Once you have determined the IF bandwidth 
from the
field of view criteria, the next step in the continuum decision tree
is to estimate the total on-source integration time
required for given sensitivity on your final image. Here you
will use the expression for the r.m.s. noise 
on an image
made with an N-antenna array:
where n is the number of independent IFs contributed to the image
per antenna (n=2 for images of Stokes I from two orthogonal
polarization states at one sky frequency, or for images of
at one sky frequency), is in seconds, and
is in MHz. In the numerator, 
for natural
weighting and is >1 for other weightings (the value depending on the
gridding and on the baseline distribution--see Lecture 8 for
VLA-specific details). 
is the single-interferometer
sensitivity per second per MHz of IF bandwidth.
The sensitivity goal will be determined by (a) the significance level you need for a detection to achieve your astronomical goals, and (b) whether the interesting emission is extended (see Angular resolution: how much is enough?). If you are doing polarimetry, calculate the sensitivity from the expected polarized intensity, not total intensity.
You have now reached a critical decision. You have an
estimate of the total on-source integration time needed for
the project as you would initially prefer to do it. Adding some
overhead for calibration (typically 10% to 20%, but see
Calibration Strategy
for details), you now have an idea of how much observing time you
would need.
If this first estimate of the total observing time is much longer than the time that the source is above the elevation limits of the telescope each day, then consider carefully whether your choices of frequency and baseline range were optimal. You may wish to re-enter the decision tree with different starting parameters before taking the proposal further.
If your is about as long as the source is visible each
day, then your project may be well suited to making a full-synthesis
image. You could then adjust the observations to fit into a single
pass of the source across the sky.
If your is much less than the time that the source is
visible each day, your observing strategy will be determined by the
need for dynamic range (i.e. for good u-v
coverage) and by whether you can merge observations of several such
sources into one program. If you need high dynamic range, or you want
to image an extended structure with less than needed for a
full synthesis, you should still sample the u-v plane as uniformly as
possible. This can usually be done satisfactorily by distributing the
observing over several short (e.g., 
-minute) scans spaced
equally through the available hour-angle range. Note however that the
dynamic range achieved in a given observation also depends on
atmospheric and ionospheric conditions, on the elevation angle range,
and on your
calibration strategy as well as
on the u-v coverage.
Two-dimensional arrays with many antennas may have good enough instantaneous u-v coverage to make a ``snapshot" mode attractive for observations of strong, compact sources when the total integration time required is small. ``Snapshot Mode'' discusses the advantages and disadvantages of snapshot mode at the VLA.
