It might seem like submarine communication is a relatively new fangled idea, but no; the first cable to link the UK and the US came into operation around 1860, carrying telegraphy.

Since those heady days, submarine cables have been the way to link callers on different continents and, more recently, to carry the vast amounts of data spawned by the internet. While communication satellites are available, they handle much less than 10% of all traffic.

But it is only in the last 30 years or so that submarine cables have come into their own and the reason is fibre optics. Before that, all data travelled over coaxial cable. Those of a certain age will recall the difficulty of making a transatlantic call and the message that 'all circuits are busy; please try again later'. The reason was that coaxial cable had a maximum bandwidth of 45Mbit/s, equating to around 6000 voice calls. The cable was already more than 40mm in diameter and increasing the bandwidth would have required it to be even larger, leading to greater cost and handling issues. Today's fibre optic cables – less than 20mm in diameter – can handle up to 200million voice circuits per fibre pair – and they may have up to eight fibre pairs.

Even with that capacity, submarine fibre optic cables are coming under pressure as the volume of data spirals.

Xtera is one of the leading companies in the field. Stuart Barnes, general manager of the company's UK operations, is a submarine communications veteran, having cut his teeth at STC Submarine Cables in Greenwich.

He said that, by 1980, it was obvious that coaxial cable had run out of steam. "We were trying to squeeze the last drop of performance from the technology," he recalled.

Meanwhile, work was well underway at STC's Research Labs in Harlow into the use of fibre optics for communications – work that would win Dr Charlie Kao the Nobel Prize in Physics 2009 'for groundbreaking achievements concerning the transmission of light in fibres for optical communication'. It made sense, in Barnes' opinion, to see whether that work could be used in the submarine world.

"There were three ways we could go," he recalled, "continue with coax, use waveguides or move to fibre. It was very apparent that optical fibre was going to be the winner."

The first subsea optical cable came into use in 1985; a 140Mbit/s link between the UK and Belgium, laid following trials of the technology across a Scottish loch.

Since then, fibre optic technology has leapt ahead. "The line rate has gone from 144Mbit/s to 155, 622, 2.5Gbit/s and 10G," said Barnes. "Today, we supply technology that supports 100G per wavelength, with fibres capable of supporting more than 100 wavelengths."

But the rapid increase in line rate – eight generations – has not been accompanied by a similar improvement in amplifier technology. "There have only been two generations of optical repeater," Barnes pointed out, "regenerative and then the optically amplified repeater."

Regenerative systems required 'three Rs' in each repeater – reamplifying, reshaping and retiming the signal. "It was an optical to electrical to optical process," he explained. Each device was spaced roughly 50km apart on a transatlantic cable and built using discrete components on rigid boards. Not only did they have to deal with the data passing through, they also had to withstand the high pressure of being on the seabed, which brought mechanical issues into play.

Tony Frisch, senior vp of Xtera's repeater business unit and another subsea veteran, added: "There were about 10 ICs per fibre pair in the regenerator, along with four lasers – two for each direction – two integrated receivers and a few discrete components. All these were mounted on specialised boards. And we also needed to handle fibres, electrical connections and power."

One thing common to cables now and then is power. Barnes said: "You need 25kV DC to power a long transoceanic cable and half of that power is lost in the cable." Frisch added: "We still push about 1A through the cable and still need several kV, depending on the length of the cable."

Barnes believes submarine fibre optic cable technology developed rapidly because all relevant parts of UK industry were pointed in the same direction. "It was the heyday of the UK telecomms industry," he asserted. "There were a lot of optical research programmes underway, plus there was serious interest in the technology from GEC, Racal, Plessey and STC." BT, for example, did some groundbreaking work on transistors and ICs.

Even so, developing the required technology wasn't easy. "We had to invent everything – from semiconductor lasers to the fibre itself – and we had to get everything right. Submarine cable technology was its own world; everything had to be done in a particular way because of reliability."

Times change, however. In the 1980s, submarine cable technology was leading terrestrial technology. "That's flipped now," Barnes claimed. "High speed terrestrial communications is now leading, driven by the needs of data companies. That's why we have moved from SDH/Sonet based rates to those more closely associated with data, such as 100G."

As Barnes noted, while there have been eight steps in data rate, there have only been two developments in amplifier technology. Frisch added: "Despite the evolutions in transmission rate, the amplifier – which is based on erbium doped fibre – is essentially unchanged; the only real developments have been gain flattening to get more bandwidth and higher power." These developments came in as systems evolved from being single wavelength to supporting more and more wavelengths.

The challenge now for those developing subsea systems is what could be broadly termed 'the laws of physics'. "There are limits on the bandwidth of simple doped fibre amplifiers," Barnes explained. "But as far as the fibre is concerned, there is still the 'third window' to be filled."

There are essentially three 'windows' in the optical silica fibre spectrum, where low attenuation enables optical communications. Barnes said: "One is centred on 850nm, using multimode fibre. The second is centred on 1310nm, using single mode fibre. The third window, again with single mode fibre, is centred on 1550nm and supports multiple channels in an optical bandwidth of 32nm. These windows have an attenuation of about 1dB/km, 0.4dB/km and less than 0.2dB/km respectively. We quickly went to single mode fibre, but were always keen to get to the third window because it brought a big jump in span length.

"With doped fibre amplifiers, we can stretch optical bandwidth to 40nm over short distances, but for long transoceanic cables, we can go to no more than 32nm – the fibre bandwidth, however, is more than 100nm."

Raman repeaters – where amplification is derived from the fibre in the cable – offer a significant increase in optical bandwidth and a way to exploit the 'third window' fully.

Subsea cables currently use coherent transmission technology and this poses another challenge. "One problem with high capacity coherent transmission is that you're limited by data converter technology; the faster the data rate, the faster the converter has to run," said Barnes, "and the availability of high speed converters is an issue. We have to work out where we go."

Today, 100G signals use two polarisations and four phases, with data symbols transmitted at a rate of around 30billion/s, which requires sampling and A/D conversion at twice that speed. "We are already seeing schemes with 16 levels and the next generation will operate at twice this speed – a major challenge for the ICs which do it."

The ultimate limit is set by Shannon's law, which tells you how much data can be sent over a link with a particular signal to noise ratio and and bandwidth. "A few years ago," Frisch noted, "I would have said the limit was a long way off. Now, it's a lot closer; for example we're squeezing the last few drops out of error correction technology."

Like almost every other technology, subsea comms is being driven by cost – in this case, cost per unit bandwidth. "Ever since deregulation, we have been driven by the customer to provide more capacity for lower cost," Barnes continued. "We are looking at all the things that can help us go to the next stage, but the development cycle for this technology takes several years."

Xtera is deploying a new optical amplifier to help attain the higher bandwidth needed to provide greater capacity. Development of the amplifier – codenamed Project Olympics – started in 2011. "Our design goal was to increase the bandwidth of EDFA repeaters to 55nm, something like a 70% increase on the bandwidth of conventional amplifiers," said Barnes.

Barnes noted that it required more complex laser pumping schemes and combinations of pump lasers. "But we can now apply what we've learned to the next generation, start to stretch the laws of physics and go for more bandwidth."

As part of this work, Xtera and Corning have used Raman technology to set a record of 607km for unrepeatered transmission of 100G. "But you don't get that performance without things like pre and post amplification and low loss fibre," Barnes observed.

Barnes also sees the opportunity for photonics to help boost performance. "I would like to improve power conversion efficiency from electrons to photons; that converts straight away into more capacity. And photonic ICs would allow us to do away with things like splices in the amplifiers, providing more room for pumps and greater bandwidth without an increase in device size. It's all about miniaturisation; the smaller the amplifier module, the smaller the canister that contains it."

The amplifier canister – made from marine grade titanium, instead of steel – is now small enough to pass through a plough during burial operations, making it easier to handle during laying and assembling.

Progress is being made in boosting performance and Barnes is confident of achieving more, but it remains an incremental process. "Once you get close to any of the limits, it gets very expensive to make progress: we need to push on doors that are more difficult to open," Barnes concluded.