2017 Volume 6 Issue 8 Pages 484-489
We studied a novel nonlinear compensation scheme using digital signal processing based on a neural network (NN). Distortion of 16QAM signals caused by self-phase modulation (SPM) was compensated for by using a three-layer NN without feedback loops. In the scheme, the input layer of the NN has feedforward tapped delay lines. Input signals of I and Q components of the 16QAM signals are fed into the delay lines. The compensated 16QAM signals of the I and Q components are outputted by two neurons in the output layer of the NN. BER and EVM performance was investigated by numerical simulations, and the EVM was improved by more than 20% by the compensation.