I. Introduction
Memory effects observed in typical RF PAs normally originate from 3 sources: self-heating [1], electron trapping [2] and baseband effects [1], [3]. Usually, these sources affect the performance of PAs simultaneously, which makes an understanding of the impact of each source difficult. Consequently, PAs with memory effects are normally analyzed and modeled behaviorally on the basis of measured data. For instance, Volterra series [4] represents a typical behavioral modeling approach. However, a behavioral modeling approach does not offer any physical explanation and understanding of the mechanism behind the memory effects.