RF Power Amplifier Modeling and Linearization with Augmented RBF Neural Networks

In this paper, we present an augmented radial basis function (RBF) neural network (ARBFNN) to linearize a wideband Doherty RF power amplifier with strong memory effects., A 51-dBm Doherty RF power amplifier and a 25 MHz mixed test signal were utilized to validate the performance of the ARBFNN nonlinear model and predistorter. The validation results illustrated that the ARBFNN nonlinear model outperforms the memory polynomial (MP) model and the real-valued time-delay neural network (RVTDNN) model with 3 and 5 dB improvements in the normalized mean square error respectively. The ARBFNN predistorter exhibits a significant improvement over the RVTDNN predistorter and MP predistorter in the suppression of the out-of-band spectral regrowth.