A Look-Up Table Assisted BiLSTM Neural Network Based Digital Predistorter for Wireless Communication Infrastructure

Neural networks are increasingly attractive for digital predistortion applications due to their demonstrated superior performance. This is mainly attributed to their ability to capture the intrinsic traits of nonlinear systems. This paper presents a novel hybrid predistorter labeled as the look-up table assisted bidirectional long-short term memory (BiLSTM) neural network (LUT-A-BiNN) that combines a neural network cascaded with a look-up table in a manner that both sub-models complement each other. The main motivation in using this two-box arrangement is to eliminate the highly nonlinear static distortions of the PA with the look-up table, allowing the neural network to focus on the compensation of the dynamic distortions. The proposed predistorter is experimentally validated using 5G test signals. The results demonstrate the ability of the proposed predistorter to achieve a 5 dB enhancement in the adjacent channel leakage ratio when compared to its single-box counterpart (BiLSTM neural network predistorter) while maintaining the signal-agnostic performance of the BiLSTM predistorter.