Behavioral Modeling Of Power Amplifier Using Recurrent Neural Networks

With the fast development of wireless communication including the applicationsand researches on3Generation and4Generation, power amplifiers are becoming moreand more important. The design of power amplifier in the system level decides theperformance of system. In order to build a complete and more accurate model for poweramplifier, we need to research behavioral modeling technique for the feature andfunction of power amplifier. ANN can produce a fast and accurate response and can betrained using the training data without knowing the details inside the power amplifier.It cuts down the time of design and useful for the next design in the system level.Recurrent neural network (RNN) is an ANN model with recursive signal back to theinputs. This paper apply new technique extracting slow changing signal for modeling.The training and test data are time-domain data. With a good training, the RNN modelcan describe the behavior of power amplifier.This paper describes power amplifier modeling technique using RNN. RNN hasthe advantages of easy simulation conversion, good nonlinear responds, and lesssimulation time.The research work of this paper mainly includes two parts：1. Theory on power amplifier modeling based on RNN techniqueThis part firstly introduce some power amplifier modeling techniques andnonlinear features of power amplifiers. Secondly, the basic theory of artificial neuralnetworks is discussed. Thirdly, I describe the detail of recurrent neural networktechnique for power amplifier modeling to show the advantages of RNN modeling.2. Testify the performance of the recurrent neural network models using two poweramplifiers examplesIn order to show the result of RNN modeling technique, I choose two examples inreal applications to demonstrate the usage of RNN. I can build the power amplifiermodel with memory effects, which shows the RNN model can produce fast and accurateresponse. The result reach the destination of my work.