[daip] DFTPL question

[Eric Greisen]

Hanno Spreeuw wrote:
> Hello, I have a question about DFTPL.
> When setting BPARM(9)=2 error bars are reported. But these seem somewhat 
> smaller, perhaps a factor 1.5 or 2 or so, than the rms noise in the map 
> from the corresponding timespan, computed, e.g., from tvwin;imstat.
> So how does DFTPL compute the error bars?
DFTPL is plotting the real part of the visibilities and producing the 
rms from the usual summing:

      AVG = Sum (Re(i)) / N

      RMS = sqrt (SUM(Re(I)**2) / N - AVG*AVG) / sqrt (N-1)

An image will have contributions from the Imaginary part as well so I 
would expect an increase of sqrt(2).  It is also possible that the 
errors are correlated so that dividing by sqrt(N) may be overly optimistic.

Eric Greisen