基于改进型径向基函数神经网络的功放线性化

提出了一种基于改进型径向基函数神经网络(MRBFNN)的数字预失真线性化模型,用于更为精确地矫正宽带射频功率放大器的动态非线性。该神经网络模型的输入层使用传统的延时抽头以补偿功放的线性记忆效应,同时对每个抽头进行级数展开用于补偿功放的非线性记忆效应,从而更好地抑制功放的动态非线性失真。文中使用WCDMA 三载波信号对一个460MHz 的Doherty 功率放大器进行数字预失真线性化实验。实验结果表明,与传统数字预失真线性化模型相比,基于改进型径向基神经网络的数字预失真线性化模型能更好地抑制宽带功放动态非线性引起的带外频谱再生,其三阶互调(IMD3)失真最多可以抑制23dB,大大提高了功放的线性度,验证了所提出的数字预失真线性化模型的有效性。

Power Amplifier Linearization with Modified Radical Basis Function Neural Networks

This paper presents a novel digital predistortion linearization model based on the modified radical basis function neural network (MRBFNN), which is more accurate for dynamic nonlinearity correction of broadband radio frequency power amplifier (RFPA). The input layer of the neural network model using traditional time-delay tap to compensate linear memory effects of power amplifier. At the same time, a series expansion is added into every tap to compensate nonlinear memory effects so as to better suppress the dynamic nonlinear distortion of power amplifier. A three carrier wideband code division multiple access (WCDMA) signal is applied to a 460MHz Doherty power amplifier for the digital predistortion linearization experiment. Compared with traditional digital predistortion linearization models, the experimental result illustrates that the digital predistortion linearization model based on the MRBFNN can suppress the out-of-band spectrum regeneration caused by the dynamic nonlinear of broadband power amplifier better, and the third-order intermodulation distortion (IMD3) can be improved more than 23dB. The linearity of the power amplifier is largely augmented. This validation results verify the effectiveness of the proposed digital predistortion linearization model.