Science News

Fibre optics rev up

Thu, 30th May 2013

Dave Ansell

Listen Now    Download as mp3 from the show Shedding light on LEDs

Whether you know it or not you are probably using fibre optics every day, they are the very  thin glass fibres which carry most of the information around the internet and telephone systems. The information travels in the form of incredibly short pulses of infra red laser light, this is picked because the glass is very transparent in this colour, and can travel tens of kilometers and still be detected, it can then be amplified and continue travelling.

The problem is that the further the pulses travel the more they tend to spread Optics and scalpelsout due to imperfections in the glass's properties, this means that you have to make them longer, so you can get fewer though per second and so you can't transmit as much data, or you need to regenerate the signal every few hundred kilometres in a very expensive process.

But Xiang Lui and collegues from Bell labs in New Jersey, have come up with a system that might help. A pulse of light will disperse and spread out along an optical fibre, but there is a pattern of light that the imperfections of the fibre will cause to have exactly the opposite effect, which is called it's phase conjugate. The group sent these two signals down the fibre optic one with a horizontal polarisations and one with a vertical one, and then recombined them at the end and reduced the distortions by a factor of nearly 30.

This has allowed them to send 400Gbits per second down an optical fibre which is 12800km long which is very impressive, and could reduce the price and increase the capacity and reliability of long data cables



Subscribe Free

Related Content


Make a comment

I'm an engineer who oversees a lot of bespoke fibre cabling and I found the article hard to understand! Essentially I'm thinking of it as analogous to the common-mode rejection used in copper balanced transmission lines. If anyone can shed any light on "phase-conjugated twin waves" I'd appreciate it.
Will it allow similar performance increase in multi-mode fibre? The recent introduction of graded-index cable for multi-mode working has increased the bandwidth-distance product four-fold but if this works with OM3 / OM4 cable then that will be very significant. FunkyWorm, Sun, 9th Jun 2013

"Researchers reporting in Nature Photonics suggest putting not one beam of light down a fibre, but a pair, each a kind of mirror image of the other. When recombined on the receiving end, the noise that the signals gather in the fibre cancels out. These paired beams can travel four times farther than a single one."

Smart. Like if you got a error checking of the 'waves', by either 'setting them' somehow at the send off, or else just comparing them at the end. Which one, or both? I don't know.

"Kerr nonlinearity imposes a limit on the achievable transmission performance and capacity of optical fibre communication links. We show that the nonlinear distortions of a pair of phase-conjugated twin waves are essentially anticorrelated, so cancellation of signal-to-signal nonlinear interactions can be achieved by coherently superimposing the twin waves at the end of the transmission line. " yor_on, Sun, 9th Jun 2013

Ahh,  "The group sent these two signals down the fibre optic one with a horizontal polarisations and one with a vertical one, and then recombined them at the end and reduced the distortions by a factor of nearly 30." That should be a analogue to using a beam splitter, getting opposite spins from it. So you would need a different transmission system than what is used today I guess, even though the fibre will be the same.

Wonder what would happen if you tapped in, on one of those signals? It should be noticeable, shouldn't it? At the error correction, or, maybe not :) yor_on, Sun, 9th Jun 2013

Without actually buying the article, the print is a bit too small to read...

The maximum speed you can achieve in a fiber is dependent on the transmit power - but if the transmit power gets too high, non-linear effects like the Kerr effect distort the pulse shape, and it is hard to unscramble the pulses.

I think what they are aiming for is a way to measure the nonlinear distortion of a fiber. They do this by sending two pulses, one the inverse of the other (phase conjugate). By comparing the two pulses, they can estimate the distortion, and reduce it. This allows them to obtain a readable signal, at a higher laser power, achieving longer range & higher transmit speed.

While they are using two polarisations (horizontal and vertical), it is not clear that this is actually part of the error-correction mechanism. Dual-polarisation transmission is a technique used to double optical fiber capacity at the 40 Gbps level and above. The new idea they claim is the dual-pulse modulation, which may also well work for a single polarisation.

They are already using some techniques to remove linear distortions, such as the fact that a pulse consists of photons with a range of different wavelengths; each wavelength travels at a slightly different speed, resulting in scrambled pulses after a few km of fiber (let alone several thousand!). However, by using a length of "anti-dispersive" fiber, where the relative velocities of the different photons are reversed, you can reconstruct "clean" pulses.

It will be interesting to see if the same mechanism can overcome other non-linear optical effects.

This technique will also be assuming the use of an error-correcting code (similar to those used on CDs, ADSL, GPON and digital TV), which become mandatory for very long fiber links, or for optical speeds at the 100Gbps level and above. evan_au, Mon, 10th Jun 2013

Channel Packing - OFDM and the like are very standard for high speed RF and copper-transmission standards but I assumed that when they hit fibre they'd start in multi-mode applications before they worked their way to single-mode. FunkyWorm, Mon, 10th Jun 2013

I had a quick look at the paper today: They are using the cross-polarisation method to obtain two simultaneous measurements of the fiber distortion using two different pulse shapes. The error correction performance drops if the measurements are too far apart in time

Optical transmission gets a lot harder at 100Gbps and higher, so researchers are trying a number of techniques like the OFDM mentioned by yor_on, which have not been necessary at lower optical speeds. In theory, it should be possible to transmit around 100,000Gbps (ie 100 Tbps) down a single fiber - and this will almost certainly require techniques like OFDM.

"Single-mode" and "multi-mode" have a specific meaning in optical fiber terminology, and it is almost a mechanical/optical specification of the fiber.

The cheaper multi-mode fiber has a physically larger diameter (which makes for simpler plugs and splices), but light can take several different paths down the fiber; each path has a slightly different length, and arrives at a slightly different time, confusing the pulses. In the past, this usually limited multi-mode fiber to medium-speed/short range applications (eg up to 500m at 1 Gbps, suitable for campus/industrial networks).

Single-mode fiber is much thinner, and light takes essentially a single path down the center of the fiber. This preserves the pulse shape over longer distances - routinely 10 to 40km, and routinely at speeds of 10Gbps (with some vendors offering 100Gbps). So if you are trying to break a distance record, you will probably try single-mode fiber first.

Having said that, some people wanted to upgrade already-installed multimode fiber from 1Gbps to 10Gbps, and this led to a recent 10Gbps Ethernet standard which measures the arrival time of all the different paths down a multimode fiber, and adds them together to produce a clean signal at 10Gbps. But this requires enormous processing power - something like 50 billion multiplications and additions per second (ie around 5 multiplications & additions for every bit received). This also requires additional electrical power which is not required for transmission of 10Gbps over single-mode fiber.

The single/multi-mode character of the fiber is fairly independent of the channel packing and modulation techniques which are used to transmit information over the fiber, like OFDM, DWDM, QPSK, or the traditional On/Off keying. The paper under discussion tested a single data stream, and also around 8 simultaneous streams on adjacent wavelengths. evan_au, Tue, 11th Jun 2013

There are quite a few modulation schemes that can be used. The more complex ones enable more bits per second to be carried withing a particular bandwidth. In fibre, keeping the bandwidth narrow is useful as it reduces the effects of dispersion. In RF it is useful mainly useful so that more of the available spectrum is utilised. In both cases, the more bits/second you send the shorter is the range however. However, if the aim is to just get as much data tranferred as possible (and the distance between repeaters is not a factor) then elaborate modulation schemes are worthwhile. It makes a lot of sense to send two signals differentially because any "common mode" distortions can be cancelled effectively improving the Signal to Noise+Distortion ratio. It has been a long time since I worked on fibre transmission so it is interesting to see the latest developments. graham.d, Tue, 11th Jun 2013

evan_au - couple of notes; although multi-mode is cheaper to manufacturer than single-mode cable (it has a 50 micron core as oppose to 9 microns) they have different application. Typically multi mode is used for SAN traffic and the like. Additionally being able to launch many modes down a cable reduced the cost of host-bus adaptors; it's how you can have a fibre card in your PC for only a few hundred quid. The SFPs are cheaper as well.
Network speeds - Gigabit was common a decade ago down the older OM1 standard but we're now moving form OM3 to OM4 (OM2 never really caught on) and so 10gig has been common since 2007 with 40gig being the new emerging standard.
Having said that quite a lot of the OM1 system I installed ten years ago happily worked to eight gigs (the ATTO Celerity standard). For sure channel packing applies to either but since their applications are different it seemed the market would implement it where it "was needed" first.
FunkyWorm, Wed, 12th Jun 2013

We should not ignore the possibility of further advances in optical fiber materials. For example Fluoride glasses offer the possibility of lower attenuation than conventional silica fibers.

But this will also require parallel developments in deployment, as often the optical splices or connectors have losses that are as high as the fiber itself.

Improvements in lasers are also desirable - currently topping the wishlist is a tunable single-frequency laser, which would allow deployment of technologies that would give every home their own wavelength of light for communication, as well as allowing much more flexibility in the network core. evan_au, Thu, 13th Jun 2013

I first read about using phase-conjugate mirrors (PCM) to amplify and clean up optical fiber light pulses some years ago (a decade? 15 years?). PCMs are pretty cool. They make use of optically-nonlinear materials (ONM) that alter their optical density in response to light. Imagine a set-up for creating a hologram but instead of photographic film, you have a block of ONM. There's an image light signal coming in which interferes with a reference beam also going into the ONM. The trick is that a second reference beam, exactly conjugate to the first one (identical but coming in from the opposite direction), is sent into the ONM. This beam interacts with the pattern of varying optical densities created by the interference of the first reference beam and the image light signal and produces light that goes back along the paths that the original signal did. If a particular ray of light enters the ONM along some path, a duplicate beam comes back out retracing that exact path.

You can even insert a plate of frosted glass between the source and the ONM and the return light rays will STILL be following the correct paths.

By increasing the intensity of the reference beams you can increase the gain of a PCM so that any reflective surface that happens to face the PCM will spontaneous acquire a laser beam bouncing between it and the PCM.

The reason PCMs are good for cleaning up light pulses is that optical fibers are highly uniform and the sort of distortions a pulse acquires going down one kilometer of fiber is pretty much the same as it would get going down any other km of fiber. So, a nice clean pulse of light has bounced around through a length of multimode optical fiber until it's a blob of light almost too messy to read. This gets redirected into a PCM which spits back a perfect duplicate of the incoming wave front, but amplified. This amplified blob of light is sent along the next length of fiber during which the wave reflections reverse themselves and from the messy blob of light, an almost clean pulse re-emerges.

Nifty, huh? AndroidNeox, Mon, 24th Jun 2013

See the whole discussion | Make a comment

Not working please enable javascript
Powered by UKfast
Genetics Society