【作者】 李爱红；

【导师】 张尔扬；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2008， 博士

【摘要】 数据中继卫星系统(DRSS)以其高覆盖率、大容量、使用灵活等优点成为航天领域的一个重点发展方向,其返向通信链路高速数据传输是它的主要业务之一。高速数据传输系统发射端的数字信号处理面临着运行速度快、计算量大等难点;且发射端射频设备的非理想因素容易导致宽带信号失真、频谱扩散、系统性能下降。为解决这些问题,论文重点研究了基带信号成形滤波、宽带正交调制器I/Q失配预校正、高功率放大器(HPA)自适应预失真线性化等关键技术。首先,论文完成了高速成形滤波器的设计和实现。从系统误比特率(BER)性能和硬件可实现性两个角度,对成形滤波器的参数进行了精心设计;在FPGA中实现了基于查找表的高速成形滤波器,并利用成形滤波器的多相实现结构完成了对I/Q两路信号的自动同步。其次,论文研究了宽带正交调制器I/Q失配预校正方法。研究了宽带正交调制器I/Q失配模型,并给出了一种时域内I/Q失配测量方法;推导了TCM-8PSK信号在正交调制器I/Q失配条件下的星座图、BER性能;研究了宽带正交调制器I/Q失配的预校正方法,并将之应用到工程系统中,利用修改成形滤波器的抽头系数实现I/Q失配的预校正,不需要占用另外的硬件资源。然后,论文研究了HPA有记忆效应的非线性特性及其线性化技术。在对HPA有记忆效应的非线性模型总结的基础上,提出了一种基于并行滤波器组的扩展Volterra-Wiener模型;完善了TCM-8PSK信号在HPA有记忆效应非线性条件下的性能分析,包括信号星座图、功率谱密度(PSD)、BER性能等;对HPA常用线性化技术进行了介绍,对每种技术的基本原理、优缺点作了详细比较。最后,论文研究了HPA基带自适应预失真线性化技术。对基于Volterra级数和扩展Volterra-Wiener模型的自适应预失真算法进行了推导和仿真,仿真结果表明这两种算法系统收敛速度较慢;提出了将离散小波变换(DWT)应用到预失真模型中的方法,以减小输入序列的相关性,提高系统的收敛速度;针对强非线性HPA提出了一种基于离散小波神经网络(DWNN)的HPA数字自适应预失真算法,充分利用了神经网络对强非线性系统很强的学习和逼近能力。仿真结果表明,基于离散小波神经网络的自适应预失真算法与基于RBF神经网络的算法相比较,大大减少了系统的计算量,并提高了系统的收敛速度和HPA线性化性能。更多还原

【Abstract】 Data Relay Satellite System (DRSS) is a promising research area in the field of spaceflight thanks to its near-world wide coverage, large capacity and agile controllability. High speed data transmission of the satellite communication in the backward channel is one major service of DRSS. There are many signal processing problems such as high speed operation, a great quantity of computation at the transmitter of high speed data transmission system. And nonideal factors in RF equipments may cause wide-band signal distortion, spectrum spead and system performance degradation. To solve these problems, the key techniques such as high speed pulse shaping filter, predistortion of wide-band I/Q modulator mismatch and adaptive predistortion of High Power Amplifier(HPA) are mainly researched in this thesis.Firstly, the design and implementation of high speed pulse shaping filter is investigated. The filter’s parameters are well-chosen based on bit error rate(BER) performance and realization possibility. The filter is implemented using look-up tables in FPGA, and realized I/Q auto-synchronization with the filter’s polyphase structure.Secondly, the predistortion technique of wide-band I/Q mismatch is investigated. The I/Q mismatch model for wide-band I/Q modulator is studied and a measure method is given. The performances of TCM-8PSK such as signal constellation and BER are analyzed in the situation of I/Q mismatch. One predistortion technique is realized in FPGA by modifing the coefficients of pulse shaping filter, which saves some handware resources.Thirdly, the nonlinearity with memory effects and the linearization techniques of HPA are inverstigated. An extended Volterra-Wiener model based on parallel filters bank is proposed after considering various nonlinearity models. The performances of TCM-8PSK such as signal constellation, power spectrum density(PSD) and BER are analyzed in the situation of nonlinearity with memory effects. The usual HPA linearization techniques are introduced, and comparing their principles, advantages and disadvantages.Finally, the adaptive predistortion techniques of HPA in baseband are inverstigated. The adaptive predistorted algorithms based on Volterra and extended Volterra-Wiener are deduced and simulated. The simulation results show that the convergence speeds of the two algorithms are slow. The Discrete Wavelet Transform(DWT) is proposed to use in the predistortion model to reduce the correlation of input sequences. The simulation results show that DWT can make the convergence speed faster. An adaptive predistorted algorithm based on Discrete Wavelet Neural Network(DWNN) is proposed for strong nonlinear HPA, which makes use of the learning and approaching capabilities of neural network. The simulation results show that the DWNN method can make the convergence speed faster, reduce the computation complexity and improve the linearization performance when compared to the RBF method.更多还原