Analysis of a generalised framework for spectral estimation. Part 2: Reshaping and variance results

The capabilities of lag reshaping for the generalised spectral estimation technique, in order to realise desirable effective windows, is illustrated for several combinations of temporal and lag weightings. Also the variance of the spectral estimate is presented and computed for some windows with very good sidelobe behaviour. It is found that, with proper overlap, if the length of the temporal weighting is somewhat larger than the length of the lag weighting, the variance is at a near minimum relative to any technique which realises the same frequency resolution with the given finite data record length.