Predicting the performance and convergence behavior of a turbo-equalization scheme

This paper proposes a simple and time non-consuming method to predict, at any iteration, the performance of a turbo equalization scheme using a soft-in/soft-out (SISO) minimum mean square error (MMSE)/interference cancellation (IC) equalizer and a SISO decoder. Gaussianity of the extrinsic log-likelihood ratios (LLRs) output by the equalizer as well as the decoder is assumed. This paper shows that the equalizer behavior may then be very reliably predicted only by calculations (no simulations are needed) whereas that of the decoder requires simulations for only one independent input parameter. Comparison between the proposed prediction method and plain simulations of the overall turbo equalization scheme shows that our method accurately determines the system performance at any iteration.