Performance of turbo codes with quantized channel measurements

The performance of turbo codes decoded by the soft-output Viterbi algorithm (SOVA) with quantized channel measurements is investigated in this paper. Implementation of SOVA, whose advantages are its simplicity compared to the MAP algorithm and its ease of implementation, only requires minor modification from the Viterbi decoder chips. However, quantisation loss due to hardware implementation has not been thoroughly addressed previously and is discussed in this paper. Due to over-estimation of reliability values by SOVA, soft-limiting (clipping) of reliability values can provide some gain which can compensate for the loss due to quantization and, in certain cases, provide performance that is close to MAP decoding.