The telecom bust of the past two years is forcing a sea change in the fiberoptics industry. Traditionally, the industry has put a premium on performance, and sought to build the biggest, fastest, and longest information pipelines. Developers succeeded all too well, and learned the hard way that you can have too much bandwidth. Long-distance carriers laid so much high-capacity fiber that most of it sits dark and empty. Like the host of a boisterous New Year's party, the post-bubble fiberoptic industry is waking up the morning after with a throbbing hangover in the midst of a big mess.
Cutting-edge research on long-haul, high-capacity systems is not going away. The postdeadline sessions at the European Conference on Optical Communications in early September included four reports of transmitting more than one terabit per second through a single fiber for distances of more than 1000 km. Other papers described high-speed and high-capacity switches, lower attenuation in photonic bandgap fiber with both solid and hollow cores, new types of fiber with low dispersion slope, and an erbium-doped fiber amplifier with output above one watt.
Yet commercial fiberoptic technology has turned in a different direction, evident on the sparsely populated show floor at the National Fiber Optic Engineers Conference in mid-September. Coarse wavelength-division multiplexing (CWDM) is "hot," not because it packs the most bits per second through a single fiber, but because it offers the most bandwidth per dollar. Broadband services are hot because they represent a real market for local telephone and cable-television companies. Tunable lasers are starting to work their way into real systems because they promise real economies in operating and capital expenses. In contrast, companies have shifted development of 40-Gbit/s technology into low gear because carriers aren't buying.
The appeal of CWDM is its ability to cut terminal costs by relaxing tolerances on transmitters and optics. CWDM channels are 20 nm wide, compared to 0.8 or 0.4 nm for DWDM channels (see Fig. 1). Distributed-feedback lasers require temperature stabilization in DWDM systems because their temperature drifts by approximately 0.08 nm/°C, but the 4-nm variation that comes with a 50°C change fits easily within a 20-nm CWDM window. The wider channels also allow the use of simpler multiplexing optics, including thin-film filters with about one third as many layers as needed for 100-GHz channel spacing. Eliminating the need for transmitter cooling and simplifying optics both help reduce costs.
The International Telecommunications Union already has standardized under Recommendation G.694.2 a CWDM grid of 18 wavelengths spaced 20 nm apart from 1270 to 1610 nm (see Fig. 2). The standard specifies transmission distances up to 50 km, and architectures including hubbed rings, point-to-point links, and passive optical networks (PONs). Each wavelength carries a separate signal that can be in a different format from other channels on the same fiber.
Those specifications put CWDM technology squarely into the range of metro and access networks, where most links are 50 km or less, and keeping terminal costs down is much more critical than in long-haul systems. The short distances greatly reduce dispersion problems. Although developers are talking of 10-Gbit/s transmission with CWDM systems, most current metro networks transmit at 2.5-Gbit/s or below, which also eases dispersion requirements.
Long-haul systems are limited to the erbium-doped fiber amplifier window by the need for optical amplification, pushing the use of DWDM. In contrast, CWDM systems normally would not require amplification, so signals could be transmitted throughout the low-loss window from 1270 to 1610 nm. Both Corning and OFS are promoting low-water fibers as offering low-loss transmission throughout that entire band.
However, the water absorption peak near 1380 nm is not an insurmountable barrier. Most standard step-index single-mode fiber has attenuation of no more than 1.5 dB/km in that band, allowing transmission over 10 to 15 km. That gives carriers the important option of using existing fibers in the water peak for shorter links, such as between adjacent central offices in metropolitan areas. Amplification is possible at wavelengths at which suitable amplifiers are available, giving added flexibility.
Developers say that the use of simpler optics and uncooled transmitters should reduce terminal costs of a CWDM system to about a third that of a DWDM system. Two years ago they formed the Full Spectrum CWDM Alliance to push standards. Their work isn't finished, but commercial products were on the floor at NFOEC from companies including Tsunami Optics. Padtec, a company based in Campinas, Brazil, demonstrated a CWDM system running over OFS fiber at the OFS booth. The hope is that CWDM will be the right technology in the right place at the right time to help grow the metro and access markets.
The state of the market for broadband Internet access is a matter of perception. Growth is well below projections by many market analysts, and Wall Street is unhappy. Yet the number of broadband subscribers continues to grow at a time when the overall economy is uncertain. The United States is in fourth place, far behind the leader, South Korea, said Greg Beck, vice president of network engineering for SBC Communications, in his plenary talk at NFOEC (see Fig. 3). A recent study says that more than half of South Korean households now have broadband service.
Cable modems lead the U.S. broadband, reaching 65% of customers, followed by DSL at 33%, and satellite and wireless terrestrial services at 1% each, Beck said. He blamed the slow spread of DSL on government regulations for sharing phone lines as well as on limited customer demand. Yet it's also clear that local phone companies share the blame, both for inadequate response to customer inquiries, and for limited network capacity.
Fiber provides infrastructure for all broadband technologies except satellites, but running fiber all the way to the home promises even greater bandwidth. Actual system installations are few and far between, but growing. As of July, the Fiber to the Home Council had counted 50 sites with fiber installations to some or all residences. Most are small towns or new developments. The council reports that the number of homes with working fiber links more than tripled (a 200% increase) over the preceding year. Most are in two market niches—new large-scale developments where utilities must be installed from scratch and rural communities where transmission lines must run long distances.
Many installations are by small companies, but large carriers are getting into the game. By the end of 2002, Beck said, SBC will have finished installing a broadband passive optical network (B-PON) in the Mission Bay community in San Francisco, CA. Beck likes the B-PON approach because it allows users to share a fiber network, cutting costs.
The biggest cloud on the broadband horizon could be efforts by the entertainment industry to block individuals from copying and sharing digital sound or video files. The industry complains pirates are robbing them, but critics say the proposals would block "fair use" such as time-shift recording of television programs. Sharing of music files accounted for much of recent Internet growth, so blocking sound and video sharing could reduce demand.
The big trend on the components front is the emergence of commercial tunable lasers. Sean Maloney, executive vice president and general manager of Intel Components, showed one in a well-received NFOEC plenary talk. Intel's temperature-tuned laser transmitter is limited to the 3.2-nm range of eight DWDM optical channels spaced 50 GHz apart, and requires about a minute to switch channels. Yet it can operate at 10 Gbit/s, and the switching time is fine for applications such as provisioning wavelength services.
Critically, system makers are starting to adapt tunable lasers. Lucent is building tunable lasers from iolon into its Lambda-Xtreme system. Each laser can be tuned across 16 optical channels spaced 50 GHz apart, so the eight lasers can cover 128 possible channels. Lightscape Networks has built a more broadly tunable laser from Agility Communications into a tunable line card for metro networks. Users are now testing the cards, which can be tune across 100 optical channels in the erbium amplifier C-band.
Although encouraging, the initial applications of tunable lasers are limited to DWDM systems in which channels are closely spaced, wavelength stabilization is necessary, and laser prices are high. Outside of the metro market, demand for those systems is limited now. Yet by simplifying logistics and reducing operating costs for both system makers and carriers, tunable lasers can carve themselves an important niche in the marketplace.
Another interesting component trend is the emergence of optical fibers with 80-µm cladding diameters, shown at NFOEC by both Corning and OFC. The earliest uses of small-cladding fibers were in couplers, but developers now report interest in small fibers to improve packing density of components, and to allow smaller bending radii in devices such as fiber amplifiers.
The biggest sore points in the fiberoptic outlook are in areas that were particularly hot during the bubble—40-Gbit/s transmission and new system designs.
The reality is that virtually no existing equipment runs at 40 Gbit/s and the demand for cramming data through fibers at the highest possible speeds has evaporated. In one sense, that's good news, because it gives developers much-needed time to deal with tough technical issues. Yet many companies have scaled back 40-Gbit/s development, and it is far from certain that carriers will deploy 40-Gbit/s systems in the near future.
Indeed, uncertainty is the watchword on the frontier of new system technologies. During the bubble, new and established system vendors pushed a plethora of new approaches and protocols for switching and transmission. All claimed their systems were more cost-effective and efficient than the venerable SONET/SDH standard. Yet established carriers were wary of change, preferring to go slow in network renovation.
That hasn't changed in the face of steep cuts in capital spending. The carriers that remain solvent want to cut operating expenses, but are confronted with a bewildering variety of system technologies. With no clear winners yet emerging from the pack, most carriers are sticking with the tried and true, or putting off purchases until system trends become clearer.
Indeed, if there's a single overarching trend in the whole fiberoptic industry, it's a tendency to wait and see what happens. That marks a dramatic change from the madness of the bubble years, when the mantra was "don't just sit there, do something." The future should benefit from the extra thought.
Jeff Hecht is contributing editor to WDM Solutions. He can be reached at firstname.lastname@example.org