New soft-output MLSE equalization algorithm for GSM digital cellular systems

We propose a new SO-MLSE (soft-output maximum likelihood sequence estimation) equalizer, which can be used in a GSM digital cellular system. It uses a complex correlation of the training sequence to obtain the channel information and the equalization is performed by MLSE using the Viterbi algorithm. In order to generate a soft-decision input to channel decoder (Viterbi decoder), the soft-output equalization algorithm is needed. The adopted algorithm doesn't require one to modify the structure of the HO-MLSE (hard output MLSE) equalizer, that is, the SO-MLSE equalizer can be implemented by adding a soft-output generation block to the HO-MLSE equalizer. This algorithm uses the outputs of a matched filter and the HO-MLSE equalizer. It turns out that the complexity of the proposed SO-MLSE equalizer is simpler than those of other SO-MLSE equalizers and its performance is almost the same as those of the others. Finally, the proposed SO-MLSE equalizer is also implemented as a prototype with the ADSP-2101 16-bits fixed point digital signal processing chip.