Optical networking's future

Shown is Professor Polina Bayvel in her lab at University College London. Bayvel gave the opening plenary talk at ECOC.

Should the industry do more to support universities undertaking optical networking research? Professor Polina Bayvel thinks so and addressed the issue in her plenary talk at the ECOC conference and exhibition held in Glasgow, Scotland, earlier this month.

In 1994, Bayvel set up the Optical Networks Group at University College London (UCL). Telecom operators and vendors like STC, GPT, and Marconi led optical networking research. However, setting up the UCL's group proved far-sighted as industry players cut their research budgets or closed.

Universities continue to train researchers, yet firms do not feel a responsibility to contribute to the costs of their training to ensure the flow of talent. One optical systems vendor has hired eight of her team.

In her address, Bayvel outlined how her lab should be compensated. For example, when a club sells a soccer player, the team that developed him should also get part of the fee. 

Such income would be welcome, says Bayvel, citing how she has a talented student from Brazil who needs help to fund his university grant. Her lab would also benefit. During a visit, a pile of boxes - state-of-the-art test equipment - had just arrived.

Plenary talk

Bayvel mentioned how the cloud didn't exist 18 years ago and that what has enabled it is optical networking and Moore's law. She also tackled how technology will likely evolve in the next 18 years.

Digital data is being created at a remarkable rate, she said. Three exabytes (a billion billion bytes) are being added to the cloud, which holds several zettabytes (1,000 exabytes or ZB) of data. By 2025, data in the cloud will be 275ZB.

The cited stats continued: 6.2 billion kilometres of fibre have been deployed between 2005 and 2023, having 60Zbits of capacity. In comparison, all data satellite systems now deployed offer 100Tb, less than the capacity of one fibre.

Moore's law has enabled complex coherent digital signal processors (DSPs) that clean up the distortions of an optical signal sent over a fibre. The first coherent DSPs consumed 1W for each gigabit of data sent. Over a decade later, DSPs use 0.1W to send a gigabit.

Data growth will keep driving capacity, says Bayvel. Engineers have had to fight hard to squeeze more capacity using coherent optical technology. Further improvement will come from techniques such as non-linear compensation. One benefit of Moore's law is that coherent DSPs will be more capable of tasks such as non-linear compensation. For example, Ciena's latest 3nm CMOS process, the WaveLogic 6e DSP, uses one billion digital logic gates.

Extra wide optical comms

But only so much can be done by the DSP and increasing the symbol rate. The next step will be to ramp the bandwidth by combining a fibre's O, S, C, L, E and U spectrum bands. New optical devices, such as hybrid amplifiers, will be needed, and pushing transmission distance over these bands will be hard.

"We fought for fractions of a decibel [of signal-to-noise ratio]; surely we're not going to give up the wavelengths available through this [source of] bandwidth?" said Bayvel.

In his Market Focus talk at ECOC, BT's Professor Andrew Lord argued the opposite. There will be places where combining the C- and L-bands will make sense, but why bother when spatial division multiplexing fibre deployments in the network are inevitable, he said.  

"It is not spatial division multiplexing versus extra wide optical comms; they can co-exist," said Bayvel.

Bayvel describes work to model the performance of such a large amount of spectrum that has been done in her lab using data collected from the MAREA sub-sea cable. Combining the fibre's spectral bands - a total of 60 terahertz of spectrum - promises to quadruple the bandwidth currently available. However, this will require more powerful DSPs than are available today.

Another area ripe for development is intelligent optical networking using machine learning.

An ideas lag

Bayvel used her talk to pay tribute to her mentor, Professor John Midwinter.

Midwinter was an optical communications pioneer at BT and then UCL. He headed the team that developed the first trial systems that led to BT becoming the first company in the world to introduce optical fibre communications systems in the network.

In 1983, his last year at BT, Midwinter wrote in the British Telecom Technology Journal that this was the year coherent optical systems would be taken seriously. It took another 20-plus years.

Bayvel noted how many ideas developed in optical research take considerable time before the industry adopts them. "Changes in the network are much slower," she said. "Operators are conservative and focus on solving today's problems."

Another example she cited is Google's Apollo optical switch being used in its data centres. Bayvel noted that the switch is relatively straightforward, using MEMS technology that has been around for 25 years.

Bayvel used her keynote to attack the telecom regulators.

"It is simply unfair that the infrastructure providers get such a small part of the profits compared to the content providers," she said. "The regulators have done a terrible job."