Singular value decomposition using an array of CORDIC processors

This paper describes a novel mode of operation of an array of CORDIC processors whereby a pair of complex vectors is made orthogonal using a simple, direct procedure in which rotation angles do not need to be evaluated. By this means, singular value decomposition (SVD) of a complex matrix can be achieved using the one-sided Jacobi method described by Hestenes. The procedure provides an efficient and economical implementation of SVD for use in MIMO precoding and other real-time signal processing tasks involving small matrices of complex elements.