Kim Roberts: The 2019 John Tyndall Award winner
A Profile
A conceptualiser, mathematician, furniture maker, prolific inventor, sushi lover, creative spirit, team leader and mentor. These are just some of the descriptors applied to Kim Roberts of Ciena by the people that know him.
Roberts has been awarded the 2019 John Tyndall Award by The Optical Society (OSA) and the IEEE Photonics Society in recognition of his “pioneering contributions to the development of practical coherent communication systems”.
“It is well deserved,” says Seb Savory, who first knew Roberts when they both worked at Nortel and who is now an academic at the University of Cambridge working on joint projects with Ciena. Ciena acquired Nortel in 2010.
“The best measure of Kim’s contributions and impact is the simple fact that the practical coherent optical communication systems that Kim pioneered are now the gold standard for high-capacity fiber-optic systems,” says Steve Alexander, Ciena’s CTO. “They have become the foundation of the fabric for how the world connects, having been deployed by nearly every large network operator on the planet.”
“Roberts was among the very first who introduced electronic signal processing in optical communications," says Professor Ioannis Tomkos at the Athens Information Technology Center (AIT). Initially, the signal processing was at the transmitter - electronic-based equalisation for pre-compensation in direct-detection systems - and then at the receiver. “Roberts’s ideas have made coherent detection practical and revolutionised the industry,” says Tomkos.
Meanwhile, his boss, Dino DiPerna, vice president, packet optical platforms R&D at Ciena, describes Roberts as having one of the most brilliant minds he has encountered.
Education
Roberts read electronic engineering with an emphasis on maths at the University of British Columbia in Vancouver, Canada. “I took courses and took extra courses to get all the requirements of an Honours maths degree,” says Roberts.
Why add maths to an already demanding degree? A recognition of the importance of having a firm grounding in maths for electronic engineering or simply a love of the subject? “It is fun,” says Roberts.
Roberts can span from the theory down to the gate level of chip design
His exploration of maths is ongoing. Most Fridays, Roberts works from home where, after catching up with email, he studies maths. “It is not related directly to engineering, I just learn a new branch of mathematics,” says Roberts. The goal is to learn something new that may or may not help trigger ideas in the future. Overall, the practice has proved very fruitful, he says.
After his undergraduate studies, Roberts gained a fellowship to pursue a Masters degree in a neurological lab developing diagnostics tests for the disease, Multiple Sclerosis.
“Signals were measured on the brain and my task was to identify whether this was normal or abnormal, indicating Multiple Sclerosis,” he says. “I baked cinnamon buns to bribe my friends to come in and be the normals; I needed 25 normals.” His resulting system outperformed a neurologist looking at the same data.
Roberts finished his Masters in the winter of 1983-84. It was a period of recession and he had two job offers: one was to continue at the lab and the second was at Nortel.
“I went off to Nortel and learnt about optics,” says Roberts.
Career
Roberts has overseen several notable projects in his career, as highlighted in the press release announcing the 2019 John Tyndall Award.
These include Nortel’s Superdecoder, terrestrial optical amplifiers for 2.5- and 10-gigabit transmissions, bringing 10-gigabit optical transport to market including developing the first WaveLogic IC, and coherent-based optical transmission.
The Superdecoder was Nortel’s internal name for the application of signal processing to intensity-modulated signals received using a direct detection-based optical receiver.
There are three critical parameters involved with intensity-modulated direct detection (IMDD), says Roberts: the gain of the avalanche photo-detector, the phase of the clock making the decision, and the ‘slicing level’ - the sampling threshold - which may not necessarily be at 50 percent.
The detection circuitry includes the slicer and a flip-flop. The slicer determines if the received bit is a 1 or 0 that is then latched into the flip-flop at the appropriate time. “If it made an error, it made an error,” says Roberts. “In a 2.5-gigabit system there was no error correction, so that was a customer error.”
Nortel’s cleverness was to add two more slicers and flip-flops in parallel to the central channel decoding the actual data. These extra detector channels could have different sampling levels that were executed at different times. The Exclusive-OR logical operator was applied to each output of the extra channels with the central channel. If they differed, a ‘pseudo-error’ was called.
While pseudo, such errors add value in that they help identify where the centre of the eye diagram is to optimise the actual data detection.
“You want it [the detection] at the optimum time and voltage,” says Roberts. “You try to reduce the number of errors to get the optimum positioning of the data channel by doing high-speed measurements in parallel.”
The project came about after a former boss took an internal course on gallium arsenide (GaAs) circuits and chose this as his course project.
“The two of us designed the GaAs circuit and I then went on to do the mathematics of how do we use this idea,” says Roberts. “It went from a course project into what we do in all our IMDD receivers.”
Optical amplifiers and 10 gigabit
Roberts led the Nortel team that developed terrestrial optical amplifiers. “Optical amplifiers had been used for transatlantic links and we developed optical amplifiers to go 2.5 gigabit and 10 gigabit,” says Roberts.
The challenges included the physics, which was new, and determining how to make optical amplification into a product that customers could engineer, install and manage.
Roberts also managed the team that did the science and the prototypes that eventually became Nortel's 10-gigabit product that started shipping in 1995.
“He was one of the big forces behind getting to 10 gigabit [optical transport],” says Savory. “Nortel was the one that went to 10 gigabit when others were saying you couldn’t do 10 gigabit.”
I like to build things and electronic engineering gives you the tools to be able to take your ideas and build them
Pre-compensation & the WaveLogic chip
In the early 2000s, Roberts started working on what became Nortel’s first WaveLogic IC. Here, electronics was used as an aid to counteract optical transmission impairments.
"He developed digital signal processing schemes for pre-distorting the signals at the transmitter, with the goal to compensate for transceiver imperfections and chromatic dispersion-induced distortions as the optical signal pulses are propagating over the transmission fibre,” says Tomkos. “Signal pre-distortion and chromatic dispersion effects on the propagating optical pulses counteract each other as they travel along the fibre so that an almost clean signal is detected at the receiver.”
“There was a realisation by our team that if we modulate the complex electrical field - not just turn the light on and off - then we could compensate anything we wanted to before we transmitted,” says Roberts. “We could do digital linear filtering for the chromatic dispersion that was going to be on the line.”
Using an IC at the transmitter meant the receiver circuitry didn’t need to change. It also meant that electronics could replace the spools of chromatic dispersion-compensation fibre that was, at the time, the solution used. Such fibre spools were costly and added optical loss.
According to Tomkos, Nortel only went public about its pre-distortion technique once similar ideas were published in a paper by an academic.
“I started going around the world preaching a new gospel that dispersion is your friend, dispersion is good,” says Roberts. “That we can fix enormous amounts of dispersion and that the dispersion helps to smear out nonlinearities.”
However, the claim was met with skepticism. The issue, says Tomkos, was the nature of the distorted transmitted pulses: they had a very high peak-to-average power ratio. “When you send very high power signals, it generates non-linear effects,” says Tomkos. Nortel’s competitors claimed the technique wouldn't work and it had no future.
“It took a few years of preaching for the community to become believers,” says Roberts.
Developing the pre-distortion chip - what became the first WaveLogic device - subsequently led to Nortel’s development of the coherent optical receiver.
Coherent receiver
The critical circuit within the WaveLogic IC was its high-speed digital-to-analogue converter (DAC) implemented in CMOS. The overall bit rate was 10.7 gigabits-per-second (Gbps) such that the DAC operated at twice the rate - over 21 gigasamples-per-second - to satisfy the Nyquist sampling theory.
Roberts has 160 patents
“Once we realised we could build these, we realised we could build the analogue-to-digital converter that could run just as fast and we could build [a] digital coherent [system],” says Roberts. Until then, most of the work had looked at analogue optical coherent which Roberts describes as very hard.
“Four of us in a room sketched through on a whiteboard the pieces we needed and we concluded we could build it,” says Roberts.
“It was the same bet-your-houses philosophy with the move to coherent [that was used by Nortel for 10 gigabit],” says Savory. “Kim was very much leading the team that put together the coherent receiver.”
People thought coherent was a nice idea but that it would never happen, says Savory, the issue being the effort required to develop the coherent digital signal processor (DSP).
The coherent receiver - the optics and the coherent DSP - not only solved the problem of chromatic dispersion but also overcame the issue of polarisation mode dispersion which, at the time, was a barrier to achieving faster optical transmission speeds of 40Gbps and higher.
By handling the enormous amounts of polarisation mode dispersion, the coherent receiver could work anywhere a 10-gigabit wavelength would work, says Roberts: “You just plugged us in; it made it very easy for customers to deploy.” And that led to coherent’s commercial success.
Savory mentions how Nortel’s coherent receiver work was submitted as a post-deadline paper at an ECOC conference and was rejected. “I remember chatting to Kim afterwards who, like me, was despondent given the technological breakthrough this represented,” says Savory.
He encouraged Roberts to submit the work to the OSA’s Optical Express journal, where it was subsequently published in 2008 with Roberts as the senior author. “It is now seen as one of the key papers,” says Savory.
Since then, a total of four generations of WaveLogic devices have been announced, the latest being Ciena’s WaveLogic Ai. Indeed, Roberts’ job title is Vice President of WaveLogic Science.
_________________________________________________________
It was my last meeting on the final day of the OFC 2010 show. I was being showcased Nortel’s first 100-gigabit coherent system in a private room reserved for prospective customers. It was also the week that Ciena closed the acquisition of Nortel.
I was talked through the various components of the system. The spectrum analyser display was also explained - the system used two tones per channel, each carrying 50-gigabits of data - while on a whiteboard, constellation points and modulation schemes were drawn to explain the theory.
Despite being exhausted after a long week and poor sleep due to the time zone difference, I left the room feeling energised and elated. I had witnessed the most impressive technological display just as the show was winding down. The 100-gigabit coherent platform had also been explained to me by a patient and clearly authoritative Nortel engineer.
That was my first encounter with Kim Roberts.
Editor, Gazettabyte.
Talents
Savory describes Roberts as a conceptualiser: “You get some people that can do the maths but don’t have the conceptual understanding.” Roberts can do both. “He is also a prolific inventor,” adds Savory: Roberts has 160 patents.
Roger Carroll, Ciena’s vice president, optical modem development, explains how Roberts can span from the theory down to the gate level of chip design. “That type of person is extremely rare,” says Carroll.
Ciena's DiPerna highlights another talent of Roberts: he is an unconventional thinker. “When you are creating, it is easy to get wrapped up in, 'Well, we can’t do that because of so and so',” says DiPerna. “Many brilliant, experienced people can fall into that trap and miss a chance to change the game.”
An example both DiPerna and Carroll cite was the development of WaveLogic 3 where Ciena first included soft-decision forward error correction (SD-FEC) and signal processing at the transmitter.
“The mathematics [of SD-FEC] had been around for a while but its implementation in a chip, at that time, that was magic,” says Carroll. “It had to fit in something that was manufacturable and with a reasonable power [consumption].”
That is what turns Kim’s crank because he sees his ideas can come to life through that iterative process with the teams
At the time, other companies were getting 100-gigabit coherent to market using conventional hard-decision FEC while the WaveLogic 3 incorporate SD-FEC and signal processing at the transmitter.
“There was a lot of pressure at the time,” explains Carroll. “We stuck to our guns, Kim stuck to his guns and the chip helped pull us ahead in the coherent game.”
Work practices and leadership skills
The way Roberts works is something he has practiced throughout the development of the WaveLogic ICs.
A colleague will come into his office and the two will spend an hour or two arguing in front of the whiteboard. “Then, they [the colleague] will take a cell phone picture of the two big whiteboards and will go away and write out the maths proofs, or do a MATLAB simulation, or both,” says Roberts. “Meanwhile, someone else has come in with a different problem and we will go off in another direction and work through that.”
DiPerna says over the years he has been told how important Roberts is and asked what he was doing to keep Roberts excited.
First, he points out the working relationship he, Carroll and Roberts share, having all worked together for over 30 years. Second, DiPerna has made sure that Roberts is surrounded with top talent in the various teams he interacts with, such as Ciena’s analogue, opto, digital and ASIC teams.
“That is what turns Kim’s crank because he sees his ideas can come to life through that iterative process with the teams,” says DiPerna.
Career choice
If Roberts were to start university today, would he still choose electronic engineering?
“I would,” he says. “There are other opportunities but I like to build things and electronic engineering gives you the tools to be able to take your ideas and build them.”
Roberts says his son has also chosen this path, having just completed a PhD in electronic engineering dealing with the optimisation of optical networks in the face of optical non-linearities. “Not much imagination there to move very far away,” quips Roberts.
Roberts’ practical nature extends to making furniture (see image). His creations have appeared as part of his presentations, props that he uses to explain the physics, says DiPerna: “It is all part of the gawky character we love so dearly.
“The guy is a brother to us,” concludes DiPerna. “We are absolutely thrilled for him with this award; it is incredibly deserving.”
Further information
Coherent optical turns 10: Here's how it was made, click here
OSA Paper: "Real-time meaurements of a 40Gb/s coherent system", 2008, click here
Has coherent optical transmission run its course?, click here
Reader Comments