Behavioral Model for RF Power Amplifiers with Dynamic Transmission Based on Pattern Sensing and Deep Neural Network

This paper proposes a novel behavioral model of power amplifiers (PAs) operated with multiple states based on a pattern sensing module and a deep neural network (DNN) module. The results of the pattern sensing module are taken as the input of the DNN. Different PA behaviors can be effectively recognized by sensing the amplitude-modulation to amplitude-modulation (AM-AM) patterns and then quickly modelled by one DNN. Experimental results show that PA working in multiple states can be modelled by the proposed method with good performance without updating or selecting model coefficients.