Experimental Evaluation of Predistortion Techniques for High-Power Amplifier

Modern communication systems suffer from the nonnegligible distortion caused by the inherent nonlinearity of high-power amplifiers (HPAs). As such, many methods have been proposed to mitigate this nonlinear effect. However, their performances are seldom compared, and their practicalities are rarely evaluated. In this paper, experimental setups to investigate two popular types of HPA systems have been developed: 1) the traveling wave tube amplifiers and 2) the solid-state power amplifiers. Experimental data were collected to evaluate four predistortion compensation techniques including the polynomial, rational function (RF), neural networks, and neurofuzzy methods. Using different performance measures, the experimental evaluation shows that the neuro approaches have a higher accuracy for some measures, but the tradeoff is a much higher computation cost. Overall speaking, the RF method seems to have the best balance between the compensation accuracy and complexity among the methods.