Generalized Nonuniform Memory Polynomial Model for Digital Predistortion of RF Transmitters

DOI number:10.1109/JMW.2026.3705081

Affiliation of Author(s):南京航空航天大学集成电路学院，中国科学技术大学信息学院

Journal:IEEE Journal of Microwaves

Funded by:NSFC 62371436, 62571234

Key Words:Basis function, digital predistortion (DPD), memory fading, memory width, power amplifiers (PAs), truncated models, Volterra series.

Abstract:A novel generalized nonuniform memory polynomial (GNMP) model is proposed for be havioral modeling and digital predistortion (DPD) of RF power amplifiers (PAs). The GNMP model uses a structure with simultaneous fading of memory depth and relative memory width. This paper proposes the concepts of “memory width” and “relative memory width” based on the close relationship between cross-terms and memory effect in Volterra basis functions. The memory width compensates for the short coming of using only the memory depth to describe the memory effect while avoiding the performance bottleneck caused by the fixed cross-term form. The even-order basis function is also introduced to improve the performance of the behavioral model with a further simplified form. The structural features of the GNMP model allow for more freedom in static nonlinearity and dynamic memory effect, making it superior in modeling and linearizing PAs with complex characteristics. Experimental tests on two different PAs show that the proposed GNMP model can achieve almost the same or better performance than the traditional model for the same complexity. More importantly, at an acceptable complexity, the GNMP model can further break the performance bottleneck of the traditional model by a few dB of typical metrics.

First Author:Renlong Han

Co-author:Falin Liu

Indexed by:Journal paper

Correspondence Author:Siping Gao

Document Code:10.1109/JMW.2026.3705081

Volume:Online

Issue:Online

Page Number:1-15

Translation or Not:no

Date of Publication:2026-06-15