Influence of quantization on the bit-error performance of turbo-decoders

Turbo-codes are under consideration for third generation mobile communication systems. For any implementation of a turbo-decoder a fixed-point representation is mandatory. Usually, the bit-width of a fixed-point implementation has to be traded off versus its decoding performance. Past approaches towards a fixed-point representation show a degradation in performance. In this paper we apply a novel quantization methodology on turbo-decoders with (Max-)Log-MAP component decoders and present simulation results for both an AWGN and a Rayleigh fading channel model. The new quantization scheme leads to fixed-point implementations that do not degrade the bit-error performance. Under certain conditions it can even be slightly improved