Shortest useful subs - how short?

[chd]

For faint and nebulous deep sky objects, is there a general guideline for how short the shortest useful sub is likely to be? I know it depends on a lot of things and I'm not expecting an exact number, just an idea, figuring an 8" f/2 telescope.

I imagine, though may be mistaken, that from a light gathering point of view you'll collect about the same number of photons in one 5-minute exposure and 5 1-minute exposures, though I realize the noise characteristics should be different. However, at some point as subs get shorter you won't collect enough photons for a particular spot to record signal above the noise at all, so stacking as many of these as you like no longer gets you anything. One 5-minute exposure is still better than 500,000 half-second exposures.

Put another way, let's say that tracking limitations limit me to no more than say 15 second exposures. At what point am I no longer gaining anything by adding more and more of them?

(The real answer is probably "go buy a good polar mount and learn to do carefully guided photography" - but my budget dictates otherwise for the time being. )

Thanks!

[sharkmelley]

The real answer is: "it depends".

Are you imaging narrowband or wide-band?  Is your camera cooled or uncooled?  Do you good or bad sky quality?  Are you using an dedicated astro-camera with high levels of read noise (10electrons) or a modern DSLR with extrermely low read noise (1.1electrons)?

Even if you are only taking 15 second exposures, there is no limit to how many frames you can add and still improve your image.  Each time you double the number of fames you improve your signal to noise ratio by a factor of 1.4

In any case, a good rule of thumb is to increase exposure length until the noise from the background sky glow drowns out the camera's read noise.  It's not too difficult to calculate this - I ought to write up the methodology some time.

F/2 is a very fast optical system so short exposures should be quite feasible.  But you might (or might not) improve image quality by using longer exposures.

Mark

[chd]

- Imaging wide-band (no filters, anyway).

- Cooled camera - ATIK 428EX CCD

- Sky quality, not sure how others would rate it; probably mid-to-poor. M31 is only rarely a naked-eye object, no perceptible Milky Way most of the time.

I have not shot any bias frames, but read noise should be included and subtracted out in dark frames, right?

The real answer is: "it depends".

Are you imaging narrowband or wide-band?  Is your camera cooled or uncooled?  Do you good or bad sky quality?  Are you using an dedicated astro-camera with high levels of read noise (10electrons) or a modern DSLR with extrermely low read noise (1.1electrons)?

Even if you are only taking 15 second exposures, there is no limit to how many frames you can add and still improve your image.  Each time you double the number of fames you improve your signal to noise ratio by a factor of 1.4

In any case, a good rule of thumb is to increase exposure length until the noise from the background sky glow drowns out the camera's read noise.  It's not too difficult to calculate this - I ought to write up the methodology some time.

F/2 is a very fast optical system so short exposures should be quite feasible.  But you might (or might not) improve image quality by using longer exposures.

Mark

[sharkmelley]

- Imaging wide-band (no filters, anyway).

- Cooled camera - ATIK 428EX CCD

- Sky quality, not sure how others would rate it; probably mid-to-poor. M31 is only rarely a naked-eye object, no perceptible Milky Way most of the time.

I have not shot any bias frames, but read noise should be included and subtracted out in dark frames, right?

Unfortunately read noise is random noise which is added to every frame you shoot.  There's no way to subtract it out.  Dark frames don't remove read noise but only the pixel to pixel variations caused by dark current.

Given what you've told me, wide-band imaging with a 428EX (read noise = 4 electrons) on f/2 optics in, say, magnitude 5 skies, you would definitely benefit from longer exposures - up to a minute or two.  But 15 sec exposures are certainly not out of the question and if you can stretch to 30 seconds without star trailing it would be better.

Mark

[chd]

On the Bortle scale (which I just looked up), I'd say the sky here is maybe about a 6, perhaps 5 on a good late night with no moon.  So far I haven't tried more than about 10 minutes worth of 20-second or 30-second exposures, so I'll try raising that number considerably and see what kind of results I can get. I may be able to increase it a little; I do find that the tracking of my altazimuth mount gets a little better after it's sat on one target for awhile. I assume that's any backlash in the gears settling out.

Even if that were perfect I'd still get trails from the field rotation at some point, unfortunately.

Thanks!

[Superdavo]

Lots of short subs will brighten any detail that's captured, but won't catch any fainter detail

David

[sharkmelley]

On the Bortle scale (which I just looked up), I'd say the sky here is maybe about a 6, perhaps 5 on a good late night with no moon.  So far I haven't tried more than about 10 minutes worth of 20-second or 30-second exposures, so I'll try raising that number considerably and see what kind of results I can get. I may be able to increase it a little; I do find that the tracking of my altazimuth mount gets a little better after it's sat on one target for awhile. I assume that's any backlash in the gears settling out.

Even if that were perfect I'd still get trails from the field rotation at some point, unfortunately.

Thanks!

Definitely keep going for longer, the more exposures you take, the fainter the details that will emerge from the background noise.  The only problem with so many exposures is that it takes quite some time to process and stack!

Mark

[dph1nm]

Lots of short subs will brighten any detail that's captured, but won't catch any fainter detail

Yes they will catch the fainter detail (unless you have the gain set so that several photons are required for 1ADU)

NigelM

[Zakalwe]

Yes they will catch the fainter detail (unless you have the gain set so that several photons are required for 1ADU)

NigelM

Are you sure? More subs will increase the SNR, but not the brightness of faint stuff. For faint stuff you need longer subs.

http://deepskystacker.free.fr/english/theory.htm#CombineHowMany

[sharkmelley]

I'm with Nigel on this.  It's a myth that signal totally lost in the noise or that signal that does not appear on all frames is lost forever.  The power of stacking allows even such low level signals to build up until it is distinct from the noise (given sufficient subexposures).

Mark

[LouisJB]

sorry for slight hijack, but conversely what are the longest useful subs. If you had guiding working perfectly and could take really long subs, at what point would it be too long? On something like M31 with a really bright core then of course it will over-expose quickly, but say M33 that's uniformly quite dim - I wonder about this. Sort of the flip side of the OP question but I guess when you understand both constraints you can decide on optimal exposure times for a given setup.

[RobH]

"the real answer"...what a douche.

I see you're new to the forum.

In case you haven't read the CoC, we do not tolerate abuse of members, no matter how mild the abuse is perceived to be by the guilty party (you).

I trust you will take this on board and not see fit to post like this again. If you have a problem with this, then your time as a member here will be very limited.

[Merlin66]

Having just discussed (at some length) the impact of SNR on faint detail. I'm with Nigel!

The fainter detail doesn't become "brighter" just that the surrounding noise is reduced.

For a max sub, I'd say go as far as light pollution/ sky glow will allow without saturating the brighter stars....

(I'd be interested in any comment/ feedback on what could be an "acceptable" limit for SNR - ie is there any real improvement between an image with SNR= 200 and one with SNR=400......)

[chd]

I'll definitely do some experiments with large numbers of shorter subs and see what I can get. That will probably also help with a different problem I have, which is that bright stars in the image tend to spread out into dinner plates (well, not quite, but they are definitely not points). Some of that may be focus inaccuracy, but I suspect some of it is oversaturated pixels leaking electrons into their neighbors. I'm definitely limited to short subs until I get a proper guiding setup anyway.

Thanks!

[chd]

Yes they will catch the fainter detail (unless you have the gain set so that several photons are required for 1ADU)

NigelM

I guess my assumption (perhaps incorrect) has been that yes, you're typically going to need more than one photon to capture any information, especially if you're not in very dark skies. If that's not actually the case - so much the better!

[Merlin66]

Hmmm

No you definately need to collect (at least one!) photon to build up the total...

Think about probablities.....

If the object is soooo faint that only one photon is received every 4-5mins (!!)

Then the probablity of recording one on any 1min sub is only 20%

With five subs there's a good probablity of recording that lonely photon...you may have four subs without it, but one with it...

As you increase the total exposure the probablity increases that you will collect more of those photons...when you achieve a "good" SNR the resulting image will start to appear above the "forest" of noise......

[Ant]

But if you have 5 subs, 4 do not have the photon and 1 does... Surely the stacking software will think that's noise and get rid?

I thought that was the whole point of the stacking process?

Ant

[Merlin66]

No, the stacking process doesn't get rid of anything (unless you go for hot pixel etc)

The photons will just continue to build up....

When the total level of this meagre photon collection reaches a "significant" level it will be high enough above the noise background to be seen.

[StuartJPP]

This topic keeps coming up so often that I am getting sick of reading them...(I keep reading them to see if there is a conclusion!)

The only way to settle this is a practical test.

Would someone like Sara or Olly, who have clear skies, please, please take some subs for people to experiment on?

These subs should be provided to us in RAW format so that we can all have a go ourselves to see how stacking stacks up. Not a competition to see who can process the image the best.

A simple starting point would be an hour's worth of subs in say 30 second subs and 5 minute subs, everything else being the same.

Failing that I have seen absolutely no evidence that taking 1 million subs at a fraction of a second produces anything useful compared to a single sub of the same exposure time.

[Merlin66]

Stuart,

Why the problem......????

A long exposure of say 60mins ( excluding sky glow and light pollution etc etc) will always produce a better image than a series of subs....

It's the additional noise you create with every sub....it, in the end, adds up and also impacts on the final result.

Don't you accept that the increasing signal (with exposure) gradually overcomes the noise "base"??????

[Physicist13]

Other than tracking issues. It only comes down to whether you are "read noise" dominated or not. If you are, you need more total exposure to get the same SNR. If not then as long as the total exp is the same, sub length doesnt matter. Plus you also dont need a signal above noise in any sub frame (obviously its  better if you do as youll need less total exp, but its not a reqirement!) Even sub 1 bit signals can be recovered! In fact, random noise if your friend!!

[ollypenrice]

This topic keeps coming up so often that I am getting sick of reading them...(I keep reading them to see if there is a conclusion!)

The only way to settle this is a practical test.

Would someone like Sara or Olly, who have clear skies, please, please take some subs for people to experiment on?

These subs should be provided to us in RAW format so that we can all have a go ourselves to see how stacking stacks up. Not a competition to see who can process the image the best.

A simple starting point would be an hour's worth of subs in say 30 second subs and 5 minute subs, everything else being the same.

Failing that I have seen absolutely no evidence that taking 1 million subs at a fraction of a second produces anything useful compared to a single sub of the same exposure time.

I don't think I can shoot in raw since I use a CCD. It would be worth doing this test, though.

What I would suggest to anyone with tracking difficulties (or the OP using an alt az mount) is a layered compromise. Usually the faint stuff has little detail to be resolved so a little trailing wouldn't matter much. So you do a short, untrailed set and a longer, rather trailed set, and then layer them together, using the hightly resolved detail from one and the faint data from both. It's fairly easy in Ps.

Olly

[Zakalwe]

I have seen a Flickr site of someone that shoots DSO using a Canon DSLR and L series lens on a fixed tripod. The user takes many hundreds of very short exposures (something like 3 seconds IIRC). The images were OK (ish), especially when considering the tripod used, but were very, very noisy. I guess that read-out noise in short exposures swamp the detail.

I'll see if I can find the site again.

[chd]

One example with short subs. And other problems; the halo-ed stars are probably a spacing problem between my camera and the lens, I think, and the distortion along the bottom is probably dew on the corrector plate that I didn't notice right away; not a great image, but I wanted to throw it out there as a sample of short subs.

This is about 60 ten-second subs.

[RobH]

For any of you who haven't seen this.....here's the definitive answer

http://stargazerslounge.com/topic/200397-many-short-exposures-or-a-few-long-ones/