Methods for Switching Between Long Term and Short Term Transmit Beamforming in OFDM

Obtaining channel state information (CSI) at the base station in an OFDM system can significantly improve both the capacity and reliability of the downlink. However, this CSI will suffer from estimation error plus possible quantization error for certain feedback methods and will also become outdated because of time variations in the channel. If the channel variations are reasonably slow and the estimation (or quantization) errors are small, then per-subcarrier maximal ratio transmission (MRT) is the optimal single-stream beamforming technique (MRT is designed to match the instantaneous channel and hence is a short-term transmit beamformer). When the estimation (or quantization) error is high, then eigen beamforming (EBF) is typically better than MRT because it uses long-term statistics that can be estimated more accurately by averaging over many samples. In addition, if the channel variations are high, then EBF is typically better than MRT because again EBF is matched to the long-term channel statistics (e.g., channel covariance matrix) that do not change rapidly. Since MRT is better than EBF under some conditions and EBF is better than MRT in others, it may be necessary for the base to choose between the two. This paper introduces three methods to switch between MRT and EBF and shows through simulation results that the methods are fairly effective at choosing the best method under various channel conditions.