Parallel interference cancellation multiuser detection: Performance and applications

This dissertation considers the performance and applications of parallel interference cancellation (PIC) multiuser detectors for digital multiuser communication systems with nonorthogonal signaling. The majority of this dissertation focuses on wireless code division multiple access (CDMA) communication systems but we also consider the application of multiuser detection and PIC to digital subscriber loop communication systems here as well.; The first section of this dissertation analyzes the performance of parallel interference cancellation detectors in the K-user synchronous, nonorthogonal, binary communication scenario. We derive exact and approximate expressions for the bit error rate of the hard-PIC (HPIC) detector where the multiuser interference estimates are formed after hard decisions are performed on estimates of the interfering users signals. We compute expressions for the exact and approximate signal to interference plus noise ratio of the HPIC detector and present an analytical comparison to the successive interference cancellation (SIC) and matched filter (MF) detectors. We also analyze the performance of the linear-PIC (LPIC) detector where the multiuser interference estimates are formed prior to the hard decisions on estimates of the interfering users signals. We show that the HPIC detector is a better estimator of multiuser interference than the LPIC detector. We derive an exact expression for the bit error rate of the LPIC detector and show that there exists a nontrivial class of operating conditions where the LPIC detector performs poorly. We present an analysis of the large-system case with random spreading sequences of length N that shows that the multistage LPIC detector does not converge to the decorrelating detector when K/ N > 0.17.; In the middle section of this dissertation, we consider two distinct techniques for improving the performance of the HPIC detector. The first technique, called partial cancellation HPIC (PC-HPIC), attenuates each user's multiuser interference estimate by a individual scalar partial cancellation factor optimized under three different performance criteria. The second technique, called soft cancellation PIC (SC-PIC) replaces the sgn(·) nonlinearity in the HPIC detector with a function that minimizes the Bayesian mean squared error of the interfering user's bit estimate. We present numerical results that suggest that both techniques yield significant performance improvements.; The final section of this dissertation considers the application of PIC multiuser detection to the IS-95 digital cellular downlink and to digital subscriber loops (DSLs). We show via analysis and numerical results with measured on-air data that PIC detection can significantly improve the fidelity of IS-95 downlink reception, especially in cases when the desired signal is received in the presence of strong out-of-cell multiuser interference. Our analysis of DSLs shows that PIC detection, as applied to the problem of crosstalk cancellation, does not provide any performance benefit in most cases. We demonstrate that more sophisticated multiuser detectors do have the potential to accurately cancel crosstalk and significantly improve the performance of DSLs.