Behavioral modeling of power amplifiers using fully recurrent neural networks

This paper describes the first implementation of fully recurrent neural networks (FRNN) for behavioral modeling of power amplifiers. The proposed recurrent neural network model utilizes global feedbacks and full interconnections between neurons in hidden layers. The model has successfully been trained with W-CDMA signal. Then it is tested with not only W-CDMA signal but CDMA-1S95 and 2-tone signals as well. Good agreements between measured and modeled results have been achieved.