Hi Dex,

correlation processing is relatively cheap, so no harm done ;-)

> No stacking allowed when message is known.

I politely disagree. A decode made after stacking blindly with constant phase and weights (or weights derived solely from noise level) hasn't made use of a priori information of the correct result. Correlations are ok, but their output should not be used to tweak the phases before stacking. As Paul mentioned earlier, stacking a large enough number of tweaked-phase data sets will eventually produce any desired decode from pure noise.