Optical Amplifier Response Estimation Considering Non-Flat Input Signals Characterization Based on Artificial Neural Networks

Optical communication systems are facing the challenge to become robust to dynamic operating conditions, thus requiring autonomous devices. Optical amplifiers are essential devices for establishing optical communication systems with good quality of transmission. However, optical amplifiers include noise and present a non-flat spectral response. As a consequence, simple regression methods are not suitable for modeling amplifier response. The amplifier output signal estimation has been studied in other works. However, all the previous works only use characterization data with a flat input signal power spectrum, which can decrease the accuracy of the previous estimators when the amplifiers are under real-world conditions. In this work, we proposed models based on an artificial neural network to estimate the optical amplifier output signal considering characterization data with non-flat signals. The results show that the previous models, proposed for flat characterization, do not perform well with non-flat data. The results obtained by the new models are promising since they returned a smaller estimation error than the previous models. We observed an error reduction of 4 dB when compared with the previous models under certain circumstances. Moreover, the proposed models are robust even when we use limited non-flat data in the training process.