DISSERTATION Digital Pre-distortion of Microwave Power Amplifiers

With the advent of spectrally efficient wireless communication systems employing modulation schemes with varying amplitude of the communication signal, linearisation techniques for nonlinear microwave power amplifiers have gained significant interest. The availability of fast and cheap digital processing technology makes digital pre-distortion an attractive candidate as a means for power amplifier linearisation since it promises high power efficiency and flexibility. Digital pre-distortion is further in line with the current efforts towards software defined radio systems, where a principal aim is to substitute costly and inflexible analogue circuitry with cheap and reprogrammable digital circuitry. Microwave power amplifiers are most efficient in terms of delivered microwave output power vs. supplied power if driven near the saturation point. In this operational mode, the amplifier behaves as a nonlinear device, which introduces undesired distortions in the information bearing microwave signal. These nonlinear distortions degrade the system performance in terms of increased bit error rate and produce disturbance in adjacent channels. A compensation of the nonlinear distortions is therefore of significant importance, not only to keep the system performance high, but also to comply with regulatory specifications regarding the maximum allowed disturbance of adjacent channels. Nonlinear equalisation at the receiver is possible but complicated due to the unknown effects of the channel. Further, this method does not reduce the disturbance in adjacent channels, thus additional analogue filters would have to be placed at the output of the power amplifier. It is therefore natural to reduce the nonlinear distortions at the point where they occur, namely at the transmitter. Different linearisation methods exist which aim to reduce the nonlinear distortions while keeping the power amplifier in the nonlinear and efficient mode. Traditionally, these techniques employ additional analogue circuitry. Linearisation by digital pre-distortion is a new method which applies digital signal processing techniques for compensating the nonlinear distortions. Digital pre-distortion splits into three tasks: modelling of the microwave power amplifier, adaptive identification of the model parameters, and development of the pre-distortion filter. These tasks are addressed in this thesis. Further, a prototype system is developed which allows to test the pre-distortion algorithm in real-time using a fixed-point environment. For the first task, measurements on microwave power amplifier were performed in order to evaluate different models. The difficulty is to find low-complex but at the same time accurate models, which describe not only the nonlinear effects, but account also for the memory effects of the power amplifier. The adaptive identification of the parameters of two nonlinear models, namely a Volterra model and a Wiener model, is presented thereafter. Gradient-type algorithms are developed and investigated with respect to stability in a deterministic context. A powerful method for the determination of the pre-distortion filter is presented next. For nonlinear systems it is in general not possible to devise analytic solutions for the pre-inverse which linearises the system for a certain class of input signals. Here, an iterative technique is presented which finds an approximate solution for the pre-inverse. Based on the developed pre-distortion algorithm, a real-time prototype system is developed. This system proves that the algorithm can be implemented with a limited amount of hardware resources. Further, measurement results show that the algorithm keeps its excellent performance also in an environment with a limited dataand arithmetic accuracy.