Design considerations for low-margin elastic optical networks in the nonlinear regime [Invited]

We demonstrate from a system design perspective that nonlinearity can be exploited to minimize the impact of system margins on system performance for both point-to-point links and elastic optical networks. A nonlinear interaction causes a 2 dB reduction in launch power to be reduced to $ \lt\!{0.25}\,\,{\rm{dB}}$ signal-to-noise ratio (SNR) penalty, and likewise, a 2 dB peak–peak (pk-pk) perturbation to the output power of an optical amplifier is reduced to $ \lt\!{0.25}\,\,{\rm{dB}}$ SNR penalty (for 5, 10, and 20 spans). Extending this to a gain ripple of 1 dB pk-pk with an internode spacing of ${5} \times {80}\,\,{\rm{km}}$, ${10} \times {80}\,\,{\rm{km}}$, and ${20} \times {80}\,\,{\rm{km}}$, the penalty is 0.4 dB, 1.5 dB, and 5.1 dB, respectively, with pre-emphasis reducing this to 0.01 dB, 0.3 dB, and 1.2 dB, respectively. In elastic optical networks, we consider the nonlinear relationship among SNR, margin, and the fraction of capacity available. We consider scaling internode distances of a 9-node German scale network (DT9), such that the initial network diameter increases from 1120 km to 6720 km (six-fold scaling). We generate 1000 different topologies based on the scaled DT9 node locations to quantify the impact of margin. For the unscaled DT9 network, a 3 dB margin results in, on average, a 21% reduction in network throughput; however, when the internode spacing is increased six-fold to a continental scale network, the network throughput is reduced by 40%, on average, for the same 3 dB margin.