We have had a problem with one of the ThAr lamps slowly dying. The scripts used for calibration exposures include a (windowed) test exposure to define the required exposure by looking at a single line. This directly serves as a `health' check of the lamp. However, it turns out that lines from different excitation levels were decreasing in intensity in a different way which both cause the problem not to be detected immediately and the overall shape of the calibration spectra to change which caused significant shifts in the resulting wavelength calibration. To avoid this, we plan to develop a quality control system that makes a more complete analysis of any exposure using a ThAr lamp (including exposures with simultaneous observations of a target and a ThAr lamps) and checks for line-ratios to avoid this problem effecting the data.
It was noted that increasing the the ThAr lamp exposure time by a factor 6 with respect to to the 'standard' exposure time the RMS around the wavelength solution can be brought down from 80 m/s to 45 m/s. See
Thomas Augusteijn 2010-02-09