To bin or not to bin?

[Robculm]

Hi, can anyone please confirm I'm understanding this correctly & your recomendation!

With my 200PDS + coma corrector / reducer, I have a focal length of 900mm.

EOS800D has 3.7um pixel size.

Gives a resolution of 0.85" per pixel according to Astronomy Tools.

The ideal pixel size for Good Seeing (1-2" FWHM) seeing is: 0.33 - 1" / pixel & for OK Seeing (2-4" FWHM) seeing is: 0.67 - 2" / pixel.

So with good seeing, probably even OK seeing, I'm in a good spot in terms of optimum 'sampling', which implies that I shouldn't bin?

Since starting with Startools, I'm generally binning (software) by 50%, mainly to reduce the 'size' & help with processing speed (I have a reasonably new laptop, but it's fairly basic & certainly struggles).

Should I be exploring a faster computer (or just be more patient) & keeping full resolution, or binning by 50%?

Don't want to risk losing good data when it's so hard to come by!

Cheers,

Rob

 

[barbulo]

As I’ve been told several times, a good resolution for beginners, like myself, is 1.5”-2”/px. With an accurate guiding and a really good seeing you could aim at 1”-1.5”. Never below. In your case, binning at 2x2 would give you a 1.7”/px, a good starting point.  
HTH

Edited October 5, 2021 by barbulo

[vlaiv]

24 minutes ago, Robculm said:

The ideal pixel size for Good Seeing (1-2" FWHM) seeing is: 0.33 - 1" / pixel & for OK Seeing (2-4" FWHM) seeing is: 0.67 - 2" / pixel.

That is far from ideal.

Almost no amateur is capable of producing images that are truly below 1"/px.

Math behind all of that is somewhat more complex than simple rule of the thumb like that. There are three major components that impact final FWHM of stars in the image: Seeing FWHM, Guiding performance and Scope aperture.

If you want things simplified - Optimum sampling rate is FWHM of stars that you actually produce divided with 1.6. For example - you shoot your image and measure FWHM of stars in it and it turns out to be 3.5" - optimum sampling rate in that case is 3.5"/1.6 = 2.1875"/px

Saying that seeing alone is responsible for star profile FWHM is incorrect. Seeing FWHM is defined as FWHM of star profile imaged for 2 seconds with very large aperture (large enough so its contributions are negligible - that means 20" or more).

If you have 2" FWHM seeing and use 60mm scope - you can't expect to have stars of 2" FWHM in your image.  This is because Airy disk of 60mm scope alone is ~4.3", and two "mix" to produce final result (they convolve in mathematical terms). Add to that - your guiding errors if you image for few minutes, as mounts don't track perfectly and smear your stars some more.

You have rather big aperture at 8" and you should be ok with 1.7"/px so yes, bin your image.

In fact - you can just image normally and calibrate and stack image normally and then in the end, while still linear before you start any processing - decide if you want to bin and by how much.

On some nights it will make sense to bin x2 but on some nights - it will make sense to bin x3.

This will also depend on what sort of coma corrector you are using. Some coma correctors introduce a bit of spherical aberration thus enlarging stars in the image - which reduces resolution further (above discussion on seeing, airy disk and guiding holds for diffraction limited optics. Coma Correctors, Field flatteners and reducers often bring down telescope below diffraction limit - which is fine for long exposure imaging, but should be taken into account - stars will be just a tad wider than diffraction limited scope).

If you are in doubt if image can be binned - there is rather simple method of testing this. Take original image, and down sample it to smaller resolution and then up sample it back to original resolution and look for differences. Here is an example. This is part of M51 image that has optimum sampling rate around 2"/px (taken with 8" RC scope on modded Heq5 with guide error in 0.5-0.6" RMS range). I actually binned it initially to 1"/px (since my native resolution is 0.5"/px ASI1600 and 1600mm FL scope)

You might think that this image is looking good and sharp - but it is not properly sampled, here let me show you.

Here it is down sampled to 2"/px using Lanczos resampling. Now that is what I would call sharp and not first image. This image looks like it is properly sampled (btw, all of this is when you observe image at 100% zoom - if you have it rescaled to fit the screen - you might not notice any of this).

I'm now going to enlarge that small image to 200%

Can you spot any differences to original image (except perhaps lower noise)? This goes to show that no detail is lost at 2"/px - and SNR is gained at 2"/px.

That is the point of binning - process your data so that no detail is lost in the image and you have best SNR possible for your data.

[Robculm]

Thanks barbulo & thanks Vlaiv for the detailed response 👍.

Now I just need to figure out how the Startools binning scale corresponds to this!

It's not clearly defined (at least to my interpretation), but I guess 50% is equal to x2 binning?

Cheers,

Rob

[vlaiv]

26 minutes ago, Robculm said:

Thanks barbulo & thanks Vlaiv for the detailed response 👍.

Now I just need to figure out how the Startools binning scale corresponds to this!

It's not clearly defined (at least to my interpretation), but I guess 50% is equal to x2 binning?

Cheers,

Rob

Yes.

[Robculm]

👍 perfect. 

Thanks again 😀