High fibre diet

A consortium of UK universities has received nearly £5m of government funding to develop technology that could make home internet at least 1,000 times faster.

Next January, optical and electronic experts from Surrey, St Andrews, Leeds, Warwick and Southampton universities, with industry partners Intel and Qinetiq, will begin advancing silicon photonics, the technology of making optical devices using silicon and complementary metal oxide semiconductor (CMOS) techniques.

By evolving this niche technology, the researchers claim silicon-based optics could become cheaper through mass production and that this would assist the widespread implementation of fibre to the home (FTTH) technology.

This is where the entire fibre optic system is installed in the home and people can receive everything from high-speed internet to high-definition television via a single fibre, at gigabit speeds.

'Interest in silicon photonics is hotter now than ever,' said project leader Prof Graham Reed of Surrey University. 'We started work in this area in 1989, when the expectation was that the technology would never really be high-speed. We were looking at smaller, lower cost applications like sensor type work and interrogating sensors.

'About five years ago, suddenly there was interest from semiconductor companies, and the first to contact me was Intel. Their interest is driven from optical interconnects, either for communicating between chips on a board, between boards, or even rack to rack.'

Smaller and faster electronics has created obstacles for microprocessor firms.

'One of the problems with modern microprocessors is they are getting faster and more complex. That has been achieved by reducing the size of the transistors and getting more into an electronic circuit. However, there is a problem of connecting circuits, because as you make all the electronic devices smaller, the resistance of the interconnect goes up and it slows everything down again,' he said.

'The fall in thje cost cost of electronics in the high street is all because of silicon. It is the best understood material from an electronic processing point of view, and because most modern electronics are made in silicon, it is natural to look at silicon photonics for this means of interconnect,' he added.

As well as developing silicon interconnects, the project will also develop optical non-linear processes, such as optical amplifiers and detectors to convert light from optical to electronic signals. But the main goal, said Reed, was to integrate optical and electronic processes on the same layer.

'The biggest aim is to first produce the devices individually and then put them together on a single chip. We are going to produce two demonstrators — a transmitter and a receiver — as part of this,' he said. 'The transmitter will have the optical coupling to a small waveguide, and it has to have a very fast modulator to convert an electronic signal into an optical one and then transmit it. We will also need to produce drivers that drive the modulators in the electronics.

'On the other side, we will have the receiver, so again there will be the optical coupling problem, and then we will have to convert the optics into an electronic signal and do some electronic processing on that chip and then extract an electrical signal.'

Reed said the growth of home internet has emphasised the need for a low-cost way to handle very fast data and ever-increasing bandwidth.

'The most interesting application is FTTH,' he said. 'At the moment, it is done electrically in the UK, which means we are limited to very slow speed. If we could get the optical signal right into the house, you would have the full bandwidth of the optical signal.

'The typical existing internet connections could not even handle HDTV very well, whereas with a fibre connection in your home you could imagine your entire entertainment system — TV, internet, phone and a whole host of other things — being delivered via a single fibre.'

The researchers are aiming for two speed targets, an integrated 10Gb/sec and 40Gb/sec for a standalone device, but admit that the integration will not be easy.

'The integration process is a huge challenge. Intel has demonstrated a 40Gb/sec modulator, but it had a very small modulation depth (signal intensity) and it was unreliable. It was more of a demonstration of a concept as it was never good enough for applicability,' said Reed.