Joint detection for UMTS: complexity and alternative solutions

Joint detection within TD/CDMA multiple access scheme for UMTS is a good technique to reduce both inter-symbol and multiple access interference. But even the simplest algorithm, ZF-BLE (zero-forcing block-linear-equalizer), leads to more than 100 million operations per second in some cases at the mobile station, even for low rate services. As most of the complexity (70%) comes from the resolution of the linear system, two alternative solutions to decrease this complexity are presented within this paper: first, an approximation based on the convergence of the Cholesky decomposition performed within ZF-BLE is described. This method is a direct consequence of the block-Toeplitz structure of the linear system matrix; a compromise between complexity and performance has to be chosen. Then the idea of leaving the block treatment and replacing it with filtering is studied. Indeed, by inverting one row of a small part of the matrix equation, a filter for each CDMA code shall be constructed. The optimal length for the filters is discussed: the best result is obtained when taking into account samples which contribute efficiently to the estimated data.