Blaze correction

To remove the wavelength dependent sensitivity or blaze from the data we need to use a quartz flat. The file quartz2dscrunch.sdf is a ready scrunched quartz lamp image.

   % display object2dscrunch mode=pe accept
   % lutgrey

The bright line towards the right of the image is a ghost. We wish to generate a smooth fit to the quartz lamp and the easiest way of doing this is to add together the rows to form a single spectrum. We use the values for the range in y-axis obtained earlier from measuring the arc lines.

   % extract quartz2dscrunch 13 493 quartzspec

This generates a spectrum quartzspec.sdf which we can view with

   % splot quartzspec whole=true autoscale=true hardcopy=false label=quartz

Figure: The extracted quartz spectrum.

As can be seen in the figure below, this spectrum has zero values at each end and the ghost line appears superimposed on the quartz lamp continuum. In order to remove these effects and obtain just the continuum we fit a polynomial curve to the data. We use the FIGARO cfit command:

   % cfit output=quartzfit

Use the cursor and mark points (using the a key) along the spectrum, including points at the start and end where we have no data and ignoring the ghost line. When you have finished this press x the exit. The fitted line is then drawn over the spectrum.

We can view this on its own with

   % splot quartzfit whole=true autoscale=true hardcopy=false label=quartzfit

We now have a smooth spectrum to divide into our data but before we do this one further stage is to normalize this curve. To do this we need to find the mean value for the quartzfit spectrum:

   % stats quartzfit

For my fitted spectrum this gave these results:

   Pixel statistics for the NDF structure
/home/mips/star/starlink/cookbook/example/quartzfit

      Title                     : QUARTZ SII FOR HONEYCOMB
      NDF array analysed        : DATA

         Pixel sum              : 1.7037235E9
         Pixel mean             : 1663793
         Standard deviation     : 511041.5
         Minimum pixel value    : 794998.6
            At pixel            : (1024)
            Co-ordinate         : (6753.693)
         Maximum pixel value    : 2492099
            At pixel            : (1)
            Co-ordinate         : (6700.157)
         Total number of pixels : 1024
         Number of pixels used  : 1024

The key number is the pixel mean value of 1663793, which we will divide the spectrum with to give a mean value of 1.

   % icdiv quartzfit 1663793 quartzfitnorm

We can now finally use this normalised spectrum to correct our observed data using the FIGARO isxdiv command:

   % isxdiv object2dscrunch quartzfitnorm object2dscrunchnorm

This command divides the quartz spectrum into each row of our data frame to produce the quartz corrected file object2dscrunchnorm.sdf (see the figure below.

Figure: The calibrated data frame, corrected for the instrumental response.

) with:

   % display object2dscrunchnorm clear mode=sc low=6 high=25 accept
   % lutgrey

With the data calibrated you have now completed the data reduction of a 2-D longslit spectral array. For analysis of this data you may want to use the TWODSPEC[23] LONGSLIT package or DIPSO[11].