The effect of imperfect SNR knowledge on multiantenna multiuser systems with channel aware scheduling

We analyze a cellular communication system in which a basestation (BTS) or access point transmits packet data to several mobile data users by means of a TDMA scheme. All users estimate their instantaneous signal-to-noise-ratio (SNR) in each slot and feed this information back to the BTS. A scheduler in the BTS then uses this information to allocate the channel resource to the user which maximizes a certain metric. We are interested in assessing the sensitivity of the system performance in terms of spectral efficiency per cell with respect to an imperfect knowledge of the multiuser channel, expressed by the estimated SNR of all users. By assuming a block fading channel model for each user, data-aided maximum-likelihood intra-slot SNR estimation can be performed if known pilot symbols are transmitted in each slot. We derive a novel SNR estimator for the block fading channel, which takes the channel statistics into account. The new estimator clearly outperforms the AWGN ML estimator in terms of system performance. Not only is the spectral efficiency larger for the system, but the optimum spectral efficiency is also achieved with fewer pilot symbols per slot.