SNR Estimation with Heterogeneous Frames

We study the problem of estimating the signal-to noise ratio (SNR) in linearly modulated systems. It is assumed that the frame of observations consists of several subframes, so that the symbols in different subframes are drawn from different constellations. This situation arises in practical systems, since preambles using simple constellations are inserted within data portions using more spectrally efficient, denser constellations. We consider both data-aided (DA) estimators and non-data-aided (NDA) moment-based estimators for these systems with heterogeneous frames. As it is not immediate how to generalize existing NDA SNR estimators derived for homogeneous frames to these settings, we propose several possible extensions in order to improve estimation performance. We derive the Cramer Rao lower bound for this problem and the analytical variances of the best of the proposed solutions. Numerical simulations in the framework of the DVB-S2 standard confirm the analysis.