Impact of nonorthogonal training on performance of downlink base station cooperative transmission

Base station (BS) cooperative transmission is a promising technique to improve the spectral efficiency of cellular systems, and by using it, the channels become asymmetric in average gain. In this paper, we study the impact of the asymmetric channel gains on the performance of coherent cooperative transmission systems, when minimum mean square error (MMSE) and least square (LS) channel estimators are applied to jointly estimate the channel state information (CSI) under nonorthogonal training. We first derive an upper bound of rate loss caused by both channel estimation errors and CSI delay. We then analyze the mean square errors of the MMSE and LS estimators under both orthogonal and nonorthogonal training, which finally reveals the impact of different kinds of training on the precoding performance. It is shown that nonorthogonal training for the users in different cells leads to minor performance degradation for the MMSE channel-estimator-assisted downlink precoding. The performance degradation induced by channel estimation errors is almost independent of the user's location. By contrast, the performance loss caused by the CSI delay is more severe for users located at the cell center than that for users located at the cell edge. Our analysis is verified via simulation results.