Abstract: A new simplified soft decision metric, without noise variance knowledge, is derived in this paper. By recognizing that it is simply a distance metric in Euclidian space, we further generalize it to p-norm. Performance comparisons are presented for LDPC coded 16QAM/64QAM with different parameters. Simulation results show that, as compared with the simplified soft decision metric reported in ref.[6], there is no performance loss with our proposed generalized soft decision metric for both AWGN and multipath fading channel. Moreover, it is unnecessary to estimate noise variance at channel output, which greatly facilitates practical implementation. Further simulation indicates that, in multipath fading channel, a performance gain can be obtained with p a bit less than 2 at high signal to noise ratio.

Key words: soft decision metric, LDPC, fading channel