BLOCK ADAPTIVE AND NEURAL NETWORK BASED DIGITAL PREDISTORTION AND POWER AMPLIFIER PERFORMANCE

摘要

The purpose of this paper is to compare the methodology and performance of two different techniques for digital predistortion. The first technique is the block adaptive digital predistorter, which operates using a modified least-mean-squares (LMS) algorithm. The second technique uses a feed-forward time-delay neural network to achieve its linearization performance. The performance of each of these techniques is evaluated for an orthogonal frequency-division multiplexing (OFDM) system with lOdB peak-to-average power ratio (PAR). For easy visual inspection of the tradeoffs, enabling preliminary analysis and teaching, a Java-DSP application has been developed. For more detailed analysis, a clustered MATLAB simulation environment has also been developed. By adding higher-ordered terms as inputs to the neural network, the authors have attained additional linearization near maximum power output for specific power amplifier models.