A technique called PCS developed by Nokia Bell Labs will make it possible to get very close to the Shannon limit, the maximum amount of data that can be squeezed into a channel.
Nokia announced a new chipset it says will make it possible to boost the spectrum efficiency of optical networks to close to theoretical limits. The increases in capacity that will result in some instances could be as much as 65 percent.
Nokia said its Photonic Service Engine 3 (PSE-3) coherent DSP is the first chipset to implement a technique called probabilistic constellation shaping (PCS), which improves the spectrum efficiency of nearly any optical network to close to the Shannon limit, while also reducing energy consumption. The figure for energy reductions is nearly as impressive as the capacity gain from the spectrum efficiency boost; it can be as high as 60 percent, Nokia claims.
The technology is just as applicable to long distance lines (undersea cables, network backbones) as it is to shorter links. Capacity in optical networks diminishes with distance. The rate/reach tradeoff is inherent in optical networking and PCS does not change that. The maximum capacity limit will therefore be different on an undersea cable (perhaps 200 gigabits per second) than a link within a metro network (up to 600 Gbps).
The shorter distances are where this new technology is likely to be used first, however, according to Kyle Hollasch, director of product marketing for optical networks at Nokia. He told EDN the reason is that it’s the web-scale companies and communications service providers who have the more acute need for capacity increases.
Companies in these categories (cloud providers, colocation companies, cable network operators, etc.) are experiencing non-stop skyrocketing traffic associated with streaming video, big data applications, and the IoT, and they are keen to contain operational costs by squeezing every bit of performance out of their expensive fiber.
The marquee benefits are speed and energy-efficiency, but almost as important, Hollasch said, is that PCS simplifies the network. For starters, a network built with systems using the new chipset will require fewer optical transponders – up to 35 percent fewer, the company claims.
A network built with PCS will also be operationally simpler. “Traditionally there have been several parameters that need to be altered to achieve that rate/reach tradeoff, and they’ve had the effect of complicating networks. There have been a lot of technologies introduced in the last 10 years that gave us more capacity, but what they did was they made things more complex,” Hollasch said.
“What PCS does is, instead of having to juggle multiple different modulation formats and multiple baud rates, you can just stick with one. You can have one knob, and one format, and one size channel, and dial that up and down from 200 gig to 600 gig. And that’s a throwback – that’s how networks used to be 10-15 years ago, but they’ve gotten much more complicated with things like flex spectrum and increasing data rates and baud rates. This technology is delivering not just capacity but also making things easier.”
Essentially, PCS is another layer of processing on top of all the other signal processing that is occurring in optical signal transmission, Hollasch said.
The modulation scheme used in optical networks is 64 QAM. Ordinarily, the 64 channels are represented as a constellation of dots in a square or, when adding the time dimension, a cube. PCS eliminates the corners by reshaping the cube into a sphere (see illustration).
“When you run the math, you can fit the same number of constellation points into a ball as you can in a cube, but with less energy. You can think of the energy of those points as how far they are from the origin. The pointy corners, they’re inefficient, and you don’t want to use them,” Hollasch explained. “They’re the first to go as you dial down the rate. As you keep dialing down the rate to 580 gig to 520 gig to 500 gig to 450 gig, the energy is getting sucked into the middle of the constellation.
“As you dial down, you use the middle ones much more frequently,” Hollasch continued. “If you look straight through the middle of the ball, that’s the densest part. It’s just a more energy-efficient way to pack bits in. It allows us to fill the capacity of the pipe.”
The introduction of PCS by no means implies that the capacity limit of optical networks has been reached. It does mean that one of the factors in the capacity equation is nearly maxed out.
“What we’ve achieved is the ultimate in performance,” Hollasch said. “Ultimate doesn’t mean just the coolest, man, it also means the last. Any more big gains in spectral efficiency aren’t really going to be achievable, and if there are improvements, there will be diminishing gains at great engineering cost. It’s not that we’re not at max capacity; we just need to find alternate routes to more capacity.”
The PSE-3 is the company’s fourth-generation PSE chipset (it didn’t number the first one). PCS, making its first appearance in the PSE-3, was developed by Nokia Bell Labs.
The PSE-3 will be available across Nokia’s packet-optical portfolio, including a new version of the 1830 Photonic Service Interconnect, a compact modular WDM platform. The 1830 PSI-M will offer both cost-optimized and high-performance modules using the PSE-3. Nokia will have it be available starting in the third quarter.
Brian Santo has been writing about science and technology for over 30 years, covering cable networks, broadband, wireless, the Internet of things, T&M, semiconductors, consumer electronics, and more.