The on-line visibility averaging time 
at the VLA may be
set to any multiple of 1.67 sec, but the even multiples are
preferable, for reasons associated with the VLA phase switching system.
In practice, averaging for only 1.67 sec may give large
( 
10 mJy) correlator offsets, so the shortest recommended
for all applications is 3.3 sec. The ``default" values of the
on-line visibility averaging time at the VLA are 10 sec for the
A and B configurations and 30 sec for the C and
D configurations.
The effects of the on-line averaging are permanent, because the averaged data are written to the archive tape created by the on-line system. It is therefore important to choose an on-line averaging time that limits time-averaging distortions to acceptable values, and is shorter than the expected atmospheric coherence time (provided the hardware and computer restrictions permit such a choice--spectroscopists must note Table 9). Longer averaging times may be used off-line, e.g., once the data have been calibrated.
For Gaussian tapering in the u-v plane, the intensity reduction due to
a finite averaging time is given by Bridle &
Schwab's Equation 13-39 with 
, which we may write as
Spectroscopists should use Equation 7,
with values of 
that
are acceptable for their experiment, to determine the maximum
acceptable averaging time.
Continuum observers may wish instead to use an averaging time
that produces the same
intensity reduction for a source near the pole as does an IF bandwidth
. This can be approximated (for small intensity
reductions) by equating the terms in 
from
Equations 6 and 7, obtaining
Equation 8 gives a reasonable criterion for
the maximum averaging time that should be used with an IF
bandwidth for continuum observations at observing
frequency 
. Notice that does not depend on
VLA configuration or on 
, owing to the first-order similarities
between the chromatic aberration and time-average smearing for a polar
source. There is an intuitive basis for
Equation 8: for a polar source, time
averaging smears the visibilities over an angle 
in the u-v plane so its effects scale as
, whereas those of bandwidth smearing
scale as 
.
You may have to exceed the value of calculated
from Equations 7 or
8 because the VLA's on-line
system cannot provide averaging times less than 1.67 seconds in any of
the normal observing modes. Note also that the VLA on-line system
averages all baselines with the same . If the calibrator
observations are only a few minutes each (as is often the case at low
frequencies), averaging times longer than 1 min are undesirable
because they permit only crude editing of the calibrator data.
is also limited in many cases by the size of the final data set,
especially for multiconfiguration continuum syntheses, or
multichannel--continuum or spectral line--observing
(Table 9). If storing the data set
requires much more than the average disk space available per user in
the off-line computers, data processing (and/ or life with one's
colleagues) becomes difficult. Averaging times are sometimes
set longer than would otherwise be desirable, just to avoid such
``excessive" data volumes. Most VLA continuum observing is done with
10 sec 
1 min because of these constraints. For line work,
Table 9 gives the limits that are
imposed by the correlator.
