A direct learning predistortion algorithm based on PSO in frequency domain

Time domain predistortion algorithms have some defects. In this paper, directly taking the outband spectrum leakage suppressing as the optimization object, we propose an algorithm for parameters identification of an predistorter by applying improved PSO (particle swarm optimization) algorithm with great generalization optimization capacity. Aiming at the problem of a large number of parameters, the proposed algorithm identifies part of the parameters in batches according to memory depth, and then optimizes all the parameters by taking the preceding results as the initial conditions. Simulation results demonstrate that ACLR (adjacent channel leakage ratio) can be suppressed below 50 dB with the predistorter designed in frequency domain to meet the requirements of 3 GPP standards.