The bootstrap algorithm: a robust multiuser CDMA detector with time delay variation

Almost all near-far resistant detectors, adaptive or nonadaptive, assume knowledge of the relative delays of the different users´ signals. Estimation and tracking errors of these delays cause errors in the correlation matrix as they all perceived by the receiver. Some linear detectors, like decorrelating detector, minimum mean-square error (MMSE) detector, or nonlinear detectors, such as zero-forcing decision-feedback (ZF-DF) detector, all directly perform some manipulation on the correlation matrix and hence they are very sensitive to delay shift or estimation error. The bootstrap detector, on the other hand, performs adaptive transformation on the output of the matched filter directly, taking into account any such errors if they exist. Therefore one may expect this detector to be more robust to delay errors. We analyse the bootstrap detector when delay errors exist. We examine analytically as well as by simulation, the BER obtained with this detector and compare it to the conventional decorrelating detector. We show that while the latter becomes non-near-far resistant when error exists, the bootstrap stays near-far resistant in the limit although with some degradation in performance when the condition number of the correlation matrix becomes large