An international team of researchers, lead by Australia's Centre for Ultra-high bandwidth Devices for Optical Systems (CUDOS), has made what appears to be a significant breakthrough in optical signal processing.
The team has built a device that extracts a single 10 Gbit/s data stream from a 640 Gbit/s signal; in other words it switches out every 64th bit of data.
The result was presented by CUDOS director Ben Eggleton in a post-deadline paper at the Opto-Electronics and Communications Conference (OECC) in Sydney earlier this month.
"Our research is really looking to the future, and seeing if we can actually build the key building blocks for a system that might operate at a terabit-per-second," Eggleton told fibresystems.org.
The device, a demultiplexer, is a key component for future high-speed optical systems based on optical time division multiplexing (OTDM). This architecture has been proposed for terabit-per-second transmission because there are no detectors or electronics that can operate at such high data rates, nor are there likely to be in the future.
OTDM interleaves multiple low-speed tributary signals in the time domain to create a high-speed composite signal. At the other end of the link, the demultiplexer extracts the signals optically, and then the rest of the signal processing on the tributaries can be carried out in the electronic domain.
"The key innovation is the material we're using," says Eggleton. "The material [chalcogenide] is highly non-linear so in many ways it is the holy grail for optical signal processing."
Chalcogenide is a an arsenic tri-sulphide composition of glass, so it's cheap and can be deposited on a silicon wafer and patterned using standard micro-electronics processing techniques. Fine-tuning the chemical recipe for etching well-defined waveguides has taken the researchers several years of work, but that's behind them now.
The 5 cm-long chalcogenide waveguide exploits the optical non-linearity to do wavelength conversion. When the 640 Gbit/s data signal overlaps a clock signal at 10 Gbit/s, it creates an "idler" signal at a third wavelength, which is tapped at the output.
The waveguide is a proof-of-concept device, and the researchers would like to build more functionality onto a chip to make the demonstration more convincing. "You could imagine building a multi-wavelength device on a single chip," says Eggleton. "I think that's the whole vision of monolithic integration — wavelengths coming in, separated spectrally and spatially and then parallel processed on that chip."
To extract all 64 tributaries from a 640 Gbit/s signal would require a 64 waveguides and 64 clocks, and that's before extra wavelengths are added. Even though the waveguides are "colourless", meaning they can operate at any wavelength, integrating them all on a single photonic integrated circuit would be a hugely complicated task. Optical clock recovery is another significant engineering challenge.
In the next stage of the work, CUDOS researchers are looking at ways to shrink down the dimensions of their waveguide by boosting the strength of the non-linearity, so they can fit more components on a chip. To explore the possibilities of integration, they will need to partner with industry, says Eggleton.
"This is not going to be an inexpensive component at the end of the day," he says. "But if you can really do 640 Gbit/s on a single chip then that's a very exciting proposition."
• The technology is the result of a scientific collaboration between CUDOS teams at the University of Sydney and the Australian National University, Canberra, together with the Technical University of Denmark in Lyngby, Denmark, and the University of Science and Technology in Huazhong, China.