12 replies to this topic

[Ben Diss]

I've been imaging narrowband with an EdgeHD 11" and struggling to get good SNR. I'm about to switch to a 6" refractor. Should I expect to get better SNR results with a refractor in general or would any observed differences be the result of my specific equipment?

 

Thanks and I'm not trying to stir up a hornets nest.

[TOMDEY]

Expect it to be 90% equipment and entire configuration related. There's no fundamental difference between dioptric vs catadioptric vs catoptric involved, provided the systems' 1st order properties are apples-apples.    Tom

Edited by TOMDEY, 15 April 2025 - 07:17 PM.

[brischy]

I've been imaging narrowband with an EdgeHD 11" and struggling to get good SNR. I'm about to switch to a 6" refractor. Should I expect to get better SNR results with a refractor in general or would any observed differences be the result of my specific equipment?

 

Thanks and I'm not trying to stir up a hornets nest.

What's the pixel size of your sensor? That might be the trick to the SNR with an Edge HD 11" since it is a light bucket, but with that unwieldy focal length. 

[555aaa]

Astronomers define SNR for stars at a given magnitude in an image, not for an image. When you use that definition and use the most optimal binning, the largest aperture always has the highest SNR. You have to bin on the C11, and BTW the SNR is strongly correlated with FWHM. For example, if you go from 3 arc seconds FWHM down to 2 arc seconds, you can reach the same SNR in the same exposure time for one whole magnitude fainter stars. Tighter stars is free aperture (or free SNR).

Edited by 555aaa, 15 April 2025 - 09:24 PM.

[charles.tremblay.darveau]

If you put the same camera on a refractor with faster F/# and same binning, then yes each pixel will accumulate photon fasters.

 

However, if you match the same sampling in arcsec then the larger aperture will win. The 6200mm is a good match for the edge 11 as you as you bin it to a reasonable sampling for you seeing. What the refractor will do is get a larger FOV with a reasonable sampling for most people seeing. But if you are interested in small target the Edge 11 will do a better job after downsampling.

 

The refractor will also have better contrast, and while it's difficult to put a number on it, the image will look cleaner for the same SNR.

[Ben Diss]

What's the pixel size of your sensor? That might be the trick to the SNR with an Edge HD 11" since it is a light bucket, but with that unwieldy focal length. 

ASI6200 is 3.76um. I'm binning 2x so 0.55 arcsec/pixel. I'm bortle 6, so maybe binning would help?

[Ben Diss]

OK, so I read a bunch of online stuff on binning CMOS. I don't see how hardware binning helps SNR, but the consensus is that Pixinsight binning could reduce background noise. I can't wrap my head around this. Can anyone point me to a thread or article that might straighten out my thinking on this topic?

[555aaa]

OK let's consider photometry, using the aperture method, because SNR is well defined since the object is known (it's a point source). In the aperture method, you have a circle of a certain angular size that you overlay on the star image. Add up all the ADU counts of all the pixels in the circle. Then, outside the circle and separated by a small gap, you make a ring (a donut) around the circle. Add up all the ADU counts of all the pixels in the donut. That provides the measure of the sky background.  Now, you know that inside the circle, you have both star flux plus sky flux. In the donut you have sky flux only. So the SNR is the ratio of the counts you measured in the circle to the square root of the sum of the circle counts plus the donut counts (also you normalize the the area of the donut and the area of the ring).

 

So the smaller you make the circle, the more you get of star only and not sky. You can do that when you have good seeing and high resolution. If your circle is five arc seconds in diameter, you are getting a lot of sky flux added in. If your circle is only one arc second in diameter, you get 25x less sky flux so your SNR is (IIRC) sqrt 25 or 5x higher.

 

Even if you're not doing photometry you can easily see this on images where when the seeing is good, faint details suddenly become visible. That is because there is more information in those images = higher SNR. The SNR of a complex image is not really well defined because you never know what the noise-free image is. But with a star you know exactly what the noise free image is and it's super easy to calculate and many tools do this for you automatically.

 

You can mostly discard the camera performance when you do not have heroically good skies or if you're doing narrowband; the sky flux is going to dominate for most of us if we are using good cameras, which we mostly have now.

 

see 6.5.2 in the paper below

 

https://www.stsci.ed...osuretime6.html

 

So you need to share what FHWM you are getting for stars, what you are binning to (either in camera or processing), and what your sky background is. ASTAP can calculate the sky background for you. A 10" SCT can go very faint under the right conditions, a lot fainter than a 6" refractor.

Edited by 555aaa, 16 April 2025 - 03:20 PM.

[RichA]

I've been imaging narrowband with an EdgeHD 11" and struggling to get good SNR. I'm about to switch to a 6" refractor. Should I expect to get better SNR results with a refractor in general or would any observed differences be the result of my specific equipment?

 

Thanks and I'm not trying to stir up a hornets nest.

Lower signal to noise ratio going to the much smaller refractor.  You cannot beat aperture if both scopes have reasonable (not great) optics.

[Jeff B]

Best asked in the imaging forum?

[charles.tremblay.darveau]

OK, so I read a bunch of online stuff on binning CMOS. I don't see how hardware binning helps SNR, but the consensus is that Pixinsight binning could reduce background noise. I can't wrap my head around this. Can anyone point me to a thread or article that might straighten out my thinking on this topic?

There is a misconception that binning won't improve SNR in CMOS camera. The reason is that there are two types of noise:
-Read noise from the camera itself: This one benefits more on CCD (1/N_bin) versus (1/sqrt(N_bin)) for CMOS. However Read noise is very low on CMOS to begin with and only dominates for very short captures.

-Shot noise from electronic, skyglow or true signal: Because you sample finite photons in finite pixels, you are susceptible to sampling noise. This is the main noise that we see in CMOS camera for DSO captures. This one averages as 1/sqrt(N_bin) for BOTH CCD and CMOS.

 

When you bin2, you effectively increase the cross-section by 4, so you improve acquisition speed by 4X and SNR by 2X relative to bin1. That means your 10hrs captures in bin2 now has the same SNR as a 40hrs capture in bin1. Quite the improvement.

 

You don't need to do this binning in hardware and can do it in SW, but the cost is a 4X increase in the storage space. If you know you will bin at the end of the day, why not do it in a efficient way to save up processing time? If you are worried about resolution drop, you can always do the exercise to compute HFR at bin1 and bin2. A 0.55arcsec resolution should be sufficient for about any site except if you live at the summit of Haleakala .

[Ben Diss]

OK let's consider photometry, using the aperture method, because SNR is well defined since the object is known (it's a point source). In the aperture method, you have a circle of a certain angular size that you overlay on the star image. Add up all the ADU counts of all the pixels in the circle. Then, outside the circle and separated by a small gap, you make a ring (a donut) around the circle. Add up all the ADU counts of all the pixels in the donut. That provides the measure of the sky background.  Now, you know that inside the circle, you have both star flux plus sky flux. In the donut you have sky flux only. So the SNR is the ratio of the counts you measured in the circle to the square root of the sum of the circle counts plus the donut counts (also you normalize the the area of the donut and the area of the ring).

 

So the smaller you make the circle, the more you get of star only and not sky. You can do that when you have good seeing and high resolution. If your circle is five arc seconds in diameter, you are getting a lot of sky flux added in. If your circle is only one arc second in diameter, you get 25x less sky flux so your SNR is (IIRC) sqrt 25 or 5x higher.

 

Even if you're not doing photometry you can easily see this on images where when the seeing is good, faint details suddenly become visible. That is because there is more information in those images = higher SNR. The SNR of a complex image is not really well defined because you never know what the noise-free image is. But with a star you know exactly what the noise free image is and it's super easy to calculate and many tools do this for you automatically.

 

You can mostly discard the camera performance when you do not have heroically good skies or if you're doing narrowband; the sky flux is going to dominate for most of us if we are using good cameras, which we mostly have now.

 

see 6.5.2 in the paper below

 

https://www.stsci.ed...osuretime6.html

 

So you need to share what FHWM you are getting for stars, what you are binning to (either in camera or processing), and what your sky background is. ASTAP can calculate the sky background for you. A 10" SCT can go very faint under the right conditions, a lot fainter than a 6" refractor.

 

Thank you for the very detailed explanation. I'll reread this a few times.

[Ben Diss]

Best asked in the imaging forum?

Yes, thanks. I'll study this material and take it to the imaging forum for follow-up.  Thanks.

 

There is a misconception that binning won't improve SNR in CMOS camera. The reason is that there are two types of noise:
-Read noise from the camera itself: This one benefits more on CCD (1/N_bin) versus (1/sqrt(N_bin)) for CMOS. However Read noise is very low on CMOS to begin with and only dominates for very short captures.

-Shot noise from electronic, skyglow or true signal: Because you sample finite photons in finite pixels, you are susceptible to sampling noise. This is the main noise that we see in CMOS camera for DSO captures. This one averages as 1/sqrt(N_bin) for BOTH CCD and CMOS.

 

When you bin2, you effectively increase the cross-section by 4, so you improve acquisition speed by 4X and SNR by 2X relative to bin1. That means your 10hrs captures in bin2 now has the same SNR as a 40hrs capture in bin1. Quite the improvement.

 

You don't need to do this binning in hardware and can do it in SW, but the cost is a 4X increase in the storage space. If you know you will bin at the end of the day, why not do it in a efficient way to save up processing time? If you are worried about resolution drop, you can always do the exercise to compute HFR at bin1 and bin2. A 0.55arcsec resolution should be sufficient for about any site except if you live at the summit of Haleakala .

Good info, thanks. This makes a lot of sense.