Jump to content


Photo

A Comparison of Exposure Lengths... XMas Tree

  • Please log in to reply
63 replies to this topic

#26 freestar8n

freestar8n

    Soyuz

  • *****
  • Posts: 3907
  • Joined: 12 Oct 2007

Posted 09 February 2013 - 04:33 AM

All I know is that the 9X20min was considerably easier to process than the 93X5min picture. I'll stick with 20min on objects like this.



Yes - if you can and there is no downside - you probably should. But I see you use HDR on objects like the Orion nebula. So - I assume you realize that "longer subs result in much better images" is not something you apply universally even to your own setup - let alone others with a range of f/ratios, read noise, and sky quality.

The point of the other thread is to provide quantitative guidelines that apply to anyone's setup, and the model used is the same one used by professionals. Both professionals and amateurs appreciate the benefit of stacking many subs, and realize that although the improvement from each added sub is less noticeable - it continues to improve the SNR.

Note that the other thread had a comparison of a narrow band shot using two different subexposures, and the end result showed little difference. Both that result and your result are consistent with predictions - assuming his exposure is sky limited and yours is read noise limited.

If you want to provide more insight into how your result is or is not consistent with the noise model - please provide data on your gain/read noise/sky background. You refer above to providing "raw" data to help people - but two independently processed images are not "raw" - and a general conclusion derived from them could in fact hurt people if they feel they need long sub exposures when they are sky limited.

Frank

#27 vpcirc

vpcirc

    Soyuz

  • -----
  • Posts: 3965
  • Joined: 09 Dec 2009
  • Loc: Merced CA

Posted 09 February 2013 - 07:42 AM

It's funny how so many of the arm chair experts never post an image themselves and yet criticize someone who is trying to help others with actual results. Thanks for your work and test Jimmy. I'm sure many have benefited from seeing the difference it makes without getting into a quickly degenerating conversation.

#28 freestar8n

freestar8n

    Soyuz

  • *****
  • Posts: 3907
  • Joined: 12 Oct 2007

Posted 09 February 2013 - 08:42 AM

I posted an image in the other thread and described how it fit with the noise model. I also cited an image in the other thread that complements this one - and is somehow being ignored since it does not support the benefit of longer images in all situations. I also cited images by the OP of this thread that show the benefit of short exposures in a bright object (Orion nebula HDR). Other images were posted in the other thread, including one by the OP of that thread, to support his conclusion.

What I have noticed in these threads is that the people who disagree with the noise model also confess that they don't understand the math behind it. It's as if they don't want to be bothered by it, or are suspicious of science itself - yet they are confident that the model is not suited for realistic imaging situations. Again, despite its being used by professional astronomers in their work.

So - is there anyone who *does* feel comfortable with the model, and *does* understand how it works - who also disagrees with it? The math is pretty simple - only involving a square root.

I think there are people involved in imaging who would be perfectly comfortable with the math and they would be "reachable" by the technical aspects of this discussion. I hope that for them the conclusions are clear and they would recognize good advice from bad in these discussions.

Frank

#29 orion69

orion69

    Viking 1

  • -----
  • Posts: 989
  • Joined: 09 May 2010
  • Loc: Croatia

Posted 09 February 2013 - 09:23 AM

Frank (and others of course), I have question that has been already addressed in this and other posts but to make things more clear (at least for me): suppose I shoot somewhat dim object and I'm using NB filters. After 10x30 min subs I've got my new stacked picture. Now I'm shooting same object, with same gear, same sky quality (completely same conditions as before) only now I'm shooting 5 min subs x times.

Question: is there some detail on first image that would not be visible on second image no matter how many 5 min subs I take?

#30 neptun2

neptun2

    Viking 1

  • -----
  • Posts: 805
  • Joined: 04 Mar 2007
  • Loc: Bulgaria

Posted 09 February 2013 - 09:28 AM

Well at least i understand it following way. It is reasonable to go as long as possible if you consider the following:

1. Not to overexpose to a point where you start to loose detail.

2. Not to make the background too bright due to a light pollution.

3. Not to start experience mount induced guiding problems.

#31 freestar8n

freestar8n

    Soyuz

  • *****
  • Posts: 3907
  • Joined: 12 Oct 2007

Posted 09 February 2013 - 09:41 AM

Hi-

Thanks for a specific question that is also in the context of this thread and how realistic imaging situations relate to the noise model.

The key issue is how much sky noise is in the sub vs. read noise, as others have mentioned above concerning the OP's images. Since you don't fully specify the f/ratio, it could be that the 5m exposure - even though it is narrow band - will have sky noise that dominates read noise - in which case there will be much less win with 30m subs.

But if we take the more interesting case, which I think you are asking about, in which the 30m exposures are way sky noise limited, while the 5m exposures are read noise limited - then there will be a large reduction in noise and a huge benefit.

But your question has an additional layer to it - which, if I am reading it right, is asking - if some faint detail is not visible in the sub at all - is it possible for it to appear after 8, 16, 200 stacked subs? The answer from the noise model, and from professional astronomers is - absolutely yes. The accumulation is a statistical process and even if you don't "see" it in the sub - even faint signal is present in that noise, and the accumulation of many subs will let the SNR of that signal go as the square root of the number of exposures. That's why the Hubble Ultra Deep Field involved stacking 288 frames for some wavelengths. They used the same model we are discussing here.

So - the idea that stacking "runs out of gas" at 16 is not supported by the model and is not supported by professional work - nor is it supported by many of the nice images in CN that show many subs over many nights stacked into a high snr result.

Frank

#32 Inverted

Inverted

    Mariner 2

  • *****
  • Posts: 210
  • Joined: 19 Jan 2013
  • Loc: LP Land

Posted 09 February 2013 - 09:43 AM

Frank (and others of course), I have question that has been already addressed in this and other posts but to make things more clear (at least for me): suppose I shoot somewhat dim object and I'm using NB filters. After 10x30 min subs I've got my new stacked picture. Now I'm shooting same object, with same gear, same sky quality (completely same conditions as before) only now I'm shooting 5 min subs x times.

Question: is there some detail on first image that would not be visible on second image no matter how many 5 min subs I take?


That's an interesting question, i keep thinking about. The models seem to assume that photons arrive instantaneously, in large quantities. So, no. However, in practice, while photon emissions rates are "astronomically" huge, what other factors influence how many hit your sensor, from the faintest regions of a nebula, I have no idea. I seem to recal that when I did this stuff more often, I could usually see the detail, in shorter subs was there,but the problem was I just couldn't stretch the image sufficiently without flooding it with noise, to get it into the final image. From my recent trials, that seems to be the case too. If there are photons that just aren't hitting our sensors over a five minute period, then we are getting to some remarkably faint detail. Clearly, whatever the reason may be, some objects do just do better with longer subs, at least if you really want to bump the faintest regions up above the noise floor.

#33 vpcirc

vpcirc

    Soyuz

  • -----
  • Posts: 3965
  • Joined: 09 Dec 2009
  • Loc: Merced CA

Posted 09 February 2013 - 09:48 AM

Frank (and others of course), I have question that has been already addressed in this and other posts but to make things more clear (at least for me): suppose I shoot somewhat dim object and I'm using NB filters. After 10x30 min subs I've got my new stacked picture. Now I'm shooting same object, with same gear, same sky quality (completely same conditions as before) only now I'm shooting 5 min subs x times.

Question: is there some detail on first image that would not be visible on second image no matter how many 5 min subs I take?


Your shorter subs will show far less detail resolution and contrast no matter how many you take. That is the point of Jimmy's experiment. Even with far less total imaging time the amount of faint data he was able to obtain is clearly visible. Secondly, with narrowband, you are never going to reach the sky limit from what Don Goldman has told me. There are bright objects that will get blown out by longer subs such as M 42, but if you stick to short subs, you'll miss this wonderful detail that exists elsewhere. Adam Block has a great tutorial on how to combine short and long subs in CCDStack using missing value to combine. Tony Hallas demonstrates another great procedure in his newest addition to Photoshop processing in Vol 5 using the reveal all layer mask in CS5 and CS6. Jimmy used the Hallas method in his M42.

#34 JWalk

JWalk

    Apollo

  • -----
  • Posts: 1360
  • Joined: 06 Jun 2010
  • Loc: San Antonio, TX

Posted 09 February 2013 - 09:49 AM

The skies I get to shoot in are very good and VERY dark. They are usually 1.5 seeing or better. In no way did I give the 20min subs more attention than the 5min subs. I gave the 5min subs more time and effort because it needed extra processing to get it to the level that I wanted it at. The 20min stack just exploded with signal after 3 levels stretches.

My thoughts are go as long as your sky allows. You want to grab the dimmest stuff you can. I applaud the guys who shoot from the backyard and have to deal with light pollution and bad seeing. I enjoy seeing great work from not so great skies. As for narrowband, I shoot HA with my canon at home and an FSQ and with the moon out at about 50% I can usually shoot away from it and go 15 min subs. That is about it before the DSLR gets blown out

Orion is shot in HDR style. I did exposures for the core to bring that out. That is a no brainer that we all know. 10min exposures worked perfect for
Me on that area of sky.

#35 Inverted

Inverted

    Mariner 2

  • *****
  • Posts: 210
  • Joined: 19 Jan 2013
  • Loc: LP Land

Posted 09 February 2013 - 10:01 AM

Your shorter subs will show far less detail resolution and contrast no matter how many you take. That is the point of Jimmy's experiment.


In practice, that may be true. as photons are emitted at a high frequency and presumably arrive at a high frequency, they should register even in short exposures though... So, if all the other noise is random, then we should absolutely be able to stack and get the same result. So, it would seem the limitation is likely non-random, error, or noise, or unwanted signal or whatever people want to call it. In that case, it may just be limitations of calibration for example. I don't know, I do know professionals, such as the Hubble operators, do stack pics and get really faint detail, so, it is possible....

#36 freestar8n

freestar8n

    Soyuz

  • *****
  • Posts: 3907
  • Joined: 12 Oct 2007

Posted 09 February 2013 - 10:06 AM

Orion is shot in HDR style. I did exposures for the core to bring that out. That is a no brainer that we all know. 10min exposures worked perfect for Me on that area of sky.



Yes - you want to expose long, but avoid saturation. The time to saturate will depend on f/ratio, camera sensitivity, and well depth. Therefore - someone else with hyperstar might be better off on the same object not going beyond 5m, and should not be given the impression his images are suffering by not being 20m.

You are fortunate in having one of the darkest spots available, along with an OSC camera that blocks a lot of sky signal. That is very different from many suburban imagers doing monochrome LRGB at around f/4, who would be quickly sky limited in short exposures by light pollution. For them, there would be little gain in increasing subexposure time because their main noise source is sky background. All they can do is stack - and they should not be given the impression that stacking beyond 16 has no benefit.

So - there is a model that works well, and the benefit of longer exposures depends on factors that are different for each imaging setup. In your case there may be a win going 5m to 20m, but for many others there would be little gain in SNR and it may make things worse due to other factors.

Frank

#37 orion69

orion69

    Viking 1

  • -----
  • Posts: 989
  • Joined: 09 May 2010
  • Loc: Croatia

Posted 09 February 2013 - 10:11 AM

Reason I asked the question is that I never tried shorter then 30 min subs with NB filters. The thing is, when my mount and camera arrived I tried 10 min subs with clear filter just to test setup and since that went well I jumped strait to 30 min subs NB.
It would be useful if I can get same detail with more shorter subs but I see that there are different opinions in that question...

#38 freestar8n

freestar8n

    Soyuz

  • *****
  • Posts: 3907
  • Joined: 12 Oct 2007

Posted 09 February 2013 - 10:25 AM

It would be useful if I can get same detail with more shorter subs but I see that there are different opinions in that question...



Well anyone who is doing perfectly fine at 30m and has no downside from saturation or lost frames has no reason at all to go shorter - as long as he exposes enough frames for a good combine.

The penalty of going to 15m exposures depends on the ratio of sky noise to read noise in the subexposure. Your SNR will go down with 15m exposures, but it will be slight if you are sky limited - but big if you are read noise limited - at 30m.

If you find the numbers for your images at 30m - read noise and sky background noise (sqrt of sky background signal in electrons) that will tell you a lot.

Either way - the model, and professional astronomers, would say the SNR goes as the sqrt of the number of subexposures. So you can achieve the same SNR with 15 as you can with 30 but it will take some additional amount of time. But this does require you to have very good master flat/dark/bias so the calibration is excellent and your stacking does behave according to the model.

Frank

#39 freestar8n

freestar8n

    Soyuz

  • *****
  • Posts: 3907
  • Joined: 12 Oct 2007

Posted 09 February 2013 - 10:38 AM

The models seem to assume that photons arrive instantaneously, in large quantities.


No - the photoelectrons arrive randomly, slow or fast, with some mean arrival rate. The resulting counts in each exposure then obey the Poisson distribution. You could have only one signal electron every 10 frames - the other 9 being zero counts at a pixel - and the signal would still be "there" - even though it doesn't even have a single electron to show for it in most subs.

The signal will nonetheless slowly accumulate with each added frame in a linear fashion - while the noise goes as a square root. Eventually the signal will win over the noise and you will start to "see" it.

That's the theory, that's the model, and that's why the Hubble ultra deep field summed 288 frames.

Frank

#40 neptun2

neptun2

    Viking 1

  • -----
  • Posts: 805
  • Joined: 04 Mar 2007
  • Loc: Bulgaria

Posted 09 February 2013 - 10:41 AM

Well thanks for the explanation. The calculation with the square root definitely makes sense. I have seen that with my own pictures - when you stack for example 5 frames the SNR improvement against the single frame is quite high - the snr is more than 2 times higher. To go further you need good amount of frames. For example to get 4 times better SNR than the single frame you will need 16 frames.

#41 vpcirc

vpcirc

    Soyuz

  • -----
  • Posts: 3965
  • Joined: 09 Dec 2009
  • Loc: Merced CA

Posted 09 February 2013 - 10:45 AM

Frank could you provide us a link to your images. I bet you do fantastic work. I'd really like to see how you apply your knowledge.

#42 freestar8n

freestar8n

    Soyuz

  • *****
  • Posts: 3907
  • Joined: 12 Oct 2007

Posted 09 February 2013 - 10:51 AM

Yes - I think when people talk about hitting a brick wall - it's simply describing the slowness of sqrt(N) as N gets big.

With one sub, you add another and get a big gain. To get the same gain again, you add 2. To get the same again, you add 4. Once you reach 16, you add 4 and - not much change. So you add 4 again and - not much change. But if you keep going from 16 to 32, you will get a net big change. Same with 64 to 128.

But as the image improves, something else happens in that you reach a limit of SNR where the image is very clean and you just don't get any improvement in appearance. But that should be when the image shows no obvious speckly pattern anymore.

But if you really do hit a brick wall, and the image is just as noisy at 16 as it is at 128, then I would look into a problem in the calibration and combining. If you see noise in the bright areas - that points to a problem with the flats. If in the dark areas, a problem with darks or bias. Anyway - it points to a problem.

But I think for most people - it is just less rewarding to have to add a bunch of exposures to get an extra bump in quality. So many people just switch to a different target and move on. But it doesn't mean the SNR isn't still improving according to the model. The model says the relative change with each added frame should decrease - and so it does.

Frank

#43 korborh

korborh

    Viking 1

  • *****
  • Posts: 800
  • Joined: 29 Jan 2011
  • Loc: Arizona

Posted 09 February 2013 - 12:23 PM

Your shorter subs will show far less detail resolution and contrast no matter how many you take. That is the point of Jimmy's experiment. Even with far less total imaging time the amount of faint data he was able to obtain is clearly visible. Secondly, with narrowband, you are never going to reach the sky limit from what Don Goldman has told me. There are bright objects that will get blown out by longer subs such as M 42, but if you stick to short subs, you'll miss this wonderful detail that exists elsewhere. Adam Block has a great tutorial on how to combine short and long subs in CCDStack using missing value to combine. Tony Hallas demonstrates another great procedure in his newest addition to Photoshop processing in Vol 5 using the reveal all layer mask in CS5 and CS6. Jimmy used the Hallas method in his M42.


Unfortunately you are not able to benefit from the discussion as you do not seem to understand the simple concept of statistical sum. If you did appreciate this, you would know why there can be really no sharp cutoffs on number of frames and SNR. I very much doubt the experts you refer to say this, and if they do they are wrong.
But one is free to shut themselves off from basic math and science, and instead base their understanding on discrete bits of information from authority they worship.

#44 vpcirc

vpcirc

    Soyuz

  • -----
  • Posts: 3965
  • Joined: 09 Dec 2009
  • Loc: Merced CA

Posted 09 February 2013 - 12:40 PM

Post deleted by vpcirc

#45 dawziecat

dawziecat

    Vanguard

  • *****
  • Posts: 2288
  • Joined: 20 Oct 2010
  • Loc: Rural Nova Scotia

Posted 09 February 2013 - 01:06 PM

Without falling into the unfortunate polarization that has fractured this and the original thread, I am curious about the choice made by the STSCI imaging team for the ultra deep field Hubble images.
The Hubble is a native f/24 system . . . VERY slow! Who of us here would even consider an f/24 OTA for doing AP? :foreheadslap:
The individual subs comprising the UDF used a typical exposure time of 1200 seconds according to this Wikipedia article.

Now, I am totally unsure if the Hubble UDF images were actually taken at f/24. If not, just what was the EFR for the exposures?

1200 seconds at f/24 is mighty short! f/24 is about 4.5 f stops slower than a more typical imaging OTA used by folks here . . . say f/5.6 just for discussion.

A 1200 sec exposure at f/24 is approximately equal to a mere 120 seconds at f5.6. Or a 1 minute exposure at f/4.0

Seems to me the STSCI folks are using mighty short subs considering the slow FR of the instrument. Of course they are using total integration times approaching two weeks!!!
More than 800 20-minute subs! :bigshock: No "16 sub rule" here!
What wuzzes we are in comparison!! :)

Why? Perhaps . . .
1/ They are limited by the observing window of opportunity, i.e., the target is not available long enough per orbit to allow longer subs. Or . . .

2/ I am completely wrong and they are NOT imaging at f/24 at all, but at a much faster FR.

I am unsure why this thread has become so contentious. The consensus seems correct to me, i.e., go as long as your gear and sky allow but, do not deny the obvious either. At some point, short and long sub results will converge so as to make the the difference in the final result imperceptible. The point at which this happens will be a factor of equipment, sky condition and the image scale at which the final image is viewed. Also exposures must not be so long that saturation and/or blooming become overly obtrusive.

The exposure limit determined by "practical" matters will, for many of us, be less than the ideal as defined by the mathematics. For some of us, it will be WAY less! :(

Insisting on going longer than your mount can track/guide well is a sure-fire way to ruin an exposure totally. Much more so than using an equivalent total exposure duration using shorter, even MUCH shorter, subs.

Many here are pretty expert imagers but I know from my own personal experience that there are people perusing this forum that are just considering CCD imaging. They may presently be using a DLSR, and contemplating making the leap.
They could get some pretty discouraging ideas here, that you just must be able to do 20 minute or longer subs to get nice narrow band images. I once thought that. If you want "world-class, APOD contenders" that may well be so. Especially with slower OTAs.

However, for "pretty pictures" that a hobbyist can take great satisfaction in and amaze his friends with, it isn't so.

#46 dawziecat

dawziecat

    Vanguard

  • *****
  • Posts: 2288
  • Joined: 20 Oct 2010
  • Loc: Rural Nova Scotia

Posted 09 February 2013 - 01:11 PM

From John Smith

I have found that prima-donnas, attention-seekers, know-it-alls and other personality types that vex the spirit are to be avoided. Rather than try to change or convert such individuals, an impossible mission in most cases, I will choose to disengage from these vexatious types as much as possible, whether individually or the venue in which they operate. They just aren't worth any of my remaining time.


OK, I'm outt'a here. This is just name-calling and hostility.

#47 psandelle

psandelle

    Viking 1

  • -----
  • Posts: 654
  • Joined: 18 Jun 2008
  • Loc: Los Angeles

Posted 09 February 2013 - 01:11 PM

Nicely said Terry - that's kind of the conclusion I've come to on all this stuff (both threads).

As for the Hubble...I haven't a clue, but MY SPACE TELESCOPE will definitely have a faster focal ratio. :)

#48 psandelle

psandelle

    Viking 1

  • -----
  • Posts: 654
  • Joined: 18 Jun 2008
  • Loc: Los Angeles

Posted 09 February 2013 - 01:14 PM

Oh, and anyone who wants to use my space telescope (once it's up) is free to.

Paul

#49 vpcirc

vpcirc

    Soyuz

  • -----
  • Posts: 3965
  • Joined: 09 Dec 2009
  • Loc: Merced CA

Posted 09 February 2013 - 01:33 PM

I don't think Hubble has an atmosphere detracting light or light pollution to deal with!

#50 orion69

orion69

    Viking 1

  • -----
  • Posts: 989
  • Joined: 09 May 2010
  • Loc: Croatia

Posted 09 February 2013 - 02:06 PM

Frank, thanks for explanation, I'll try your advices in practice when my new astrograph (F/4.5) arrives. It's very possible that I would need to shorten my subs, at least for LRGB filters. I intent to replace my Baader NB filters with 3nm Astrodon's so hopefully I wouldn't need to shorten subs for NB.






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics