Skip to Main Content
For frequency selective channels, conventional equalizations based on a whole vector space enhance the noise power at randomly distributed spectral nulls. A signal subspace detection technique is proposed for single carrier (SC) block transmission with unique words (UW). Different from conventional Fourier basis, a new orthogonal basis to orderly separate the space into signal subspace and noise subspace is adopted. In this way, known symbols of UW can compensate the noise subspace and facilitate data recovery by a maximum likelihood (ML) detector in the signal subspace, which is robust to the ill-conditioned channel matrix. The proposed technique does not require any feedback from the receiver to the transmitter. A near optimal performance is achieved with its computational complexity similar to the time domain equalization.