Ideally, the averaging time 
for spectroscopy will be
determined by considering (a) the distortion produced by time-average
smearing
(Equation 2), and (b) the atmospheric coherence
time 
. Because narrow-band channel images are relatively free
from chromatic aberration, it rarely makes sense to choose for
high-resolution spectroscopy by matching time-average smearing to
chromatic aberration, as in some continuum experiments.
In practice, however, observations with narrow channel
bandwidths may also require long total integration
times 
to reach sufficient
signal-to-noise in the individual channel images. This interacts with
the choice of , because a long averaging time may be needed
simply to restrain the number of independently recorded visibilities.
If is limited by the data volume,
Equation 2 should be used to check the
distortion at the edge of the interesting field. If a that
keeps the data volume manageable also produces significant distortion,
you may have to balance using (a) long averaging times (image
distortion), (b) wider channel bandwidths 
(poorer velocity
resolution), (c) fewer channels
(poorer velocity span), or (d) fewer antennas N
(poorer image quality) to restrain the data volume ( 
).