Metrics and Methods for Benchmarking of RF Transmitter Behavioral Models With Application to the Development of a Hybrid Memory Polynomial Model

Abstract

This paper presents a study of performance evaluation metrics for behavioral models of power amplifiers and transmitters. A novel normalized absolute mean spectrum error criterion is proposed as a performance evaluation metric along with a method for accurate benchmarking of behavioral models and their ability to predict the in-band response, static nonlinearity, and memory effects of the device under test. The proposed metric and method are validated with a study of different memory polynomial based models, focusing on the model accuracy, complexity, and identification robustness. This experimental validation highlights the robustness of the proposed metric and its ability to accurately quantify the performance of several behavioral models in predicting the static nonlinearity and memory effects of the device under test for several test conditions. In addition, the results of the comparative study between the memory polynomial models are used to propose a hybrid memory polynomial model. The superiority of the proposed model is assessed by comparing its performance to that of the studied memory polynomial models.