Near-optimum iterative estimation of dispersive multipath channels

Correlation-based channel estimation techniques offer low complexity, but suffer a performance penalty relative to that achievable using the minimum-variance unbiased estimator. This paper describes two iterative estimation algorithms, called sequential and strength-ordered, respectively, that exploit the known properties of the channel sounding sequence to reduce the bias of an initial correlation-based estimate. Performance is compared with both the sliding correlator technique and the Cramer-Rao performance lower bound in the setting of discrete-time FIR, channels. The sequential algorithm leads to a pipelined implementation that is amenable to on-line estimation, while the strength-ordered algorithm can be used for off-line improvement of initial estimates obtained using a correlator.