Nerual Network Modeling Of RF Power Amplifier And Its Implemen-Tation Of Software And Hardware

Power amplifier's (PA) linearization is one of the key technologies for wider bandwidth applications such as wideband code-division multiple access (WCDMA). At present, in contrast with feedback and feedforward, digital baseband predistortion is a highly cost-effective way to improve power amplifier's nonlinear characteristic. Modeling power amplifier is one of the key technologies when we design the predistorter. The PA system is complex, usually nonlinear, time-varying, and has memory effect. To realize real-time predistorter that performs well, cascade-correlation algorithm is used for online modeling the nonlinear characteristic of PA. First, residual correlation method is used to determine the structure and initial parameters of the neural network model for power amplifier system, then the model parameters are online adjusted with forgetting factor recursive least square algorithms(FFLS). Application shows that the neural network model reaches the performance index satisfactorily.Firstly the development background of the project is introduced and several means of power amplifier's linearization are discussed, then the related information of System Identification and Neural Network is introduced, and then modeling RF PA which is the key part of this paper is discussed, then hardware implementation of system with DSP is introduced. At last the summary of all the work is given, and besides, some beneficial advices are discussed.