Proc. of IEEE Milcom, Boston, MA, Oct . 13-16, 2003.
Multiple-input multiple-output (MIMO) spatial multiplexing wireless systems achieve high spectral efficiencies by demultiplexing the incoming bit stream into multiple substreams. Spatial multiplexing is of practical importance because the multiple substreams can be decoded using linear receivers. Unfortunately, this reduction in complexity degrades the probability of error performance.Ê To overcome this difficulty, error rate performance of spatial multiplexing systems can be improved by sending fewer substreams than the number of transmit antennas by linear precoding. Criteria have been proposed for designing these precoders when complete channel knowledge is available to the transmitter. The assumption of complete channel
knowledge is often unrealistic in many communication systems such as those with low rate feedback channels. Thus a quantized precoding scheme is proposed where the receiver sends back a fixed number of bits to the transmitter. This bit pattern corresponds to an index within a finite set of precoding matrices. A previously proposed criterion is used to determine the matrix in this precoder codebook to choose. A design method for these codebooks using techniques from Grassmannian subspace packing is presented. Simulation results show this technique outperforms typical antenna selection.that antenna selection represents a special case of constrained precoding. Simulation results show this technique outperforms typical antenna selection.
This paper is available on .IEEE Xplore .