Lick Infrared Camera User's Guide
Reduction Technqiues
A Suggested Reduction Sequence
The usual goal of the reduction process is the removal of dark current,
cosmetic defects, pixel-to-pixel response variations, and sky contribution
from
individual images. Targets may be photometrically calibrated. Finally,
individual images are combined. Most reduction strategies suggested by
various
`experts' for the treatment of near-IR array data follow the same general
theme as
the one presented below.
- Dark Subtraction A master dark frame is
created by
averaging a number of dark frames of a given exposure time. Dark frames
may be scaled with caution, remembering not to scale the bias if it is
present.
The master dark frame is subtracted from all source and calibration frames
with the same exposure time.
- Bias Subtraction If baseline subtraction was
disabled
(default condition for LIRC-II), dark frames will have included the ADC bias,
and it will be automatically removed in step one, above. Separate bias
subtraction may be done, using a master bias frame made by averaging a series
of `zero' second darks. If baseline subtraction was
enabled, the bias will have been removed from the raw frames at the time
they were made.
- Removing Cosmetic Defects Hot or
dead
pixels should now be removed from both sky and source frames. If allowed
to remain, spurious counts in bad pixels may affect the flat-field
renormalization. The customary procedure is to interpolate over bad pixels.
A .i.bad pixel map; can be generated from a single suitable sky frame at the
beginning of the data reduction.
- Flat-fielding Flat field frames, be they
twilight or night-sky
frames, are averaged, using a median filtering or sigma clipping algorithm to
remove any stars. It is assumed here that individual images, or sets of
images, were offset from one another, so that the same star falls on different
pixels in different frames. Each source frame is divided by the flat-field,
and
renormalized by multiplying by the mean of the flat-field frame.
- Sky Subtraction Master sky frames are
made for each observation by averaging and filtering its associated set of sky
frames. The master frame is then subtracted from each source frame. Bear in
mind that in many cases, source frames, taken at slightly different positions,
can serve as one another's sky frames.
- Coaddition The individual,
reduced source frames are now combined,
registering them as necessary, to create a final source frame with the best
possible signal to noise. The use of a clipping algorithm here will eliminate
cosmic ray events.
- Flux Calibration For absolute photometric work, the
standard-star frames are reduced exactly like source frames. The data are
then
calibrated using the known standard star magnitudes and converting them to
observed counts as in visual photometry.
Quick Looks
The foregoing reduction recipes are not suited for quick looks at new images
while at the telescope. Simply subtracting a raw sky frame from a raw image
frame of the same exposure time provides an adequate first look. Where the
signal-to-noise in individual images is low, it may be necessary to coadd and
average a series of sky-subtracted, co-aligned frames.
Lirctop provides a sky-subtraction
procedure under its `quick-reductions'
option.
Registering Images
A single image in its final form is usually the sum of a number of reduced
source frames. Source frames are likely to be offset from one another due to
small errors in telescope tracking, deliberate offsets, or mosaicing to cover
regions of sky larger than the detector format. Images must be properly
registered when being coadded or fitted into a mosaic.
One or more bright point sources, common to adjacent frames, serve well as
registration marks. Frames without such reference points must be registered
on the basis of telescope coordinates. It is therefore critical to have
accurate records
of the telescope's position for each offset. Note that the FITS header
includes the
telescope position at the time of the observation, so long as `telco
access' is
enabled in the data-taker. While absolute positions recorded in this way are
only as good as the pointing calibration of the telescope, relative positions
within a limited field are quite accurate.