Statistical nonidentifiability of close emitters: Maximum-likelihood estimation breakdown

For closely spaced signals, we demonstrate that the minimum signal-to-noise ratio (SNR) at which they can be resolved (and their parameters reliably estimated) strongly depends on the a priori source-power model. If no restrictions are made on the admissible powers, then there will exist an “ambiguity region” that encompasses scenarios with completely erroneous direction-of-arrival (DOA) estimates in addition to the true scenario with closely spaced emitters. We implement a global maximum-likelihood (ML) search to support the Cramér-Rao bound (CRB) predictions.