An Optimized Digital Pre-distortion Method Based on Neural Network

Lie Zhang,Yan Feng
DOI: https://doi.org/10.3969/j.issn.1000-2758.2014.06.025
2014-01-01
Abstract:An optimized digital predistortion (DPD) approach of radio frequency (RF) power amplifier (PA) is proposed;It employs real-valued neural network model based on low-order generalized memory polynomial, which is optimized with genetic algorithm.It cascades genetic algorithm optimized low-order generalized memory polynomial and neural network model to increase matched degree between correction model and distortion of the PA. It can not only improve the correction ability of the model but also accelerate convergence speed of the model.The 60MHz LTE 3 carrier signal is employed to do measurement.Results show that the proposed approach is 6 dB better than real-valued focused time delay neural network model in ACLR and gives faster convergence speed.
What problem does this paper attempt to address?