Study On Behavioral Modeling And Predistortion Techniques For Power Amplifiers In Broadband Communications

Power amplifiers(PAs) are essential components in overall wireless communication systems and the inherently nonlinear characteristics are the major sources of nonlinear noise. Meanwhile, the PA efficiency characteristic directly limited the performance and the communication quality of wireless base station. When a non-constant envelope modulation signal goes through a nonlinear PA, spectral regrowth (broadening) appears in the PA output because of the extremely serious intermodulation distortion, which in turn causes adjacent channel interference (ACI). It also causes in-band distortion, which degrades the error vector magnitude (EVM) performance and consequently affects the accurate demodulation of the receiver signal. Stringent limits on the ACI are imposed by modern wireless communication systems, and thus the extent of the PA nonlinearity must be controlled in order to reduce the distortions. Therefore, PA behavioral modeling and linearization are often the necessary methods to evaluate the effects of the nonlinear distortion on communication systems, suppress spectral regrowth and adjacent channel interference, reduce the bit error rate (BER), and enhance the PA efficiency.In this dissertation, some theories and methodologies of nonlinear system identification and adaptive signal processing are applied in order to deeply study and investigate some problems of the behavioral modeling and digital predistortion of PA with memory in broadband communication systems. The main results and contributions of the dissertation are as following:1. The measurement setup used for model extraction and validation is established. The measurement results of nonlinear characteristics of PAs directly determine the precision of the extracted model. The measurement setup used for behavioral modeling extraction and verification under the broadband digital modulation signal excitation is constructed after an overview of a typical measurement setups used for amplifier characterization in this dissertation. The common extraction process and verification means of the off-line and adaptive behavioral modeling is provided based on the measured data. 2. A complex radial basis function (RBF) network approach is developed for dynamic behavioral modeling of nonlinear power amplifier with memory. In this approach, the complex RBF network with tapped delay lines is applied to construct the behavioral modeling of PAs with memory. And the complex QR-decomposition based recursive least squares (QRD-RLS) algorithm, which is implemented using the complex Givens rotations, is employed to update the weighting matrix of the complex radial basis function (RBF) network. The QRD-RLS algorithm has the characteristics of good numerical robustness and regular structure, can significantly improve the complex RBF network modeling accuracy and well suite for hardware implementation. The simulation results demonstrate that the model has advantages of fast convergent speed and good numerical robustness.3. A novel multichannel IQRD-RLS algorithm used for behavioral modeling of PAs is proposed. The model coefficients are extracted by the multichannel inverse QR decomposition recursive least-squares (IQRD-RLS) algorithms. The IQRD-RLS approach solves directly for the time-recursive least squares weight vector, while avoiding the highly serial backsubstitution step required in previously derived QRD approaches. Furthermore, the approach employs Givens rotation operations to recursively update the model coefficients and is suitable to be implemented using systolic array to achieve fast convergence and good numerical properties. The simulation results indicate that this model is capable of satisfactorily describing the nonlinear dynamic behavior of PAs.4. A digital predistorter based on nonlinear auto regressive moving average (NARMA) model is constructed to linearize PAs with memory. And the Gauss-Newton algorithm is used for update the predistorter parameters dynamically. The result of computer simulation shows that the method has advantages of high computational precision, less computational efforts, fast convergent speed and good numerical robustness.5. A predistortion method for 16QAM signal in satellite communication system is presented. The general memory polynomial (GMP) model and the indirect learning architecture are employed to construct the predistortion system, and the QR decomposition recursive least squares (QRD-RLS) algorithm is used for update the predistortion parameter dynamically. According to the problem of the QRD-RLS algorithm operating without explicitly updating the coefficient vector, a new method named by"weight flushing"is proposed. The result of computer simulation shows that the method converges fast and robust and well effective compensation, and well suite for hardware implementation of VLSI.6. The design scheme of the hardware implementation of the digital predistortion for PAs is provided. The design scheme has been put in practice and a part of the project have been accomplished. Finally, the performance of the described digital predistortion is assessed through experimental verification.