An Artificial Intelligence Based Efficient Digital Predistortion Method for 5G NR Communication Systems

Efficiency of 5G New Radio base stations depends on the nonlinear structure of power amplifiers on radio frequency transmitters. The effect of this restraint is greater on the wide frequency bands which are used on orthogonal frequency division multiple access based 5G New Radio wireless communication systems. On this paper, a digital predistorter system with power amplifier to raise the energy efficiency of radio unit on 5G New Radio base stations is proposed. The system is based on modelling the behavior of power amplifier with least squares algorithm and using this model for the purpose of designing digital predistorter with artificial neural network. The link level simulations are performed with using orthogonal frequency division multiplexing symbols of 10 MHz frequency bands to analyze the performance of designed digital predistorter. According to the simulation results, approximately %89 improvement on error vector magnitude and 45dB gain for normalized mean square error are achieved compared to the system which only artificial neural networks are used. Additionally, linearization values of input and output signals of the systems are compared since these values affect the efficiency of power amplifiers.