Efficient Joint Detection Techniques for TD-CDMA in the Frequency Domain

Third generation mobile radio systems will employ TD-CDMA in their TDD mode. To increase the capacity and performance of this system, the receiver will contain a joint detector. Joint detection is equivalent to solving a least squares problem, which represents a significant computational effort because of the amount of data that is involved. Therefore, algorithms and implementations must be developed that lower this complexity as far as possible without degrading the performance of the joint detector. This paper presents an algorithm that is based on the idea of extending the system matrix of the least squares problem to a block-circulant matrix. It is then possible to block-diagonalize the matrix with Fast Fourier Transforms. In addition, overlapsave techniques are presented that reduce the computational complexity further. The resulting algorithm is well suited for the implementation on parallel architectures.