Industry View: Stefan Spälter

Stefan Spälter is head of optical engineering at Nokia Siemens Networks in Germany. His team is part of the R&D organization of the business line that builds dense wavelength-division-multiplexed optical communication systems, the equipment that allows fibers to transmit light of many wavelengths, in which information is encoded.

“The focus of my team is optics: we assess innovative optical technologies and integrate them into our products. The time horizon extends from four to five years, before product maturity until product implementation.”

“Today, system vendors like Nokia Siemens Networks offer a data transport capacity of 3–4 Tbps. That's the total capacity of a single fiber, transporting many channels of light with a per-channel capacity of 40 Gbps. Nokia Siemens Networks is the market leader for high capacity, 40 Gbps-based deployments.”

“In our next product generation, each channel will transport 100 Gbps. By multiplexing 96 channels per fiber, that amounts to nearly 10 Tbps of total fiber capacity.”

“With some carriers requiring the 3–4 Tbps capacity already today, there will be a need for capacities even beyond 10 Tbps in the future. You could gain a factor of two by using a broader wavelength range. The 96 channels I mentioned before are located in the so-called C band. In addition, the longer wavelength L band can be used, ending up with 20 Tbps total.”

And that is, Spälter explains, about as much as present technology will be capable of.

“If you want to go even further in capacity, another level of sophistication will be required. From today's perspective, there are basically three options. First, you can use wavelength bands other than the C and L bands; second, you can increase the spectral density by packing more channels into the same wavelength range; or, finally, you can increase the data rate within each channel. Now, from today's perspective the first option is not really there. That's because erbium-doped fiber amplifiers, which are used typically every 80 km to boost the signal in the fiber, only amplify signals in the C and L bands. To accommodate other wavelength bands, you would need different amplification technologies, which have not yet reached maturity. Moreover, additional equipment would be required, which makes this option very costly, both from an equipment cost and from an operational expenditure point of view. Also, because signals would then need to be processed separately from those in the C and L bands, it would diminish the flexibility in combining or separating data streams. In short, from an operational perspective, it would be somewhat similar to using several fibers and equipping them individually.”

Spälter thinks that there may be a way out for the first option, and that is by complementing or even replacing the existing erbium-doped amplifiers. Spälter: “We can do that by Raman amplification. The ‘Raman effect’ is the transfer of energy from a high-frequency pump source to a lower-frequency signal. The bandwidth over which Raman amplification works is about 13 THz in optical fibers, which exceeds the combined width of the C and L bands.”

“But I am not a strong believer in that option, for two reasons. First, although you get to use a wider wavelength range, the overall loss characteristics in the optical fiber are worse. That is because the minimum fiber loss is at a wavelength of about 1,550 nm. The farther away the channels are from this optimum, the higher the fiber loss and the lower the overall reach are for these channels. Second, the Raman effect also acts among the signal channels: short wavelength channels transfer power to the long wavelength channels, resulting in another transmission-reach penalty. So there is a trade-off: what you gain in capacity, you lose in optical reach performance.”

Of the two remaining strategies, the second one – packing more channels into the available wavelength bands – has that same drawback to some extent, says Spälter. “For instance, you can increase the spectral density, using new, higher-level modulation schemes. But because you bring the channels together, you will get more linear and nonlinear crosstalk between them, so again what you gain in capacity, you lose in reach. Moreover, this option suffers from not being compatible with an infrastructure previously deployed with a grid containing larger spacing between channels.”

“Which leaves us with the third option, which I consider most promising: to keep the spacing between the channels, but to increase the per-channel data rate. It is this path that the industry has followed up to now: we went from 2.5 Gbps to 10, then to 40, and now to 100 Gbps. It would be logical to extend this, based on higher-level modulation formats similar to those used today in radio communications.”

“Since there is a reach penalty as well when moving from lower to higher per-channel data rates, at first sight it looks like we are going to face the same trade-off as with the other two options. However, I expect that mechanisms will be developed which mitigate the reach penalty. Such mechanisms are likely to be based on electrical pre- and post-processing of the optical signal at the transmitter and receiver, respectively.”

That means that the 20 Tbps threshold is not really a fundamental limit, but we require a new step, by means of new technology, to pass it. Spälter: “That is both optical and electronic technology. In the end, what matters to the customer is total cost of ownership, which is affected by capacity and reach, but also power consumption and footprint. We will have to offer small-footprint, low-power consumption solutions with high-reach-capacity product. As for power consumption, although data throughput will increase, the roadmaps of all major chip companies show a steady increase in gate density, which helps to reduce power consumption per bit. I see two focal points for research here. First, work should be performed on more advanced algorithms for digital signal processing, having the potential for improving the reach-capacity product. Second, optoelectronic integration technology should be further developed, that is, integration of electronic and photonic components – lasers, photodiodes, transistors, and so forth. Optoelectronic integration results in lower power consumption, and at the same time in reduced footprint. Note that footprint is a big issue: our customers, that is to say the carriers, will mostly want to deploy new technology in existing sites, for example the small huts they use along the fiber routes. Building new, larger huts is not an option.” Spälter expects that the migration from electronic switching to all-optical packet switching, if this happens, still lies much further ahead: "What is technically feasible today is optical burst switching in metro ring topologies: switching a large number of packets at the same time. That relaxes the time constraints for switching. Given the current status, I don't see all-optical-packet or burst-switching technologies to mature into the broad market within the next ten years."

Spälter has come to value his collaboration with COBRA in recent years. “The researchers that come to us from Eindhoven are exceptionally skilled. One of the professors is now doing his sabbatical for a couple of months with us. Two former PhD students got permanent positions. I hope our collaboration will continue to focus on this highly important research question: how can we tackle the 20 Tbps threshold?”

Stefan Spaelter
Coriant GmbH & Co. KG
St. Martin Str. 76
81541 Munich 
Phone: +49 89 87806592