On the benefits of subspace dimension reduction for Grassmannian multiuser beamformers

Block diagonalization (BD) is a linear precoding technique to eliminate inter-user interference (IUI) in multiple-input multiple-output (MIMO) multiuser wireless communication systems. In this paper, a limited feedback-based BD algorithm with Grassmannian codebook is studied. When the number of bits available for feeding back channel state information (CSI) from multiple terminals to the base station (BS) is limited, we demonstrate feeding back singular vector corresponding to the largest singular value provides a better capacity performance as compared to feeding back the whole signal subspace.