If you want to switch optical signals – for example, adding traffic to a stream in a multiplexer and getting that traffic back off the optical network – the world has a tried-and-true approach: turn the optical signal into an electrical signal, do the switching, and turn the signal back into light for transmission across fibres.
The problem with this approach is that speed is limited by how long this three-stage process takes, because even our fastest silicon operates far below the potential speed of optical networks. Light operates at frequencies so high that it's easier to talk about wavelength – 1330 nanometres is much easire to work with than hundreds of thousands of Gigabits per second. In the silicon world, we're still just beginning to work at terabits per second.
So there's a huge benefit to be had if the world could start working directly with light, instead of electricity.
To some degree, we can already do so. Devices can be built which switch light from one fibre between several other fibres. The next step in development of optical switching, however, takes things a step further, by using a light beam as a switch to control the behaviour of other light beams.
That's where research such as from CUDOS, which a fortnight ago made headlines with its optical switching development, comes in.
Professor Ben Eggleton, head of the Centre for Ultrahigh Bandwidth Devices for Optical Systems, explained that the work by CUDOS researchers, presented at the Opto-Electronics and Communications Conference in mid-July, represents a new way of handling optical switching.
The basis of the work is a material – chalcogenide glass – in which an incoming light beam distorts a cloud of electrons, down at the molecular level. This can then be used to control a second light beam, something akin to how one voltage can switch another voltage in electrical devices such as transistors.
“The chalcogenide glass that CUDOS has developed over the past five years contains some fairly nasty chemicals, but it is quite stable and robust,” Eggleton said.
“The key to the interaciton is that the control beam is distorting the electron cloud in this material. That distortion is instantaneous, and that manifests in the signal beam changing properties.”
The result is that a switch built this way can operate on individual photons in a stream.
Today, optical multiplexers generally work in two ways: either by progressively aggregating electrical signals into ever-faster optical streams (for example, in optical SDH add-drop multiplexers), or by aggregating light streams using different wavelengths (in wavelength division multiplexing, or WDM).
It's the aggregation of signals into streams that the CUDOS work seeks to address.
For example, the optical networks that connect Australia to the outside world are built on a hierarchy of speeds: 155 Mbps streams are put together into 622 Mbps streams, which are aggregated upwards towards 40 Gbps speeds. The next step in traditional networks will be the 100 Gbps-plus systems now emerging from carrier-scale network vendors.
However, this approach depends on several layers of aggregation – all of which happens as electricity.
A wholly-optical switch changes the game. Not only is it faster, but as Professor Eggleton explained, it also opens the way to manipulating the data streams on a much more granular basis.
Using a freeway as an analogy, Professor Eggleton said today's switches take entire lanes from the freeway, and gradually break them down to smaller and smaller streams of traffic. Instead, devices such as the CUDOS switch could “take one car off the freeway, instead of a whole lane of traffic”.
Professor Eggleton said commercialisation of the switches will foucs on how to ensure reliable fabrication and packaging, and how best to integrate the devices into larger-scale switches.
While the switch demonstrated in Sydney operates at 640 Gbps, Professor Eggleton said “terabit per second per second, per channel capacity is something we can think seriously about”.