Structured multiuser channel estimation for block-synchronous DS/CDMA

Uplink channel estimation for a block-synchronous chip-asynchronous DS/CDMA system as proposed for the time-division duplex option of third-generation cellular systems is considered. Training midambles are employed for joint channel estimation of all users. The standard unstructured approach based on modeling the effective user channels as unknown FIR filters is compared with two structured methods that exploit a priori knowledge about the user channels such as the maximum delay-spread, the transmit chip-shaping pulse and the path delays. Since these are usually unknown, a low-complexity estimator for the path delays of all users is derived from a maximum-likelihood approach. For all channel estimators, optimal sets of training sequences based on perfect root-of-unity sequences are found. For these optimal sets, it is shown that the reduction in channel estimation mean-squared error of the structured estimator versus the unstructured estimator is exactly the ratio of the number of structured parameters to unstructured parameters. Simulation results show that structured channel estimation provide advantages up to 4 dB in terms of output signal-to-interference plus noise ratio with respect to unstructured estimation, for linear RINSE detection. In contrast, for conventional single-user matched filtering, unstructured estimation proves to be sufficiently good.