Reducing a Flat

To reduce a FLAT, Portable--CGS4DR reads the raw data from store in $idir/ and FITS header information from store in $odir/, applies a bad pixel mask, optionally subtracts a reduced BIAS frame, linearises the data (if required), optionally subtracts a reduced DARK (with the same on-chip exposure time), normalises the data (if required) and writes the reduced data parameters into the index file and stores the reduced FLAT in $rodir/ as shown in figure .

For normalisation a 1-D spectrum is extracted from within the window defined by the bad pixel mask and a polynomial is fitted to that spectrum. The order of the polynomial is defined by the user and the polynomial is grown along the slit to fill the array. Finally, the original flat field is divided by the polynomial frame to produce a normalised flat field. Normalisation also removes the spectrum of the black body calibration source used to generate the flat field.

An alternative method of normalising the data using a smoothing box is also available.

For spectral configurations, the flat field algorithm pre-supposes that the illumination from the black body source is uniform along the slit. This does mean that if a very accurate photometric comparison of spectra at different points along the slit is required, the flat field will need to be checked for variations in illumination and refined to remove this variability.

It has also been discovered that at some wavelengths, the CVF acts like a Fabry-Perot and produces diagonal stripes in observations. These should be removed using oversampled FLAT fields although this has yet to be tried. An alternative strategy to remove these stripes is to take an OBJECT observation of the black body source at the same oversampling factor and normalise this particular observation to create a pseudo-FLAT field. Subsequently divide your real observation by this observation.

Finally, if the black body has not been allowed to warm up sufficiently, or the detector array temperature varies, or the detector is allowed to saturate resulting in ghost images and so forth, the spectrum generated may be too noisy. Observers should pay due care and attention to acquiring the data.

  
Figure: Reducing a FLAT