Abstract— The implementation of advanced DS–CDMA re-
ceivers based on multiuser detection principles is becoming a
reality thanks to the combination of an improved understanding
of the theoretical basis of multiuser detection and advances in
digital, mixed-signal, and RF technologies. Due to their lower
complexity, subtractive interference cancellation approaches are
attractive for the practical implementation of multiuser detection.
In a parallel interference cancellation receiver, it is practical
to use the soft outputs of a matched ?lter bank for amplitude
estimation. A bias arises in the decision statistics, however, due to
imperfect estimation and interference cancellation. In this paper,
the source of the bias is explicitly recognized, and a partial inter-
ference cancellation scheme that mitigates the negative effects of
biased estimation and signi?cantly improves system performance
is proposed. A practical real-time algorithm that signi?cantly
reduces the implementation complexity of this scheme without
sacri?cing performance is then derived. To facilitate a software
radio implementation, the signal processing complexity of the
approach is characterized. The real-time processing algorithm
is tested via implementation in software on a ?oating-point
general-purpose DSP. The prototype includes a ?exible software-
based architecture which performs IF sampling and uses digital
downconversion prior to baseband processing. The hardware test
setup is described, and the results are presented and compared
with simulation and analytical results. The experimental results
con?rm the simulation and analytical results which show large
performance gains over the conventional matched ?lter.
Index Terms—CDMA, interference cancellation, joint detec-
tion, multiuser detection, software radios, spread spectrum.