Implementation of Neural Network Adaptive Digital Pre-distortion for Wireless Transmitters
Abstract
An Artificial Neural Network, more precisely Real Valued Spatiotemporal Neural Network (RVSNN) based real time adaptive digital pre-distorter (DPD) is proposed and implemented on FPGA for the linearization of nonlinear dynamic wireless transmitter. Power amplifier is the core component of wireless transmitter, and is the source of all the nonlinearities and distortions. To alleviate these distortions, DPD, designed based on the inverse characteristics of power amplifier, is the key technology in 3G and beyond wireless communications. Though off-line DPD ameliorates system performance considerably, it is still dependent on changes in system temperature, voltage, load mismatch and average signal power. In this regard, real time DPD provides increased stability along with the standard linearity and inter-modulation distortion requirements. But the proposed online RVSNN model is very sensitive to hardware delay and hard to realize, thus offline RVSNN model is implemented on FPGA which provides identical performance to its MATLAB counterpart.
Description
Keywords
Engineering--Electronics and Electrical
Citation
Hasan, M. M. (2015). Implementation of Neural Network Adaptive Digital Pre-distortion for Wireless Transmitters (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/26468