Performance Comparison of Feasible MMSE-PIC Algorithms with Channel Estimation for HSUPA

In this paper, we study the speech user performance of selected parallel interference cancellation (PIC) algorithms in a single-cell high speed uplink packet access (HSUPA) system. To combat intersymbol interference (ISI), single-user chip-level minimum mean-squared error (MMSE) equalizer is used to detect the data users. PIC is applied to suppress the interference from data users to speech users, and RAKE to detect the speech users. As an input to the PIC stage, we may use the hard or soft symbol decisions (HD, SD) from the data users' demodulator or decoder outputs. The latter requires re-encoding of the decoded symbols. This adds to complexity, but increases also the reliability of the interference estimates. Results and analysis are provided for both ideal and MMSE channel estimation (CE). We show that the combination of PIC and RAKE outperforms the simple MMSE equalization in all studied cases. Using re-encoded soft symbol decisions for interference estimate calculation always leads to the best performance. However, in some scenarios the performance loss due to using the soft symbol decisions from the demodulator output is almost negligible.