Back to forums

Bin or No Bin?

SetiAstro Statistical Stretch Celestron EdgeHD 8" Processing Binning Deep Sky
23 replies•650 views
Derek Vasselin avatar
Derek Vasselin avatar
Here's the scenario:
  • I'm playing around with a new Celestron EdgeHD 8"
  • Running it at native f/10 on an EQ6-R mount and OAG
  • Guiding RMS is usually around 0.5"
  • Seeing is usually average
  • Using an ASI2600MC, so that puts my pixel scale at 0.38 arcseconds/pixel


Image details (images linked below):
  • Per PSFImage, my FWHM (pre-BXT) is about 5.80"
    • My FWHM (post-BXT) is about 2.05"

  • About 5 hrs of integration (300s subs)
    • Note: I have 30 hrs of data, just haven't stacked yet

  • Process:
    • Crop > DBE > SPCC > BXT the Bin 1 image
    • Cloned the Bin 1 image
    • Used IntegerResample to get 2x2 binning of the clone
    • Used SetiAstro Statistical Stretch for both with default settings


Here is my ultimate question: Do I have to bin? 

Everything says I should, technically speaking.

Comparing the two visually, I see the Bin 1 image has better resolution (as expected), and the Bin 2 image has better signal, but the difference isn't massive, IMO.

I also imagine with 30 hrs of data, the Bin 1 would be fine in terms of SNR?

With this in mind, is there a specific reason I should use 2x2 binning, or is it ultimately up to personal preference?

Personally I like the better resolution and am willing to get more data to compensate for reduced SNR with 1x1 binning.

Images:
https://drive.google.com/drive/folders/1T1aAunfpEsRarPruAjC7QeYKUV1MdNSq?usp=drive_link
Helpful Engaging
Dark Matters Astrophotography avatar
Dark Matters Astrophotography avatar
No reason to other than storage space. The binning on the camera is software binning so it’s the same thing as you using Integer Resample in post. Might as well shoot bin 1 and decide later what you want to do in terms of sampling changes. Unless of course storage or bandwidth are considerations.
Helpful Concise
Tony Gondola avatar
Tony Gondola avatar
Exactly, there's no reason to make the original data lower in information when you can make that choice on the processing side. Storage is super cheap so even with a 2600 I wouldn't bin at capture.
Himanshu Pandey avatar
Himanshu Pandey avatar
I do the same, 1x1 binning with C8, in my mind I think that I'll get lucky here and there when the seeing is good and get some really good frames in the mix by keeping it at full resolution.
Carastro avatar
Carastro avatar
I almost always bin the colour or Oiii Sii as most of the details is in the Ha or Lum. 

l can then do 600 sevs in 300secs binned.  Spreds up the capture.  

one or two exceptions to this would be where the feature is in the Oiii fir example the Squid.
Richard Carande avatar
Richard Carande avatar
I have an Edge-11 and faced this issue a  few years ago when i got it and was using my legacy 2600.  I decided to get a camera with larger pixels for the Edge.  The only one really available was the full-frame ASI-2400mc-pro, with pixel size of 5.94um.  That pixel size is sort of like doing a bin 1.57x on the 2600 (3.76um pixels), but with better SNR, with the pixel area is 2.5x as big.  I was able to pick one up used and have been happy with it.

An added benefit is also the increased field of view the full-frame provides. 

However, if you're using an off-axis guider, as I am, you'll find it more difficult with the full frame camera since you've got less space to put the pick-off prism without eclipsing the light going to the main camera, and since you're on the very edge of the the larger FOV with less light power, good guide stars can sometimes be challenging to find.  (I'm typically using with the .7x reducer on the scope.)

And finally, if you're interested in mono cameras, you're out of luck; as far as I know there is no mono version of the 2400.  (I am shooting OSC obviously.)
Helpful Respectful Engaging Supportive
Willem Jan Drijfhout avatar
Willem Jan Drijfhout avatar
Your question is exactly the one I had a while ago. I did some testing and found that against the general rules of thumb, modern AI-driven tools such as BXT actually pull a tiny bit more detail out of an oversampled image. In follow up discussions, this behaviour was confirmed by others. If you're interested, you may want to check out this blog-post. 
So unless you're short on storage space, or you want to save on processing time, the best option is to stick to Bin1, also with a pixel scale of 0.38".
Well Written Helpful Insightful Concise
Dave Rust avatar
Dave Rust avatar
I have the Celestron 925 and shoot 1:1 with the 2600. Maybe 40% of the time, I'm able to get slightly more detail than BIN2, otherwise about the same as BIN2. It takes exceptional nights of calm air and good seeing to stay within the mount's guiding ability (best nights it guides .35-.45 across all subs). I'll post online the BIN2, but I keep the BIN 1 because it looks just a tad smoother with less aliasing when printed. I consider it the "master."
Helpful Insightful Concise
Derek Vasselin avatar
Derek Vasselin avatar
Appreciate the feedback.

I'll clarify I'm not concerned about storage space and I always shoot Bin 1 by default. If I do Bin 2, it would only be post-stacking.

My curiosity is solely about final image quality.
Well Written
Derek Vasselin avatar
Derek Vasselin avatar
Dave Rust:
I have the Celestron 925 and shoot 1:1 with the 2600. Maybe 40% of the time, I'm able to get slightly more detail than BIN2, otherwise about the same as BIN2. It takes exceptional nights of calm air and good seeing to stay within the mount's guiding ability (best nights it guides .35-.45 across all subs). I'll post online the BIN2, but I keep the BIN 1 because it looks just a tad smoother with less aliasing when printed. I consider it the "master."

I like this idea of a Bin1 "master"
John Stone avatar
John Stone avatar
Why not do your gradient removal and blurX processing on Bin1 then "software bin" a.k.a. Resample down to FWHM/3.33"-per-pixel?

This way you're getting all the signal at the same time as preserving all the resolution your seeing allows?
Ben Koltenbah avatar
Ben Koltenbah avatar
I also recommend imaging in Bin 1 with CMOS as there is no benefit to Bin 2 or higher other than reducing storage.  Storage is relatively cheap these days.  As others have said, processing afterwards in software gives you the choice of binning should you wish to downsample.  Deconvoluion typically works better on oversampled images, so that would be a reason to image in Bin 1.

With CCD sensors, the "to bin or not to bin" question is more nuanced.  With hardware binning on CCD, there is a read noise reduction of a factor of two going from Bin 1 to Bin 2.  I used to choose CCD hardware binning with my color channels because I didn't need the resolution for the color and benefited from the noise reduction.  CMOS doesn't provide such benefit through on-board software binning, of course.
Well Written Helpful Insightful
Alex Nicholas avatar
Alex Nicholas avatar
You don't get the same increased SNR by binning CMOS cameras that you would have with CCD's back in the day. The result will be the same wheterh you shoot Bin 1 and scale the image down 50% at the end, or shoot Bin 2. 

If I were you, shoot bin 1 all the time, you get the benefit of detail, and if you really need to smooth the data out - resize the image down 50%…
Insightful
Rick Krejci avatar
Rick Krejci avatar
And note that going full resolution opens up the possibilities to resample in finer increments than bin 2,4,etc.  If 75% resample looks good, then stop there. You will still improve noise.
Clayton Ostler avatar
Clayton Ostler avatar
I've been on the fence about this. My average exposure times are around 180 seconds, so I end up with lots of images, 

For me storing them is cheap, but transferring 500 images off my asiair then staking them, and editing them takes a lot of waiting even with a good computer. 

Question. I understand that software binning in cam for the asi2600mc is not near as good as what the ccd cameras do, or even as good as what mono cameras do.

But if I am happy with that image size/resolution as an output, am I gaining or losing anything by bin2 the image? When I say say size/resolution I mean pixels for example 3600x3600

Let me throw one more in here that I cant seem to understand…

Giant Pixels…. Either from Binning or Native Bigger….

Lets say I have big pixels and I am greatly undersampled.  I understand the blocky stars, (which I think drizzling seems to fix, combined with BlurrXterminator), but beyond that, I dont see the downsides. 

With larger pixels I could theoretically increase the gain (within reason) and capture faster.

But I must be missing something. If a 3600x3600 image is full of 8um pixels or full of 3um pixels, It will largely show the same details unless I am blowing the image up really large. But if my goal is to share online, I dont that I can see a difference, but I am not an expert, hence the questions. 

This whole conversation stems from me looking at a new camera that has 5.7um vs my existing 3.76, but my 3.76 are already undersampled with wide field, but the new camera is higher pixel count, (forget the FOV part), Is the new pixel size going to hurt of help in widefield?

Caviat, These are usually never viewed on anything larger than 40 inch monitor.
Tony Gondola avatar
Tony Gondola avatar
If you are down sampling your image to display it on a monitor that has fewer pixels than your image than yes, you will never see the difference in terms of resolution. For instance, a 4K monitor is 3840x2160. A 6200 camera is 9576x6388 and so is down sampled for display.
Helpful Concise
Clayton Ostler avatar
Clayton Ostler avatar
Tony Gondola:
If you are down sampling your image to display it on a monitor that has fewer pixels than your image than yes, you will never see the difference in terms of resolution. For instance, a 4K monitor is 3840x2160. A 6200 camera is 9576x6388 and so is down sampled for display.

I think considering a target/native resolution for viewing might help me determine if its worth it or not.
Tony Gondola avatar
Tony Gondola avatar
Yes, for viewing or printing if you ever do that.
Clayton Ostler avatar
Clayton Ostler avatar
Tony Gondola:
Yes, for viewing or printing if you ever do that.

Most my images arent worth the paper they are printed on.
Alberto Lopez avatar
Alberto Lopez avatar
I have same doubt. My equipment is a Meade ACF f/8 with Asi 2600MC Pro that provides native resolution of 0.32 arcsg/px. My target are small objects.

My options are: b1 (0.32), bin2 (0.64), or a focal reducer that provides with b1, 0.5 arcsg/px.

The conclusion I am reaching is that the final result depends of each object. Of course, for larger or brighter nebulae is more comfortable b2. But for small and dim galaxies, it is not so clear for me.
Clayton Ostler avatar
Clayton Ostler avatar
Try it out
Alberto Lopez avatar
Alberto Lopez avatar
Comparison between binning 1 and 2 on Stephan's Quintet (only Luminance).

First photo of 60s x 122 = 2h 2min.
https://moneda-hispanica.com/astrofotos/b1-60sX122-integration-DBE-BXT-Hist.jpg

Second photo of 300s x 21 = 1h 45 min.
https://moneda-hispanica.com/astrofotos/b2-300sX21-integration-DBE-BXT-Hist.jpg

Applied integration, DBE, BXT and stretching, trying to make both photos comparable.
Without removing noise, which is shown greater for binning 2. Please ignore vignetting, artifacts, etc.
Which photo do you think offers more detail?

Greetings,
dkamen avatar
dkamen avatar
Alberto Lopez:
Comparison between binning 1 and 2 on Stephan's Quintet (only Luminance).

First photo of 60s x 122 = 2h 2min.
https://moneda-hispanica.com/astrofotos/b1-60sX122-integration-DBE-BXT-Hist.jpg

Second photo of 300s x 21 = 1h 45 min.
https://moneda-hispanica.com/astrofotos/b2-300sX21-integration-DBE-BXT-Hist.jpg

Applied integration, DBE, BXT and stretching, trying to make both photos comparable.
Without removing noise, which is shown greater for binning 2. Please ignore vignetting, artifacts, etc.
Which photo do you think offers more detail?

Greetings,

I think they are about the same, if you scale down the bin 1 (or up the bin 2). How can bin2 be so noisy though? Supposedly you get a 2X boost from binning, a 2.23 boost from increased exposure and a 4.58 boost from integration = 20X the SNR of a single 60X bin 1 sub. Whereas the bin1 integration should have 11X the SNR of a single bin 1 sub, i.e. should be twice as noisy.

There must have been other differences, e.g. in the sky conditions.
Tony Gondola avatar
Tony Gondola avatar
That was my impression too. The biggest difference was in the noise. To be fair, that kind of a subject probably isn't the best candidate for this kind of test. Something like M1 or M82 would be a lot better. You need something with a lot of fine detail to really show off any differences.
Well Written