Neural Network Predistortion Technique for Nonlinear Power Amplifiers with Memory

Digital predistortion is the most promising technique to overcome the nonlinearity of power amplifier (PA). The early work on the field of predistortion is mostly limited to memoryless nonlinear. However, memory effects are typically observed in PA for wideband applications. In this paper, a double-input double-output (DIDO) forward neural network predistorter combined a tapped delay line is proposed to compensate the nonlinear and memory effects of PA simultaneously. The proposed predistortion scheme based on in-phase and quadrature components of transmitted signal can reduce the additional computations of rectangular-to-polar conversion or avoid the complex signal operation as in previous publications, and even can correct any nonlinearity in the conversion process. The adaptive predistorter is realized using indirect learning architecture associated with the backpropagation algorithm. Simulation results using 16-QAM modulated signals demonstrate an outstanding linearization performance.