Prior Knowledge Input Neural Network for Microwave Power Amplifier Modeling

In this paper we present a novel neural network based behavioral level model for a microwave power amplifier (PA) utilizing a prior knowledge technique. Model parameters are extracted and verified using independent wideband digitally modulated signals in the time domain. Both quasi-memoryless and neural network models without any prior knowledge inputs are also extracted and tested with the same sets of data. Comparison of the various methods using goodness of fit statistics, show the new model presented here performs better and has improved generalization capabilities compared to models without prior knowledge inputs