Gain prediction of dual-pump fiber optic parametric amplifier based on artificial neural network

Abstract

Optimized parameters of dual-pump fiber optic parametric amplifier (FOPA) to give optimized FOPA gain can be obtained through optimization techniques. However, it is complicated to determine the multi-objective functions (gain, bandwidth and flatness), multi decision variables and multiple global solutions. Optimization works only considered undepleted pump configuration or pump depletion but without fiber loss. Recently, a machine learning approach was applied to design a Raman amplifier. Thus, this study intends to design a desired dual-pump FOPA gain utilizing an artificial neural network (ANN) to predict pump powers and pump wavelength by considering pump depletion and fiber loss. First of all, the FOPA training gain data were obtained through the 6-wave model and supplied into the ANN to learn the relation between the gains with their pump wavelengths and pump powers. Once the smallest mean square error (MSE) between input and target was obtained, the ANN model was saved. The ANN model can be used to predict the desired pump wavelengths and pump powers if the desired gain is given. The desired gains of constant values from 10 to 45 dB over 1540–1589 nm for optical communication are predicted very well with mean absolute error (MAE) of 1 dB variations.