基于神经网络的射频功放非线性模型研究

发布时间：2018-02-28 11:05

本文关键词： 非线性功率放大器记忆效应行为模型径向基函数　出处：《宁波大学》2015年硕士论文　论文类型：学位论文

【摘要】：由于其优良的逼近能力,神经网络正逐渐成为一种越来越有吸引力的解决功率放大器建模的方案。在不清楚功率放大器内部结构的基础上,人工神经网络也可以根据采集得到的功放输入以及输出数据,来学习到该电路的内部行为。因此,本文首先从单波段射频功率放大器的非线性特点着手研究,然后介绍了功率放大器的记忆效应,对该效应引起的原由进行了分析,并介绍了几种常见的功率放大器记忆模型,以及这些模型之间的关系;随后介绍了神经网络相关理论知识,利用神经网络中的径向基函数(RBF)神经网络对射频功放进行行为模型建立;最后对双波段功率放大器的非线性特性进行理论研究以及相关的一系列实验,并对双波段射频功放进行建模。主要工作包括:(1)研究单波段射频功放的非线性特性理论,阐述了给射频功放带来这种失真的根源,以及非线性失真对信号放大的一些影响;接着对记忆效应进行分类以及分析其产生的缘由,其中记忆效应与输入信号的带宽密不可分,带宽越宽记忆效应越深越明显,此时,AM/AM以及AM/PM曲线发散的越厉害,所以记忆效应对建模精度造成的影响不容小觑。本文后面的建模等相关工作都是在该理论的指导下所进行的。(2)研究了单波段射频功放的行为模型,提出了两种基于神经网络的单波段射频功放行为模型:第一种是交叉实数时延径向基函数神经网络,第二种是递归型径向基函数神经网络。第一种模型在传统RBF神经网络模型基础上,在输入端加入了多项式交叉项和延迟项,交叉项用于模拟功放的强非线性,延迟项用于模拟功放的记忆效应。实验结果表明:该模型可以适合射频功率放大器行为模型,并且效果非常好,与反向传播(back propagation,BP)网络的实数延时神经网络相比,收敛速度快,大大缩短了训练时间。为了进一步提高建模的精度,接着提出了递归型径向基函数神经网络模型,该模型在前一个模型的基础上将输出项反馈到模型的输入端进行功放建模,实验证明,对于强非线性的射频功率放大器的建模,模型效果很好,拟合精度高,且收敛速度更快。(3)研究了双波段射频功放的行为模型,首先分析了双波段的非线性特性,进而阐述了双波段对于现代通信的重要性,接着阐述了双波段信号如何同步以及获取问题,最后提出了基于径向基函数网络的射频功放双波段行为模型。
[Abstract]:Because of its excellent approximation ability, neural network is becoming a more and more attractive solution to the modeling of power amplifier. The artificial neural network can also learn the internal behavior of the circuit according to the input and output data of the power amplifier. Therefore, the nonlinear characteristics of the single-band RF power amplifier are studied in this paper. Then, the memory effect of power amplifier is introduced, and the original cause of this effect is analyzed, and several common memory models of power amplifier are introduced, as well as the relationship between these models. Then the related theory knowledge of neural network is introduced and the behavior model of RF power amplifier is established by using the radial basis function (RBF) neural network in the neural network. Finally, the nonlinear characteristics of dual-band power amplifier are studied theoretically and a series of experiments are carried out, and the model of dual-band RF power amplifier is established. The main work includes: 1) studying the theory of nonlinear characteristics of single-band RF power amplifier. The source of this distortion to RF power amplifier and the influence of nonlinear distortion on signal amplification are expounded, and then the memory effect is classified and analyzed, in which the memory effect is closely related to the bandwidth of input signal. The wider the bandwidth, the deeper the memory effect is, and the more the AM / AM and AM/PM curves diverge, Therefore, the effect of memory effect on modeling accuracy is not to be underestimated. In this paper, modeling and other related work are carried out under the guidance of this theory. (2) the behavior model of single band RF power amplifier is studied. Two kinds of radio frequency power amplifier behavior models based on neural network are proposed. The first one is crossover real number delay radial basis function neural network. The second is the recurrent radial basis function neural network. The first model is based on the traditional RBF neural network model, and the polynomial crossover term and the delay term are added to the input end. The crossover term is used to simulate the strong nonlinearity of the power amplifier. The delay term is used to simulate the memory effect of power amplifier. The experimental results show that the model is suitable for the behavior model of RF power amplifier, and the effect is very good. Compared with the real number delay neural network of back-propagation propagation, the convergence speed of the model is faster than that of BPN. In order to further improve the accuracy of modeling, a recurrent radial basis function neural network model is proposed, in which the output term is fed back to the input side of the model for power amplifier modeling. Experimental results show that the model is effective, accurate and convergent, and the behavior model of dual-band RF power amplifier is studied. Firstly, the nonlinear characteristics of dual-band RF power amplifier are analyzed. Then, the importance of dual-band for modern communication is expounded. Then, the synchronization and acquisition of dual-band signals are discussed. Finally, a dual-band behavior model of RF power amplifier based on radial basis function network is proposed.
【学位授予单位】：宁波大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN722.75;TP183

【参考文献】