[Back]


Diploma and Master Theses (authored and supervised):

H. Ruotsalainen:
"Investigation of Orthogonal Basis Expansions for Adaptive Wiener Models";
Supervisor: R. Wichman, M. Rupp, R. Dallinger; Institute of Communications and Radio-Frequency Engineering, 2010.



English abstract:
In modern wireless networks, radio frequency (RF) power amplifiers (PA) are essential components which are inherently nonlinear. However, the use of spectrally efficient modulation techniques and densely populated transmission bands require stringently linear behaviour of the PAs. Currently, commercial RF PAs which are operating in their linear region, show poor power efficiency. In order to increase the efficiency, linearization methods are necessary. Predistortion in the digital baseband is considered to be an efficient method to compensate for the nonlinear effects of PAs and can also be combined with other linearization methods.

In this thesis, static nonlinear models based on orthogonal polynomials are presented. It is investigated whether one orthogonal polynomial basis outperforms other bases in the context of PA identification, with respect to digital predistortion. With increasing transmission bandwidth, the memory effects become more prevalent which requires the use of dynamic nonlinear models. This can be achieved in a simple way by cascading a static nonlinearity with a leading linear filter. If such a structure, known as the simplified Wiener model, is used to model the PA, the predistorter (PD) is obtained as a Hammerstein model.

Additionally, the characteristics of the PAs change over time due to changes in temperature and aging. Therefore, the identification of the PA and the PD needs to be performed adaptively. In this thesis, three architectures for adaptive PDs are considered: the direct learning architecture, the indirect learning architecture and the direct learning architecture based on the nonlinear filtered-x least mean squares algorithm.

The work first analyses the numerical properties of different orthogonal polynomials. Then, the adaptative identification of the models, with and without memory, is investigated based on their convergence behaviour. Finally, the performance of the PD architectures is evaluated by simulations and burst measurements. For a commercial PA, the results of the study demonstrated that it was possible to increase the power efficiency by 55%.

Keywords:
Wiener model, Hammerstein model, orthogonal polynomials, basis expansion, adaptive predistortion, indirect learning, nonlinear filtered-x least mean squares algorithm


Electronic version of the publication:
http://publik.tuwien.ac.at/files/PubDat_185139.pdf



Related Projects:
Project Head Markus Rupp:
Signal and Information Processing in Science and Engineering - Entwicklungsmethodik


Created from the Publication Database of the Vienna University of Technology.