A variety of operations need to be performed on a CCD frame before the
reduction process described in Figure
can commence. These include reorienting the image,
subtraction of the CCD bias, removal of cosmic rays, and
division by a flat field frame to remove pixel to pixel gain variations.
The FIGARO conventional orientation requires that orders lie
horizontal with wavelength increasing to the right, and with bluer orders at
the bottom. AAO CCD frames must be rotated and flipped:
$ ROTATE image=object output=obj_rot
$ IREVY image=obj_rot output=obj_final
The bias level can be obtained from the overscan region of the CCD
frame using ISTAT, and removed using ICSUB.
If time--variations of the bias occurred during readout, the overscan rows can
be EXTRACT-ed, SFIT with a suitable polynomial, normalized with
ICDIV, and ISXSUB-tracted.
If the CCD has a small non-linearity when the intensity level
is near saturation (65535 ADU), after removing the bias
the non-linearity can be removed using CCDLIN. This requires
knowledge of the non-linearity parameter
, which is known for AAO CCDs
but is negligible for most spectroscopic data.
Cosmic rays can be eliminated using CLEAN, BCLEAN, or
BCFCLEAN algorithms.
The cosmic ray may be a small fraction of the signal (but not of the noise)
and difficult to see. It is worth temporarily removing the inter-order gaps and
the intensity variations along the slit before cleaning the image. A
simple procedure for doing this is as follows:
$ IXSMOOTH IMAGE=object SIGMA=100 WIDTH=50 OUTPUT=smooth $ IDIV IMAGE=object IMAGE1=smooth OUTPUT=temp $ CLEAN IMAGE=temp OUTPUT=temp $ IMULT IMAGE=temp IMAGE1=smooth OUTPUT=object
If several similar images are available, say from a series of flat field or stellar exposures, then one of the images can be used instead of the smoothed image as a reference. Cosmic rays on the denominator image will appear as dark spots on the image to be cleaned but can be ignored in the cleaning process.
Flat field frames are obtained by taking repeated exposures of a
continuum source and then adding them. The effects of non-uniform illumination
of the detector are removed by dividing the frame by a smoothed reference frame.
Overscan columns can cause errors in the smoothed flat field near the edge of
the frame. To overcome this
problem it is necessary to first subset the image to remove the bias columns
and then superset the frame later to restore the proper dimensions. The
following sequence illustrates the principal. In this sequence it is
assumed that the individual flat field frames have been reoriented, cleaned and
added into the final frame named `flat'. In this example, the original frame
has dimensions of x= 584, y=386.
$ ISUBSET image=flat ystart=1 yend=386 xstart=1 xend=576 output=flat_sub $ IXSMOOTH image=flat_sub sigma=100 width=50 output=smooth $ IDIV image=flat_sub image1=smooth output=flat_sub $ ISUPER image=flat_sub xsize=584 ysize=386 xpixel=1 ypixel=1 output=flat