Parameterized basis functions for numerically stable behavioral modeling of RF power amplifiers

Abstract

In this paper, we propose a new set of parameterized basis functions suitable for behavioral modeling of RF power amplifiers exhibiting memory effects. Depending on the input signal distribution, the proposed basis functions can be optimized for numerically stable model identification procedure. The efficiency and robustness of the resulting polynomial model are demonstrated and compared to the ones of previously published polynomial models. The obtained results revealed an exceptional numerical stability regardless of the input signal statistics, making the proposed new model suitable for multimode and broadband nonlinear wireless transmitters.

Modeling of Deterministic Output Emissions of Power Amplifiers Into Adjacent Receive Bands

“…This fact suggests a way to enhance the robustness by removing some terms in the model. The condition number has been used as a metric for model performance, e.g., in [23], and also as a criterion to find basis functions for modeling the main band and around [24]. However, we use it here as a criterion for model pruning, and more specifically, for pruning the RxBN model.…”