Resolution limits of blind multi-user multi-channel identification schemes-the bandlimited case

Blind space-time equalization and separation of multi-user digital communication signals presumes that the number of antennas M and the oversampling rate P is sufficiently large to be able to detect the number of sources and all channel lengths, and that the channel matrix is sufficiently well conditioned to allow inversion. A singular value analysis of the channel matrix for bandwidth limited signals provides necessary conditions for sufficient resolution, and guidelines for the selection of suitable M, P and equalizer lengths in relation to the bandwidth