Iterative processing for joint detection in multiuser systems

Iterative (turbo) processing of multiuser detection of CDMA signals is studied. The factor graph representation of a CDMA system is used to motivate an iterative joint detection algorithm. The algorithm reduces complexity by replacing the large sums that appear at the multiple access nodes by simple averages. This system shows surprisingly good performance achieving single user performance, whenever the iterations can start successfully. It is shown how this is related almost exclusively to the behavior for the error control codes at low signal-to-noise ratio values. Simulation and analytical results on the performance of these detectors are shown to illustrate complexity/performance trade-offs.