Radio system characterisation and optimisation using non-linear power amplifier models

Abstract

Performance evaluation tools are becoming considerably important due to their usefulness in adaptive link deployment. Developing accurate and fast converging evaluation tools able to account for all hardware and channel impairments is not a straightforward task. Particularly, the performance characterization of the PA non-linearities is of notable significance as they greatly degrade the QoS. One of the aims of this thesis is to study the impact of the PA non-linearities in the system performance. An evaluation tool based on a Monte Carlo simulation which computes the BER over a scenario which considers an AWGN channel and the presence of a PA is used to carry out the study. In order to account for the PA behaviour in the proposed tool, several state-of-the-art memoryless methods are investigated: Saleh’s, Ghorbani’s, Rapp’s and the Bessel-Fourier. In particular, original propositions to improve the accuracy of the Bessel-Fourier model, in order to capture operating scenarios that reach beyond the device characterization, are presented together with first convergence studies and experimental verification of the technique for scenarios more complex than two-tone tests. The distortion generated by the PA may be mitigated by the use of pre-distortion. This thesis also proposes a novel data pre-distorter which enables compensation for the PA distortion occurring at APSK systems by providing the optimum value of the constellation ring ratio under PA compression. The accuracy of this methodology is tested by the use of the mentioned evaluation tool and a testbed. Finally, the PA load variation presented by a particular radiating element in an antenna array, due to the mutual coupling between all elements, is also studied in this thesis as this may compromise the effectiveness of pre-distortion techniques. To do so, a methodology to characterize the beamforming degradation due to mutual coupling in large arrays of active antennas is presented as well as measured results showing the output impedance variation impact on the performance of two GaN PAs: a GaN Class-B and a GaN Class-AB