Jump to content


Photo

How much difference do more subs make?

  • Please log in to reply
71 replies to this topic

#1 mpgxsvcd

mpgxsvcd

    Surveyor 1

  • -----
  • Posts: 1951
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 21 December 2013 - 08:03 PM

So I wanted to see where the point of diminishing returns comes into play. I had stacked 79 6 second ISO 6400 images of M42 before.

I decided to see what 60, 28, and 16 looked like instead. I stacked the 60, 28, and 16 in the exact same manner and saved the output of DSS without any edits.

Here are what they looked like. I also have the original 79 image file. However, the processing on that was slightly different.

I think that 30-60 images is sufficient. 16 images definitely shows some difference.

60 Images

Attached Files



#2 mpgxsvcd

mpgxsvcd

    Surveyor 1

  • -----
  • Posts: 1951
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 21 December 2013 - 08:03 PM

28 Images

Attached Files



#3 mpgxsvcd

mpgxsvcd

    Surveyor 1

  • -----
  • Posts: 1951
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 21 December 2013 - 08:03 PM

16 Images

Attached Files



#4 mpgxsvcd

mpgxsvcd

    Surveyor 1

  • -----
  • Posts: 1951
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 21 December 2013 - 08:04 PM

Here is the 79 images but with slightly different processing.

Attached Files



#5 Tom and Beth

Tom and Beth

    Soyuz

  • -----
  • Posts: 3786
  • Joined: 08 Jan 2007
  • Loc: Tucson, AZ

Posted 21 December 2013 - 09:58 PM

Interesting. undoubtedly a significant jump between 16 and 28, and at least to me another significant jump going to 60 looking only at these down-sampled examples to fit CN guidelines.

In particular the "V" shape just off center at the 11 o'clock position

#6 jbalsam

jbalsam

    Viking 1

  • *****
  • Posts: 519
  • Joined: 06 Jul 2012
  • Loc: Spencerville, MD

Posted 21 December 2013 - 10:22 PM

The 60-frame image certainly looks the best. I don't think you've hit the point of diminishing returns yet because you can definitely see a difference between 28-60. Try 120 and 250 frames. 6s is a short exposure time so they're easy to capture.

In other news, what are you using for a coma corrector?

#7 Samir Kharusi

Samir Kharusi

    Vanguard

  • *****
  • Posts: 2016
  • Joined: 14 Jun 2005
  • Loc: Oman

Posted 21 December 2013 - 11:34 PM

I doubt that there is EVER a point of diminishing returns. The longer the integration time, the fainter/dimmer the wisps you can dig out at a good Signal-to-Noise-Ratio. But one should not think of x frames more, think of doubling the integration time for each step. After 79 think of 160, then 320, etc. Here's what I mean, small crops displayed at almost 1:1 from APS-C stacks, identical processing for all, 2-minute subs at f4 at a dark site:
Posted Image

#8 Tonk

Tonk

    Fly Me to the Moon

  • *****
  • Posts: 7141
  • Joined: 19 Aug 2004
  • Loc: Leeds, UK, 54N

Posted 22 December 2013 - 05:26 AM

Yes - I don't count number fo subframes but the total integration times.

Put it this way, the Gendler's, Croman's and Hallas's of this world do 20 to 30 hours integration. A few DSLR folks here have run up to 20 hours with stunning results.

Personally I find things look very good after 10 hours worth.

#9 garret

garret

    Ranger 4

  • -----
  • Posts: 332
  • Joined: 26 Jul 2009
  • Loc: Netherlands

Posted 22 December 2013 - 07:41 AM

Multi hour integration time? not needed for a brite object:

Here is a crop of my imaghe of M17, only 14 subs of 60 seconds, canon 5 dmk2, F 3.6 telescope, m17 was only 19 degrees above the horizon.
In the image the right part is noise free more subs is not needed, left could have more subs.

Garret van der Veen

Attached Files



#10 Starhawk

Starhawk

    Space Ranger

  • *****
  • Posts: 5614
  • Joined: 16 Sep 2008
  • Loc: Tucson, Arizona

Posted 22 December 2013 - 09:03 AM

Additional frames improve signal to noise ratio proportionally to the square root of the number of subs. So, if you take 4, you get twice the signal to noise ratio of one. Remember, this is statistical improvement. To get 4 times the improvement, you need 16 subs.

Adding up the integration time grossly overestimates the improvement. With 100 subs, you would predict the image was 5 magnitudes deeper than one sub. In reality, the difference is closer to 2.5 or less. The only time this is valid is when it is all in the same exposure. So, twice the exposure length gives twice the signal to read noise.

I've ended up doing 16 sub frames and putting as much integration time into them as possible.

-Rich

#11 Starhawk

Starhawk

    Space Ranger

  • *****
  • Posts: 5614
  • Joined: 16 Sep 2008
  • Loc: Tucson, Arizona

Posted 22 December 2013 - 09:05 AM

Note, there are other sources of noise besides read noise, so the real improvement from stacking isn't quite as much as the square root of the number of subs.

-Rich

#12 jbalsam

jbalsam

    Viking 1

  • *****
  • Posts: 519
  • Joined: 06 Jul 2012
  • Loc: Spencerville, MD

Posted 22 December 2013 - 09:44 AM

There's always a point of diminishing returns. In this case, it's when you don't notice a difference in the image quality by adding a significant amount more data. It probably takes a large amount of data to accomplish this for a low SNR object (20-30 hours as someone mentioned) where the shot noise of the object is significant.

#13 Cotts

Cotts

    Just Wondering

  • *****
  • Posts: 4946
  • Joined: 10 Oct 2005
  • Loc: Toronto, Ontario

Posted 22 December 2013 - 10:52 AM

I am very interested in this 'minimalism' discussion. I will be travelling to Chile in March for 11 nights at Atacama Lodge. I plan to devote about 5 of the nights to astrophotography through an AT65EDQ/Canon60Da combination. (The other nights will be visual observing with the 24" dob while shooting wide-angle stills and time-lapses..)

There are waaaay more targets than I will have time for so I will have to compromise on S/N ratio by shooting some bare minimum of subs, flats, darks etc. ISO 1600 works well with my camera...

The big question is: How many subs is the minimum acceptable number? And how long should they be? Which would be better, e.g., 2x6min, 3x4min, 4x3min, 6x2min or 12x1min?

I have a 'wish list' of about 25 targets with the AT65 so, with my planned schedule, I'll do about 5 targets a night -- about 1.5 hours maximum per target. Whatever shall I do?

Thanks

Dave

#14 Hilmi

Hilmi

    Soyuz

  • *****
  • Posts: 3814
  • Joined: 07 Mar 2010
  • Loc: Muscat, Sultanate of Oman

Posted 22 December 2013 - 01:26 PM

Dave,

You are sometimes limited by saturation in the stars. If stars get saturated they look bad post processing, so my personal advice is to take an image at the maximum exposure at which the stars don't saturate and work on from there.

Hilmi

#15 Footbag

Footbag

    Fly Me to the Moon

  • *****
  • Posts: 6275
  • Joined: 13 Apr 2009
  • Loc: Scranton, PA

Posted 22 December 2013 - 01:31 PM

I am very interested in this 'minimalism' discussion. I will be travelling to Chile in March for 11 nights at Atacama Lodge. I plan to devote about 5 of the nights to astrophotography through an AT65EDQ/Canon60Da combination. (The other nights will be visual observing with the 24" dob while shooting wide-angle stills and time-lapses..)

There are waaaay more targets than I will have time for so I will have to compromise on S/N ratio by shooting some bare minimum of subs, flats, darks etc. ISO 1600 works well with my camera...

The big question is: How many subs is the minimum acceptable number? And how long should they be? Which would be better, e.g., 2x6min, 3x4min, 4x3min, 6x2min or 12x1min?

I have a 'wish list' of about 25 targets with the AT65 so, with my planned schedule, I'll do about 5 targets a night -- about 1.5 hours maximum per target. Whatever shall I do?

Thanks

Dave


Just expose until your histogram is separated from the left by 1/3. How many? That's up to you. How noise free do you want your images to be?

#16 Tonk

Tonk

    Fly Me to the Moon

  • *****
  • Posts: 7141
  • Joined: 19 Aug 2004
  • Loc: Leeds, UK, 54N

Posted 22 December 2013 - 02:03 PM

+10 ^

#17 nofxrx

nofxrx

    Vendor (HyperCams & Mods)

  • *****
  • Vendors
  • Posts: 5241
  • Joined: 12 Jul 2005
  • Loc: Palm Bay,Florida

Posted 22 December 2013 - 03:28 PM

But do you use the BOC histo to determine the 1/3 rule, or use an unstretched histo?

Also,that 1/3 rule is really dependant on the model used... For instance my 5D2 could EASILY take that to 1/2(center) or even slightly more to the right (like 9/16th to 2/3 from left)
But my old XS could only do 1/3 to slightly closer to center. ?.
The T2i was closer to 1/2 as well without issues..

And this was using the BOC histo. (Which is stretched non linearly (IIRC))

If you look at the Raw images on a PC they are MUCH darker...
So you would have to expose for nearly 2/3++ to get those to look similar to the stretched histo. Just not sure if there is any benefit our negatives in doing so(Though logically would think that the images would be completely washed out with LP... which would make removal near impossible... Again just a guess..

#18 Cotts

Cotts

    Just Wondering

  • *****
  • Posts: 4946
  • Joined: 10 Oct 2005
  • Loc: Toronto, Ontario

Posted 22 December 2013 - 03:41 PM

Adam, you said, "Just expose until your histogram is separated from the left by 1/3. "

Can I observe my histogram 'on the fly', as I shoot or must I shoot a 'guess frame' and then look at the histogram then adjust exposure?

Thanks,

Dave

#19 petemumbower

petemumbower

    Surveyor 1

  • -----
  • Posts: 1701
  • Joined: 12 Jul 2009
  • Loc: West Michigan

Posted 22 December 2013 - 03:41 PM

Agree with above too, we all want the awesome images we see posted, but rarely do many of us put in the integration times needed to achieve those results. But we all got to decide if we want a handful of great images or lots of ok/good ones.

#20 Footbag

Footbag

    Fly Me to the Moon

  • *****
  • Posts: 6275
  • Joined: 13 Apr 2009
  • Loc: Scranton, PA

Posted 22 December 2013 - 03:46 PM

Adam, you said, "Just expose until your histogram is separated from the left by 1/3. "

Can I observe my histogram 'on the fly', as I shoot or must I shoot a 'guess frame' and then look at the histogram then adjust exposure?

Thanks,

Dave


You will need to take a guess frame. The exposure length will depend on sky brightness, F-Ratio, iso, etc...

At Atacama, I suspect you will be using longish exposures, unless you are imaging with very fast optics.

#21 Wmacky

Wmacky

    Surveyor 1

  • *****
  • Posts: 1933
  • Joined: 24 Nov 2007
  • Loc: Florida

Posted 22 December 2013 - 09:22 PM

I hate to ask, but a google search comes up with nothing.

Whats BOC?

#22 pfile

pfile

    Gemini

  • -----
  • Posts: 3169
  • Joined: 14 Jun 2009

Posted 22 December 2013 - 09:29 PM

the 1/3 rule is for back of camera ("BOC") histogram. or the histogram that BYE shows you. those histograms represent a default stretch (the same stretch that the camera would apply if you let it create a jpeg for you)

the 1/3 rule is sometimes also given as "20%" or "well-detached".

note that if you are at a very dark sky site it may be impossible to push the histogram that far to the right without really long sub exposures.

rob

edit: dave you will have to take test shots; the histogram is not displayed until the exposure ends.


#23 Wmacky

Wmacky

    Surveyor 1

  • *****
  • Posts: 1933
  • Joined: 24 Nov 2007
  • Loc: Florida

Posted 22 December 2013 - 09:36 PM

the 1/3 rule is for back of camera ("BOC") histogram. or the histogram that BYE shows you. those histograms represent a default stretch (the same stretch that the camera would apply if you let it create a jpeg for you)

the 1/3 rule is sometimes also given as "20%" or "well-detached".

note that if you are at a very dark sky site it may be impossible to push the histogram that far to the right without really long sub exposures.

rob

edit: dave you will have to take test shots; the histogram is not displayed until the exposure ends.


Thanks! I have one other quick question. When taking a exposure test shot. Is there a quick way to tell if you have over saturated the brighter stars, other than a visual check? The histo turn into a hard to see thin line as it approaches the right hand side.

#24 Jerry Lodriguss

Jerry Lodriguss

    Vendor

  • *****
  • Vendors
  • Posts: 3703
  • Joined: 19 Jul 2008
  • Loc: Voorhees, NJ

Posted 23 December 2013 - 09:31 AM

I am very interested in this 'minimalism' discussion. I will be travelling to Chile in March for 11 nights at Atacama Lodge. I plan to devote about 5 of the nights to astrophotography through an AT65EDQ/Canon60Da combination. (The other nights will be visual observing with the 24" dob while shooting wide-angle stills and time-lapses..)

There are waaaay more targets than I will have time for so I will have to compromise on S/N ratio by shooting some bare minimum of subs, flats, darks etc. ISO 1600 works well with my camera...

The big question is: How many subs is the minimum acceptable number? And how long should they be? Which would be better, e.g., 2x6min, 3x4min, 4x3min, 6x2min or 12x1min?

I have a 'wish list' of about 25 targets with the AT65 so, with my planned schedule, I'll do about 5 targets a night -- about 1.5 hours maximum per target. Whatever shall I do?


Hi Dave,

The brighter objects, like the Eta Carina Nebula, won't need as much exposure as the fainter objects.

Prioritize your list by which objects you really want the most, and then research how faint they are. Give more exposure to the faint objects.

Under true dark skies at altitude, you should get excellent results with shorter total integration times than at a typical observing location in North America.

You will typically need longer exposures for fainter extended objects, and shorter exposures for clusters.

Jerry

#25 mpgxsvcd

mpgxsvcd

    Surveyor 1

  • -----
  • Posts: 1951
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 23 December 2013 - 09:48 AM

I doubt that there is EVER a point of diminishing returns. The longer the integration time, the fainter/dimmer the wisps you can dig out at a good Signal-to-Noise-Ratio. But one should not think of x frames more, think of doubling the integration time for each step.


The equation for Stacking says otherwise. It varies with the Square root of the number of subs. It has nothing to do with total time at all.

Direct quote from the DSS site.

http://deepskystacke...lish/theory.htm

"The more, the better but above some threshold it is less efficient.
The signal to noise ration in increasing with the square root of the number of combined frames regardless of the exposure time of each frame.
This is true with all the combining methods (average, median, kappa-sigma clipping, auto-adaptive weighted average, ...) except entropy weighted average since this one in using the entropy to weight each pixel and thus is increasing the noise that is a big entropy contributor.."








Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics