Channel state information based LLR clipping in list MIMO detection

Suboptimal detection schemes, such as list MIMO detection, often face the challenge of having to ldquoguessrdquo at the decision reliability for some of the detected bits. A simple yet effective way of doing this is to set the maximum magnitudes of the associated log-likelihood-ratios (LLRs) to a certain predefined value: LLR clipping. However, the choice of the clipping level has a significant impact on the system performance. A majority of prior approaches attempted to determine appropriate clipping levels by manual optimization. In this work we propose to use an SNR-aware approach for calculating the LLR clipping levels in list MIMO detection. The proposed scheme exploits knowledge of the channel state information to determine the instantaneous bit error probability of the list detector, and from this an appropriate level for clipping of the LLRs. Simulation results show that this strategy outperforms schemes using a fixed clipping level.