Comparison of turbo-code decoders applied to short frame transmission systems

Recently, a novel class of binary parallel concatenated recursive systematic convolutional codes termed turbo-codes, having amazing error correcting capabilities, was proposed. However, the decoding of turbo-codes relies on the application of soft input/soft output decoders. Such decoders can be realized either using maximum a posteriori (MAP) symbol estimators or MAP sequence estimators, e.g., the recently proposed a priori soft output Viterbi algorithm (APRI-SOVA). In this paper, the structure of turbo-code encoders as well as of turbo-code decoders shall be described. In particular, four different decoder structures shall be illustratively characterized and their error rate performance capabilites shall be compared in both additive white Gaussian noise (AWGN) as well as flat Rayleigh-fading channels based on extensive simulation results for short frames used for speech transmission in the uplink of a recently proposed digital mobile radio system applying code division multiple access and joint detection [12], [18]. The decoders are investigated as follows: 1) the MAP symbol estimator-based approach used by Berrou [1], 2) the MAP symbol estimator-based approach used by Robertson [2], 3) a new reduced complexity MAP symbol estimator-based approach [3], and 4) an APRI-SOVA based approach used by Hagenauer and Papke [4].