Skip to Main Content
This paper presents an alternative derivation of the Gaussian noise (GN) model of the nonlinear interference (NLI) in strongly dispersive optical systems. The basic idea is to exploit an enhanced regular perturbation expansion of the NLI, which highlights several interesting features of the GN model. Using the framework, we derive a fast algorithm to evaluate the received NLI power spectral density (PSD) for any input PSD. In the paper we also provide an asymptotic expression of the NLI PSD which further speeds up the computation without losing significant accuracy. Moreover, we show how the asymptotic expression can be used to optimize system performance. For instance, we are able to prove why a flat spectrum is best to minimize the NLI variance when the input is constrained to a fixed bandwidth and power. Finally, we also provide a generalization of the NLI GN model to arbitrarily correlated X and Y polarizations.
Date of Publication: Nov.15, 2013