Out of the Lab: 100 Tbit/s
The universities -- St. Andrews, Bristol, Glasgow, Heriot-Watt, and Imperial College, London -- have already secured US$18 million (£12.5 million) of government funding for the six-year project.
And now the UPC is trying to figure out how it's going to achieve its 100-Tbit/s target.
So far, only one thing's for sure. The UPC is going to use femtosecond pulses. These mind-bogglingly short bursts of light are about 10,000 times shorter than the pulses in a 10-Gbit/s signal. Therefore, on a single channel, they would provide 10,000 times more information-carrying capacity, or 100 Tbit/s. (To put this in perspective, there are one thousand million million femtoseconds in a second.)
Of course, if it were easy to carry traffic with such ultrashort pulses of light, folk would be doing it already. They're not doing it because light sources, modulators, and detectors that can operate at such phenomenal speeds don't exist. And then there are things like dispersion to consider. Femtosecond pulses behave in a non-linear way when they travel down an optical fiber, or any other material. In other words, their behavior is difficult to predict.
Carrying all the information on a single channel is one possible scenario, but it's not the only one. In its research proposal, the UPC outlines an alternative method of exploiting femtosecond pulses, which it terms "spectral slicing."
This has similarities with wavelength-division multiplexing (WDM) and takes advantage of the fact that femtosecond pulses are composed of a broad range of wavelengths. By shining the pulses on the equivalent of a prism -- possibly an arrayed waveguide grating or other grating -- it is possible to split each pulse into a rainbow of its constituent wavelengths. Next, groups of wavelengths can be modulated separately. It sounds complicated, but effectively the pulses are acting as a single, broadband source with a total system bandwidth of, say, 200 nanometers.
"It is reasonable to suppose that this type of source will provide the potential for ten separate 2.5 Tbit/s channels thus constituting a total transmission rate of 25 Tbit/s," the research proposal states.
Bristol University's Ian White, who directs the systems aspect of the UPC work, points out that individual channel speeds of 2.5 Tbit/s aren't significantly higher than what's already been accomplished in the lab. A single-channel bit rate of 1.28 Tbit/s has already been achieved by NTT Corp., according to a paper it presented at the European Conference on Optical Communications last September.
However, the key here is not really the speed of the individual channels; it's the broadband nature of the system. It's no good maximizing bit rate without also considering system bandwidth, because there is a tradeoff between the two (see Essex Claims 4000-Channel DWDM). As a rule of thumb, the maximum attainable bit rate is roughly equal to the channel spacing. By that reckoning, each 2.5-Tbit/s channel requires at least 20nm of bandwidth to itself.
"That's why we're exploring materials that will provide broad bandwidths, such as quantum dots [tiny semiconductor particles] and polymers," says White (see Zia Laser's Not-so-Dotty Idea).
The quantum dot work at Imperial College focuses on wavelengths around 1300nm, while the polymers being developed at Bristol University, Imperial College, and St. Andrews operate at visible wavelengths (400 to 750 nm). The UPC isn't tied to any one wavelength regime or technology at this stage. It's more a case of backing lots of wild horses and hoping that one of them finishes the race.
The universities' efforts are being backed by commercial component vendors that have signed up as project partners. They include Agilent Technologies Inc. (NYSE: A), Kymata Ltd., JDS Uniphase Inc. (Nasdaq: JDSU), Marconi Communications PLC, (Nasdaq/London: MONI), Nortel Networks Corp. (NYSE/Toronto: NT), Sharp Corp., and Vitesse Semiconductor Corp. (Nasdaq: VTSS).
– Pauline Rigby, senior editor, Light Reading http://www.lightreading.com