Abstract:
Remote Radio Head of the C-RAN base station architecture was further simplified based on the proposed ROF fronthaul system. However, the ROF fronthaul system downlink includes nonlinear devices such as power amplifiers, so it still faces nonlinear distortion. It is a very important step to obtain the nonlinear characteristics through behavior modeling for building a predistorter for the whole system. In this paper, based on the modeling and analysis of the downlink behavior model of ROF fronthaul system, an improved complex RBF Neural Network (ICRBFNN) is proposed based on the radial basis function neural network, and using 4 MHz bandwidth LTE signal as a test signal. The paper compares it to the traditional power amplifier model, such as General Memory Polynomial (GMP), Augmented General Fractional Order Memory Polynomial (AGFMP), Real-Value Time -Delay Neural Network (RVTDNN), Real-Valued Time-Delay Radial Basis Function (RVTDRBF) and other power amplifier behavior models, to verify the validity of the model. It is verified by simulation experiments that ICRBFNN has achieved more than 3 dB modeling accuracy improvement compared with the traditional power amplifier model.