We have developed a highly automated IRAF-based data reduction package, CIRDR,
that can be used for both quick look data inspection, and generation of
astrometrically calibrated science quality 40964096 images based on a
contiguous 4
4 mosaic.
CIRSI, the Cambridge Infra-Red Survey Instrument, is a new near-infrared camera
that was built at the Institute of Astronomy. It
consists of four 10241024 pixel Rockwell HAWAII HgCdTe
(Mercury-Cadmium-Telluride) detectors, arranged in a sparse-filled
mosaic--making it the largest infrared camera in the world to date. CIRSI's
unprecedented wide field of view means it is ideally
suited for survey work. The previous generation of infrared detectors (typically
256
256 pixels) tended to have too limited a field of view for panoramic
surveys. This was unfortunate, as the near-infrared is very desirable for such
survey work. For example, when searching for high-redshift clusters of galaxies
in optical wave-bands, one finds that beyond about
, the contrast against
the very high field galaxy counts makes detections less and less reliable.
Additionally, the k-correction tends to dim the
light of distant early-type galaxies, which kills the cluster contrast at
optical wavelengths.
Thus we are able to combine the advantages of searching in the near-infrared wave-bands and CIRSI's ability to cover a large solid angle, an essential feature for finding rare objects like rich clusters of galaxies.
Most of our observations have been carried out on the 2.5m Isaac Newton Telescope on the Canary Islands. So far this has consisted of four observing runs, totaling 35 nights.
Table 1 lists our most important cluster observations.
At a scale of 0.45''/pixel on the INT, an image from a single chip has a 7.7'
field of view. Hence a filled 44 mosaic (see Figure 1) is just
over 30'
30' in size.
The camera is typically read out in non-destructive read mode (NDR)--see Figure 2 for an illustration.
The controller is a standard LSR-AstroCam 4100 CCD controller. Each quadrant on
a chip is read out individually, giving
images for each read. Hence,
for a typical observation, which would be a 40-second exposure consisting of 3
reads, the data rate would be
0.5MB | [size of a 512![]() |
![]() |
[number of quadrants] |
![]() |
[number of reads] |
giving 24MB in 40 seconds, i.e. 36MB a minute which comes to over 2GB an hour at a sustained rate.
![]() |
Each observation pointing is then repeated, offset by a small amount (typically 15''). Usually 9 or so of these dithers then make up a pointing.
In order to deal with such large volumes of data, we employed an observing system consisting of several computers and running a variety of operating systems, linked by fast 100Mb Ethernet. Storage was provided by several 16GB hot-swappable hard drives, in addition to multiple SCSI disks on the individual machines. Data would be written to DDS-3 tapes during the daytime.
The user interface is provided by a GUI, a program called PixCel, running on a Microsoft Windows 95 PC. The raw data is copied across automatically to Linux machines for reduction.
We have developed CIRDR, a highly automated IRAF-based piece of software, for reducing our data. This was initially intended as a quick look data inspection facility, but has since evolved into a fully-fledged data reduction package. The requirements for such a package are manifold:
For these reasons, and to maintain backward compatibility with previous software, we chose to base our package in IRAF. CIRDR is essentially a collection of CL scripts, with certain (speed-critical) tasks written in SPP.
The multiple tasks required for reducing are as follows:
![]() |
Even though these steps are all performed by individual sub-packages and tasks (which has the advantage that they are also usable individually, and could be re-used in separate packages), there exists a monolithic blanket task which runs through a large part of them fully automatically after initial input from the user.
Since earlier this year, with the start of CIRSI observations at the 2.5m DuPont Telescope in Chile, things have changed somewhat. Data is now read out in read-reset-read mode (RRR), and the images are written to disk pre-assembled and reset-corrected. Sky subtraction (and/or flat-fielding) is performed, after which the objects are located and subsequently matched with the on-line APM catalogue. This gives the exact offsets for co-addition, and for mosaicing if it is required. This new version of CIRDR is now written in a hybrid of CL and C.