基于BP神经网络的透明转发卫星功放预失真  

BPNN Pre-distortion Algorithm for Bent-Pipe Satellite High Power Amplifier Distortion Compensation

作　　者：杨茂强[1] 郭道省[1] 潘小飞[1] 

机构地区：[1]解放军理工大学通信工程学院,江苏南京210007

出　　处：《计算机仿真》2013年第9期177-181,共5页Computer Simulation

基　　金：国防预研基金（9140A22031010JB3801）

摘　　要：研究卫星通信功放性能优化问题,传统的预失真技术通常用来补偿地面功放的非线性失真或仅考虑卫星功放的失真补偿,线性化性能有限.为解决上述问题,提出了一种适合透明转发卫星的星地一体BP神经网络预失真算法.改进算法的学习结构同时考虑了卫星地球站固态功放和透明转发卫星功放的记忆非线性特性,利用带抽头延迟的BP神经网络作为预失真器,并结合收敛速度较快的Levenberg-Marquardt算法对其权值和阈值矢量进行自适应更新.仿真结果表明,经过神经网络预失真的星座图误差矢量幅度改善了84.67％,输出信号功率谱带外再生抑制提升近了13 dB,线性化效果十分显著.The traditional pre-distortion techniques are generally employed to linearize high power amplifiers （HPA） settled in ground-station or on board satellite merely,which limits the linearization performance.In this paper,a new pre-distortion algorithm,incorporating the inherent characteristics of satellite channel,was put forward for bent-pipe satellite HPA.In the improved learning architecture,the memory nonlinear distortion of earth station and bent-pipe satellite HPAs were both taking into consideration.Back propagation feed-forward neural network with tap delay line in input layer was utilized as pre-distorter and Levenberg-Marquardt was adopted for pre-distorter coefficients updating due to its fast convergence speed.The simulation results demonstrate that the error vector magnitude （EVM） of 32APSK signal constellation is significantly improved by 84.67％,the out-band power spectrum density （PSD） of signal achieves about 13 dB constraint,and ideal system linearization performance is acquired with predistortion.

关 键 词：卫星通信 高功率放大器 记忆非线性失真 反向传播神经网络 预失真 

分 类 号：TN911[电子电信—通信与信息系统]