Estimation of Memory Length for RF Power Amplifier Behavioral Models

In RF power amplifier (PA) behavioral modeling, special emphasis has been put on memory-effects since they can have a determinant impact on the performance of the model. If too short a memory length is considered, it is not possible to fully characterize the PA, whereas using an excessively long memory length results in a slower, less efficient model. Therefore, it is vital to find the optimum memory length during model extraction. In this paper we present a novel approach to estimating the optimum memory length required for accurate behavioral modeling of RF power amplifiers. For example, using this method, we can directly find the optimum memory length needed in a time delay neural network or a Volterra series based model from the measured input and output signals of the PA in the time domain. Experimental examples are presented here for the case of discrete Volterra series based models