Two-stage non-data aided adaptive linear receivers for DS/CDMA

Closed-form steady-state performance analysis of the signal-to-interference plus noise ratio (SINR) at the output of well-known adaptive implementations of the linear minimum mean-square error receiver for DS/CDMA show that non-data aided schemes may suffer from a considerable performance degradation with respect to their data-aided counterparts. Motivated by this fact, we propose a new two-stage non-data aided scheme where symbol-by-symbol pre-decisions at the output of a first adaptive stage are used to train a second stage. We derive closed-form steady- state performance analysis for both the two-stage and classical decision-directed schemes, taking into account detection errors in decision-directed adaptation. Our analysis shows that the SINR of the two-stage algorithm is close to optimal over a large range of values, while the SINR of the decision-directed scheme is far from optimal when the optimal SINR is small. Finally, we consider the case of time-varying fading channels. We derive modified RLS and LMS adaptation schemes by considering SINR maximization rather than mean-square error minimization (that is useless under the assumption of zero-mean random channels). The resulting two-stage receiver shows good tracking properties in heavy near-far conditions (at least for moderate normalized Doppler bandwidth), while the decision-directed receiver may easily loose tracking after deep fades.