This research work focuses on improving the performances of digital predistorters while maintaining low computational complexity for mobile and wireless communication systems. Initially, the thesis presents the fundamental theory of power amplifiers, overview of existing linearisation and memory-effects compensation techniques and reveals the current issues in the field. Further, the thesis depicts the proposed solutions to the problems, including the developed in-band distortion modelling technique, model extraction methods, memoryless digital predistortion technique based on distortion components iterative injection, baseband equalisation technique for minimising memory effects, Matlab-ADS co-simulation system and adaptation circuit with an offline training scheme. The thesis presents the following contributions of the research work.
A generalized in-band distortion modelling technique for predicting the nonlinear behaviour of power amplifiers is developed and verified experimentally. Analytical formulae are derived for calculating predistorter parameters.
Two model extraction techniques based on the least-squares regression method and frequency-response analysis are developed and verified experimentally. The area of implementation and the trade-off between the methods are discussed.
Adjustable memoryless digital predistortion technique based on the distortion
components iterative injection method is proposed in order to overcome the distortion compensation limit peculiar to the conventional injection techniques.
A baseband equalisation method is developed in order to provide compensation of
memory effects for increasing the linearising performance of the proposed predistorter. A combined Matlab-ADS co-simulation system is designed for providing powerful
An adaptation circuit is developed for the proposed predistorter for enabling its adaptation to environmental conditions.
The feasibility, performances and computational complexity of the proposed digital predistortion are examined by simulations and experimentally. The proposed method is tuneable for achieving the best ratio of linearisation degree to computational complexity for any particular application.