Clearly there is some limit to the interval of time over which you can
sensibly interpolate or extrapolate gains. If the gap between the secondary
observations is too large, then the interpolation of the gain solutions may
have no resemblance to the true gains at a given time.
The software that interpolates or extrapolates between the
gain solutions has a tolerance which limits
the gaps in time that it will permit.
If the time differences between a source observation and two
gain solutions bounding it are both less than the tolerance, then the
program source gain is the interpolation of the two gain solutions.
If there is only one gain solution within
the time tolerance, then this is used without any
interpolation. If there are no gains within the time tolerance, the
corresponding program source data are marked as flagged.
This interpolation tolerance is stored as the interval item
the visibility dataset. It is given as a double precision number with
units of days. Task gpcal
sets its value as half a day (interval=0.5). Though this is excessively generous, this will not be a
problem unless there are large gaps between calibrator scans (perhaps because
some calibrator scans had to be discarded). You can set the set the
interpolation tolerance using puthd. For example
||Set the interpolation tolerance of
||Set the tolerance to 2.4 hours (0.1 days).