Star size vice exposure and binning

[menardre]

Reviewing the 'literature' I found different suggestions concerning taking a few long exposures or several shorter exposures. A while ago I did my own test in which it was evident that faint nebula's require longer exposures rather than many shorter exposures in order to pull out details and color.

But... what about star size? What is the best setting if your interest is trying to keep pin-point stars?? Last week I decided to run my own test. I took exactly one hour of exposures of a star cluster with 4 different settings:
1) 60, 1 minute exposures binned 1x1
2) 60, 1 minute exposures binned 2x2
3) 30, 2 minute exposures binned 1x1
4) 30, 2 minute exposures binned 2x2

All of these were with 11 inch SCT and ZWO ASI2600, and all in same night. With this setup my image scale is about 0.396 arc-sec/pixel. With 'seeing' at about 1 arc-sec/pixel, I should be able to bin 2x2 or even 3x3 without noticeably degrading the image. So, let's see.

The results are interesting but not always intuitive.
I processed the images exactly the same using Pixinsight and only doing the processing required to stretch the image. Then I evaluated the image statistics:
1) Eccentricity doubled with doubling the exposure but was independent of binning. The eccentricity of all 1 minute images regardless of binning was about 0.3 and the eccentricity of the 2 minute images regardless of binning was about 0.6.
2) Number of stars found increased with binning. In both the 1 minute and 2 minute sets, the number of stars found when binned 1x1 was significantly lower than when binned 2x2. I did not expect this.
3) No appreciable difference in FWHM.
4) Signal to noise ratio increased considerably when going from binned 1x1 to binned 2x2 in both the 1 minute and 2 minute sets. This was expected since binning 2x2 reduces read noise.

Attached are the images with minimal processing - just to the point of stretching. You can draw your own conclusions.

At first glance the images look close to identical. You can only see differences when you magnify the images. To my eye the 1x1 binned images look ever so slightly more 'focused'...


Roger

60 -1 minute exposures binned 1x1

NGC 659 1 min 1x1 Master_ABEdiv_Autocolor_BN_STF_NXT.jpg (12.9 MiB) Viewed 343 times

60 - 1 minute exposures binned 2x2

NGC 659 1 min 2x2 Master_ABEdiv_Autocolor_BN_STF_NXT.jpg (3.51 MiB) Viewed 343 times

30 - 2 minute exposures binned 1x1

NGC 659 2 min 1x1 Master_ABEdiv_Autocolor_BN_STF_NXT.jpg (12.58 MiB) Viewed 343 times

30- 2 minute exposures binned 2x2

NGC 659 2 min 2x2 Master_ABEdiv_Autocolor_BN_STF_NXT.jpg (3.5 MiB) Viewed 343 times

[mark.m]

Roger, I'm glad you did this experiment. With some exceptions, it confirms things that theory predicts:

• Yes, eccentricity gets worse with longer exposures. Stars will look more round by stacking many short exposures than by stacking a smaller number of long exposures. (This is due to imperfect guiding/tracking that gets "fixed" during stacking.)
• Number of stars being found is purely a function of the algorithm used for finding stars. (The theoretical performance is the same.) If the algorithm properly tunes itself to the image characteristics, there won't be any difference. (On the other hand, the software may be (should be?) much slower when analyzing the 1x1 binned image than the 3x3 binned image. In my case, the slowdown is something like a factor of 20x slower.)
• FWHM should be the same, as long as you're measuring FWHM in arcseconds and not pixels.
• SNR should be the same as you change binning. If it isn't, this says more about the software than about the image. (You also need to be careful about the definition of SNR; I'm using the definition that is used for star analysis, which may be different than the definition of SNR used for general astrophotography?) SNR should improve slightly (maybe unmeasurably) with longer subexposures. With my version of this camera, star SNR (for faint stars) is dominated by background image noise caused by light pollution; read noise is too small compared to light pollution noise to have much of an effect. But correct computation of the SNR requires the software to grab (and use!) the system gain information (probably in the image header), and the system gain can be different when binned vs unbinned (maybe -- depends on how the binning was done: average vs. sum).

Good experiment!!

[Pete]

Fascinating Roger. Thank you for running this study.

To my eye the 2X is noticeably brighter and also a bit more bloated. Not sure that this correlates with theory for CMOS.

When you bin 2X with a CCD camera your pixel is four times the size and this larger pixel captures 4X the photons of a single pixel. And the ASI2600MC-P does this for 2X binning, but only in the RAW 8 format. For RAW 16 images it simple averages the 4 pixels and there is no improvement in sensitivity, only in smaller file size that we've discussed before.

The size of the stars should theoretically be limited by seeing and not binning. Local seeing is usually 2 to 5 arc seconds. Need to review. But at this time, like Mark, I suspect that the difference is in the processing software and the brighter larger stars are equivalent to having cranked up the gain, allowing dimmer stars and dimmer regions around bright stars to show - same as what happens when while processing one stretches the image.

I've saved board posted prior experiments. WHAT IS THE BEST EXPOSURE TIME? from 8 Dec 2020 and WHAT IS THIS THING CALLED GAIN? from 6 Jan 2021 and from TO BIN OR NOT TO BIN, THAT IS THE QUESTION posted 31 Dec 2020 This last one is found at viewtopic.php?f=32&t=6672&p=51960&hilit ... bin#p51960 but there is followup at viewtopic.php?f=32&t=7051&p=53725&hilit ... bin#p53725