Dynamic behavior modeling and nonlinearity pre-compensation for broadband transmitters

The dynamic behavioral model of a transmitter is a useful tool in the system simulation for designers of communication systems since the behavioral model can greatly reduce complexity and time-consuming for design and optimization procedures. In addition, the behavioral model can provide a fast and effective approach to validate the performance of the different predistortion schemes during the predistorter parameter extraction process, and can be helpful for determining the topology of the predistorter for pre-compensating the dynamic nonlinearity of the broadband transmitter. In this thesis, a novel two-box model based dynamic behavioral model for characterizing a broadband transmitter and the corresponding predistorters for pre-correcting the dynamic nonlnearity of the transmitter is reported. Moreover, a real-valued time delay neural network is presented for modeling the dynamic nonlinear behavior of a 3G base-station power amplifier.; At first, a dynamic exponential weighted moving average algorithm is developed to establish a LUT-based nonlinear Wiener model for memoryless wideband transmitters. To improve the relatively limited accuracy of the conventional Wiener model, a new augmented Wiener model is proposed. The superiority of the augmented Wiener model to the conventional Wiener model is validated by comparing the measured spectra of the two behavioral with that of the practical transmitter.; Then, an augmented look-up-table-based Hammerstein predistorter is proposed for the first time to further improve the pre-compensation performance of the traditional Hammerstein predistorter for the broadband transmitters. A wireless transmitter prototype, which includes an L-band push-pull GaAs FET 60-Watt peak-envelope-power amplifier, is utilized to evaluate the performance of the newly proposed predistorter. The pre-compensation performance of the proposed augmented predistorter in suppressing the spectrum regrowth will be illustrated by comparing the output spectra of the transmitter linearized by the different predistorters with that of the transmitter without predistortion.; Finally, a novel real-valued time delay neural network is also put forward to construct a dynamic behavior model for 3G base station power amplifiers (PA). Compared with the previously published neural network based PA models, a significantly reduced complexity and shorter processing time in the analysis and training procedures is obtained with this RVTDNN model. After training the RVTDNN with the measured baseband data, the RVTDNN behavioral model of the PA is obtained and different test signals are applied to this model to validate its accuracy and generality. The time-, frequency- and power-domain validation results are presented to demonstrate the accuracy of the RVTDNN model in predicting the memory effects.