Zwo Asi2600mc binning issue

It's always helpful to post some images displaying the issue you talk about. So some subs and the final images would help us a lot to support you.
For sure you need binned calibration frames. I wonder whether APP is up to the task to perform the proper demosaicing of the binned images.

You may find this article about binning color sensors interesting: Analysis and processing of pixel binning for color image sensor | EURASIP Journal on Advances in Signal Processing | Full Text (springeropen.com)

Clear skies
Wolfgang

It's always helpful to post some images displaying the issue you talk about. So some subs and the final images would help us a lot to support you.
For sure you need binned calibration frames. I wonder whether APP is up to the task to perform the proper demosaicing of the binned images.

You may find this article about binning color sensors interesting: Analysis and processing of pixel binning for color image sensor | EURASIP Journal on Advances in Signal Processing | Full Text (springeropen.com)

Clear skies
Wolfgang

Thanks Wolfgang for the link, I will check it out.

  Cheers
   Wozza

Why CMOS binning is a "bad idea" is you don't get any real benefits other than smaller files sizes.  With CCD it would actually pool pixels and create mega pixels, meaning each pixel's well would be combined.  In CMOS, this is not true.  Each pixel is still isolated.  So if you bin in post-processing it's the same effect, though as noted the larger file size.  Sounds like in your case there are other effects on the color.

Why CMOS binning is a "bad idea" is you don't get any real benefits other than smaller files sizes.  With CCD it would actually pool pixels and create mega pixels, meaning each pixel's well would be combined.  In CMOS, this is not true.  Each pixel is still isolated.  So if you bin in post-processing it's the same effect, though as noted the larger file size.  Sounds like in your case there are other effects on the color.

There are a lot of benefits to binning! It's Just not a good idea to do it on chip due to 16bit limitation. You'll still increase your signal and the noise is already so low on these new CMOS that its not a concern. 20k well becomes 80k at 2xbin. 


https://astronomy-imaging-camera.com/tutorials/everything-you-need-to-know-about-astrophotography-pixel-binning-the-fundamentals.html

Matthew Proulx:
20k well becomes 80k at 2xbin.

For a CCD, yes. For CMOS each well (pixel) remains independent. Therefore binning while shooting is the same as binning while processing. Tradeoff is file size, you do not get the benefits CCD provide as it's a different architecture.

Wrong. Read the link I provided. Even look at the graphic. 4 pixels combined as 1 pixel increases dynamic range and well depth, thereby increasing SNR. There’s so much info out there on this.

for ccd binning provides 4x SNR, for cmos the read noise doubles because the values are combined in digital after the readout, not analog before the readout, thereby cutting SNR to 2x. There’s still the benefit of 2x SNR. Why would you throw away signal? 

the reason it’s not a good idea to do it on chip with a cmos is because 4 16bit integers added is actually 18bit, but the camera only outputs a 16bit value. So you would likely clip your bright values. If you were doing short exposures this might not be an issue.

https://www.photometrics.com/learn/imaging-topics/binning-2#:~:text=CMOS%20Binning&text=In%20CMOS%20the%20binning%20occurs,for%20that%20resulting%20super%E2%80%91pixel.

andrea tasselli:
To add further to what Matthew said, I do not know of a procedure (in PI or Siril) that does the binning at the RAW level hence doing that in-situ, so to say, has non-negligible benefits.

Integer resample.

andrea tasselli:

Matthew Proulx:
Integer resample.

It doesn't do binning at the RAW (Bayer Matrix) level, AFAIK.

makes no difference as far as I’ve tested. do it after stacking to a linear image. 

I’ve done it to images after stacking and each individual mono image before stacking, the SNR is on par. It’s less time consuming to do it after.

andrea tasselli:

Matthew Proulx:
makes no difference as far as I’ve tested. do it after stacking to a linear image.

It does not do what binning at the camera level do, which is to average RGB levels (4*R, 8*G and 4*B) in a 16 pixel square and put them in the right order for a Bayer matrix scheme, WITHOUT interpolation, which is key to preserve the integrity of the original data.

It does close to the same thing though.



There’s some degradation but does it matter when A. you’re oversampled already and B. you’re using a OSC ? I’d rather bin than have a large oversampled image that is soft and lacking signal regardless of OSC or mono.

Not sure if this thread is dead, but ZWO recommends SW binning afterward, not in camera, please read the manual for the 2600MC. The only advantage is faster transfers, if you don't need that they recommend doing it later in SW.

I ran the SharpCap the sensor analysis for both 1x1 and 2x2. It is true the S/N and dynamic range is better for HW binning, but it is also true for SW binning later, so there is no advantage there. And remember if you HW bin in camera, the 4X increased signal gets crammed into the same 16 bits, so you lose bit resolution in your signal that you download from the camera, as each bit represents more signal. On the other hand, when you stack and register in DSS for example, it does stacking at 32 bits, so there is no issue there with scrunching your raw data into your final stacked FITS files. I presume other stacking software will similarly guard the dynamic range of your data, only downgrading carefully to 16 bits at the very end, where you will do the least harm.

Finally, why would you do something in-camera that is totally irreversible to your raw data. And there is no way to know ahead of time if you keep the data at 1x1, or should bin 2x2, or even more binning. Absolutely no way to know. Every target and every night is different, unless you are under uniformly dark skies with great seeing and no haze all time–judging from Astrobin, almost no one has these conditions.

Keep your raw data–as your experience increases, or you get the newest and greatest SW package, you can potentially recover more and more from the raw data. If you do something to the raw data now in camera, you can never recover.

There is a lot of mythology out there which is  quite true for CCD cameras, but CMOS plays by different rules. So what is a duh for CCD may be a poor choice for CMOS.

Maybe you can tell, I am a retired scientist, and as such, I hate to see anyone compromising their raw data.
Rick

andrea tasselli:

Rick Veregin:
Not sure if this thread is dead, but ZWO recommends SW binning afterward, not in camera, please read the manual for the 2600MC. The only advantage is faster transfers, if you don't need that they recommend doing it later in SW.

I ran the SharpCap the sensor analysis for both 1x1 and 2x2. It is true the S/N and dynamic range is better for HW binning, but it is also true for SW binning later, so there is no advantage there. And remember if you HW bin in camera, the 4X increased signal gets crammed into the same 16 bits, so you lose bit resolution in your signal that you download from the camera, as each bit represents more signal. On the other hand, when you stack and register in DSS for example, it does stacking at 32 bits, so there is no issue there with scrunching your raw data into your final stacked FITS files. I presume other stacking software will similarly guard the dynamic range of your data, only downgrading carefully to 16 bits at the very end, where you will do the least harm.

Finally, why would you do something in-camera that is totally irreversible to your raw data. And there is no way to know ahead of time if you keep the data at 1x1, or should bin 2x2, or even more binning. Absolutely no way to know. Every target and every night is different, unless you are under uniformly dark skies with great seeing and no haze all time--judging from Astrobin, almost no one has these conditions.

Keep your raw data--as your experience increases, or you get the newest and greatest SW package, you can potentially recover more and more from the raw data. If you do something to the raw data now in camera, you can never recover.

There is a lot of mythology out there which is  quite true for CCD cameras, but CMOS plays by different rules. So what is a duh for CCD may be a poor choice for CMOS.

Maybe you can tell, I am a retired scientist, and as such, I hate to see anyone compromising their raw data.
Rick

ZWO most enphatically does NOT recommends software binning afterwards, they recommend doing the binning in software in-camera rather than using the hardware binning mode of the 2600MC. As I wrote in my previous replay you can't bin an OSC camera for any software that I know of.

My objection to binning SW or HW in camera is that you can't go back. I would rather use another SW that I can control the binning exactly, rather than have to accept an arbitrary bin that I have to choose somehow before I see what the image looks like. I would never recommend a fixed bin for any particular setup--each target could very well require different bin. My setup is oversampled for deepsky targets, for good nights with targets with fine detail that need sharpening (Registax), I don't bin, or I even drizzle in SW later (just as one would for planetary to get the best resolution). For worse nights and/or objects that I can't get enough integration time I bin to varying degrees in SW once I have all my data, and can make the best tradeoff of noise and resolution on the final image. So I can try different binning to find the best one--which you can't do if you bin at acquisition.

As for programs that bin, DSS has a superpixel mode, so you can do binning in stacking, though I haven't used this, as to try different bin options you need to rerun the stack, a bit tedious. Typically I bin very effectively using Startools bin module with my OSC 2600MC, with improved S/N. You can do this on the final processed image, moving a slider on the bin amount to find the optimal setting. Also, when you reduce image size in Photoshop it also bins (averages--you have a few options on how to do the average). So I can and have binned in PS as well as part of my final processing. Those are the only programs I use, so can't comment on what other programs can do or can't do with binning.  I agree SW binning later is obviously not the same, one is done before debayer and one after--so one debayers the average, the other averages the debayer, but this will yield very similar if not nearly identical results once you have done both. Both work well for me, but again, since I don't know ahead of time the optimal bin for any particular image, or if I should bin at all, I do it after.

There is more than one way of doing many things that work well. Best advice I can give anyone is try different methods yourself and see what works for you with your setup, and don't expect every target has the same requirements for sampling.
Rick

Why CMOS binning is a "bad idea" is you don't get any real benefits other than smaller files sizes.  With CCD it would actually pool pixels and create mega pixels, meaning each pixel's well would be combined.  In CMOS, this is not true.  Each pixel is still isolated.  So if you bin in post-processing it's the same effect, though as noted the larger file size.  Sounds like in your case there are other effects on the color.

Pixel Size, FWC, and Dr increase while binning cmos. This comes at the expense of combining readnoise with the pixels : read noise increases. https://astronomy-imaging-camera.com/tutorials/everything-you-need-to-know-about-astrophotography-pixel-binning-the-fundamentals.html  

IMO if you choose to bin, the workflow is easier to do during acquisition. Up front just say I am sacrificing the extra pixels and shoot lights and calibration data binned 2x2.   

Unless you are sampling at or above 1" in above average seeing I don't see the point in binning.