Jump to content


Photo

Diminishing returns on subs?

  • Please log in to reply
12 replies to this topic

#1 Mike I. Jones

Mike I. Jones

    Gemini

  • *****
  • Posts: 3214
  • Joined: 02 Jul 2006
  • Loc: Fort Worth TX

Posted 25 December 2013 - 10:35 AM

Apologize if this has come up before. Is there a point of diminishing returns regarding the number of subs taken? I see pictures that are sometimes the composite of 200 or more subs. Don't they become unnecessary at some point, and in fact begin to actually degrade the picture quality? I realize this is a strong function of temperature, fixed vs. uncorrelated noise patterns, ISO setting, focal ratio, guiding accuracy, seeing, etc. Still, is there a point where enough is enough, and more is too many?

Thanks,
Mike

#2 Tonk

Tonk

    Fly Me to the Moon

  • *****
  • Posts: 7085
  • Joined: 19 Aug 2004
  • Loc: Leeds, UK, 54N

Posted 25 December 2013 - 12:05 PM

Very recent thread only a few stops below here

How much difference do more subs make?

#3 Jerry Lodriguss

Jerry Lodriguss

    Vendor

  • -----
  • Vendors
  • Posts: 3566
  • Joined: 19 Jul 2008
  • Loc: Voorhees, NJ

Posted 25 December 2013 - 12:18 PM

Hi Mike,

Some people argue that you reach a point of diminishing returns, but I don't think there is ever a point where more subs hurt.

Personally, I don't even agree with the point of diminishing returns philosophy. More photons are always going to give you a better signal-to-noise ratio because the signal can only come from the photons, and the only way to collect more photons is by collecting more subs.

In the April 2012 issue of Astronomy Magazine, Tony Hallas argued that you reach an asymptotic boundary at about 25 frames and that shooting more frames was basically a waste of time.

But this is not true. He even mentions that the signal-to-noise ratio improves with the square root of the total number of frames, but then seems to ignore it.

For example 4 frames will give you twice the s/n of 1 frame.

Likewise, 9 frames will give you 3x.

16 frames will give you 4x.

25 frames will give you 5x.

36 frames will give you 6x.

So going from 25 to 36 frames you have needed about 50 percent more total exposure time, buy your s/ has only gone up 20 percent.

Still, it has gone up.

If you follow the logic of 25 frames as a practical limit, you would never want to attempt any deep-sky objects from light polluted areas, yet there are plenty of people doing exactly that.

Look at it like this... 25 frames is 5x better than 1 frame. But, 100 frames is 10x better than one frame, and twice as good as 25 frames.

It certainly would be nice if the improvement was linear, but you still get an improvement by shooting as many frames as you can.

Now, you won't see that much improvement with 26 frames vs 25 frames. But you certainly will with 100 vs 25. The s/n will be twice as good. Is that "worth it?" Only the individual photographer can answer what his/her own level of acceptable quality is.

And this is especially important under light-polluted skies because an individual sub exposure will be limited in duration. So you will need a "hella lotta" (scientific term) subs under those conditions.

I have a shot of M27 from my suburban driveway that is a stack of 90 one-minute exposures at f/8 at ISO 800. And 90 frames certainly was not too many.

Jerry

#4 Falcon-

Falcon-

    Aurora

  • *****
  • Posts: 4869
  • Joined: 11 Sep 2009
  • Loc: Gambier Island, BC, Canada

Posted 25 December 2013 - 12:21 PM

Yes there are diminishing returns as you add more subs.

This page here (from Deep Sky Stacker's reference manual) goes over the math including examples. (The mouse-over image comparisons can take a bit to load at first so if at first the mouse-over does not work just try again in a few seconds.)

They key aspect is this:

The signal to noise ration in increasing with the square root of the number of combined frames.

To quote from the DSS page:

"This means that if your base SNR is 1, when you combine 10 images the SNR increases by 3.16 (square root of 10). For 30 images it is 5.47, for 50 images it is 7.07, for 100 images it is 10, for 300 images it is 17.32."

Put another way the steps to double your Signal to Noise Ratio (SNR) go like this:

- To increase SNR by 2x you need 4 frames
- To increase SNR by 4x you need 16 frames
- To increase SNR by 8x you need 64 frames
- To increase SNR by 16x you need 256 frames
- To increase SNR by 32x you need 1024 frames


So ya... to double your SNR from 4 frames is not very hard, an extra 12 frames (to get 16) is achievable - but to double your SNR from 64 frames requires 192 extra frames! Not impossible, but a lot harder.

In my mind there are two reasons to shoot lots of frames:

1) You imaging in a light polluted area. This means you have to use short exposures and as a result a single exposure does not have much signal to start with. Thankfully since you are using short exposures it is also easier to shoot LOTS of exposures.

2) You are imaging in a dark-sky area, but the target you are trying to go after is VERY faint. With super-faint objects photon shot-noise (the randomness in the arrival of individual photons from your target) is enough of a factor to make it hard to pick out faint details. The solution here is both long exposures *and* LOTS of those long exposures. Much harder to build up into huge numbers of exposures (much time required) but this is how you get dim detail!

#5 Falcon-

Falcon-

    Aurora

  • *****
  • Posts: 4869
  • Joined: 11 Sep 2009
  • Loc: Gambier Island, BC, Canada

Posted 25 December 2013 - 12:25 PM

Well I guess I took too long to type my response, the others covered it well while I was typing!

Some people argue that you reach a point of diminishing returns, but I don't think there is ever a point where more subs hurt.


Absolutely, Jerry is right here. I especially agree that the thing to keep in mind is the part I made bold from his quote. :)

#6 Charlie B

Charlie B

    Viking 1

  • -----
  • Posts: 960
  • Joined: 22 Mar 2008
  • Loc: Sterling, Virginia

Posted 25 December 2013 - 12:26 PM

There is a diminishing return since signal-to-noise ratio increases as the square root of the number of subs, but sometimes it's necessary. For instance, this wide-field of the Orion region is 180 20-second subs. Total time 60 minutes. This yielded ~13 times the SNR of a single sub. Ninety subs would be about 9 times the SNR and 360 subs would be about 19. So doubling my time would only give me about 50% more SNR.

In my case, I kept the time short to keep from saturating the bright stars in M42 and, since the region has a fairly large signal, I got reasonable results from a lot of subs. Others often take short subs for the bright stars and long subs to get the dimmer nebulosity, and then combine the short and long subs into a single image. Because of the wide field, I decided that a lot of short subs were better.

Each method has its advantages.

Regards,

Charlie B

#7 Charlie B

Charlie B

    Viking 1

  • -----
  • Posts: 960
  • Joined: 22 Mar 2008
  • Loc: Sterling, Virginia

Posted 25 December 2013 - 12:29 PM

There were no replies when I started, and I agree with all the above.

Charlie B

#8 Alex McConahay

Alex McConahay

    Vanguard

  • -----
  • Posts: 2377
  • Joined: 11 Aug 2008
  • Loc: Moreno Valley, CA

Posted 25 December 2013 - 12:41 PM

Both sides are right on this point.

Fact is, you reach the point of diminishing returns on your SECOND sub. (I know that is extreme, but it is true. You have doubled the time of exposure, but the amount of noise has only decreased by roughly the square root of 2, so you are already getting less of a reduction in noise than you are costing in doubling the exposure times. --Your returns on the second exposure have already diminished (considering the SNR v. # of Exposures.) However, I do not think any of us would or should stop at one exposure.

THe issue is whether the returns are no longer worth the added investment in time/equipment/ and what else you could be imaging.

Take Jerry's example:

>>>>>>So going from 25 to 36 frames you have needed about 50 percent more total exposure time, buy your s/ has only gone up 20 percent.

>>>>>Still, it has gone up.

YES, the SNR has gone up, but AT WHAT COST?

Can you see the difference in your images? Or was something else controlling the overall quality?

Could you have more wisely spent that time collecting data on another object?

In short, there is no one answer to your question. It really does depend on your standards, on what your limiting factors are, etc.

Tony Hallas may come up with 25 based on a dark sky site with some really good equipment. I may have a different answer based on my little scope in suburbia.

And as Tonk has pointed out by directing you to another thread----even with scientific evidence regarding the noise level at various exposures, there is no simple answer.

Alex

#9 mmalik

mmalik

    Fly Me to the Moon

  • *****
  • Posts: 5360
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 25 December 2013 - 03:29 PM

My take on a related subject here.... Regards

#10 Tonk

Tonk

    Fly Me to the Moon

  • *****
  • Posts: 7085
  • Joined: 19 Aug 2004
  • Loc: Leeds, UK, 54N

Posted 25 December 2013 - 03:48 PM

In the April 2012 issue of Astronomy Magazine, Tony Hallas argued that you reach an asymptotic boundary at about 25 frames and that shooting more frames was basically a waste of time.


Then why does the same Mr Hallas produce images with total integration times of 20 to 30 hours?? The only way I can reconcile his statement with what he actually produces is that each of his subframes was ~1 hour ;) :lol:

(but remember he has to sum separate LRGB frame sets being a CCD guy so his integration times do get inflated)

#11 Alex McConahay

Alex McConahay

    Vanguard

  • -----
  • Posts: 2377
  • Joined: 11 Aug 2008
  • Loc: Moreno Valley, CA

Posted 25 December 2013 - 06:57 PM

I cannot recall TOny's complete presentation on the idea, but I believe it was not that there is no more improvement, but that there may be better ways to clean up the signal, or remove the noise, than just taking lots and lots more exposures.

And remember, everybody's techniques evolve over time.

Alex

#12 whwang

whwang

    Ranger 4

  • -----
  • Posts: 343
  • Joined: 20 Mar 2013

Posted 25 December 2013 - 08:46 PM

Hi,

A lot of what you guys talked about in this thread and in the other recent thread is just a mental effect.

The feeling about how much time we spend in imaging (and post processing) is linear. We feel 10 hr is a lot harder than 1 hr, and 100 hr is much much harder than 10 hr.

On the other hand, S/N goes with square-root of total exposure time. It is not linear in our mental time clock. The square-root is in power space, but our feeling about how much time it takes is in linear space. Doubling the exposure time doesn't double the S/N. This is true going from 10 hr to 20 hr, but also true going from 1 min to 2 minutes. If both are true (10 hr to 20 hr, 1 min to 2 min), why most of us almost always feel OK to increate the exposure from 1 min to 2 min, but feel kind of painful when we think about adding 10 hr to an existing 10 hr of exposure? This is just something in our brain, not something in our image.

We always feel progressively harder and harder to improve S/N once we pass the boundary of a few hours. But, feeling harder and harder doesn't mean there is a point of diminishing returns. As long as the images are well calibrated and the camera is doing what it's supposed to do, we can always improve S/N by adding more integration time.

Here is another example. People (like me) who do not have a dome and have to use portable telescopes, start to feel things get harder once we reach a few hours of exposure. Those who do have domes (and even remote control observatories) may image more comfortably, and they can spend tens of hours on one target. If there is a point of diminishing returns, do you think such a point changes according to whether the telescope is in a dome or not? I don't think so. This is just about how we feel about using our time on imaging.

Astrophotography is fun. Spending more time is always rewarding. The only caveat is that the rewarded S/N goes with square-root of time, not linearly with time.

Cheers,
Wei-Hao

#13 Alex McConahay

Alex McConahay

    Vanguard

  • -----
  • Posts: 2377
  • Joined: 11 Aug 2008
  • Loc: Moreno Valley, CA

Posted 25 December 2013 - 09:03 PM

I think you have it right.

I am about to collect another ten hours of narrowband data with my backyard observatory. I will be adding it to the ten hours I already collected. And then tomorrow night I will probably add another ten hours so that I have something like ten hours per channel. 20 subs per channel. Thirty hours total. I might even add another night if the weather holds.

The cost for me.......about fifteen minutes of work per ten hours. There is really no reason I should not do it. I will be sleeping through the whole thing.

However, when I am at a remote site at a hard to get to southern latitude, I would probably limit each object to a couple of hours so that I could get more while away.

Alex






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics