On-the-Fly Calibration Correction
Many tasks will apply calibration corrections `on-the-fly'. This means
that often there is no need to form a calibrated dataset. Normally this
calibration correction step is performed by default. However, the correction
can be disabled using some fairly standard switches in the
options parameter. These switches are:
- Do not apply antenna gain and delay
- Do not apply polarisation leakage corrections.
- Do not apply bandpass function corrections.
For example, to disable gain and bandpass correction in uvspec,
% uvspec options=nopass,nocal
It is always possible to form a calibrated dataset by using either of the
visibility copying tasks (uvcat
or uvaver). Whether or not
you choose to do this may be often a matter of personal taste. However
there are three VERY important situations where you should form a
calibrated copy of your dataset (i.e. where `on-the-fly' calibration is
- A still significant, but decreasing, number of tasks do not
perform `on-the-fly' calibration. Tasks
which do apply calibration corrections will invariably have nocal,
nopol and nopass options. Additionally all tasks which are
performing `on-the-fly' calibration will issue messages when they are
performing these steps.
- Many Miriad tasks allow `on-the-fly' averaging of spectral
channels (using the line parameter with a width greater than 1 -
see section 5.4) in addition to bandpass correction. However
the technique used when simultaneously bandpass correcting and averaging
is only approximately correct. For arcane reasons, rather than applying
the bandpass and then averaging the channels, the tasks correct the
average of the channels with the average of the bandpass. This can be
significantly in error if the bandpass functions vary significantly over
the range of channels being averaged. Tasks performing this dubious
operation issue warning messages to alert you of your possible folly.
If this is a real problem, then it is best to form a copy of the
bandpass-corrected data, and then use these corrected data for channel
- The self-calibration tasks produce gain tables in the same format as
the calibration tables derived from observations of calibrators. Thus if
you self-calibrate a dataset containing your initial calibration, you
will overwrite your initial calibration - a generally undesirable step.
Additionally the self-calibration tasks do not apply some calibration
corrections (e.g. bandpass and sometimes leakage corrections).
Generally it is best to
self-calibrate a dataset which has had the initial calibration applied.