Wavelength Division Multiplexing
Bandwidth in communications is like closet space in your home-you can never have enough. And Internet traffic is making the demand for communication capacity grow faster than the wardrobe of a teenager with a no-limit credit card. Bandwidth-hogging megabytes of animated graphics are replacing compact e-mail messages. Data, video and voice signals crowd transmission systems that had ample space just a few years ago. The communications industry needs room to breathe.
That’s exactly what a new generation of fiber-optic technology is bringing to networks such as the aptly named Project Oxygen. Neil Tagare, founder of the CTR Group in Woodcliff Lake, N.J., picked that name for the global network because he considered the tremendous bandwidth offered by the new technology to be as vital for telecommunications as oxygen is to life itself. By sending signals at 16 different wavelengths through each of four pairs of optical fibers, Project Oxygen will carry 640 gigabits per second (Gbit/s) across whole oceans. That’s the equivalent of 10 million simultaneous telephone conversations-enough for every person in Hungary or Belgium to call the United States at the same time.
The technology that makes this new bandwidth possible is called wavelength division multiplexing, or WDM, and it represents the second major fiber-optic revolution in telecommunications. The first came during the 1980s, when telephone companies laced the United States and other countries with fibers to create a global backbone of information pipelines that could carry vastly more data than the copper wires and microwave links they replaced. WDM takes this advantage a giant step further-multiplying the potential capacity of each fiber by filling it with not just one but many wavelengths of light, each capable of carrying a separate signal.
Wavelength division multiplexing has emerged “quite conveniently, as older fiber cables were getting filled,” says Richard Mack, vice president at KMI Corp., a Newport, R.I.-based market analyst firm specializing in fiber optics. Taking advantage of WDM, long-distance carriers such as AT&T and MCI have been able to avoid laying expensive new cables; instead, they simply pump additional wavelengths through existing fibers.
The WDM revolution has arrived with unanticipated swiftness. A decade ago, Mack points out, “people said there was a glut of fiber capacity.” To allow room for expansion, phone companies had laid cables containing 24 to 36 fibers, many held in reserve as “dark fibers.” Each fiber carried hundreds of megabits per second at a single wavelength. Since then, carriers have raised data rates to 2.5 Gbit/s and lit most of the dark fibers. But the tremendous increase of traffic has crowded these cables that once seemed so voluminous. The closets, it seems, are rapidly being packed to the rafters-and stuff is spilling out onto the floor. Telephone usage accounts for some increase, including the spread of fax machines and mobile phones. But the most dramatic growth has been from Internet traffic, which roughly doubles each year.
What’s also clear is that there’s no end in sight to the soaring demand, especially if, as many experts believe, two-way video communication becomes more common. “The communications industry is undergoing a transition that in a few years shall bring us digital video for our everyday use at home and at work,” says Shahab Etemad, who heads WDM transmission development at Morristown, N.J.-based Bell Communications Research, or Bellcore. (Initially the research arm of the local and regional phone companies, Bellcore now operates as a network management consultancy with a variety of corporate clients.) Etemad expects the change from voice telephony to digital data heavy with video to require multiplying backbone transmission capacity by about a factor of 200-and, he insists, WDM “has to play the leading role” in meeting that expanded demand.
Thanks to advances in WDM methods, fiber has done a good job in keeping up with this explosion in demand. According to David Clark, senior research scientist at MIT’s Laboratory for Computer Science, “The ability to get bits down a fiber is growing faster than Moore’s Law,” which predicts the doubling of computing power every 18 months. At the moment, Clark notes, the carrying capacity of fiber is doubling every 12 months.
Doing It with Erbium
The term “wavelength division multiplexing” reeks of engineering jargon, but the concept is simple: simultaneously send separate signals through the same fiber at different wavelengths. Essentially the same idea forms the foundation of radio and television broadcasting, where each station sends its signal out on an assigned wavelength in the radio-frequency spectrum. Of course, most people think in terms of frequency instead, but the two values are inextricably bound by their relationship to the speed of light. (For instance, 100 megahertz on the FM dial corresponds to a wavelength of about 3 meters.)
The same principles work for the light going through an optical fiber as for radio waves transmitted through air. Optical fibers transmit best at the invisible, near-infrared-light wavelengths between 1.3 and 1.6 micrometers-roughly double the wavelength of red light.
If WDM is both straightforward and an idea that’s been around-why has it only recently become practical? The biggest obstacle has been the lack of suitable amplifiers. Light signals traveling through even the most transparent optical fibers fade to undetectable levels after a couple hundred kilometers. For most of the time fiber optics have been in place, the only way to span fibers longer than that was to regenerate the signal through an optoelectronic process: A photodetector would convert the stream of weakened light pulses into a voltage signal that could be amplified electronically; this boosted signal modulated a laser transmitter.
The problem is that light detectors don’t discriminate between wavelengths-they scramble signals at different colors, much the way your ears have trouble discerning what is being said if two people talk at once. For optoelectronic systems to work with multiple wavelengths, they must have a way to separate the wavelengths optically, using filters or other similar elements, enabling each signal to pass through its own regenerator. Until recently, though, that has proved impractical.
This limitation disappeared with the invention of a technique for boosting the signal light’s intensity directly, without the need for an intermediate electronic step. The key piece of technology is something called an “erbium-doped fiber amplifier.” These devices, developed in the late 1980s, made the WDM revolution possible.
Unlike a regenerator, a fiber amplifier operates directly on light. Light in the feeble input signal stimulates excited erbium atoms in the fiber to emit more light at the same wavelength. Chains of optical amplifiers can combine to carry signals through thousands of kilometers of fiber-optic cable on land or under the ocean-without regenerators. Because they preserve the wavelength of the optical signals, erbium fiber devices can amplify several different wavelength channels simultaneously without scrambling them. Erbium amplifiers work well across the near-infrared region of the spectrum at which fiber-optic systems operate.
On Land and in Sea
Long-distance telephone companies were the first to realize that wavelength division multiplexing could cut the cost of bandwidth. Compared with the alternative of adding new fiber, WDM technology provides “a much more effective way to add capacity,” according to Dana Cooperson, optical network analyst for RHK Inc., a market consultancy in South San Francisco. Laying new cable is expensive and time-consuming. And burying new cable along the same route already occupied by an older cable is risky-new excavation invites cable breaks that could put the whole system out of service.
The telecommunications carriers’ desire to save time and money has driven a rapid development in WDM techniques. In the mid-1990s, the carrier companies began using systems transmitting at four wavelengths, and soon upped the count to eight. Developers quickly sliced the spectrum even more finely to squeeze 16 wavelength channels through a single fiber for what has become known as “dense” WDM.
When the carriers saw the need, manufacturers were equally quick to sense the market. Lucent Technologies of Murray Hill, N.J., adapted technology developed at its Bell Labs subsidiary. Ciena, a Linthicum, Md., company founded in 1992, charged ahead faster, delivering its first commercial 16-channel system in 1996-at nearly the same time as the AT&T spinoff. Other telecom giants around the world followed, including Nortel, Alcatel, Pirelli, NEC, Hitachi, Fujitsu and Ericsson. Over the past two to three years, several companies-including Ciena, Lucent and Nortel of Saint-Laurent, Que.-have begun to market systems that slice the erbium-amplifier spectrum into 32 or 40 slivers, each only 0.8 nanometer wide. Last September, Lucent delivered its first 80-channel system to AT&T. Pirelli Cable of Lexington, S.C., followed by promising a 128-channel version, but had not delivered hardware as of mid-January.
Telecommunications carriers don’t need all those channels today-and thanks to WDM’s inherent modularity, they don’t need to buy more channels until they’re ready. A carrier installing a WDM system can start with only the transmitters and receivers needed for the few initial channels. Later, as demand for capacity grows, additional equipment can be plugged in to open up new wavelengths.
Taking full advantage of WDM often requires upgrading older cables by adding components that compensate for a troublesome effect called chromatic dispersion. This refers to the tendency of a short light pulse to stretch out as it travels through a fiber owing to the fact that some wavelengths travel faster than others. Dispersion smears light pulses together and therefore limits transmission speed. Avoiding this phenomenon is especially important in submarine cables, where light signals must travel through several thousand kilometers of fiber from shore to shore. New installations can exploit fibers designed for optimum WDM performance, recently developed both by Lucent and by Corning (Corning, N.Y.).
Last year, the first big submarine cable designed for multiwavelength operation-called Atlantic Crossing 1-began sending 2.5 Gbit/s at four wavelength channels on each of its four fiber pairs. The capacity of this system can be upgraded to 16 wavelengths per fiber at that speed, says Patrick R. Trischitta, director of technical marketing at Tyco Submarine Systems Laboratories in Holmdel, N.J. That promises a total of 160 Gbit/s through the cable, a loop connecting the United States with Britain, the Netherlands and Germany.
Project Oxygen raises the bar. Newer WDM technology will carry 10 Gbit/s at each of 16 wavelengths across the ocean in four fiber pairs, a total capacity of 640 Gbit/s per cable. That’s more than 1,000 times the capacity of the first transatlantic fiber-optic cable, which began service just a decade ago. The whole system will ultimately include 168,000 kilometers of cable-enough to circle the globe four times. Other groups are planning more submarine cable systems, although none is quite so ambitious. It’s no wonder MIT’s Clark predicts, “We’re going to drown in fiber.”
On land, regional telephone companies have just begun to adopt wavelength multiplexing. Last year, Bell Atlantic began testing WDM on a 35-kilometer cable between Brunswick and Freehold, N.J., says Robert A. Gallo, the Bell Atlantic engineer in charge of the trial. Four channels each carried signals at speeds to 2.5 Gbit/s-the top rate between company switching offices-and the Ciena-built system has slots for up to 16 wavelength channels. Bell South tested three of 16 channels in a similar system on a cable spanning 80 kilometers between Grenada and Greenwood, Miss. The economics are clear: “It’s cheaper to add WDM capacity than to add new fiber,” says RHK analyst Cooperson.
Different rules apply to the shorter cables linking switching offices to major business customers. Here, in the so-called “metro” market, “the cost of increasing fiber count is not as big an issue because the runs are so much shorter,” Cooperson explains. Still, WDM improves signal transmission in other important ways. One is by carrying signals in their original digital formats rather than converting them into the digital coding used within the telephone network. Because such conversion requires costly electronics, it can be cheaper to dedicate a wavelength for end-to-end transmission in the original format.
The ability to sort signals by wavelength should streamline the operation of future fiber-optic networks. Traditionally, phone companies organize digital signals in a hierarchy of bit rates, merging many low-bit-rate tributaries into mighty digital rivers carrying gigabits per second. This packs bits efficiently onto transmission lines, but requires unpacking the whole bit stream to extract individual signals. If the signals are organized by wavelength, however, simple optics can tease out the desired wavelength channel without disturbing the others. Engineers speak of adding a new “optical layer” to the telecommunications system. Customers might lease a wavelength in this optical layer instead of leasing the right to transmit at a specific data rate. A television station, for instance, could reserve one wavelength from its studio to its transmitter and another to the local cable company-and transmit both signals in digital video formats not used on the phone network.
The Ultimate Squeeze
Since the demand for bandwidth shows no sign of slowing down, the developers of WDM systems are already thinking about how to pack more wavelengths into the same fiber. At the moment, there are two basic approaches being investigated-and limits to both are apparent.
One approach is to reduce the “space” between wavelengths, by choosing wavelengths that are closer together to carry the multiplicity of signals. Packing wavelengths closer works well up to a point, but it ultimately clashes with basic physics. As bit rates increase, optical pulses get briefer, and-following the dictates of Heisenberg’s Uncertainty Principle-this shortening forces the light signal to spread over a broader range of wavelengths. This spreading can cause interference between closely spaced channels. Lucent’s highest-capacity system handles 10 Gbit/s on wavelength channels separated by 0.8 nanometer but only 2.5 Gbit/s when channel spacing is halved. And few experts think channels can be squeezed much tighter. Among major vendors, only Hitachi Telecom of Norcross, Ga., talks about modulating individual channels at 40 Gbit/s-and admits that those signals could span only limited distances.
Prospects look better for the second option: expanding the range of transmission wavelengths. Pirelli, for example, uses three erbium-fiber amplifiers, optimized for separate bands between 1,525 and 1,605 nanometers, to squeeze 128 wavelength channels at 10 Gbit/s each into a single fiber. Lucent has demonstrated erbium amplifiers covering a similar range in the laboratory, and last year introduced a new optical fiber that opens up a long-neglected block of the spectrum around 1,400 nanometers. Good optical amplifiers are not yet available for other wavelengths.
For WDM to reach its full potential, though, more will be needed than simply packing in additional wavelengths. It will also be necessary to develop better equipment for switching and manipulating the various wavelengths after the signal emerges from the optical “pipe.” Optical switches “are getting close to practical commercial applications,” says analyst Mack of KMI. He adds, however, that “to fully emulate what happens in digital cross-connects, you need to reallocate and reassign wavelengths.” It’s impossible to allocate the same wavelength to one customer throughout an entire system because the huge network has far more customers than it has wavelengths.
The illustration below shows how signals from San Francisco and Cupertino arrive in Palo Alto at the same wavelength, both bound for San Jose. The Palo Alto node must convert one signal to a different wavelength for the final leg of its trip, so that the messages they carry aren’t hopelessly confused. Wavelength conversion now must take the same brute-force approach as regenerators, converting the optical signal to an electronic one that can drive a transmitter at the output wavelength. All-optical conversion approaches, while demonstrated in the lab, have yet to reach commercial practicality.
Even if these technical problems are solved, however, that won’t be enough for the technology to really spread its wings. For that, the price will also have to come down-a trajectory that insiders say has already become apparent. Adel Saleh, head of the broadband access research department at AT&T Labs in Red Bank, N.J., projects that cost per network node will drop by a factor of 10 every five years, starting at $1 million in 1995. Through the next year or two, he says, WDM will be economical only for backbone networks. Once cost drops to $100,000 a node, the technology will make sense for metropolitan and regional networks, starting with service to large businesses. Saleh expects that residential access in large apartment buildings will follow after costs drop to $10,000 a node in about 2005, with WDM reaching individual homes once costs decline to about $1,000 in 2010.
The real strength of WDM lies in how it expands the optical airways so that everyone can inhale more of the oxygen of information. At the dawn of the radio era, each transmitter screamed across the whole radio spectrum, blocking other signals for the duration of its broadcast. Then engineers learned to build circuits that tuned each transmitter to its own frequency, opening the radio spectrum to the many stations we can hear today. In much the same way, WDM replaces a single stream of black-and-white bits with a multitude of different-colored signals.
WDM is creating huge new information pipelines that will bring better service at lower cost. But the real information revolution won’t come until cheap WDM pipelines reach individual residences. Today’s modem connections remain bottlenecks, forcing us to sip the torrent of data through the electronic equivalent of a thin plastic straw. But get ready: As fiber reaches the home, your very own wavelength could deliver a bubbling fountain of bits.
The inside story of how ChatGPT was built from the people who made it
Exclusive conversations that take us behind the scenes of a cultural phenomenon.
Sam Altman invested $180 million into a company trying to delay death
Can anti-aging breakthroughs add 10 healthy years to the human life span? The CEO of OpenAI is paying to find out.
ChatGPT is about to revolutionize the economy. We need to decide what that looks like.
New large language models will transform many jobs. Whether they will lead to widespread prosperity or not is up to us.
GPT-4 is bigger and better than ChatGPT—but OpenAI won’t say why
We got a first look at the much-anticipated big new language model from OpenAI. But this time how it works is even more deeply under wraps.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.