Dynamic Convolutional Neural Network for Digital Predistortion in Multistate PA Modeling
Abstract
This letter presents a dynamic convolutional neural network (DYCNN) digital predistortion (DPD) model for power amplifiers (PAs) under various operating states. The proposed model employs a convolutional neural network (CNN) as its backbone and incorporates expert convolutional kernels to model PA under multistate operating scenarios through linear weighting. This design achieves robust linearization performance and reduced model complexity without the requirement for real-time parameter updating. Measurement results demonstrated that, compared to state-of-the-art multistate neural network (NN) models, the proposed model exhibits the best PA linearization performance and reduces the total number of parameters by over 64%.
Fields of Study & Subjects
Advanced Power Amplifier Design
PAPR reduction in OFDM
Radio Frequency Integrated Circuit Design
Publication details
- Journal / source
- IEEE Microwave and Wireless Technology Letters
- Publisher
- Institute of Electrical and Electronics Engineers (IEEE)
- Volume
- 36
- Issue
- 6
- Pages
- 983–986
- Publication date
- 10 February 2026