Digital Baseband Modeling and Correction of Radio Frequency Power Amplifiers

University dissertation from Stockholm : KTH Royal Institute of Technology

Abstract: Aspects related to behavioral modeling and correction of distortions for radio frequency (RF) power amplifiers (PAs) are treated in the thesis.   When evaluating the performance of a behavioral model it is important to, first of all, use an evaluation criterion, and second, make sure that the criterion actually tells something about the behavior one wishes to describe. This is used in the weighted error spectral power ratio (WESPR) criterion by means of a frequency dependent weighting function. When the parameters of the models are estimated a suitable error criterion should likewise be used. The frequency weighting function in the WESPR is used in the extraction of the parameters. It is shown on two types of PAs that the model performance measured as WESPR can be improved while the model complexity is reduced compared to the standard frequency neutral criterion.   When building a model of a system it is advantageous to take account of the physical structure and incorporate this knowledge into the model. It can improve model performance and possibly reduce the number of parameters. By starting from a physically motivated nonlinear model of a RF PA, the commonly used memory polynomial (MP) models are derived. Additionally, three novel MP model structures are derived. Using data measured on a PA it is found that two of these model structures have lower model errors while using fewer parameters than models previously published in the literature.   Methods to increase the power efficiency and the linearity of RF PAs have been investigated. One of these methods is digital predistortion (DPD) which improves the linearity, thus facilitating operation at higher power levels which improves power efficiency. The other method is a signal shaping method that makes the signal more favorable to the PA by reducing the highest peaks to lower values. It is experimentally shown that the combination of DPD and signal shaping results in an increase of power efficiency in the order of 2-4 times. An instability in the feedback loop that updates the parameters of the DPD was also identified and two solutions were proposed. One solution changes the parameters of the DPD in such a way that the instability is avoided and the other changes the signal to avoid high amplitudes.   The nonlinear effects of class-D outphasing amplifiers are considered. Four model structures are proposed and evaluated on data measured from two amplifiers. In order to reduce the distortions in the output signal from the amplifiers an algorithm using constant envelope amplitude (purely phase-modulated) signals is proposed. The DPD is evaluated and found to reduce the distortions in a state-of-the-art 32 dBm class-D outphasing PA to make it fulfill the linearity requirements for downlink signals used in the universal mobile telecommunications system (UMTS).

  This dissertation MIGHT be available in PDF-format. Check this page to see if it is available for download.