When planning an observation, it is important to consider how the data will be calibrated. Usually, the calibration process begins with the frequent observation of sources of known flux density, structure, and position. The calibration of spectral line data can be split into two orthogonal components: the overall complex gain calibration and the bandpass calibration. The former is predominantly a function of time, the latter of frequency. We discuss the bandpass calibration in the next section. The complex gain can be broken into two parts, the amplitude and the phase, and one must calibrate both of these components3.1. It is recommended that one observe the flux density calibrator at least twice during an observation, to check for consistency and to ensure that at least one succesful measurement is recorded in case one of the observations is corrupted. The phase calibrator should be observed more frequently (every 5-40 minutes depending upon the configuration, weather, and band) and should be as close to the target as possible (P-band observers should see VLA Memo # 159 written by Rick Perley). In general, the lower the frequency the longer one can wait between calibrator scans.
The overall complex gain calibration is carried out on the channel 0
data, and is very much like the one performed on continuum data. The
calibration procedure is discussed in detail in the AIPS Cookbook.
One has to be aware, however, that the bandwidth of channel 0 is
usually quite a bit narrower than in the continuum case and,
therefore, the signal-to-noise ratio is lower. Typically, one would
want a signal-to-noise ratio on a single baseline of at least 5 for
a calibrator scan. Such a signal-to-noise ratio will generally
result in 3-4% rms gain variations and a 5
rms error in
phase. For a more accurate calibration one must, of course, achieve a
higher signal-to-noise in the calibrator scan. Note that the
calibrators must be observed at the same frequency as the target,
although there is a few MHz leeway. For example, if one is observing
several galaxies at different velocities the calibrators must also be
observed at each of those velocities.
When observing Galactic HI, one has to be aware that there will be contamination, both absorption as well as emission, affecting the amplitudes as well as the phases of the calibrators. Experience has shown that the phase calibrators can be observed at the same frequency as the target fields. In the calibration, the shorter baselines of the phase calibrators, out to about 300 m, can simply be excluded, exploiting the redundancy of the VLA when it comes to determining the antenna-based complex gain solutions. Regarding the flux density calibrators, which often double as bandpass calibrators, it is best to split observations in two equal parts. These parts should be offset by plus and minus a few MHz with respect to the frequency at which the target fields are observed, ensuring that the entire band is moved outside of the range where Galactic HI can be expected. In the calibration one can then average each pair of observations. Note that Galactic HI fills the beam and significantly contributes to the system temperature. Section 3.7 deals with this issue in detail.