Experiments in photometry

[xilman]

The last half of 2017 had too many distractions for me to do anything on the photometry project but things are progressing nicely again now. In particular I now have a plausible explanation for the peculiar behaviour of the magnitude-vs-residual plots posted earlier in this thread.

First task was to take a series (66 in fact) of 1/4096 second exposures in as near complete darkness as possible and co-add them with averaging to form a master bias image. Contrary to a naive view the image is not black but a very dark and slightly noisy grey overlaid by an irregular collection of lighter or darker, vertical and horizontal lines and speckled with a few slightly lighter and very small spots. The spots are hot (warm might be more accurate) pixels which have a relatively high dark current. The stripes are artifacts of the scanning electronics and the non-zero offset is, I believe, inserted by the camera to allow for negative-going pixels when it (the camera) is performing fancy image processing --- which it isn't in this instance.

The bias image will be subtracted from every science image to remove the scanning artifacts and reduce the influence of the warm pixels. The latter will still show themselves as a slightly enhanced noise level in the processed image. Two pixels are sufficiently hot that I may decide to mask out that area of the images.

Second task was to build 2 master flat images. I set the desktop background colour on my monitor to white, iconified everything, moved the mouse pointer out of the way and then clipped on a sheet of translucent white plastic to cover the screen. This formed an entirely adequate light box. Then, with the camera tripod-mounted and viewing the screen from ~40cm away, I set it to f/4, 200mmFL, focused to infinity and took a series of snaps with exposures 2^-n seconds for 0<=n<=12. The 1/8s image had a grey-level of ~40K when converted into a 16-bit FITS file. Fifteen more images were taken at 1/8s with the above settings. Another 16 were taken at 80mm focal length. The two sets were separately co-added with averaging to produce a pair of light grey images with barely any discernible detail other than some narrow (a few pixels at most) black areas at all four edges. Histogram equalization, however, showed what was really going on. As expected with a wide-open lens, there is significant vignetting. Also as expected there were a dozen or so largish darker circular patches --- out of focus dust grains somewhere in the optical train.

All those non-linearities will throw out precision photometry unless corrections are made for them. The science images will be divided by the appropriate flat image before measurements; the result should then have a linear photometric response though, inevitably, the background noise level will vary over the image.

Tom almost certainly did not apply bias and flat corrections to his images and I suspect that the strange behaviour shown in the aforementioned plots arises from dusty and/or vignetted areas.

Now waiting for clear skies to be able to take preliminary science images. It's been cloudy for the last two weeks or so.

An additional correction which should be made but often need not be is to take a series of "dark" images, which are taken in complete darkness just before and just after the observing run at the same exposure time as the science images. The dark images are co-added and then subtracted from the science data before division by the flat image. The idea this time is to compensate for changes in camera response as a function of exposure time, temperature, and so on. Experimentation will needed to see how desirable the process may be.

[Spherical Cow]

Very useful discussion- thanks.

Norm