一种基于直接学习结构的数字预失真方法

A digital predistortion method using the direct learning architecture

针对宽带信号功率放大器(PA)的非线性效应和记忆效应,提出了一种基于直接学习结构的数字预失真(DPD)方法。该方法结合牛顿法进行参数提取,降低了参数迭代次数和运算量。以20 MHz带宽的64QAM信号作为输入信号,采用记忆多项式(MP)模型的预失真器以及Wiener功放模型进行仿真。仿真结果表明,该方法能有效补偿放的非线性失真,系统经过6次迭代后,其归一化均方误差(NMSE)可达-65.83 d B,误差矢量幅度(EVM)降低到0.06%,邻道功率比(ACPR)可达-45.33 d Bc。

To compensation the nonlinear distortion and memory effects of the wideband power amplifiers,a digital pre-distortion method based on direct learning is proposed. Combined with the Newton algorithm,this method can reduce the iteration numbers and the amount of calculation. The simulation is proceeded using a 20 MHz 64QAM signal,taking the memory polynomial model for predistorter,and the Wiener model for power amplifier. Simulation results show that the method could achieve an outstanding performance after the 6 iterations,the system＇s normalized mean square error（NMSE）can reach-65.8 dB,the error vector magnitude（EVM）could reduce to 0.06% and the adjacent channel power ratio（ACPR）can reach -45.33 dBc.