Since we have now some more spectra, (thank you to all contributors!) we started some work on automating the measurements.
In particular we are working on a ESO-MIDAS script for radial velocitys, fwhm and line minima/maxima (for calculating ratios → Teff). It is developed for working on single lines. Maybe in the future we want let it measure more lines in one run. But first we have to test the quality of the data output!
We did some comparisons between our by eye and hand measured data and data out of the script. It may be that the script is working at a relative good level.
Here are some comparison graphs:
image (1).png [ 30.74 KiB | Viewed 255 times ]
We have implemented two options: One is to measure/read the line depth by the lowest pixel found with the MIDAS command statis/imag. The other is to use the gaussian fit. The second method we assume will work slightly better in spectra with low snr or emission lines which are disturbed by absorption (tellurics...).
image (2).png [ 31.91 KiB | Viewed 255 times ]
The rv measurement is also based on gaussian fits and can be adjusted in the integration range. It seems that using the whole line (about factor 1) leads to the most little scatter. (We used the FeII 6443 for this test. The also by script measured Si II 6347 looks less disturbed in the spectra.)
But if lines are disturbed there is of course more scatter. Like you can see at begin and end (time scale) at the graph.
Do anyone have ideas to improve the accuracy on disturbed lines (Due to star spots… etc)? Would for example a x-correlation on that small wavelength range with only one line help? As template maybe could serve a good (Bisector-method tested) line from our data sample?!
Malin and Christoph