基于神经网络的TWTA非线性校正方法

The Method of Nonlinear Correction of TWTA Based on Neural Network

用于卫星转发器的高功率行波管放大器(TWTA)是非线性放大器,由于常工作在饱和状态导致非线性失真很严重.针对这种非线性变换特性,提出了一种基于神经网络的新的校正方法,主要步骤包括样本数据的预处理、网络结构的设计、采用LM算法进行网络训练,以及利用训练好的模型进行AM/PM或AM/AM校正等.通过仿真实验和对比分析表明,基于神经网络的校正方法优于传统校正方法,校正结果与真实值吻合程度高.

TWTA is a nonlinear amplifier applied in satellite translator, which is susceptible to nonlinear especially on the saturation point. Aiming at the property of nonlinear transformation, a new correction method based on neural network is presented. The main process is to preprocess samples, design the network structure, train the network with LM algorithm, correct AM/PM or AM/AM with the trained network and so on, The simulation experiment and comparison analysis show that the correction method based on LM algorithm is superior to traditional method, and the correction result is consistent exactly with the true value.