The Convergence Properties of Turbo Equalization Based on Matrix Perturbation and Measured Updated Source Entropy

This paper investigates the convergence properties of turbo equalization over an intersymbol interference (ISI) channel based on new methods of matrix perturbation and average updated source entropy from constituent SISO equalizer and SISO decoder, respectively. The simulation results show that the two pragmatic approaches from different angles basically draw the similar convergence performance for both medium and long information blocks. These results are also compared with those obtained by the traditional extrinsic information transfer (EXIT) chart technique. The proposed methods outperform the conventional one by two facts. One is using relatively less complexity real time processing. The other is that they are more suitable for short and medium blocks since there is no need for the estimation of histograms for conditional probability density functions that are necessary by the EXIT chart method