[time-nuts] The cross-correlation catastrophe
magnus at rubidium.dyndns.org
Tue Jan 26 02:07:52 EST 2016
Over a year and a half or so, we have been starting to grasp the various
aspects of the cross-correlation catastrophe. This consists of two
properties, one is the insertion of anti-correlated noise into the two
channels, and then when cross-correlated it produced a subtracting
"energy" reducing the level and even inverting the sign. Thus, there is
a bias in the estimation and overoptimistic or even seemingly physically
impossible (until you understand the mechanics). This have been
occupying the mind of some of the best folks we have in the field.
Just interacting and discussing this with some of them have been a great
learning experience. Anyway, here is a write-up that covers it sufficiently:
So, now you know why I look suspiciously at too low noise-levels.
More information about the time-nuts