Performance Analysis of `ROCSTAR´ Algorithm for CDMA Forward Link Code Detection and Cancellation under Stressing Conditions

In CDMA systems, the mobile on the synchronous forward link has knowledge only of its own spreading code; therefore, blind interference suppression (IS) algorithms are required to cancel the multiple access interference (MAI). A recently developed technique applies auto-regressive spectrum estimation (ARSE) in reduced rank to allow a mobile to estimate the other users\´ codes by producing a "code spectrum", analogous to the conventional Fourier Transform in frequency analysis. The algorithm, known as ROCSTAR (Reduced Order Code Spectrum Transform using Auto-Regression), enables the mobile to detect the other codes efficiently and with few samples, allowing adaptation to a changing environment. ROCSTAR has been previously shown to provide better probability of detection and lower false alarm rate than full rank ARSE by reducing spurious peaks in the code spectrum with less computational complexity. Code estimation allows interference cancellation (IC) to be done, improving link performance significantly. In this paper, we analyze ROCSTAR\´s performance under heavy loading, severe multipath spreading, low sample support, and low Eb/N0. ROCSTAR is shown to outperform ARSE, producing high code detection rates (PD) and low code false alarm rates (PFA). Channel estimation using a pilot is performed, and it is shown that performance degradation due to channel estimation errors is less than 3 dB. The dependence on rank is examined and found to be related to the number of multipath components and largely independent of the number of users.