Pruned Basis Space Search for Digital Predistortion of RF Power Amplifiers

Abstract

A novel behavioral modeling technique called pruned basis space search (PBSS) is proposed for digital predistortion (DPD) of RF power amplifiers (PAs). The PBSS finds the optimal DPD model by basis function search in the pruned basis space (PBS). The PBS is obtained by sparsifying the basis space comprising a wide variety of basis functions, while the basis function search is implemented based on heuristic algorithms. A basis function multiplexing-based complexity identification algorithm is proposed to improve the fitness calculation so that the basis function search can balance the performance and running complexity of the behavioral model. The PBSS model avoids the shortcomings of traditional truncated models and various popular Volterra series-based behavioral modeling approaches and thus offers superior performance. The experimental part performs behavioral modeling and linearization tests on two different PAs. The experimental results confirm that the PBSS model can achieve a better tradeoff between linearization performance and complexity than the state-of-the-art Volterra series-based model.