No Information? Delay Estimation Below the Threshold SNR

Abstract

This paper questions the appropriateness of the mean squared error (MSE) as dominant performance metric for the maximum-likelihood delay estimator. The departure of the MSE from the Crambound is seen as an incidental effect, whereas knowledge of the frequencies of nominal estimates versus outliers leads to better statistics. Simulations are used to evaluate an alternative scheme for robust acquisition of signal delay, based upon binning statistics. This scheme converges on the correct estimate at signal-to-noise ratios that are well below the MSE threshold and into the "no information" region. I. INTRODUCTION