Behavioral Modeling of Radio Frequency Transmitters

Licentiatavhandling, 2009

The dependence of modern wireless communication systems on the fidelity of the transmitter has increased the importance of modeling these components. Behavioral modeling is a commonly used technique to find the input-output relationship of a system, without the need for the knowledge of the specific components. There have been many models proposed in the literature, which has made it difficult to choose models suitable for the application at hand. By analyzing and categorizing these models depending on the type of distortions they can describe, it becomes possible to understand their distortion handling capabilities. Such an analysis led to the design of two new behavioral models, one customized for power amplifiers and the other for modulators. Once different models are analyzed, their usefulness is determined in an experimental setup. Different models have different modeling capabilities. While some may be able to model any nonlinear function with high accuracy, others may be capable of modeling at low computational cost. The accuracy/complexity tradeoff for some commonly-used behavioral models is analyzed in this thesis. In Paper [A], a new behavioral model is proposed for power amplifiers that combines two commonly used modeling techniques. The performance of this model is shown to be better than the two models it is based on, and issues in identification are also discussed. In Paper [B], a new dual-input model is constructed for modulators. This model has the ability to describe nonlinear imbalance in a modulator. Different variants of this model are also proposed, to reduce the computational complexity. In Paper [C], a detailed analysis on the accuracy/complexity tradeoff for some commonly-used power amplifier behavioral models is presented. After finding the computational complexity in floating point operations per sample, an experimental setup is used to show that among models studied, the generalized memory polynomial model has the best accuracy/complexity tradeoff.