A Multilevel Artificial Neural Network Nonlinear Equalizer for Millimeter-Wave Mobile Fronthaul Systems

We propose and validate a complex-valued multilevel artificial neural network nonlinear equalizer (ANN-NLE) for single-carrier 16QAM and 64QAM signals transmissions in the 60-GHz radio-over-fiber (RoF) transmission system. First, we analyze the nonlinearities in the fiber-wireless channels. Then, we introduce the principles of the multilevel ANN-NLE proposed for the millimeter-wave RoF systems. From the analysis of nonlinearities in a single-carrier optical transmission channel, it is essential to design a complex-valued ANN-NLE to mitigate the cross-modulation (XM) effects between the in-phase (I) and quadrature-phase (Q) components. Finally, we report on single-carrier signaling at 16QAM and 64QAM on 60 GHz, demonstrating the ability of the proposed ANN-NLE approach to minimize the nonlinear compression in an RoF system. As an added benefit, the complex-valued ANN-NLE has a great tolerance to phase rotations. The convergence time, activation function, step size, and number of taps are discussed in Section IV.