一种基于神经网络的非线性卫星HPA预失真器  

Neural network based pre-distorter for linearization of HPA nonlinearity in satellite communications

作　　者：徐勇[1] 童新海[1] 刘渊 

机构地区：[1]解放军理工大学通信工程学院,江苏南京210007 [2]解放军防化学院,北京102205

出　　处：《解放军理工大学学报（自然科学版）》2011年第6期569-573,共5页Journal of PLA University of Science and Technology(Natural Science Edition)

基　　金：江苏省自然科学基金资助项目(BK2009057)

摘　　要：卫星通信中放大器使用的行波管(TWT)会引起信号的非线性畸变,对通信质量造成较大影响。为解决这种不利影响,在放大器前端采用非线性预失真,能较理想地消除放大器的非线性。由于神经网络能够对非线性函数进行较好地拟合,可以将其引入预失真器的设计。为简化神经网络中的LM算法,提出了一种采用LMS算法的系统模型,并建立新的自适应预失真器的结构模型,大大降低计算复杂度,有利于系统性能的提高。仿真表明,采用含有一个隐层(9个神经元)的神经网络模型来设计预失真器,能够达到较好的预校正效果。The travel-wave-tube（TWT） used in the HPA in satellite communications may unexpectedly cause nonlinearity to the signals.A pre-distortion will reduce these distortions.The pre-distorter is based on a feed-forward neural network.To simplify the traditional Levenberg-Marquardt（LM） Algorithm for neural-network,a new least mean square（LMS） algorithm based model was proposed in which way the complexity of the algorithm was reduced.Simulation results confirmed that the proposed pre-distorter based on neural network consisting with one hidden layer and nine neurons produces a good performance improvement of quality of the transmission.

关 键 词：神经网络 预失真 高增益放大器 行波管放大器 最小均方算法 

分 类 号：TN927.2[电子电信—通信与信息系统]