Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Many short subs vs fewer long subs

  • Please log in to reply
20 replies to this topic

#1 joelin

joelin

    Vanguard

  • *****
  • topic starter
  • Posts: 2063
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 08 November 2019 - 10:27 PM

I found this video comparing the two interesting

https://youtu.be/EMdEhQD2WxY

TLDR: a single long exposure beats out any subdivisions

However I’m not sure why. The RASA he uses is f/2 so the case where he did 10 second exposures should be plenty long to swamp read noise. After all, 10 seconds at f/2 is the same as 40 seconds at f/4 and 160 seconds at f/8.
  • rekokich likes this

#2 fmeschia

fmeschia

    Apollo

  • *****
  • Posts: 1074
  • Joined: 20 May 2016
  • Loc: Mountain View, CA

Posted 09 November 2019 - 01:21 AM

Looks like the short exposures have a ton of fixed pattern noise, compared to the long one. Wonder whether that’s why...



#3 joelin

joelin

    Vanguard

  • *****
  • topic starter
  • Posts: 2063
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 09 November 2019 - 01:29 AM

so it seems like the comparison was not a good one because he did not dither

 

if he dithered all subs, then should the 10x30 equal the 60x5 and the 300x1?

 

another thought i had....if the helix was low in the sky for the 10x30, that would also explain it...however, he lives in southern hemisphere so the low declination of the helix would work in his favor as it would be nearly overhead so im guessing thats not the problem


Edited by joelin, 09 November 2019 - 01:31 AM.


#4 fmeschia

fmeschia

    Apollo

  • *****
  • Posts: 1074
  • Joined: 20 May 2016
  • Loc: Mountain View, CA

Posted 09 November 2019 - 01:38 AM

Even at f/2, 10 seconds seems short to swamp read noise for an ASI1600 if the sky was any good (>20 mag/arcsec2).



#5 joelin

joelin

    Vanguard

  • *****
  • topic starter
  • Posts: 2063
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 09 November 2019 - 01:59 AM

this chart seems to say https://www.cloudyni...d-maybe-qhy163/

 

just 30 seconds for a very dark sky 21.5 mag, 139 gain and f/4

 

f/2 would be 1/4th of that ...so 7.5 seconds 



#6 sharkmelley

sharkmelley

    Vanguard

  • *****
  • Posts: 2432
  • Joined: 19 Feb 2013

Posted 09 November 2019 - 02:20 AM

One thing he said was that you lose brightness and luminosity with the short (30sec on a RASA) exposures and he showed this difference in his side by side comparison.   I'm sorry to say that is nonsense.  He would have collected the same number of photons from the short exposures and the long exposures because the total integration time is identical.  The difference in brightness and luminosity he is seeing is simply caused by lack of understanding of how to process the data.

 

There is a whole lot more that he says in that video which is just totally wrong (again demonstrating a lack of understanding) but I've got better things to do with my time than to review the mistakes in YouTube channels so I'll finish there.

 

Mark


Edited by sharkmelley, 09 November 2019 - 02:24 AM.

  • psandelle, Jon Rista, PartlyCloudy and 1 other like this

#7 gezak22

gezak22

    Gemini

  • *****
  • Posts: 3074
  • Joined: 15 Aug 2004
  • Loc: On far side of moon. Send help.

Posted 09 November 2019 - 03:18 PM

If the exposure is too short, read noise will dominate. There are tons of calculators online that will help you determine the minimum sub exposure time in order to reach a certain noise level. Use them!
  • psandelle likes this

#8 joelin

joelin

    Vanguard

  • *****
  • topic starter
  • Posts: 2063
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 09 November 2019 - 09:23 PM

Yes and after using the calculator, it seems that even 7.5 seconds is sufficient in a dark sky at f/2 and unity gain.
  • sharkmelley likes this

#9 gezak22

gezak22

    Gemini

  • *****
  • Posts: 3074
  • Joined: 15 Aug 2004
  • Loc: On far side of moon. Send help.

Posted 10 November 2019 - 12:46 PM

Yes and after using the calculator, it seems that even 7.5 seconds is sufficient in a dark sky at f/2 and unity gain.

Yes, and exposing for longer will reduce the contribution of read noise even further (if your skies allow) hence result in better images as long as you are still stacking 25+ frames.



#10 joelin

joelin

    Vanguard

  • *****
  • topic starter
  • Posts: 2063
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 10 November 2019 - 12:57 PM

yes but in the sub exposure table i linked, it says "The value from the table is the shortest sub length that you can use (for that scope, sky and gain) and still be shot noise limited (5%RN criterion)"

 

 

5% seems awfully small for it to have any meaningful contribution to overall noise.. and thats at the 7.5 seconds super dark sky exposure time i specified...

 

but in the youtube video the impact on noise seems much more than 5%



#11 gezak22

gezak22

    Gemini

  • *****
  • Posts: 3074
  • Joined: 15 Aug 2004
  • Loc: On far side of moon. Send help.

Posted 10 November 2019 - 04:10 PM

yes but in the sub exposure table i linked, it says "The value from the table is the shortest sub length that you can use (for that scope, sky and gain) and still be shot noise limited (5%RN criterion)"

 

 

5% seems awfully small for it to have any meaningful contribution to overall noise.. and thats at the 7.5 seconds super dark sky exposure time i specified...

 

but in the youtube video the impact on noise seems much more than 5%

1. Perhaps the noise you see in the youtube video is not read noise.

2. You really cannot estimate noise from a stretched image.

 

All I am saying is that longer exposures will always help with SNR (unless you are saturating your detector), and the nebula in the youtube video is nowhere near being saturated.


  • Galaxyhunter and psandelle like this

#12 Galaxyhunter

Galaxyhunter

    Mercury-Atlas

  • *****
  • Posts: 2599
  • Joined: 02 Jan 2006
  • Loc: Northern Illinois

Posted 10 November 2019 - 05:40 PM

1. Perhaps the noise you see in the youtube video is not read noise.

2. You really cannot estimate noise from a stretched image.

 

All I am saying is that longer exposures will always help with SNR (unless you are saturating your detector), and the nebula in the youtube video is nowhere near being saturated.

Absolutely true.

For stellar objects like Clusters, it is not quite as critical.  For Nebulas,  You need exposure time if you want to capture the faint wispies of a Nebula.



#13 joelin

joelin

    Vanguard

  • *****
  • topic starter
  • Posts: 2063
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 10 November 2019 - 08:59 PM

That goes against everything else I’ve read which is that if you’ve exposed sufficently enough to overcome read noise then many short exposures is the same as a longer one...as long as total integration time is the same

#14 gezak22

gezak22

    Gemini

  • *****
  • Posts: 3074
  • Joined: 15 Aug 2004
  • Loc: On far side of moon. Send help.

Posted 10 November 2019 - 09:24 PM

That goes against everything else I’ve read which is that if you’ve exposed sufficently enough to overcome read noise then many short exposures is the same as a longer one...as long as total integration time is the same

Well, this article suggests that N exposures with a subexposure time T can be traded off for 2N exposures with T/2, but that does not mean that they will give equal SNR. There is SNR in the exposure itself (swamping read noise), but another associated with data rejection from many frames (more = better). I admit that I do not know what the optimum is when the total exposure time is constant.

 

Another point is that "exposed sufficiently enough to overcome read noise" is an ambiguous statement. If read noise is X, then imaging the North America nebula will require T minutes to overcome X, while it may require 3T to reach enough signal from a much fainter nebula. If you are imaging Messiers and NGCs, it may not be obvious what wonders long exposures do. But in the community of people who chase nebula discoveries, this "trick" (along with binning, just for discovery purposes) has been known for a long time as demonstrated by the discovery of the squid nebula and by the numerous Hubble deep fields.


  • psandelle likes this

#15 Galaxyhunter

Galaxyhunter

    Mercury-Atlas

  • *****
  • Posts: 2599
  • Joined: 02 Jan 2006
  • Loc: Northern Illinois

Posted 10 November 2019 - 10:13 PM

That goes against everything else I’ve read which is that if you’ve exposed sufficently enough to overcome read noise then many short exposures is the same as a longer one...as long as total integration time is the same

Well then do your own research.  Image the Horsehead.  Sample 1, take 3600 1 sec. exposures.   Sample 2 take 15 -  4 minutes exposures or 30 - 2 minute exposures.  The total integration time will be the same.  You will be amazed the difference in images quality. Then you will know.


  • Swordfishy likes this

#16 joelin

joelin

    Vanguard

  • *****
  • topic starter
  • Posts: 2063
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 10 November 2019 - 10:39 PM

Well then do your own research.  Image the Horsehead.  Sample 1, take 3600 1 sec. exposures.   Sample 2 take 15 -  4 minutes exposures or 30 - 2 minute exposures.  The total integration time will be the same.  You will be amazed the difference in images quality. Then you will know.

well its unlikely that 1 second exposures would overwhelm read noise...but i think the people who've read the sub exposure chart i linked for ASI1600 seem to all agree in the accuracy of the chart 



#17 joelin

joelin

    Vanguard

  • *****
  • topic starter
  • Posts: 2063
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 10 November 2019 - 10:41 PM

Well, this article suggests that N exposures with a subexposure time T can be traded off for 2N exposures with T/2, but that does not mean that they will give equal SNR. There is SNR in the exposure itself (swamping read noise), but another associated with data rejection from many frames (more = better). I admit that I do not know what the optimum is when the total exposure time is constant.

 

Another point is that "exposed sufficiently enough to overcome read noise" is an ambiguous statement. If read noise is X, then imaging the North America nebula will require T minutes to overcome X, while it may require 3T to reach enough signal from a much fainter nebula. If you are imaging Messiers and NGCs, it may not be obvious what wonders long exposures do. But in the community of people who chase nebula discoveries, this "trick" (along with binning, just for discovery purposes) has been known for a long time as demonstrated by the discovery of the squid nebula and by the numerous Hubble deep fields.

the sky background signal is whats constant between a bright object and a faint object... using the sub exposure table i linked to...the read noise is just 5% of the total noise... and it doenst matter if its north america nebula or NGC or whatever...the exposure time varies by the sky background



#18 gezak22

gezak22

    Gemini

  • *****
  • Posts: 3074
  • Joined: 15 Aug 2004
  • Loc: On far side of moon. Send help.

Posted 10 November 2019 - 11:15 PM

well its unlikely that 1 second exposures would overwhelm read noise...but i think the people who've read the sub exposure chart i linked for ASI1600 seem to all agree in the accuracy of the chart 

Carl's point was not that 1 s exposures will overwhelm read noise. His point was to do your own comparisons (fixed total exposure time, varying subexposure length), so you know what works for you - where the point of diminishing return is. You seem to hold the table in very high regard, but it seems to be in conflict with the youtube video. I would trust data (the youtube video in this case) over a chart that has just numbers. But it's also perfectly valid to question the way the data in the youtube video were collected, hence the recommendation to collect your own data to convince yourself what the truth is. There are only a few things more rewarding than proving/disproving a hypothesis with data.

 

 

the sky background signal is whats constant between a bright object and a faint object... using the sub exposure table i linked to...the read noise is just 5% of the total noise... and it doenst matter if its north america nebula or NGC or whatever...the exposure time varies by the sky background

Correct. But longer exposures will take your noise level to 1% (for example) thus revealing the fainter parts of the nebulae. Keep in mind that the 5% value is something taken out of thin air, and once you add nonlinear histogram modifications,a 5% noise source can be huge.


  • Galaxyhunter and psandelle like this

#19 sharkmelley

sharkmelley

    Vanguard

  • *****
  • Posts: 2432
  • Joined: 19 Feb 2013

Posted 11 November 2019 - 01:24 AM

But longer exposures will take your noise level to 1% (for example) thus revealing the fainter parts of the nebulae. Keep in mind that the 5% value is something taken out of thin air, and once you add nonlinear histogram modifications,a 5% noise source can be huge.

A 5% read noise source will still be swamped by the sky background noise and will never be noticeable.  But you mention non-linear stretching and that might be the key to what went wrong in that YouTube video.

 

The YouTuber applied different processing to the stack of shorter exposures and that led him to draw the incorrect conclusion that the shorter exposures lost brightness and luminosity.  It might also have led to the incorrect conclusion that the stack of shorter exposures was noisier.

 

Maybe he was also operating his camera at gain 0 where the read noise is highest.  There are just too many unknowns in that video.

 

Mark


  • Jon Rista likes this

#20 Galaxyhunter

Galaxyhunter

    Mercury-Atlas

  • *****
  • Posts: 2599
  • Joined: 02 Jan 2006
  • Loc: Northern Illinois

Posted 11 November 2019 - 07:49 AM

Galaxyhunter, on 10 Nov 2019 - 9:13 PM, said:

    Well then do your own research.  Image the Horsehead.  Sample 1, take 3600 1 sec. exposures.   Sample 2 take 15 -  4 minutes exposures or 30 - 2 minute exposures.  The total integration time will be the same.  You will be amazed the difference in images quality. Then you will know.

 

well its unlikely that 1 second exposures would overwhelm read noise...but i think the people who've read the sub exposure chart i linked for ASI1600 seem to all agree in the accuracy of the chart 

I guess the difference is that I'm using a CCD vs CMOS. I have been at this ( off & on ) for quite a few years.  I have played the short exposure game.  It wasn't till I joined CN & was looking at the images posted on the CCD imaging forum & looking at the image specs, there is when I learned that longer exposure time was the ticket to better pics.  If you want to live & die by a chart, have at it.

 

A question to you,  Do you think that Hubble takes 10-15 sec images?.  I mean there is no sky glow to worry about.



#21 gezak22

gezak22

    Gemini

  • *****
  • Posts: 3074
  • Joined: 15 Aug 2004
  • Loc: On far side of moon. Send help.

Posted 11 November 2019 - 01:43 PM

I highly recommend an experiment on a target with a wide dynamic range. The North America region, not framed with the Pelican, but instead framed with Sh2-119 will show a clear improvement in SNR on Sh2-119 with longer exposures, while the improvement in SNR will not be as apparent on the North America nebula. I am sure a similar target exists for winter.

 

Edit: NGC 3628 with its tidal tail is also a good candidate for the spring.


Edited by gezak22, 11 November 2019 - 02:02 PM.



CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics