Enhancing DPD of Wideband Power Amplifiers Based on LSTM and BiLSTM with Integrated Self-Attention Layers

As 5G waveforms evolve towards larger bandwidths and higher peak-to-average power ratio (PAPR), the issue of nonlinear distortion introduced by high power amplifiers (HPA) in radio frequency (RF) channels becomes more prominent. Digital predistortion (DPD) is a commonly used technique to mitigate this nonlinearity. Due to the electrical memory effect of HPA, various neural network models for time series prediction have been applied to DPD. Meanwhile, in recent years, the attention mechanism has been widely used in the field of deep learning. This paper integrates the attention mechanism into the classic long short term memory (LSTM) and bidirectional long short term memory (BiLSTM) neural networks, and conducts experiments on a 5G Test Model (TM) signal with a bandwidth of 200MHz using adjacent channel leakage ratio (ACLR), error vector magnitude (EVM) and normalized mean square error (NMSE) as evaluation metrics. The results demonstrate that DPD based on this method exhibits good linearization performance.