The following recipe describes a simple set of data reducton steps that correct for gain, bias, pixel-to-pixel sensitivty, cross-talk and mosaic the RSS amplifiers into a single image extension. Once the data have been mosaiced, external iraf tasks, e.g. noao.imred.kpnoslit may be used to extract and calibrate spectra. It is assumed that the user has some IRAF experience. Manuals and cookbooks are available from the IRAF document archive. For the sake of brevity, the recipe is written as it would appear on the command line of the IRAF cl environment, but a full list of task parameters may be inspected using 'lpar' or 'epar' inside the cl and online help is available by typing 'help taskname'. The help will provide more detail on the functionality of parameters listed below.

The image data itself is organised by one FITS file per exposure. Each file contains an empty primary HDU with scalar data recorded in keyword headers. The CCD images are stored in six subsequent HDUs, one for each amplifier in the detector.

The SALT RSS package can be accessed from inside the IRAF cl environment by typing 'salt' followed by a carriage return and then 'pfis'.

In the future, this step will add data for bad pixel maps and multi-object spectroscopy (MOS) masks to the raw data. Currently it purpose is to check the raw file for completeness and compatibility with the IRAF software, adding a few keywords required by the tools:

• pprepare inimages='P*.fits' rawpath='.' outpref=p fl_gain-
  

This step calls a calibration file stored in SALT's IRAF package that describes pre-calibrated gain correction factors that depend on readout speed and gain modes and applies them uniformally across the images.

• pgain images='pP*' gaindb='pfis$data/PFISamps.dat' fl_update+ fl_mult+
  

After subtracting a polynomial fit to the overscan strip of bias frames and trimming the overscan and underscan regions away, this step combines the residual images into a master bias frame which can be subracted from science data to remove bias structure. bias.lis is an ascii list of gain-corrected bias frames:

• pbias inimages='@bias.lis' outbias='bias.fits' fl_over+ fl_trim+ fl_inter- median- function='chebyshev' order=4 low_reject=3 high_reject=3 niterate=10 combine='median'
  

After debiasing flat field images, this tool will fit splines across the images and use them to normalize out any broad response and calibration features in the field. The resulting images can then be used to divide out pixel-to-pixel variability in science data. flat.lis contains an ascii list of flat field images from a calibration lamp.

• psflat inflats='@flat.lis' specflat='flat.fits' fl_over+ fl_trim+ fl_bias+ bias='bias.fits' fl_dark- fl_inter- fl_detec+ function='spline3' order=5 low_reject=3 high_reject=3 niterate=10 combine='average' reject=avsigclip lsigma=3 ovs_fun='chebyshev' ovs_ord=4 ovs_low=3 ovs_hig=3 ovs_nit=10
  

With reduced bias frames and flat fields in hand, this task performs cleaning of science data - debiasing, flatfielding and cross-talk correction: src.lis is an ascii list of source and arc lamp exposures:

• psreduce inimages='@src.lis' outpref='r' fl_over+ fl_trim+ fl_xtal+ fl_bias+ fl_pscr- fl_dark- fl_flat+ fl_pmos- bias='bias.fits' flatim='flat.fits' xtalkfi='pfis$data/PFISxtalk.dat' ovs_med- ovs_fun='chebyshev' ovs_ord=4 ovs_low=3 ovs_hig=3 ovs_nit=10
  

This step can be performed optionally. The alternative would be to treat each amplifier as a separate image, extract and calibrate the six separate spectra before combining them as a final step. However to mosaic the six amplifiers into a single 2-dimensional image use pmosaic. src2.lis is a list of cleaned science data.

• pmosaic inimages='@src2.lis' outpref='m' geomfil='pfis$data/PFISgeom.dat' fl_fixp+

The fl_fixp boolean is a flag to determine whether count levels are interpolated across the chip gaps or not. The advantage of doing this is in helping the spectrum extraction tools, e.g. kpnoslit, to trace the target in the dispersion direction. However the flag must be set to 'no' when mosaicing arc calibration spectra in order to avoid interpolating between widely-separated emission lines.

Form this point onwards you are in good shape to use your favourite (IRAF or otherwise) sotware to complete the reduction and analyze your data.

1. PyRAF versions of the IRAF tools.
2. Cosmic ray rejection.
3. Multi-Object Spectroscopy.
4. Bad pixel masks.
5. Polarimetry mode tools.
6. Fabry-Perot mode tools.
7. Slot and Frame Transfer mode compliance.