A new upper bound (UB) on the information rate (IR) transferred through the additive white Gaussian noise channel affected by Wiener's laser phase noise is proposed in the paper. The bound is based on Bayesian tracking of the noisy phase. Specifically, the predictive and posterior densities involved in the tracking are expressed in parametric form, therefore tracking is made on parameters. This make the method less computationally demanding than known non-parametric methods, e.g. methods based on phase quantization and trellis representation of phase memory. Simulation results show that the UB is so close to the lower bound that we can claim of having virtually computed the actual IR.
We consider a line with noise in the simplest case. Loss does not add noise. Amplification via phase insensitive amplifiers do add noise. A lower bound of this capacity is the quantum analog to the Shannon capacity of a linear channel with additive white Gaussian noise, namely the difference of the Von Neumann entropy of the signal plus noise at the output of the line and the entropy of the noise alone. We show that this expression is indeed the capacity for the case of an amplifier with infinitesimal gain
G=1+ε, and for a cascade of an amplifier with arbitrary gain and a large loss, such that the overall gain of the cascade is infinitesimal.