A new technique to reduce the effect of noise in fibre-optic cables, by adapting signals to compensate for it in advance, has been demonstrated by researchers in the US. The team has increased the maximum power – and consequently, the distance – at which optical signals can be sent through the fibres. Its technique could potentially be implemented in existing optical fibres that carry information all over the globe, thereby significantly enhancing their capacity, and could be a first step toward a faster Internet.

While there are many benefits of using fibre-optic fibres to transmit data, one of the most significant is that photons do not significantly interact with one another, so multiple optical signals can travel down the same line. However, as the signal propagates, noise gradually creeps in, and this creates "crosstalk" between the signals. To reduce the crosstalk and prevent the signals becoming lost in noise, engineers cannot squeeze different signals too close together within a single channel, or from spacing the channels too close in frequency, limiting the capacity of a fibre.

Noisy crosstalk

The noise comes from two main sources. First, from amplifiers spaced regularly along the fibres to regenerate the signal as it propagates. If this were the only noise, one could simply turn up the signal power, reducing the number of amplifiers needed and the resulting noise they inject. However, another type of noise is caused by the fact that the refractive index of silica is not perfectly linear, which creates high harmonics of the signals that interfere with one another, creating crosstalk. This type of noise gets worse at higher power levels, so above a certain threshold, turning up the power actually increases the overall amount of noise contaminating the signals.

Unlike the random noise from amplifiers, however, this nonlinear interaction noise is deterministic. This makes it possible, in principle, to modulate the signals in advance to compensate for it. However, various groups' efforts to achieve this had met with little success. "A system typically has about 30 to 200 optical channels, each transmitted by an independent laser," explains optical physicist Nikola Alic, of the University of California, San Diego. "These lasers are by no means perfect: their exact frequency has a certain tolerance, and furthermore they wander in frequency over time." This uncertainty in the relative frequencies of the channels makes it impossible to calculate the exact nonlinear interaction noise and compensate for it effectively. Instead, in August last year, Alic and colleagues proposed using a frequency comb – in which one laser produces a series of pulses with equally spaced frequencies that act as a "ruler" – and transmitting each channel using a different tooth of the comb. If the fundamental laser frequency changes, all of the teeth move in step, so the relative frequency does not change and the nonlinear interaction is unaffected.

More power, less noise

The researchers have now put their proposal into practice, sending signals down three separate frequency channels in over 1000 km of standard optical fibre, and using computer software, have attempted to compensate for the nonlinear interaction noise. When the three channels were transmitted by three separate lasers, the researchers found that, despite their attempts to compensate for it in advance, there remained significant crosstalk between the signals, and as a result, signal clarity declined if they put more than 200 μW of power into each signal. Conversely, when the channels were transmitted via the teeth of a frequency comb, the researchers could reduce crosstalk far more effectively. They could therefore make the signals 10 times as powerful before nonlinear noise became a problem. If implemented in practice, this would mean fewer amplifiers needed, less random noise injected into the signals, and much greater fibre capacity.

Govind Agrawal, of the University of Rochester in New York, who was not involved in the current work, told physicsworld.com that "frequency combs have been used for metrology and other applications, but this paper is showing, for the first time, that they can have a big impact on telecommunications." Peter Andrekson, of Chalmers University of Technology in Sweden, who was also not involved, agrees that the work is "significant", although he cautions that "if you want to apply this in a real live information traffic situation, you will have to implement real circuits that do this pre-compensation on the fly. That will require the development of very sophisticated, expensive, power-hungry and high-bandwidth circuits." Alic agrees, but he remains optimistic: "Yes, a push will be necessary from both researchers and industry," he says, "but I think the rewards will be worth it."

The research is published in Science.