TOMKOS, and ADAM FILIOS, Corning Inc.

While advanced metropolitan networks borrow heavily from concepts and practices originally developed and evaluated for their long-haul cousins, there are significant differences between the two application spaces. Among these is the necessity to support a broad spectrum of services and data rates in the metro while keeping costs as low as possible.

The system houses that have been successful in the metropolitan market have obviously been cognizant of these different considerations, as evidenced in their felicitous choices of system architectures, components, and performance specifications. As for manufacturers of optical fiber, however, most have entered the metropolitan environment late with fibers designed and intended for use in the decidedly different long-haul space. However, metropolitan-optimized fibers specifically engineered to address the burgeoning demand for metropolitan bandwidth and the associated complexity and cost have now appeared.Th 81716

Figure 1. Dispersion characteristics of negative-disperion fiber, standard singlemode fiber (ITU G.652)/low water peak fiber, and non-zero dispersion-shifted fiber.

The dispersion characteristics of various fiber types are illustrated in Figure 1. Negative dispersion fiber (NDF), such as Corning's MetroCor fiber, is notable in comparison to conventional non-zero dispersion-shifted fiber (NZDSF), including Lucent's TrueWave-RS fiber, and ITU G.652-compliant standard singlemode fiber for its zero-dispersion wavelength outside the usable bandwidth regime (1280-1625 nm), consequently displaying consistently negative dispersion across the full usable spectrum.

A significant consideration in metropolitan DWDM networks is the total transmission path that can be satisfied by system optics without resorting to the cost and complexity of either electrical regeneration or dispersion compensation. The common and outdated misconceptions that metropolitan/regional networks are almost exclusively small in scale (e.g., <100 km) is clearly refuted, for example, by the increasing adoption of extended-reach OC-48 transmission optics into metro system houses' product lines.

This extended-reach capability comes at significant cost increases, however. At higher data rates (such as OC-192), current reach capability over standard singlemode fiber still falls well short of that achievable at lower speeds. It's important to be mindful that the shortest-reach transmitter dictates the need for costly electrical regeneration and/or dispersion compensation in a system. In a mixed-rate DWDM system, necessary compensation for the higher-data-rate traffic effectively nullifies the advantage otherwise promised by lower-speed extended-reach transmitters.Th 81717

Figure 2. Experimen tally measured eye-diagrams of transmission over 300-km standard singlemode fiber (a) and 300-km negative-dispersion fiber (b).

Although the cost model for dispersion-compensation modules is more attractive than that for electrical regeneration, their introduction nevertheless introduces significant cost into the fiber plant. Furthermore, their use adversely affects noise figures, as well as introduces considerable complexity in DWDM configurations where multiple signals on a single fiber can encompass a wide range of transmission distances due to origination at varying locations on the interoffice ring.1

The most cost-effective solution to enable longer-reach (>200 km) metropolitan systems is to leverage the inherent characteristics of the lowest-cost optics, thereby foregoing expensive transmitters, regeneration, and dispersion compensation. With NDF, the interplay between the fiber's negative dispersion and the positive chirp fundamentally inherent in all directly modulated distributed feedback transmitters (DFB DMLs) and VCSELs dramatically extends the signal reach by compressing the signal pulses as they propagate.

A general comparison of transmitter type characteristics is detailed in Table 1.Th 81722

A chirped signal effectively has a time-varying change in its central wavelength over the duration of the pulse, which will effect distortion in the overall pulse shape as the corresponding frequency components at different points travel at different effective velocities. This phenomenon is shown in Figure 2, with measured results indicating positive chirp interaction with positive dispersion (standard singlemode fiber) and negative dispersion (NDF).

In laboratory evaluations of nearly 100 randomly selected off-the-shelf C-band (1530-1565-nm) OC-48 DMLs from multiple manufacturers, all lasers easily satisfied a reach exceeding 300 km over NDF, with a significant percentage surpassing 600 km.1 These transmitters were all specified by the manufacturers for 80- to 100-km reach over standard singlemode fiber.

In the L-band (1565-1625 nm), NDF shows an even more pronounced reach advantage due to a smaller magnitude of dispersion, while standard singlemode fiber and NZDSF each have increasing values of positive dispersion at the higher wavelengths (Figure 1). Any standard-reach OC-48 DWDM transmitter will show at least a 3x to 4x reach improvement transmitting over NDF compared to standard singlemode fiber, without the need for additional inventory screening. Experimental evaluation of multichannel transmission incorporating 32 OC-48 DML channels has also demonstrated performance beyond 300 km over NDF, results unattainable with standard singlemode fiber (Figure 3).1Th 81718

Figure 3. Comparison of measured Q-factors after transmission of 32 2.5-Gbit/sec directly modulated lasers over 300 km of negative-dispersion fiber and standard singlemode fiber.

Extended-reach DMLs at 2.5-Gbit/sec rates are now beginning to emerge as an option, though at a significantly higher cost. These transmitters also perform considerably better over NDF. However, the fact that standard-reach DMLs offer ample reach at the lowest possible cost obviates the need for these extended-reach devices.

Nascent development of long-wavelength VCSELs in both the C- and L-bands further exploits the advantage of NDF, where significant cost reductions enable improvements in economics and system performance. Initial evaluations of 2.5-Gbit/sec C-band VCSELs have shown uncompensated reaches over NDF in excess of 800 km.2

Thermally induced performance variability of DWDM DML transmitters in concert with NDF is negligible, as all DWDM sources are cooled to ensure wavelength stability in compliance with ITU-grid specifications. Aging-induced effects on transmission over NDF are expected to be no more significant than those in an uncompensated standard singlemode fiber network.

With higher-data-rate DWDM transmitters (e.g., OC-192), NDF maintains a consistent advantage in extended reach. The approach to 10 Gbits/sec taken by metro system houses to date has been either through the use of electroabsorption modulated lasers (EMLs) or externally modulated LiNbO3 (Ex-mods). Both transmitter types can be operated in two regimes, where the chirp is biased to zero or is optimized for the particular fiber type deployed. In either instance, NDF significantly outperforms standard singlemode fiber, through the combination of low and negative fiber dispersion. The reach advantage in NDF over standard singlemode fiber ranges from 50-100%, depending on the transmission approach chosen by a particular system house.

A significant development offered by using NDF as the transmission medium is the ability to deploy directly modulated 10-Gbit/sec transmitters for reaches comparable to those achieved by the best externally modulated sources over standard singlemode fiber. As with OC-48 DMLs, the interplay of inherently positive chirp and low negative fiber dispersion allow for significant increases in uncompensated reach, here extending to an improvement of 8x to 10x.3

The true advantage lies in the significant cost reductions enabled by the relatively simple design of a DML, along with other traits detailed in Table 1. That is of particular interest in considerations of 10-Gigabit Ethernet, where the promise of low-cost transmission and maintenance can be further bolstered by utilizing significantly lower cost optics to easily meet and exceed standardized 40-km reach re quirements.4 Additionally, the inherently higher output power afforded with 10-Gbit/sec DMLs can enable significant improvements in required optical signal-to-noise-ratio (OSNR) values when compared with external-modulation options. Th 81719

Figure 4. Comparison of uncompensated reach with varying transmitter types over standard singlemode fiber (ITU G.652) and negative-dispersion fiber (NDF) within the C-band. (L-band performance improves with NDF, degrades with standard singlemode fiber.)

An overview of uncompensated-reach comparisons between NDF and standard singlemode fiber with all transmitter types is summarized in Figure 4.

With serial 1310-nm distributed-feedback transmitters operating at OC-48 rates, NDF also demonstrates beneficial pulse compression performance similar to that in the C- and L-bands (Figure 5). Higher fiber attenuation at 1310 nm (0.4-0.5 dB/km), inherent in all available fibers, sets a universal limit on 1310-nm reach capability. VCSELs at 1310 nm should perform similarly.

Again reflecting a long-haul-focused mindset, discussions around non-linear effects in the metro in support of NZDSF solutions are often misleadingly couched within the context of transmitter launch powers not commonly seen even in long-haul configurations with larger span losses and tighter requirements on noise considerations. These analyses bring into question maximum allowable channel counts and minimum channel spacing.

Properly considered treatments of metropolitan networks' non-linear effects, with shorter extent and closer spacing of nodes, do not map directly from the (often misrepresented) long-distance transport regime. The minimum required per-channel launch power (Pchan) in an amplified optical network can be calculated easily using the widely accepted formula below, where the driving consideration is acceptable OSNR:5

OSNR=58 dB+Pchan-NFEDFA-Lspan
-10log10(Nspan).

Assuming a 200-km ring with six equidistant optical add/drop multiplexer (OADM) sites and typical values shown in Table 2, plus a span-loss allowance of 6 dB for additional system impairments, required minimum per-channel power is roughly 0 dBm. Indeed, optical powers in DWDM systems designed for metropolitan environments are typically on the order of 0 to 3 dBm per channel, sufficiently low to avoid significant non-linearity-induced degradation over NDF. Emergence of forward error correction in some metro systems enables marginal increases in uncompensated reach but more significantly relaxes OSNR requirements, allowing for larger NDF-based networks with additional add/drop site capacity while satisfying noise requirements.Th 81720

Figure 5. Perform ance of 1310-nm distributed-feedback transmitters over negative-dispersion and standard singlemode fiber.

NDF has reduced dispersion in the infrequently deployed L-band and thus has greatest sensitivity there to four-wave mixing (FWM) effects. NZDSF has reduced dispersion in the universally deployed C-band, thereby increasing its susceptibility to FWM penalties within the most commonly used metropolitan DWDM transmission window. As discussed, the effect of FWM on system performance with NDF is negligible, as power levels are commonly low. Indeed, NDF sees significant transmission reach improvement within the L-band due to reduced dispersion, ensuring in creased compatibility with future operating regions. NZDSF has higher dispersion in the L-band and thus an associated decrease in uncompensated reach for all transmitter types.

Full-spectrum (1280-1625-nm) coarse WDM (CWDM) continues to foster growing interest as a cost-effective means by which to enable efficient bandwidth usage without the complexity and tight tolerances on optics associated with higher-capacity DWDM systems. The ability to have future CWDM compatibility built into all segments of the network reflects intelligent and proactive planning.

Deploying NDF with reduced water peak attenuation (<0.4 dB/km at 1383 nm) ensures compatibility with both CWDM and DWDM implementations. NZDSF, as shown in Figure 6, is wholly incompatible with full-spectrum CWDM due to very high attenuation (1 dB/km) within the E-band, thereby eliminating future consideration of CWDM in any network containing conventional NZDSF.6

Raman amplification has been speculated as a potential development in the metro, but attention given to it by metro system houses has been to date minimal at best. NDF is nevertheless well prepared for Raman by displaying markedly improved Raman efficiency compared to both standard singlemode fiber and NZDSF.Th 81723

Regardless of fiber type, universally short dispersion-limited reaches at 40 Gbits/sec will necessitate compensation at intervals likely to be smaller than common internodal spacing, significantly complicating installation and maintenance of facilities. The costs and technical challenges associated with high-speed electronics, the complexity of compensation for both chromatic and polarization-mode dispersion, and the concurrent increase in network design complexity present pronounced hurdles to acceptance within the metropolitan environment.

Nevertheless, mindfulness of compensation requirements, should 40 Gbits/sec become a viable metropolitan option, NDF will be positioned for future capability regardless of the direction in which systems evolve. Building on the concepts and mindset already in place within metropolitan system houses, taking a banded wavelength approach to compensation maintains consistency with the banded OADM approach used in most systems.1Th 81721

Figure 6. Attenuation characteristics of non-zero dispersion-shifted fiber, negative-dispersion fiber, and low-water peak fiber.

The incorporation of NDF into a DWDM metropolitan network offers unparalleled performance advantages, us ing the lowest-cost optical components available. Considerable savings can be realized by reducing component cost, eliminating the complexity and management requirements of engineering compensated links, and eliminating the need for regeneration and/or dispersion compensation.

Due to the disparate application spaces subsumed within the "metropolitan" moniker, from access to core to regional, the utility of a dual-fiber cabled design approach in addressing the varying requirements in some networks is obvious. The decision remains as to whether a hybrid cable is the best means by which to enable maximum system capacity and lowest cost on an individual basis, but the larger issue is the determination of which fibers are to be included within the cable.

In a dual-fiber solution, standard singlemode fiber or a more versatile low-water peak fiber each present an easy choice for flexibility and integration. The primary consideration, though, is determining the fiber best suited for implementation in the high-capacity and ultra-cost-conscious core and regional segments (80 km to >300 m). The obvious choice is the fiber designed to extend the capability of existing and forthcoming optical technologies to extend transmission reach, minimize equipment costs, and reduce (or altogether eliminate) complexities otherwise necessary to mitigate degrading effects. A metro-optimized negative-dispersion NZDSF meets these requirements. Following are a few of the existing and forthcoming issues addressed with NDF:

  • Significant uncompensated-reach advantage (3x-4x) with all off-the-shelf inexpensive directly modulated lasers at 2.5 Gbits/sec.
  • Significant uncompensated-reach advantage (1.5x-2x) with off-the-shelf and optimized externally modulated transmitters at 10 Gbits/sec.
  • Ability to viably deploy low-cost directly modulated transmitters at 10 Gbits/sec with significant performance ad vantages (8x-10x) compared to standard singlemode fiber.
  • Compatibility with legacy serial 1310-nm DFB transmitters and VCSELs.
  • Compatibility with long-wavelength fixed and tunable VCSELs.
  • Adaptability to DWDM and CWDM applications with reduced attenuation at the water peak.

Service providers installing NDF in their networks position themselves to maximize revenues by reducing costs and management complexity. In anticipation of future technological trends, they can be comfortable with the knowledge that their investment in NDF will ensure maximum flexibility and versatility.


Andrew Woodfin is strategic alliance manager, metro/access, for Corning Inc. (Corning, NY). Ioannis Tomkos and Adam Filios are senior research scientists for Corning's Somerset, NJ, office. Woodfin can be reached by e-mail at woodfinad@corning.com or by phone at 607-974-4961.


  1. I. Tomkos, et al., "Demonstration of Negative Dispersion Fibers for DWDM Metropolitan Area Networks," IEEE Journal of Selected Topics in Quantum Electronics, May/June 2001.
  2. I. Tomkos, A. Filios, "Transmission Performance of 1.5-m, 2.5-Gbit/sec Directly Modulated Tunable VCSEL over Negative Dispersion Fiber," OFC 2002 (pending publication).
  3. I. Tomkos, et al., "Transmission of 1.55-µm 10-Gbit/sec directly modulated signal over 100 km of negative-dispersion fiber without any dispersion compensation," OFC 2001, TuU6, 2001.
  4. IEEE draft 802.3ae, "Media Access Control (MAC) Parameters, Physical Layer, and Management Parameters for 10-Gbit/sec Operation," August 2001.
  5. J. Zyskind, et al., I. Kaminow and T. Koch, ed. "Optical Fiber Telecommunications IIIB," 1997, Academic Press.
  6. 2492C, "Lucent TrueWave Single-Mode Fiber Product Information," Issue 8, June 2000.
More in Network Design