On Reliability Metrics for Soft-Input Decoding in Presence of Channel Estimation Errors

The use of generalized logarithmic likelihood ratio (GLLR), as a metric for soft-input decoding, was shown to provide gains over the conventional logarithmic likelihood ratio (LLR). However, computational complexity of the GLLR is greater than in the case of LLR, so the choice of GLLRs must be justified by a notable performance improvement. In this paper, we investigate the judiciousness of the adoption of GLLRs or LLRs for soft decoding analyzing the performance of both metrics in two different scenarios. Metrics expressions are derived and channel estimation errors are considered in analysis and simulation. Numerical results show that GLLRs outperform LLRs but, in general, the achieved improvement is relatively small. Therefore, application of GLLR metrics requires cautious analysis