Re: How to keep FITS header when stacking?

Posted by karenacollins on Aug 14, 2017; 11:25pm
URL: http://astroimagej.170.s1.nabble.com/How-to-keep-FITS-header-when-stacking-tp725p747.html

Hi Luca,

That's actually quite a nice light curve. If I bin the exposures by 6 to simulate a minute exposure, your RMS is right
at 1.0 mmag, which is on the better end of ground-based results.

I understand now why your error bars are so underestimated. The photon noise contribution is tiny because the target
star (HD189733) aperture has nearly 10^7 counts. In this extreme case of very high photon count, and short exposures,
the AIJ errors will be way underestimated (as you have discovered) because of scintillation (3x as you have pointed
out), and there are other sources of noise in that regime that will probably be hard to pinpoint.

There are also systematics (the red noise you mention I assume) that are causing excursions from the model, which could
be caused by changes in the sky background, changes in FWHM, or movement of the field on the detector, etc. It's hard
for me to assess the movement on the detector since apparently the images were aligned before running photometry. It
would be best to remove any known sources of systematics using detrending.

If I detrend using airmass, time, sky_background, and FWHM, the fit cleans up a good bit. I wonder if being able to
detrend on x- and y- centroid would resolve some more. Also, I noticed that some of your exposures are very near
saturation. Even if you are not saturating, you are likely into the higher end of the non-linear regime. If your
detector is like most I have worked with, I'd try to keep the peak pixel counts to 40-45K or less (or even 30K if the
star is not so bright).

The bottom line is that I agree that having AIJ handle the scintillation errors would dramatically help in this special
situation. I will indeed add this feature when I can find time to do so. By the way, I understand that making the
calculations should be an easy addition, but the biggest part of the work is tying it into the user interface and
testing to make sure the results are correct and that I haven't broken other stuff in the process.

Karen