System-independent threshold and BER estimation in optical communications using the extended generalized gamma distribution

Two important tasks with respect to the optimized configuration of an optical communications system are those of the performance evaluation and the receiver decision threshold estimation. In this paper, a new training-based BER and threshold estimation technique is proposed relaxing the assumption of Gaussian distributed received signals. The proposed method is similar in philosophy to the Gaussian Approximation one, and is system-independent and simulation-based. This means that the probability density function (pdf) of the sampled electrical current is estimated based on training data provided via simulations without any assumptions on the specific configuration of the communications system under consideration. The novelty of the paper is that for the first time a combination of a generalized form of the gamma distribution together with the noncentral chi-square distribution have been used for the modeling of the pdfs of the spaces and the marks, respectively.