The new FORS imaging calibration paradigm
We have shown
( (Møller et al., 2005;
Freudling et al., 2006)
that false illumination of the FORS1 CCD has been
the defining limit of its photometric calibration accuracy. The first
step to improve the FORS calibration is therefore to obtain data to
allow a computation of the illumination correction across the field.
This false illumination problem, which is inherent in all calibration
employing ``flat field correction'', has been a known problem for
several years, but its solution is mostly viewed as too complex and too
observing time costly to be addressed in a general way.
In very general terms the solution can be obtained via a set of
dithered and rotated exposures of the same stellar field. With the
same star observed in several different positions on the CCD, and with
a large number of stars, we obtain a number of linear equations which
is large enough to strongly over-constrain the determination of
relative magnitudes of all objects, of relative extinction of each
exposure, and of a set of parameters describing the 2D correction
field across the CCD.
Another limitation on the calibration accuracy has been the brightness
of the Landolt standards which have forced the use of extremely short
exposures introducing small but non-negligible shutter-time errors
and also resulting in a very small number of useful standards per
frame. The second generation FORS calibration paradigm is designed
to solve both of those problems with the same set of calibration data.
The new calibration plan does not require more time for calibrations
than previously, it merely requires the data to be taken in a different
way, and also requires a much more sophisticated pipeline.
The second generation FORS imaging calibration plan
We are in the process of selecting 8 Landolt/Stetson standard fields
(see here for current status).
Within each of those fields 3 different but largely overlapping
pointings will be chosen. Each pointing shall be carefully selected
to fulfill the following three conditions:
(1) It must include a large number of Stetson standards suitable for
observation with FORS for at least 10 seconds without saturating
under typical conditions,
(2) it must include at minimum 2 Landolt stars with published U band
(3) it must not contain any stars bright enough to heavily bleed during
a 10 second exposure in any band.
Finding pointings that fulfill all three criteria is not a trivial
task and in a few cases we have had to relax (2) above. Since the
project is designed to bootstrap itself across the sky over a two
year time span, this relaxation is not critical in any way.
Each of the chosen pointings will be observed at 4 different
rotator angles (separated by 90 degrees). Each night one of
the FORSes goes on sky the night will start with the selection
of a pointing and a rotator angle to observe at. Those will
be following a sequential pattern such that 12 nights will ensure
that all three pointings are observed at all 4 rotator angles.
At the end of night another field, at different airmass, will be
observed following the same scheme. At the end of a sequence of
12 observing nights we shall then have a complete set of dithered and
rotated frames for a hugely overconstrained solution, and since the
selection of pointings and rotations is sequential the same will be
true for all following nights.
The on-line pipeline
The new FORS imaging pipeline currently under construction will
allow a real time solution of the linear equations. For each
night we can therefore use the current nights observations
together with those of the previous 11 observing nights to obtain a
complete solution for current extinction and illumination correction.
The only assumption used for this scheme is that the pixel-to-pixel
variations on the CCD, i.e. its underlying physical properties, does
not change significantly on a timescale of 12 observing nights.
At each pointing and for each filter we shall initially obtain two
exposures, one of length minimum 10 seconds (depending on filter)
suitable for the large number of Stetson standards and long enough to
ensure that shutter time errors can be ignored, and one short exposure
suitable for the Landolt standards which allows us to add the missing
bands to the Stetson standards, and also to provide an independent
link of Stetson standards to Landolt standards. The short exposures
will be terminated when our secondary standard set is deemed complete
enough to continue the bootstrapping without the Landolt.
Global properties such as "FORS above atmosphere zero-points",
system colour terms, magnitudes of observed standard stars, Paranal
extinction curve, can also be obtained the same way on a nightly
basis and their variation can be traced. However, better values can
be obtained from post-processing of a larger dataset on a less