Measurement-Based Optimization of Channel Powers With Non-Gaussian Nonlinear Interference Noise

We present two enhancements that expand the utility of convex optimization of channel powers in wavelength-division-multiplexed systems with nonlinear interference noise (NLIN). First, we present a first-order perturbation theory-based model that can quantify NLIN without assuming any particular noise distribution, such as circular Gaussian. The model supports a convex form for the problem of maximizing the minimum channel SNR in the presence of NLIN. Second, we describe how to perform the channel power optimization based on the measurements by receivers in a live system, rather than the numerical modeling of the system. Such measurement-based optimization avoids impairments caused by the modeling errors or incorrect choice of model parameters. We verify our approach through split-step Fourier simulations of a system with significantly noncircular NLIN. We obtain a gain of 0.25 dB in minimum margin over an optimized one-dimensional power allocation in a system with 24 channels with alternating 16-QAM and 32-QAM modulation and a tilted noise spectrum.