基于神经网络的宽带功放动态非线性行为建模

Neural Networks Based Nonlinear Dynamic Behavior Modelling for Broadband Power Amplifiers

文章主要讨论了如何利用神经网络对宽带功放进行动态非线性行为建模的问题。首先简述了功放的动态非线性特性及行为建模的方法。然后回顾了基于实数时延前馈神经网络、径向基函数神经网络等浅层神经网络构建的功放动态非线性行为模型。在此基础上,针对5G/6G宽带功放具有更强的记忆效应的问题,重点分析了如何使用长短期记忆(LSTM)神经网络对功放的动态非线性进行精确的行为建模。最后展望了构建具有普适性的功放非线性行为模型将是5G/6G通信时代功放非线性建模的一个重要发展方向。

This paper mainly discusses how to use neural networks to build behavioral models for dynamic nonlinearity of broadband power amplifiers. Firstly, the dynamic nonlinear characteristics of power amplifiers and the method of behavior modeling are introduced. Then the dynamic nonlinear behavioral models of power amplifier based on the shallow neural networks such as the real-valued time-delay feedforward neural networks and the radial basis function neural networks are reviewed. On this basis, aiming at the problem that 5 G/6 G power amplifiers have stronger memory effects, this paper mainly focuses on how to use the Long Short-Term Memory(LSTM) neural networks to accurately model the dynamic nonlinearity of 5 G/6 G broadband power amplifiers. At last, it is expected that building a universal nonlinear behavior model of a power amplifier will be an important development direction of nonlinear modeling for power amplifiers in 5 G/6 G communication era.