基于RBF神经网络的射频功放行为模型研究

摘要

In order to design and optimize high-linearity Power Amplifier(PA) and communication subsystem,it is very important to build correct PA behavioral model in system-level simulation.Applying actual transistor testing board of PA, a great amount of input-output data of PA are collected from ADS simulation, it constructs a behavioral model based on RBF (Radial Basis Function) neural network.The structure design and two kinds of learning algorithm of RBF network are presented, and the model is tested.Results show that the PA behavioral model based on RBF has efficient precision,it has better approach activity and speed compared with the BP neural network.%为了设计和优化高线性功率放大器和通信子系统,在系统级仿真中,构建功率放大器精确的行为模型是极为重要的.应用实际功率放大器晶体管测试板,通过ADS(Advanced Design System)仿真得到大量功放输入输出数据,建立了一个基于RBF(Radial Basis Function)神经网络的行为模型,给出了RBF神经网络的结构设计及K-均值聚类算法和共轭梯度优化算法,并进行了模型检验.结果表明,基于RBF神经网络的功放行为模型具有较高的精度,相对于BP神经网络模型具有更高的逼近能力和速度.