SNR estimation techniques for low SNR signals

Radio receivers contain a set of adaptive algorithms that estimate the received signal´s unknown parameters required by the receiver to demodulate the signal. Often missing from the standard parameter list is Signal to Noise ratio (SNR) or equivalently Eb/No. The SNR estimate is the ubiquitous scale factor associated with all maximum likelihood estimators. The SNR qualifies the signal quality letting the estimator algorithms know whether the observables are reliable, hence should make significant contribution to the estimate or are unreliable and should make limited contribution to the estimate. Error correcting algorithms also use SNR to set soft decision probabilities and likelihood ratios. Many SNR estimates are accurate at high SNR when we really don´t need the estimates and are inaccurate at low SNR when we have most need for them. This paper discusses two SNR estimator techniques which maintain estimation accuracy down to very low SNR values.