Are short subs a barrier to extracting faint detail?

[Martin Meredith]

I've often heard the view that long subs are required to image faint objects. However, my understanding of the theory is that if sky brightness is sufficient to overcome read noise -- a situation that faces many of us much of the time -- there is no advantage in using long subs. Short subs (when combined with live stacking) are of interest in near real-time viewing and simplify the acquisition setup (no need to guide, use of alt-az mode or perhaps cheaper mounts), so I'm intrigued to know how much detail can be picked up by stacking lots of short subs.

Motivated by this thread I chose M51 as a test object since it contains some extremely faint extensions -- the outer halo -- which are not generally picked up on short (total) exposures.

Sky conditions: SQM 20.2

Equipment:  SW Quattro 8" f/4, AZ-EQ6 mount in alt-az mode, Lodestar mono X2 camera, no filters.

Novice post-processing in Nebulosity 3: dark frame (20x15s) subtraction, mean stacking, digital development.

Here's a stack of just over 1 hour (250 x 15s)

Some details of the outer halo are coming through, I think, although by comparison with much longer total exposures there are still huge swathes not captured. By continuing to stack 15s subs I wonder how much more of the halo would show up or whether the process will hit a wall? Do I have the patience to find out?    

cheers

Martin

[ultranova]

Very Interesting Martin, Have often thought about this myself,

Having a side by side comparison would be the way to do it,

250 subs at 15 seconds or say roughly 12, x 5 minute subs,

Im not sure what would be better as they are both catching photons,

the photons you don't want are the sky glow ones created by street lights,

On the plus side of taking 250 subs it should be good for getting rid of any

noise, on the negative side you would need a huge amount of disc space to put all the subs

on, plus stacking that lot together would take a while if you have a big megapixel camera,

Will be watching this thread with interest

 

Nice pic of M51

 

Paul

[carastro]

Looks like you have fast optics, put that together with dark skies and it will make a lot of difference of course.

Carole 

[ultranova]

Yes you are right Carole ,Plus the Lodestar mono X2 camera

is a very sensitive camera, I cam't wait to see what a hour look's

like with longer subs.

Paul

[Martin Meredith]

Very Interesting Martin, Have often thought about this myself,

Having a side by side comparison would be the way to do it,

250 subs at 15 seconds or say roughly 12, x 5 minute subs,

Im not sure what would be better as they are both catching photons,

the photons you don't want are the sky glow ones created by street lights,

On the plus side of taking 250 subs it should be good for getting rid of any

noise, on the negative side you would need a huge amount of disc space to put all the subs

on, plus stacking that lot together would take a while if you have a big megapixel camera,

Will be watching this thread with interest

 

Nice pic of M51

 

Paul

Thanks Paul. I agree that a side-by-side would help. Its just that when getting into these long total exposures the 'near real-time' observer in me starts to object… but certainly I'd like to continue to experiment. 

You're right of course to point out the implications for disk space and stacking time, although my main interest is in determining if the details can be extracted at all using short subs. If so, my intention is to use short subs in combination with live stacking, where there are no disk space/stacking time implications. In fact, I was live stacking while collecting these subs. Here's an example of what LodestarLive produces on the screen in real time (this is the 52nd sub i.e. 13 minutes in). The only processing is online adjustments and rotation to match the other pic. 

cheers

Martin

[ollypenrice]

I'm one of the arch-advocates of the long sub but I work from a dark site. I have no doubt at all that long subs go deeper and have plenty of back to back comparisons to persuade me of this.

However, you have caught some of the outer halo. The trouble is that gradient is swamping lower left of the chip. It just might be possible to lose this in DBE without killing the buried halo data that will be there but it will be a tall order.

It's an interesting test, certainly.

Olly

[Martin Meredith]

Yes, theory agrees with you Olly that longer subs are definitely the way to go at a dark site (which sounds somewhat counter-intuitive but its just the read noise > sky noise point again I guess).

Gradients are an issue and I see them all the time in live stacking due to the extreme stretching I sometimes engage in 'online'. It would be great to know how to reduce them both at source (sensor orthogonality? stray light? I didn't use my normal light guard nor primary mirror light shield on this occasion) and also 'live' by some kind of online processing. BTW no flats were used either. Not sure if that would help?

Martin  

[Martin Meredith]

I just wanted to update this thread not with my images but with some from Emil Kraaikamp who shows how lots of 1s (yes, one second!) luminance subs and 4s RGB subs can produce some pretty amazing images (taken on a 16" dob with an equatorial platform). They key is the low read noise camera. 

M51

NGC891

Martin

[russellhq]

I just wanted to update this thread not with my images but with some from Emil Kraaikamp who shows how lots of 1s (yes, one second!) luminance subs and 4s RGB subs can produce some pretty amazing images (taken on a 16" dob with an equatorial platform). They key is the low read noise camera. 

M51

NGC891

Martin

Wow! That's really impressive for 1s subs! At 1s, do you think you can start to do what the planetary images do with "lucky imaging" to reduce the impact of seeing?

[Martin Meredith]

Wow! That's really impressive for 1s subs! At 1s, do you think you can start to do what the planetary images do with "lucky imaging" to reduce the impact of seeing?

I believe he is rejecting plenty of subs -- though probably nothing like as many as in planetary work --  and perhaps not so much based on seeing but to reduce the impact of say gusts, tracking issues etc. There's more info in this thread.

Martin

[Stub Mandrel]

<boggle> 3300 1s exposures for L, although he did 100 4 second exposures for each of RGB

Over my DSLR and F8 6" scope he has a few advantages, but it does make me glad I invested in a timer remote.

[sharkmelley]

I've often heard the view that long subs are required to image faint objects. However, my understanding of the theory is that if sky brightness is sufficient to overcome read noise -- a situation that faces many of us much of the time -- there is no advantage in using long subs.

Your understanding of the theory is absolutely correct.  Once the noise from the sky brightness (including any thermal noise) is a few times greater than the read noise then there is vary little to be gained by increasing the length of the subs further.  Using my Sony A7S on my Tak Epsilon (F2.8) with 30second subs I typically see read noise of 1.3e and sky noise of 9e (SQM 20.8) meaning that 30sec subs are more than adequate.  The only problem with short subs is that you end up needing hundreds of them for a typical deep sky image, which causes an obvious storage problem for the data and a processing problem for the large number of subs.  From a practical point of view it therefore makes sense to take longer subs.

For completeness, I should also point out that when doing narrowband imaging on a noisy mono CCD (though not all are very noisy) then it will require very long subs for the noise from the sky brightness to exceed the read noise.  This is especially true under excellent sky conditions. This is probably why you have often heard the view that long subs are required to image faint objects.

Mark

[nmoushon]

Though those pic are nothing short of Impressive the biggest thing going for them are the 16" of light gathering power. That's a lot of light that most imagers don't get.

I started my own experiment a couple years ago where I used 2min subs on m42 for a total of 8hrs of data. At about 6hrs of data I saw a very noticeable drop in how deep the image was going for how much data I was collecting. By the time I hit 8hrs I could barely tell any difference btwn the two. In depth that is. Noise was better of course but not depth. I was then going to test at which point would 5min subs equal the depth of the 8hrs of 2min subs. Then see how much deeper I got when I got to 8hrs of 5min subs. But then several life events happened and I had to put a long hold on my AP ventures. So never got to finish it .

But my point is that no matter your skies longer subs will go deeper than shorter sub. Your skies only limit how long your subs can be. I'm positive that my personal experiment would've proven this if I ever got to it. Now by no means am I saying short subs can't produce amazing pictures. Those one second subs prove that. There's a lot of variables that have to come into play with imaging that will give different results to everyone. The best thing is to figure out what the best combo is for your equipment and location.

[Knight of Clear Skies]

I just wanted to update this thread not with my images but with some from Emil Kraaikamp who shows how lots of 1s (yes, one second!) luminance subs and 4s RGB subs can produce some pretty amazing images (taken on a 16" dob with an equatorial platform). They key is the low read noise camera. 

M51

NGC891

Martin

Very interesting thanks.

Though those pic are nothing short of Impressive the biggest thing going for them are the 16" of light gathering power. That's a lot of light that most imagers don't get.

Yes, but most of that light isn't reaching the small chip. It's the focal ratio that counts and his telescope is only f5.

[sharkmelley]

Regarding your other question, you can certainly carry on adding more and more subs and continue to extract fainter and fainter detail.  There is no "wall" that blocks this.  For a given length of sub, every time you double the number of subs then you increase the final signal-to-noise ratio by 1.4 (the square root of 2).  What this means is that if you already have 30 minutes of data then adding another 30 minutes of data will give a noticeable difference but if you already have 10 hours of data then you will need to add another 10 hours of data to see a similar increase in quality and so on.  This is what is meant by the law of diminishing returns.  At some point you will give up and say that your time is better spent choosing a different target than continuing to add data to the existing one. 

However, look at deep-sky competition winners - you'll find they often have over 50 hours of exposure!  I've never done more than 50 hours for a single image but it was still quite noticeable that 50 hours was better than 25 hours which was better than 12.5 hours which was better than 6 hours which was better than 3 hours etc.

Mark

[nmoushon]

Regarding your other question, you can certainly carry on adding more and more subs and continue to extract fainter and fainter detail.  There is no "wall" that blocks this.  For a given length of sub, every time you double the number of subs then you increase the final signal-to-noise ratio by 1.4 (the square root of 2).  What this means is that if you already have 30 minutes of data then adding another 30 minutes of data will give a noticeable difference but if you already have 10 hours of data then you will need to add another 10 hours of data to see a similar increase in quality and so on.  This is what is meant by the law of diminishing returns.  At some point you will give up and say that your time is better spent choosing a different target than continuing to add data to the existing one. 

However, look at deep-sky competition winners - you'll find they often have over 50 hours of exposure!  I've never done more than 50 hours for a single image but it was still quite noticeable that 50 hours was better than 25 hours which was better than 12.5 hours which was better than 6 hours which was better than 3 hours etc.

Mark

So you're saying that a single 30min exposure would not be any deeper than any other combination of exposure totaling up to 30min?

[sharkmelley]

So you're saying that a single 30min exposure would not be any deeper than any other combination of exposure totaling up to 30min?

Exactly.  As long as all the exposures are sky noise limited i.e. the sky noise in each exposure is sufficiently larger than the read noise.

With my Sony A7S on a fast scope (Tak Epsilon F2.8) under my typical sky conditions (SQM 21.0) I could actually get a good quality image with 1 second exposures:  sky noise 1.7e and read noise 0.5e

You've probably seen my real time video of nebulosity:  http://stargazerslounge.com/topic/249751-sony-a7s-on-takahashi-epsilon/    That was 4 frames/sec.  It would even be possible to integrate that into a half way decent image.

Mark

[StuartJPP]

So you're saying that a single 30min exposure would not be any deeper than any other combination of exposure totaling up to 30min?

Exactly.  As long as all the exposures are sky noise limited i.e. the sky noise in each exposure is sufficiently larger than the read noise.

Mark

I'd like this to be backed up by a practical example, maths and theory aside. I am not disputing it, just want to see it proved, which is the acid test.

[nmoushon]

Exactly.  As long as all the exposures are sky noise limited i.e. the sky noise in each exposure is sufficiently larger than the read noise.

With my Sony A7S on a fast scope (Tak Epsilon F2.8) under my typical sky conditions (SQM 21.0) I could actually get a good quality image with 1 second exposures:  sky noise 1.7e and read noise 0.5e

You've probably seen my real time video of nebulosity:  http://stargazerslounge.com/topic/249751-sony-a7s-on-takahashi-epsilon/    That was 4 frames/sec.  It would even be possible to integrate that into a half way decent image.

Mark

Sorry but I'm finding this hard to believe. Maybe I've just been reading miss-informing topics for so long that its hard to see. I've always read that there is a "wall" that you hit at a certain total exposure time for the given single exposures length. Similar to me falling for the "f/ratio myth" for the longest time.

And it makes since in my head too. That a single longer exposure will go deeper than a single short exposure so when you stack them at the end the longer exposure will go deeper. 

But if this is not true the only real reason people do long exposures is that is cuts down on your total time need to image. So the only reason people get crazy expensive mounts for better tracking/guiding (for longer exposures) is to cut down on the total integration time and not to actually acquire deeper data?

[ollypenrice]

Those images are, obviously, deeply impressive. The M51 doesn't go deep so far as the halo is concerned but, then again, the FOV doesn't really let this happen anyway.

In terms of giving general advice to practical imagers, though, there will always be people doing things that other people cannot do. In terms of a discussion of the theory these exceptions are thought provoking but in terms of giving you a direction to follow, maybe not? Am I tempted to ditch my present setups and go for an equatorial Dob and 1 to 4 second exposures? I can't say that I am, but hats off to Emil Kraaikamp. Wonderful stuff, SIr.

Olly

[sharkmelley]

I'll do a practical experiment next time the new moon comes round.

Actually, I've just found an interesting thread on SGL where someone has already compared 1x30sec with 30x1sec:  http://stargazerslounge.com/topic/245183-to-stack-or-not-to-stack-30-x-1s-1-x-30s/   admittedly under brighter skies.

Longer exposures don't necessarily cut total image time.  What I'm arguing is that for the same total image time, short exposures will give (more or less) the same result as long exposures as long as all of them (both short and long) are sky noise limited.

So why do people spend crazy money on expensive mounts?  It's because when you are imaging narrowband under very dark skies with a relatively slow scope (F8 or F10) then you need the really long exposures for the sky noise to drown our the CCD read noise - especially on some of those large Kodak sensors where the read noise is quite high.

I don't know where the idea of a "wall" comes from.  I think it is a subjective thing that varies from person to person.  Some people would never spend more than 1 hour on a single image whilst others would draw a limit at 100 hours.  But the 100 hour image would have 10x the signal to noise ratio of the 1 hour image.

Mark

[sharkmelley]

There was a gap in the clouds that allowed me to do a quick experiment.  Sky quality was only 20.6 because of the moon.  I used the Sony A7S on the Tak Epsilon 180ED.  A quick 1 second test exposure showed the sky noise to be roughly 1.8e.  I wanted to use an ISO of 100000 where the read noise is approx. 0.6e but a 30 second exposure saturated the sensor.  I settled on an ISO of 8000 where the read noise is approximately 1.1e   For the 1sec exposures then, the sky noise is not completely dominating the read noise and so, mathematically, we know in advance that 30x1sec will be noisier than 30sec.

So here is a 1:1 size crop of part of the Andromeda Galaxy.  I have applied identical processing to the 30sec exposure and the stack of 30x1sec exposures.

The 30sec exposure is a certainly cleaner and makes it possible to resolve more detail in the dark lanes but the difference is not vast.  Download them and blink them against each other and see what you think.

In the end though, taking 1 second exposures is not a practical way to perform deep-sky astrophotography.

Mark

[Stub Mandrel]

I think the message for someone like me who is starting out and is hoping for recognisable images rather than spectacular ones, the message is simple: Don't panic if you can only get 1-2 minute subs.

So far I have been getting up to 20 subs, but on reasonably bright subjects. For dimmer objects, the challenge of longer subs has meant more rejects and ending up with relatively few to stack.

Rather than trying to push for longer and longer subs and getting more and more rejects, now I have an electronic interval timer I will experiment to find out what is an exposure that consistently gives me good stars, then set it up the timer to take regular shots and leaving it for an hour or two.

Following advice elsewhere I shall also try using a higher ISO and get the best possible flats and darks.

I will report back...

[Martin Meredith]

Mark, I have to agree that this is not necessarily a practical route for AP. On the other hand, it is a valuable approach for near real-time camera-assisted observing where subs are not saved but simply stacked on the fly. The combination of live stacking and low read noise is very powerful. Combine that with software field derotation and we can get away with alt-az mounts too, and get to use otherwise inappropriate AP setups like big dobs on equatorial platforms. I love seeing the image develop progressively over the course of several minutes observing, especially for faint galaxy clusters where the initial sub might show nothing at all. 

Martin

[nmoushon]

I think the message for someone like me who is starting out and is hoping for recognisable images rather than spectacular ones, the message is simple: Don't panic if you can only get 1-2 minute subs.

I will report back...

This is most definitely true! I spent a year imaging unguided and was able to achieve 2min subs with my setup at the time (which is still in my sig though I have stopped imaging for now). You can achieve a lot and produce some great quality items without 30min subs.