Linearization Schemes for Radio Over Fiber Systems Based on Machine Learning Algorithms

Abstract

This work is regarding the concept and implementation of pre- and post-distortion schemes idealized for radio-over-fiber (RoF) systems. Analog RoF systems have been considered potential to increase the capillarity of future mobile networks. However, the integration of the radiofrequency (RF) and optical systems might introduce nonlinearities that increase the in-band and out-band interferences. The proposed schemes employ a multi-layer perceptron (MLP) artificial neural network (ANN) linearization for reducing the signal distortions. We have applied our linearization schemes to orthogonal frequency division multiplexing (OFDM) signals for investigating its performance, in terms of root mean square error vector magnitude (EVM <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RMS</sub> ), normalized mean square error (NMSE) and adjacent channel leakage ratio (ACLR). Numerical results demonstrate promising linearization performance, since the EVM <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RMS</sub> has been kept as low as 3%, attaining NMSE and ACLR lower than below −30 dB and −35 dB, for input RF power up to 23 dBm.