Not all optical fibers were created equal
Network designers now have several choices of fiber types that have been optimized to meet the needs of their networks today and position them for tomorrow.
Patrick Emery and Carol Cowley
Lucent Technologies
Until very recently, fiber-optic transport networks were composed of conventional singlemode fiber. Regardless of whether a carrier was building a long-distance or metropolitan network, when it came time to pick the fiber, the choice was simple. In the last several years, transmission systems have moved to higher bit rates, more wavelengths, longer lengths, and greater power levels. Optical-fiber technology began to evolve to keep pace with these changing transmission innovations.
The days of one-fiber-fits-all networks are gone. New breakthroughs in optical-amplifier and optical-fiber design and technology have resulted in new types of fibers with carefully optimized characteristics. While carriers certainly don`t want to manage a proliferation of different fiber types in their networks, they also don`t want to limit themselves--either in terms of capacity or ease of expansion. Therefore, network designers need to take a close look at the types of fiber and transport technologies available before designing a network that will keep pace with evolving technology and remain cost-competitive for years to come.
Network planning for capacity demand was never an easy job, but the complexity of this function has risen to unprecedented levels in recent years. The reason is simple: Voice traffic no longer dominates our traffic flow. This factor is unfortunate from the planning perspective because voice traffic has a nice characteristic--it`s fairly predictable, since it is based on tele-densities, call durations, and calling patterns. Data traffic, the new dominant traffic form, is significantly less predictable. For example, how many planners actually projected the data capacity impact of the Internet? Who foresaw the magnitude of electronic mail and data transfers? Ten years ago everyone was predicting that video would be the major capacity driver, but video today is an insignificant amount of total traffic.
How important is the impact of network capacity in designing future systems? Consider one fact: Fiber placed in networks today will still be in place in 2023 and beyond. What type of data or video applications will influence capacity demand? One thing can be said with certainty--network capacity will be needed for applications we haven`t even considered. System operators must therefore factor the unknown into their network design or upgrade equation. The fiber selected today must be capable of cost-effectively handling data rates that are multiplying every year, be sufficiently flexible to accommodate network growth to support different service options, and forward compatible with several generations of transmission equipment encompassing more wavelengths or different operating ranges.
The question is, which fiber is best suited for which application? Because the requirements for fiber for long-haul networks are very different from those for metropolitan networks, the goal should be to choose the fiber design that has been optimized for a carrier`s particular application and evolution strategy.
Long-haul networks commonly operate at distances of at least 100 km with little or no need to add or drop traffic. Voice and data traffic are collected in large nodes in one city and transported to nodes in other cities via large-bandwidth pipes. To achieve the lowest possible cost, multiple wavelengths (up to 40) frequently are transmitted over the same fiber via dense wavelength-division multiplexing (dwdm) and operate at the highest possible bit rate per wavelength. Speeds of 10 Gbits/sec are common in the United States today for new networks. Such systems are expected to cost up to 60% less per bit than alternative 2.5-Gbit/sec multiple-wavelength systems.
The fiber best suited for long-haul conditions is one in which dispersion is kept within a tight range, being neither too high nor too low. If the dispersion is too high, pulse distortion becomes excessive and expensive, and complex dispersion-compensation techniques are required. Too little dispersion also is undesirable because it allows four-wave mixing and cross-phase modulation interactions between signals when multiple wavelengths are used.
New fiber development
The ability of fiber manufacturers to alter dispersion has led to a number of important fiber developments. In the early 1980s, the first generation of singlemode optical fibers had zero fiber dispersion at 1310 nm (also known as the second operating window), with lasers optimized to operate in this wavelength range. Photonic research progressed and systems were designed for the third spectral window, a region around 1550 nm, where fiber loss was at its lowest and erbium-doped fiber amplifiers (edfas) could effectively operate and extend transmission distance to many hundreds of kilometers without electrical regeneration. The problem with this third operating region was that the dispersion of the first-generation singlemode fiber was so large that long-length, high-speed transmission required significant compensation for high levels of accumulated dispersion.
This led to the development of dispersion-shifted fiber (dsf), which had a 1550-nm zero-dispersion wavelength. Even though dsf required more-sophisticated manufacturing technology, this fiber was used in applications requiring single-wavelength transmission over long lengths, such as in undersea cables.
When dwdm emerged as the solution for extending the capacity of a single fiber, dsf quickly became outdated, since its low level of dispersion created crosstalk. Fiber design for long-haul routes now had different imperatives: dispersion low enough to need minimal or no compensation, yet high enough to suppress crosstalk. This realization led to the invention of non-zero dispersion fiber (nzdf).
Opening a fourth window
Of course, technology doesn`t stand still. The next development, which is still unfolding, is the opening of the fourth operating window. As the result of improvements in edfa technology, it is becoming technically feasible to use a broader wavelength band, which will increase capacity. Whereas early edfas had relatively narrow operating windows--at most from 1540 to 1560 nm--today`s edfas have much wider, 35-nm, windows that range from 1530 to 1565 nm. Research into L-band (ultra wide) amplifiers that extend the upper wavelength to 1620 nm promises even greater operating ranges within the fourth operating window.
Fiber design imperatives, therefore, have changed yet again. Now dispersion must not only be low enough to minimize compensation and high enough to suppress crosstalk, but it must also be controlled enough to work with the new L-band amplifiers. Since the dispersion value changes as the wavelengths change, the control of dispersion refers to minimizing the dispersion slope (S0) of the fiber. Different fiber designs can have large differences in S0 even though their nominal dispersion at a particular wavelength is the same. When fibers are used in a long transmission system, the effect of the dispersion slope can be substantial. Because of dispersion slope, the difference between dispersion values at the edge channels (1530 and 1565 nm for the C-band, for example) will grow with distance, eventually causing a wide enough difference in fibers with high dispersion slope that dispersion compensation becomes exceedingly complicated and expensive.
Controlling or decreasing the dispersion slope has two major advantages. It increases the power that can be used in wavelengths at the lower end of the wavelength band, and it decreases the cost and complexity of dispersion compensation in longer systems in both the C- and L-bands. nzdf fibers with high dispersion slope tend to have too little dispersion in the lowest wavelengths around 1530 nm to reliably suppress crosstalk. These fibers keep dispersion low in this range to avoid having too much dispersion at the higher end of the wavelength range. Depending on the design of the transmission equipment (power levels, channel spacing, and other factors), this lower dispersion can limit the power (thus the distance) usable in the low wavelengths.
High dispersion slope has two major effects on dispersion-compensation cost for nzdf fibers: First, when compensation is required in long networks, use of a single compensation module to compensate the entire band causes over-compensation at one end of the band and under-compensation at the other end. Thus, each band may have to be split into two or more parts for accurate compensation. This increases costs by requiring more or different compensation modules. Second, for wavelengths in the L-band, the magnitude of dispersion in high-slope fibers requires more-frequent dispersion compensation and more-expensive modules.
Accommodating metropolitan growth
Metropolitan networks have requirements significantly different from those of long-haul networks. Distances in these networks are typically less than 80 km, so optical amplification is used infrequently, and dispersion is not the primary concern. Since these networks are designed to support large end-users, the adding and dropping of traffic is common. Positioning these networks for future evolution and data capacities will involve systems that can handle possibly hundreds of wavelengths, and in many instances, support a strategic evolution to long-distance services.
What fiber characteristics are best suited for these metropolitan networks? The ideal fiber for these networks is one that would enable metropolitan carriers to use the entire wavelength region from 1280 to 1625 nm. Current singlemode systems operate in either the 1310-nm window (1280 to 1325 nm) or the 1550-nm window (1530 to 1565 nm). The wavelength region between 1350 and 1450 nm has not been used because of higher attenuation over much of this region. This higher attenuation is the result of an intrinsic characteristic of silica glass, in which the presence of water, generally incorporated in the glass in the production process, leads to the absorption of light in the 1385-nm region. A manufacturing process whereby the incorporation of water into the fiber is greatly reduced has recently been developed. The result is a fiber with essentially only the attenuation intrinsic to pure silica glass, which makes it ideal for metropolitan applications.
The performance of such metro-optimized fiber is identical to that of singlemode fiber in the 1310- and 1550-nm regions. Such fiber is matched clad with identical splicing and operational characteristics. It also uses the same transmission equipment. The difference is that metro-optimized fiber opens a previously unusable window in the fiber spectrum, which provides several significant customer benefits. For example, metro-optimized fiber has the potential to
carry significantly more traffic,
support longer distances and higher bit
rates,
enable more-sophisticated network
management approaches,
create a more economical end-to-end
solution while maintaining compatibility with all existing transmission equipment.
Because the fiber spectrum usable for dwdm transmission is increased by 100 nm (the 1350- to 1450-nm region), more spectrum is made available to carriers. This equates to 150 or more new wavelengths (at 100-GHz spacing) that can be available for dwdm applications. To provide this same capacity, two to four times as many conventional fibers would have to be used. Just assuming twice as much fiber would be required, the incremental cost of using conventional fiber versus metro-optimized fiber would be doubled.
The window opened up by metro-optimized fiber has exciting, system-enabling characteristics. In this 1400-nm region, metro-optimized fiber has lower loss than conventional fiber in the 1310-nm region. Considering that dispersion in the 1400-nm region is about 50% lower than dispersion in the 1550-nm region, unique low-loss and low-dispersion transmission options become available.
The additional spectrum provided and the unique qualities of the 1350- to 1450-nm region also enable a "banding" approach to network management, where different service types can be grouped together on the same fiber and allocated to the wavelength bands where they are most suited. For example, one fiber might carry wdm analog video in the 1310-nm region, high-bit-rate data traffic (up to 10 Gbits/sec) in the 1350- to 1450-nm region, and lower-speed dwdm traffic (up to 2.5 Gbits/sec) in the region above 1450 nm. Then, management systems can be tailored to monitor the specific services as if they were operating on separate fiber systems while gaining the economies provided by using a single fiber.
Finally, the extra spectrum opened by metro-optimized fiber could enable the use of less-expensive lasers and other components. dwdm signals, for example, could be spread over a broader range of wavelengths, allowing the use of directly modulated lasers instead of the more expensive externally modulated lasers currently needed in systems with closely spaced wavelengths. Cost reductions in other component areas, such as multiplexers, demultiplexers, and wavelength add/drop devices, also can be expected when metro-optimized fiber is used.
Making an informed decision
In today`s highly competitive telecommunications environment, carriers have no choice but to build the lowest-cost network possible if they hope to compete successfully. Yet, at the same time, they need to find a way to stay ahead of the technology curve. The good news is that the evolution of optical fiber is more than keeping pace with the rest of the industry. Choosing the fiber that is best suited for a particular application, however, is no longer a simple, straightforward task.
A network`s inherent characteristics must be taken into account before an informed decision can be made. Long-haul applications, for example, usually require an infrastructure that uses dwdm with high-bit-rate signals and a limited number of wavelengths where traffic is transported from point A to point B. Metropolitan applications, in contrast, use lower-speed signals and hundreds of wavelengths although they, too, operate over a dwdm infrastructure. Thus, we have two very different applications that require two different types of solutions (see figure).
For long-haul networks, nzdf and 10-Gbit/sec systems together are widely recognized as delivering the lowest cost per gigabit in the industry. By the end of 1998, half of the embedded base of fiber in U.S. long-distance networks is projected to be nzdf. And, just as long-distance carriers have recognized that using a fiber that is optimized for their network`s particular characteristics is advantageous, metropolitan carriers are likely to find that metro-optimized fiber will be unsurpassed in cost-effectively meeting their present and future capacity needs. u
Patrick Emery is technical manager, fiber-optic systems engineering, for Lucent Technologies, Bell Laboratories (Holmdel, NJ). Carol Cowley is offer manager for fiber-optic systems for Lucent Technologies (Warren, NJ).