Efficient Behavioral Modeling Approaches of Radio Frequency Power Amplifiers for Bit Error Rate BER Computations and Digital Predistortion DPD Design

Abstract

The proliferation of high data-rate wireless communication systems with large operational distances, has increased the requirement for power during signal transmission. Radio frequency power amplifiers (RFPAs) which are present at the front end of these systems, fulfills this requirement of providing adequate power for signal transmission. However, operating the PAs in the nonlinear region for achieving high power efficiency distorts the signal causing degradation in the bit error rate (BER) performance of wireless systems. Modeling the nonlinear behavior and memory effects of the PA serves as a crucial step in evaluating the impact of PA on the performance of wireless systems in terms of BER and in identifying algorithms that will linearize and compensate it. Nonlinearity compensation based on input and output waveforms of the PA using digital predistortion (DPD) in a simulation framework significantly reduces the burden of complex circuit level design required for the linearization. newlineThe thesis presents a novel baseband generation method for modeling RF PAs that enables us to specify operational power range and bandwidth criteria of the PA, yielding a generic power amplifier model. The proposed modeling approach is useful for BER computations with PA distortions for any digitally modulated signal. Estimated BER using the proposed PA model for the test case of QPSK and 16-QAM, shows a significant difference in BER at high input back-offs due to memory effects, compared newlineto the conventional non-linear PA models. The thesis also investigates and presents the optimal parameter identification of DPD models for RF PAs using a modified differential evolution (MDE) based optimization algorithm. It was found that, applying MDE based optimization technique saves lot of computer resources and computational time involved in searching through all possible combinatorial models and enables an expedited method for deriving optimal DPD models. Further, when weighted information criteria is incorporated as objective function..