A generalized model based on canonical piecewise linear functions for digital predistortion

摘要

A generalized model based on canonical piecewise linear functions is proposed for digital predistortion systems. Unlike conventional models using n parallel mapping from ℝ1 to ℂ1, the generalized model maps ℝn to ℂ1. Inheriting all the analytical and computational advantages that the CPWL functions have, it is a minimal formulation and only involves calculating inner products and absolute values. Furthermore, the model could be easily extended to multiple-band DPD systems or DPD systems with long term memory effects modeling. Our tests show even with less number of coefficients the CPWL-based DPD model achieves 2 ∼ 6 dB better linearization performance than the GMP model.