Implementation of Neural Network Based Digital Pre-distorter for Power Amplifier Linearisation

Digital pre-distortion refers to the use of digital signal processing to address nonlinearities caused by an analog RF front-end in a wireless transmitter. These nonlinearities can lead to nearby channel interference, reduce the transmitted signal's error vector magnitude, and often require the transmitter to operate at a less efficient but more linear region of the device. This work aims to implement a neural network-based digital predistortion algorithm. This pre-distortion algorithm is modelled using a time-delay kernel ridge regression (KRR) algorithm. This is an advanced machine-learning algorithm that considers memory effects. Both software and hardware platforms have been used for this application to produce effective results and discover a comparative comparison between the performance of software and hardware based on some parameters like accuracy, resource utilization, and operation frequency. The training and testing data of an already existing, simulated model is taken, and hardware implementation for the same is carried out using the Verilog HDL programming using the Xilinx Vivado tool.