Joint-Encoding-Based Multimodel Optimization for Digital Predistortion of RF Power Amplifiers

In digital predistortion (DPD) of power amplifiers (PAs), the intricate nonlinear characteristics of PAs make it a critical challenge to construct a high-performance basis function space and to design an efficient basis function search algorithm. To address this challenge, this article first introduces the threshold amplitude polynomial (TAP) model and formulates it within a novel genetic algorithm (GA)-based model encoding framework, in which the basis function structures of the TAP model are sequentially encoded through binary bits. Building upon this framework, a piecewise polynomial modeling-based cascaded encoding strategy is further introduced to enhance the flexibility of the DPD scheme, where all encoded basis functions from the multiple submodels are concatenated to form a unified cascade chromosome. Subsequently, a joint-encoding-based multimodel optimization (JEMO) method is proposed, which leverages the encoded basis function structures within submodels through iterative GA operations to provide feedback for adjusting segmentation boundaries, thereby tightly integrating adaptive segmentation rules with the structural optimization of multiple submodels. To validate the effectiveness of the proposed model, experiments are conducted on two different Doherty PAs. The results demonstrate that the proposed JEMO model exhibits a significant advantage over other models in achieving the tradeoff between performance and complexity.