- stack-comparison_8-100-200-frames.png (324.23 KiB) Viewed 2015 times
I settled for 200 luminance frames.
- M27-final.JPG (209.94 KiB) Viewed 2015 times
Dave
Stacking Comparison
Stacking Comparison
M27 captures with Luminance filter - 8, 100 & 200 frame stack side-by-side with 20 darks applied.
I settled for 200 luminance frames.
Dave
I settled for 200 luminance frames.
Dave
Re: Stacking Comparison
Excellent post, Dave.
According to my math the SNR of the right-most image is 5 times the SNR of the left-most image.
Brian
According to my math the SNR of the right-most image is 5 times the SNR of the left-most image.
Brian
Re: Stacking Comparison
I have a request to show the math:
So the relationship is this: to double the SNR you need to quadruple the frame count. So if you start with 8 frames and quadruple it to 32 frames then you have doubled your SNR. If we denote the beginning frame count as 'frameCount0' and the ending frame count as 'frameCount1' then the increase in SNR is given by SQRT(frameCount1 / frameCount0). In Dave's case it is SQRT(200 / 8) = 5.
Brian
So the relationship is this: to double the SNR you need to quadruple the frame count. So if you start with 8 frames and quadruple it to 32 frames then you have doubled your SNR. If we denote the beginning frame count as 'frameCount0' and the ending frame count as 'frameCount1' then the increase in SNR is given by SQRT(frameCount1 / frameCount0). In Dave's case it is SQRT(200 / 8) = 5.
Brian
Re: Stacking Comparison
FYI:
Brian
- M27-final.JPG (123.82 KiB) Viewed 2007 times
-
AussieHowie
- Posts: 25
- Joined: Fri Aug 11, 2017 12:23 am
Re: Stacking Comparison
I just learnt the maths as ... whatever the SNR is in just one frame at your exp time and gain ... if you stack N frames then sqrt of N is the new SNR.
So stack of 8 = sqrt of 8 = 2.8 times the signal to noise ratio of just one frame.
And stack of 200 = sqrt of 200 = 14.1 times the SNR of just one frame.
How many more times is 14.1 better than 2.8 .... 14.1 / 2.8 = 5
Its an often asked question ... In SharpCap (or any stacker) ... how do I know how many frames to stack? The answer I usually say is as many as you are willing to wait for! But if you do the maths you will find it is a diminishing return. ie Longer and longer waits to increase the SNR by another factor of 1.
IE If you were shooting 60 sec like the OP then ...
4 minutes for 4 frames for 2x the SNR in one frame
9 minutes for 9 frames for 3x SNR in one frame
16 mins for 16 frames for 4x SNR
25 mins for 5x SNR
36 mins for 6x
49 for 7x
64 for 8x
So stack of 8 = sqrt of 8 = 2.8 times the signal to noise ratio of just one frame.
And stack of 200 = sqrt of 200 = 14.1 times the SNR of just one frame.
How many more times is 14.1 better than 2.8 .... 14.1 / 2.8 = 5
Its an often asked question ... In SharpCap (or any stacker) ... how do I know how many frames to stack? The answer I usually say is as many as you are willing to wait for! But if you do the maths you will find it is a diminishing return. ie Longer and longer waits to increase the SNR by another factor of 1.
IE If you were shooting 60 sec like the OP then ...
4 minutes for 4 frames for 2x the SNR in one frame
9 minutes for 9 frames for 3x SNR in one frame
16 mins for 16 frames for 4x SNR
25 mins for 5x SNR
36 mins for 6x
49 for 7x
64 for 8x
-
AussieHowie
- Posts: 25
- Joined: Fri Aug 11, 2017 12:23 am
Re: Stacking Comparison
Great image Dave .... !
And interesting info Brian re exoplanet.
Cheers to both posters
And interesting info Brian re exoplanet.
Cheers to both posters
Re: Stacking Comparison
AussieHowie,
Great point about diminishing returns but when you look at AstroBin it is what makes a Top Pick. The judges are swayed by total integration time.
We are all beholden to our f/ratio, Bortle class, and QE. If you want to play the game it seems like you need to spend more money and/or move to the mountains! The longest I've gone is 11.6 hours for M81. It's exhausting but worth the effort. After that I had to take a break from AP to do Photometry in order to get my juices flowing again
Brian
Great point about diminishing returns but when you look at AstroBin it is what makes a Top Pick. The judges are swayed by total integration time.
We are all beholden to our f/ratio, Bortle class, and QE. If you want to play the game it seems like you need to spend more money and/or move to the mountains! The longest I've gone is 11.6 hours for M81. It's exhausting but worth the effort. After that I had to take a break from AP to do Photometry in order to get my juices flowing again
Brian
Re: Stacking Comparison
On the point of stacking and diminishing returns, I did come across this. I had 200 L frames and 30 each of RGB. This resulted in an SNR 'imbalance' across the 4 stacks (figures from Brian, @ooopfan).
200x60s L bin1 = SNR 42.8
30x86s R bin2 = SNR 21.7
30x56s G bin2 = SNR 21.5
30x78s B bin2 = SNR 21.5
The result was I had to do some 'pixel bending' in post processing to get the colours balanced (via histogram), resulting in the image having a 'strained' look. Brian came up with a revised L frame count where the SNR is similar across all 4 stacks.
51x60s L bin1 = SNR 21.6
30x86s R bin2 = SNR 21.7
30x56s G bin2 = SNR 21.5
30x78s B bin2 = SNR 21.5
The result was after the colour combine in Astro Pixel Processor I had a colour balanced histogram. Upon completing APP processing and loading into GIMP, I still had a colour balanced histogram.
The comparison of 200 v 51 L frames. The star colours are somewhat diminished in the JPG comparisons. In the higher quality PNG version, coloured stars (yellow and blue) can clearly be seen. With the 51 L frames, I did achieve star colour without any processing effort indicating the the 4 LRGB stacks were in balance.
Which demonstrates some thought needs to be given to 'how many to stack'. To use 100 L frames, I will need to acquire an additional 30 frames each of RGB to achieve a colour balanced image without resort to pixel mangling. Of course stacking is only part of the story. With my recent deep sky I have been having some success with 60s exposures (M27 a bright object in a rich star field) and 90s (M81, M51, M101 which are dimmer objects). All my recent April/May images have been calibrated with 20 darks, 50 bias, 50 flat frames. The other overlooked factor is the Bortle value of the sky. I have an M27 (60x60s) captured with a 183C at a Bortle 4 and Bortle 6 site - I was stunned by the difference. Both images are in here https://www.astrobin.com/users/turfpit/ . Some stacking comparisons here viewtopic.php?f=18&t=1808 .
I had a look around my file store and came up with some old M27 efforts. Almost 4 years to the day between the left and right images.
Dave
200x60s L bin1 = SNR 42.8
30x86s R bin2 = SNR 21.7
30x56s G bin2 = SNR 21.5
30x78s B bin2 = SNR 21.5
The result was I had to do some 'pixel bending' in post processing to get the colours balanced (via histogram), resulting in the image having a 'strained' look. Brian came up with a revised L frame count where the SNR is similar across all 4 stacks.
51x60s L bin1 = SNR 21.6
30x86s R bin2 = SNR 21.7
30x56s G bin2 = SNR 21.5
30x78s B bin2 = SNR 21.5
The result was after the colour combine in Astro Pixel Processor I had a colour balanced histogram. Upon completing APP processing and loading into GIMP, I still had a colour balanced histogram.
- LRGB-processing-histograms.JPG (37.7 KiB) Viewed 1968 times
The comparison of 200 v 51 L frames. The star colours are somewhat diminished in the JPG comparisons. In the higher quality PNG version, coloured stars (yellow and blue) can clearly be seen. With the 51 L frames, I did achieve star colour without any processing effort indicating the the 4 LRGB stacks were in balance.
- 200v51-luminance-frames-stack.JPG (121.81 KiB) Viewed 1968 times
Which demonstrates some thought needs to be given to 'how many to stack'. To use 100 L frames, I will need to acquire an additional 30 frames each of RGB to achieve a colour balanced image without resort to pixel mangling. Of course stacking is only part of the story. With my recent deep sky I have been having some success with 60s exposures (M27 a bright object in a rich star field) and 90s (M81, M51, M101 which are dimmer objects). All my recent April/May images have been calibrated with 20 darks, 50 bias, 50 flat frames. The other overlooked factor is the Bortle value of the sky. I have an M27 (60x60s) captured with a 183C at a Bortle 4 and Bortle 6 site - I was stunned by the difference. Both images are in here https://www.astrobin.com/users/turfpit/ . Some stacking comparisons here viewtopic.php?f=18&t=1808 .
I had a look around my file store and came up with some old M27 efforts. Almost 4 years to the day between the left and right images.
- M27-progress.JPG (63.37 KiB) Viewed 1968 times
Dave