基于Volterra级数的功率放大器的建模与预失真技术研究

发布时间：2018-04-24 00:32

本文选题：数字预失真 + Volterra模型　；参考：《北京邮电大学》2015年硕士论文

【摘要】：随着无线通信系统的应用范围愈来愈广,无线通信频谱资源变得越来越拥挤,需要采用非恒包络调制方式提高频谱使用效率,这就需要对功率放大器进行线性化处理。数字预失真因其高效率、高灵活性、低成本等优势成为功放线性化应用的重点。基于Volterra级数的预失真器模型因其灵活性强,成本廉价且实现结构简单而普遍应用。本文主要针对基于Volterra级数的功率放大器的建模与预失真技术进行深入研究。主要工作和创新点包括： 1.本文介绍了准确复杂度减小的简化Volterra级数新模型。SV(Simplified Volterra)模型是Volterra级数模型的简化模型,当预失真模型的非线性阶数和记忆深度值较大时,模型复杂度较大。针对SV模型系数数目较大这一缺点,将SV模型进行改进为ACR-SV (Accurate Complexity-Reduced Simplified Volterra)模型。计算原始SV模型系数时,SV模型的模型系数数量为模型无记忆非线系数与模型记忆非线性系数相乘,致使模型复杂度较高。而ACR-SV模型在计算模型系数数目时,将模型的无记忆非线性与记忆非线性分开考虑,ACR-SV模型的模型系数为模型无记忆非线性系数与记忆非线性系数的和,在保证了模型的精确度,减小了SV模型的模型复杂度,实验验证ACR-SV模型优于MP (Memory Polynomial)、SV和ACR-GMP模型。 2.另外,本文又介绍了改进广义记忆多项式模型。针对于GMP (Generalized Memory Polynomial)模型的模型复杂度高的缺点,将GMP模型改进为MGMP (Modified Generalized Memory Polynomial)模型。MGMP模型是在GMP模型的基础上,改变了GMP模型三个子模型中各个记忆深度值对应的最大非线性阶数的值,去除GMP模型中对模型精确度影响较小的多项式项,推导得出了MGMP模型,并通过与MP模型与GMP模型作对比,验证了MGMP模型在模型复杂度和模型精确度上较MP模型和GMP模型有优势。
[Abstract]:With the wide application of wireless communication system, the spectrum resource of wireless communication becomes more and more crowded, which requires the use of non-constant envelope modulation to improve the efficiency of spectrum use, which requires linearization of power amplifier. Because of its high efficiency, high flexibility and low cost, digital predistortion has become the focus of power amplifier linearization application. The predistorter model based on Volterra series is widely used because of its high flexibility, low cost and simple structure. In this paper, the modeling and predistortion technology of power amplifier based on Volterra series are studied. Key areas of work and innovation include: 1. In this paper, a new simplified Volterra series model, I. e., simplified Volterra series model with reduced accurate complexity, is introduced. It is a simplified model of Volterra series model. When the nonlinear order and memory depth of the predistortion model are large, the complexity of the model is higher. Aiming at the disadvantage of large number of SV model coefficients, the SV model is improved to ACR-SV Complexity-Reduced Simplified Volterra model. When calculating the coefficients of the original SV model, the number of the coefficients of the SV model is multiplied by the non-linear coefficients of the model without memory and the nonlinear coefficient of the memory of the model, which results in the higher complexity of the model. When the ACR-SV model calculates the number of model coefficients, the model coefficients of ACR-SV model are considered as the sum of the model memoryless nonlinear coefficients and memory nonlinear coefficients separately from the memory nonlinearity, which ensures the accuracy of the model. The complexity of SV model is reduced, and the experimental results show that ACR-SV model is superior to MP memory model and ACR-GMP model. 2. In addition, this paper also introduces the improved generalized memory polynomial model. In view of the high complexity of the GMP generalized Memory model, the GMP model is improved to the MGMP modified Generalized Memory model .MGMP model is based on the GMP model. The maximum nonlinear order corresponding to each memory depth in three sub-models of GMP model is changed, and the polynomial terms in GMP model which have little influence on model accuracy are removed, and the MGMP model is derived. Compared with MP model and GMP model, MGMP model has advantages over MP model and GMP model in complexity and accuracy.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN722.75

【参考文献】