Study And Implementation Of Blind Equalization Technology For High-Order QAM Signals

This thesis focuses on a study on blind equalization technology used for high-order QAM signals and the corresponding design and implementation based on FPGA. The work finished in this thesis is a part of the overall task targeting at a key project undertaken by the laboratory the author works with. High order QAM, due to its high bandwidth efficiency, has already been widely used in modern high speed communication. Blind equalization is just one of the crucial technologies used for high order QAM.The main work and innovative achievements obtained in this thesis are summarized as follows:1. Targeting at the fact that the conventional weighted multi-modulus algorithm is only suitable for 16QAM and 64QAM signals, this thesis proposed a modified weighted multi-modulus algorithm suitable for cross QAM signals. The modified algorithm uses fully the symbol statistical characteristics of cross QAM signals, acquiring an improved performance due to much better match between error models and cross constellations. Based on above achievements the algorithm is successfully further expended to 256 QAM signals.2. Targeting at the fact that the current dual mode multi-modulus algorithm based on error selection is vulnerable to noise and not reliable during the initial convergence stage, this thesis proposed a modified dual mode multi-modulus algorithm based on the change of mode-switching. The modified algorithm uses the estimated MSE value with forgetting factor for judgment, guaranteeing reliable convergence in the initial phase, speeding up the convergence, improving the steady-state performance, and reducing the sensitivity to noise as well. The computational complexity of the modified algorithm is also reduced obviously compared with the conventional weighted multi-modulus algorithm.3. Based on above achievements a fractionally spaced predictive decision feedback blind equalizer suitable for high order QAM in severe fading channels is proposed. The algorithm adopts the modified dual mode multi-modulus algorithm mentioned above, and updates the forward and feedback filter coefficients independently, acquiring reduced steady-state errors. The fractionally spaced forward filter of the structure can avoid the spectrum overlap caused by symbol rate sampling, while the feedback filter with a predictive decision feedback structure is able to avoid the false-convergence solution caused by error propagation that is inevitable with conventional decision feedback equalizer.4. Finally, an implementation scheme of the above suggested fractionally spaced predictive decision feedback blind equalizer based on FPGA is provided in detail. Simulation and test results with satisfactory recovered constellations based on microwave channel have proved the feasibility and validity of the FPGA-based test system.