论文部分内容阅读
提出一种三层前馈输入延迟神经网络模型用于建立射频功率放大器有记忆的非线性行为模型.反向传播算法用来训练神经网络以提取神经网络模型参数.在仿真中,提出的模型可以通过典型的偶次多项式有记忆功率放大器模型来证明,并且比较了不同数目的延迟单元和隐层神经感知器的模型结构下的性能.为了用实验验证模型的有效性,建立了能达到60×106sample/s采样率数字测试平台用于采集功率放大器的输入和输出数据.选用矢量信号源产生的3·75MHz16-QAM信号作为功放输入信号来测试功放的动态AM/AM和AM/PM特性.通过分析比较时域和频域仿真结果和实验测试结果,模型在收敛性、精度和执行效率方面都达到很好的效果.
A three-layer feedforward input delay neural network model is proposed to establish a memory nonlinear behavioral model of RF power amplifier.The back propagation algorithm is used to train the neural network to extract the neural network model parameters.In the simulation, the proposed model can Through the typical even order polynomial memory power amplifier model to prove, and compares the different number of delay units and hidden layer neural sensor model structure performance.In order to verify the validity of the model experiment, the establishment of a can achieve 60 × 106sample / s sampling rate digital test platform for collecting the input and output of the power amplifier data. The use of vector signal generator 3.75MHz16-QAM signal generated as the amplifier input signal to test the dynamic AM / AM and AM / PM amplifier features through By analyzing and comparing the time domain and frequency domain simulation results and the experimental test results, the model achieves good results in terms of convergence, accuracy and execution efficiency.