Optical performance monitoring in 40-Gbps optical duobinary system using artificial neural networks trained with reconstructed eye diagram parameters

A technique using artificial neural networks trained with parameters derived from reconstructed eye diagrams for optical performance monitoring in 40-Gbps optical duobinary (ODB) system is demonstrated. Firstly, the optical signal is asynchronously sampled by short pulse in the nonlinear medium such as semiconductor optical amplifier and highly nonlinear fiber, the sampled and collected data is then processed by improved software synchronization algorithm to obtain reconstructed eye diagrams without data clock recovery. Secondly, the features of the reconstructed eye diagrams are extracted to train the three-layer preceptor artificial neural network. Finally, the outputs of trained neural network are used to monitor multiple optical signal impairments. Simulation experiments of optical signal noise ratio (OSNR), chromatic dispersion (CD) and polarization mode dispersion (PMD) monitoring in 40-Gbps ODB system is presented. The proposed monitoring scheme can accurately identify simultaneous impairment with the root-mean-square (RMS) monitoring error less than 3%.