Systematic estimation of memory effects parameters in power amplifiers' behavioral models

This paper deals with systematic behavioral modeling of power amplifiers through the study of the parameters involved in the memory effects phenomenon and the appropriate method for their estimation. The gained knowledge is integrated in both memory polynomial and real-valued time-delay neural network models; and, their linearization capability is investigated and compared to their empirical non-system based counterparts. According to the measurement results, the memory polynomial was required to be over dimensioned to achieve the same linearization performance obtained using a system memory parameters based one. It is also shown that the integration of prior knowledge of system to be modeled reduces the complexity and improves model robustness.