Thanks Robin for the good info on creating a comparison quality assessment. I will use this as a framework to build my capability to the level needed. I found that I had several issues with the initial setup of Demetra, and a misunderstanding on how to use the reference stars to create a response curve.
Among the learning curve issues that I have resolved or managed to understand
– I had incorrectly set the zones in the target images for the spectrum response (too wide of a strip), and the background noise (too small of a region of the image). Once I reset those, my SNR improved significantly and the results improved.
– Calibration – still having problems in getting Demetra to ‘automatically’ recognize the calibration image and to properly assign the right spectral line. When I use ‘automatic’ I cannot get better than 2.4 RMS, where if I go in an manually set the lines to the image (5852, 6562, 6965, 7067, 7383) my calibration gets a very good .2-.8 RMS. The software may not be at fault. I develop platforms, and told one user ‘it’s idiot proof’… he confidently told me ‘you dont have the right idiots’…..
– Response: I was mistaken on how to use /implement the response curve. I thought I was to take an image of a ‘reference star’ after imaging each ‘target star’. Then use that image to create a response curve. I thought I was supposed to use the MILES (or Pickles) db to create a ‘response curve’ from the image of the ‘Reference Star’ by smoothing it from the MILES/Pickles db. Then apply that curve to the image of my ‘Target star’. I could not figure out how to apply that response curve to the ‘Target star’ in Demetra. However after some great info from the team here on BAA, I have been using just the B/A type reference star (derived from the spreadsheet) to create that smoothed ‘response curve’ directly from the MILES db onto my ‘Target star’.
Although I am not to the high quality results I am wanting yet, I am consistently starting to see results close to those on the BAA site.
Still learning a lot, and not quite to the levels to make any good sense from a test as you propose, as I want my results on a singular basis to be angstrom level accurate before I move to the next step (some targets <.5a others >2a depending on where in the spectrum)… So working on this repeatability now.
Always appreciative of any new suggestions, will keep pushing on this path.
Thanks
Keith