Towards the All-Digital C-RAN: Centralization of DPD Using All-Digital Receivers and Neural Networks for Dynamic Nonlinear PA Behavior
Domingues, J. M. D. D.
; S. Pereira, S.
; Almeida, L.
; Silva, H.
;
Oliveira, A.
;
Carvalho, N.B.C.
Towards the All-Digital C-RAN: Centralization of DPD Using All-Digital Receivers and Neural Networks for Dynamic Nonlinear PA Behavior, Proc IEEE SBMO/IEEE MTT-S International Microwave and Optoelectronics Conference (IMOC), Campina Grande - Paraíba, Brazil, Vol. , pp. - , November, 2025.
Digital Object Identifier:
Abstract
This paper introduces a digital predistortion (DPD) solution that is located at the central unit (CU). This solution uses a field-programmable gate array (FPGA)-based all-digital receiver (ADR) in the feedback loop of the DPD, allowing for a cost efficient and scalable approach to move the processing tasks from the remote radio head (RRH) to the CU. The DPD system uses neural networks (NNs) combined with signal features to enable real-time adaptation to the dynamic nonlinear behavior of power amplifiers (PAs), eliminating the need for frequent updates to the DPD model’s coefficients. The effectiveness of this method is confirmed through experimental validation, which demonstrates successful linearization across varying signal power, bandwidth and modulation schemes.
