Complexity-Reduced Neural Network for Behavioral Modeling and Digital Predistortion of RF Wireless Transmitters

DOI number:10.1109/TIM.2025.3574905

Affiliation of Author(s):中国科学技术大学信息学院

Teaching and Research Group:电子工程与信息科学系

Journal:IEEE Transactions on Instrumentation and Measurement

Funded by:NNSF 62371436

Key Words:Digital predistortion (DPD), neural network, piecewise function, power amplifier (PA).

Abstract:Neural networks (NNs) are promising for behavioral
 modeling and compensation of complicated nonlinearities in 5G
 transmitters with broadband high-efficiency structural power
 amplifiers (PAs). The high computational complexity of NNs,
 however, poses a serious challenge to their practical imple
mentation. In response to this challenge, a complexity-reduced
 NN (CR-NN) approach is proposed in this paper, which builds
 on the relationship between data features and model capacity.
 Considering the memory fading property of RF PAs, the CR
NN first utilizes post-filtering to significantly reduce the input
 features of the NN body. This is followed by the employment
 of adaptive non-uniform piecewise linear unit to improve the
 model capacity without increasing the complexity. In order to
 validate the proposed method, the experiments are carried out
 based on a two-stage Doherty PA and a GaN-based Doherty PA.
 Experimental results show that the proposed CR-NN method can
 suppress the strong nonlinearity of PA from-23.24/-24.34 dBc to -51.18/-50.72 dBc with a computational complexity comparable to
 that of the linear-in-parameters models, thus demonstrating that
 the proposed method can significantly improve the performance
complexity tradeoff of NN-DPDs and contribute to the linearization of 5G and future systems.

First Author:Chengye Jiang

Co-author:Qianqian Zhang,Junsen Wang,Junning Zhang,Kunfeng Zhang

Indexed by:Journal paper

Correspondence Author:Bo Tang,Falin Liu

Document Code:10.1109/TIM.2025.3574905

Discipline:Engineering

Document Type:J

Volume:74

Issue:2532615

Page Number:1-15

Date of Publication:2025-05-31

Included Journals:SCI