[daip] question about weights in FITS-IDI
Eric Greisen
egreisen at nrao.edu
Fri Mar 21 17:21:14 EDT 2008
Walter Brisken wrote:
> The weight vector that accompanies a visibility record stores weights on a
> per IF / per polarization basis. The documentation I have suggests that
> this weight must be between 0 and 1.
>
> Because of some subtleties in the DiFX software correlator, it is possible
> for the data weight to be slightly greater than 1 (with a hypothetical
> maximum of about 2). This occurs when the integration time is not an
> exact multiple of the subintegration time. In these cases, some records
> may have slightly less than full weight and some may have a bit more than
> full weight. The mean weight would be 1 in the case of perfect data
> playback.
>
> Will these weights that can exceed 1 cause any unexpected problems within
> AIPS?
Weights in AIPS have any value whatsoever from enormous to very small in
absolute value with negative or 0 implying flagged. I do not know if
FITLD cares other than to flag/delete all below some weight when reading
in FITS IDI format. It does not examine the value otherwise so weights
larger than 1 will not cause a problem. The wtthresh cannot be set > 1.
Eric
More information about the Daip
mailing list