Behavioral Modeling of Wideband RF Power Amplifiers Using Complex-valued Wavelet Networks

RF power amplifiers are important sources of nonlinearities in communication systems. Based on complex-valued wavelet networks, a novel behavioral model for wideband RF power amplifiers exhibiting memory effects is proposed as an improvement to existing feed-forward neural network models. The complex backpropagation algorithm is applied to training the network so as to extract the model parameters. The performance of the presented model is evaluated by a comparison with a feed-forward neural network model. The results in the time domain and frequency domain illustrate that the proposed behavioral model provides a faster convergence rate and more accurate approximation, when characterizing wideband RF power amplifiers