CSIT Group - Research Areas


Low complexity multi-user detectors for TD-SCDMA Systems


 In 3G hybrid CDMA systems with antenna arrays at the receivers, space-time (S-T) multiuser detection (MUD) algorithms are adopted at the base station to reduce both inter-symbol (ISI) and multiple access (MAI) interference. For each data-packet or block, in the up-link scenario, linear MUD solves a linear system where the system matrix has a specific block-Toeplitz structure.

Exact MUD computation exhibits a high computational load for large matrix size (i.e. high number of users, antennas and/or symbols per block), as it involves the inversion of a large correlation matrix. Exact inversion of the system matrix is not feasible due to large matrix size and real-time constraints. In fact, in real-time mobile radio systems, computational resources are limited and only reduced complexity algorithms can be employed. Nevertheless, when computational complexity is strongly reduced, sub-optimal algorithms might introduce performance degradation and other unwanted effects (e.g. near-far problems).


Two main MUD techniques are available in literature: block-based MUD and one-shot MUD also known as sliding windows detector (SWD). Both families may be implemented using real-time hardware. In practice, algorithm selection is performed taking into account various aspects such as performance, complexity and implementation issues. However, when computational power requirement and system complexity are the key constraints, some performance degradation have to be tolerated.

According to these considerations, a new MUD detector scheme based on the block-Bareiss (BB) algorithm has been introduced. The proposed algorithm, derived from the plain Bareiss factorization technique, combines good performance and low computational load with simple hardware implementation. For a specific hybrid CDMA radio system (i.e., China Wireless Telecommunication Standard or CWTS for short), the BB detector compares well with the reference direct inversion methods and with other known MUD algorithms namely the block-Levinson algorithm (BL) and the block-Fourier Transform (BFT) algorithm.

Finally, a floating-point implementation of the BB processor may be easily designed using FPGA (Field Programmable Gate Array) devices and it has good performance and complexity figures with respect to the equivalent BFT one. In addition, it does not suffer from near-far effects that plague very low complexity implementation such as low order BFT detectors. It is also worth noticing that the BB detector is also suitable for hardware parallel implementation. However, for perfect power control radio systems, the BFT remains the most effective MUD algorithm in terms of complexity and implementation feasibility.




Related Publications:

•    V. Rampa; Design and Implementation of a Low Complexity Multiuser Detection for Hybrid CDMA Systems; Journal of Communications Software and Systems (JCOMSS), Vol. 1, No. 1, pp. 42-50, Sept. 2005, ISSN 1845-6421.
•    A. Bifano, V. Rampa; Multiuser detector for hybrid CDMA systems based on the Bareiss algorithm; IEEE Proc. of International Conference on Acoustics, Speeach and Signal Processing (ICASSP’05), Vol. 3, pp. 909-912, Mar. 2005.
•    A. Bifano, V. Rampa; Low Complexity Multiuser Detectors for TD-SCDMA Systems: Design and Implementation; Proc. of SoftCOM 2004, pp. 444-448, Split, Oct. 11-13, 2004.
•    A. Colamonico, M. Nicoli, V. Rampa, U. Spagnolini; Architectures for sliding window multiuser detection in TD-SCDMA systems; Proc. of 2002 European Conference on Wireless Technology (ECWT 2002), pp. 67-70, Milan, Sept. 2002.
•    M. Beretta, A. Colamonico, M. Nicoli, V. Rampa, U. Spagnolini; Space-Time multiuser detectors for TDD-UTRA: design and optimization; Proc. of IEEE Semiannual VTC-2001 Fall, pp. 375-379, Atlantic City, Oct. 7-11, 2001.