The design of accurate estimators which also consider the noise term in low SNR scenarios is paramount for achieving optimal solutions and obtaining precise symbol detectors. Particularly, this paper estimates the propagation delays focusing on asynchronous DS-CDMA systems. The proposed minimum conditioned variance (MCV) is the choice in noisy environments, implementing the best linear detector of the transmitted symbols under a minimum mean-square error criterion. The result is an estimator that improves the conditional ML (CML) solution when noise is not negligible, and attains the derived Gaussian unconditional Cramer-Rao bound (UCRB) in the whole EbNo range as classical Gaussian unconditional ML (UML) does. Consequently, the proposed MCV estimator, becomes an optimal quadratic solution achieving similar features than UML in a straightforward way, and with no assumptions on the signal statistics.

Peer Reviewed