About SVD Based MIMO, Imperfect Channel Estimates, and Cross-Talk Interference

This paper deals with MIMO methods for which the singular value decomposition of the channel matrix is calculated and exploited as part of the transmission and reception process. For these methods it is shown that a severe inter-sub-streams cross-talk effect arises due to imperfect channel matrix estimates. It is shown that the mean cross-talk SNR has a simple functional dependency on the singular values of the channel matrix, and that there exist singular values for which the mean cross-talk SNR is significantly higher than for others. A pre-equalization method is proposed to transform a random original channel into a modified ´virtual´ channel with superior mean cross-talk SNR. The penalty associated with this pre-equalizer operation is described along with a possible optimization strategy to select an optimal modified channel. Comparative performance simulation results are presented. Significant performance gain of the proposed method, relative to methods where no such pre-equalization is applied, is demonstrated. Symbol processing complexity is shown to consist of merely a single matrix-vector multiplication.