Wavelength calibration

  • Extract the arc lamp spectra non-interactively, re-using the fibre traces from the flat as a reference. Data quality propagation is enabled here just because it results in better chip gap interpolation.

    gfreduce prgS20140919S0080 reference=prgS20140919S0060_init fl_addmdf- fl_over- fl_trim- fl_bias- fl_extract+ fl_gsappwave+ fl_wavtran- fl_skysub- fl_fluxcal- trace- recen- fl_vardq+ fl_inter-
    gfreduce prgS20140919S0081 reference=prgS20140919S0061_init fl_addmdf- fl_over- fl_trim- fl_bias- fl_extract+ fl_gsappwave+ fl_wavtran- fl_skysub- fl_fluxcal- trace- recen- fl_vardq+ fl_inter-
    • You may first want to compare the bias-subtracted arc and flat with gdisplay and check that the latter is actually an accurate tracing reference, with no major difference in flexure. Misalignment is likely to affect the wavelength accuracy only slightly, due to minimal fibre-to-fibre differences.
  • Make a copy of the GMOS CuAr line list that we can edit as needed for this dataset:

    copy gmos$data/CuAr_GMOS.dat scripts$CuAr_F110.dat

    It can be helpful to remove lines outside the wavelength range of your spectra (in this case 560-700nm) from the list at this stage.

  • Set some parameter defaults to be more appropriate for calibrating 2x1-binned IFU spectra (rather slit spectra):

  • Calibrate the arcs in wavelength. The results must be checked interactively, but you can again ignore the second exposure until you finish iterating on the first one.

    gswavelength eprgS20140919S0080 order=5 fl_inter+
    gswavelength eprgS20140919S0081 order=5 fl_inter+
    • The first plot to appear shows the spectrum of the middle fibre with automatic line identifications overlaid (assuming they were found automatically).


      Final line identifications for the middle fibre of slit 1.

      You can use the + and - keys to cycle through line identifications and see their matching list wavelengths (3rd number).

      It’s best to delete the faintest lines (eg. < 100-200 counts), as long as you have at least (say) 10-20 remaining over the spectral range. You can zoom with w e e (as for extraction) and comment out any unwanted entries in the line list.

      If no line identifications are shown, you will have to mark 2 lines manually and enter their wavelengths, then press l to match the rest of the line list (press ? for further information).

    • Pressing f shows the fitted wavelength solution (see ? for more options). Any line measurements excluded by the rejection algorithm are shown with a box around them; you may want to return to the first plot by pressing q and have a closer look at these, commenting them out from the line list if they look blended or border a chip gap etc.


      Wavelength solution for the middle fibre of slit 1.

      Here, I have commented out lines at 5606, 5772, 5802, 6155, 6172, 6364, 6369, 6466 & 6483A. The RMS of 0.07A seems perfectly reasonable for 2x1 binning (at about 1/14th of a pixel). It will vary a bit for the other fibres.

    • Press q (twice if in fitting mode) to move onto the next fibre spectrum. You can then generally type NO, to skip all the other fibres – just make sure the first one plotted looks good and that the status lines printed for the others show most of the same arc lines being found and generally reasonable RMS values.

  • Resample the calibrated arcs onto linear wavelength co-ordinates, to check the solution (ie. that the lines are straight):

    gftransform eprgS20140919S0080 wavtraname=eprgS20140919S0080
    gftransform eprgS20140919S0081 wavtraname=eprgS20140919S0081
    displ teprgS20140919S0080[sci] 1  # (for example)

    There is often some sub-pixel fuzziness, due to small differences in the solutions for individual fibres (especially at the very ends, where the fits are underconstrained), but it’s not practical to tweak every fibre individually. There are also some darker rows, since the data have not been flat fielded, as well as cosmetics etc. You may see a few second order or ghost lines superimposed, which are curved, diffuse or appear only in one slit; these can be ignored here, as long as they did not cause misidentifications in gswavelength.

    This particular spectrum exhibits faint interpolation “ringing” (dark columns) around the brightest lines, due to the 2x1 binning.