Longer or shorter exposures with these warm summer nights?

[Physicist13]

Hi Sara. The quick reply would be yes. Just depends how sophisticated you want to model/describe the problem. The so called CCD equation (poor naming but there you go) accounts for LP, faint object, chip/ccd performance, read noise and shot noise. It could be further develped to include seeing. My point is that I believe far too much emphasis is being put onto the perceived need to do long exposure subs with the requirement of guiding. This ramps up the cost and complexity of AP and I believe is putting many off. If you have that ability to do long exp then do so. But it is not a defacto essential. If, for example (and I know this is not a realisitc example) read noise were zero than its only the total exp time that matters. ie 1x100mins=10x10mins=100x1min. Its only because of read-noise that his breaks down. you need to understand the actual read noise of your system and then you can make a judgement. The larger the read noise the longer the total exp time needs to be to obtain the final SNR desired..but that may be a lot easier to do and considerably cheaper than guiding.

[ollypenrice]

But read noise is precisely the point about long subs. The very faint signal (in practice often the narrowband detail) has to climb above the read noise. How can it do that is short subs?

Olly

[Tinker1947]

While i don't image i would have to take Olly's side, "Practice vs Theory" as practice has been done as shown by the many splendid images posted by Olly and the many other fine images on these forums, Theory on the other hand is just that figures on a piece of paper, how many Theories get modified as time goes bye.......

[IanL]

While i don't image i would have to take Olly's side, "Practice vs Theory" as practice has been done as shown by the many splendid images posted by Olly and the many other fine images on these forums, Theory on the other hand is just that figures on a piece of paper, how many Theories get modified as time goes bye.......

Are you seriously asserting that practice is better than theory based on theory rather than practice? (Sorry couldn't resist ).

[ollypenrice]

Scientific theory is susceptible to a single contrary indicator from observation. I believe I have made that contrary observation many, many times. I must also say that The Physicist is the only theorist I know to assert that multiple short exposures match an equivalent time in fewer longer ones! All the other imaging theorists that I know say that only long subs will defeat the read noise, though that's anecdotal since I've no means of knowing how many of each persuasion might be out there. I'm not a theorist but I do have a passing interest in the tools I use.

One thing is quite certain, though. My long subs won't be struck off the workflow in the foreseeable. It has taken me far too long to push my subs up to half an hour. Let's not forget the astonishing images Tim has posted with subs far longer than mine. I must credit Yves, a hardware and software engineer, for pushing me into these longer subs.

Here's another question. Why isn't Bin 2X2 four times faster than unbinned? It isn't. Plenty of people have measured it and it is about 1.6 to 2x faster. I don't know why, but I know why I never bin any more. (With the refractors I'm already undersampling and the camera in the long FL rig makes a hash of binning.)

Olly

[Gina]

From my experience I have to agree with Olly and I have a degree that includes maths and physics and know a bit about the theory. What has to be realised with theory is that it is based on assumptions and whilst these may apply quite well to terrestrial physics, astronomical matters can be somewhat different. I believe I have reasonable experience of DSO imaging though nothing like Olly and co. as I generally take every opportunity of practice.

[johnrt]

I've read through this thread and I have to say I'm firmly in the "lots of 2 minute subs will not get you the same result as fewer longer subs" brigade.

The longer subs will detect signal that the shorter ones won't, so you can stack all the 2 minutes subs you like to better your SNR but if the signal isn't there in the first place you are wasting your time.

I base this on having gone outside and done it and looked at the results.

Of course on top of the maths are lots of factors to take in to account, pixel size, f/ratio, transparency and light pollution, but the fact remains a stack of 2 min subs will be blown away by a stack of 10, 20 or 30 minutes ones.

[Shibby]

Back on the temperature thing... it's not true that you can't take long exposures with a DSLR on warm evenings. Chances are, the sky backround will limit you sooner than the thermal noise becomes an insurmountable.

For example, I took this image of the Crescent Nebula mid-July last year using 10min subs. It goes much deeper than possible with shorter subs; and the hot pixels have stacked out no problem.

[dph1nm]

But read noise is precisely the point about long subs. The very faint signal (in practice often the narrowband detail) has to climb above the read noise. How can it do that is short subs?

You have a bright sky (some of us do!) or a large aperture scope (with sensibly-sized pixels)

NigelM

[dph1nm]

The longer subs will detect signal that the shorter ones won't, so you can stack all the 2 minutes subs you like to better your SNR but if the signal isn't there in the first place you are wasting your time.

This is a misunderstanding of the physics - the same number of photons will be detected from an object in 10x2mins as in 1x20mins (providing you don't use a gain/ISO which causes quantisation of course).

NigelM

[dph1nm]

What has to be realised with theory is that it is based on assumptions and whilst these may apply quite well to terrestrial physics, astronomical matters can be somewhat different.

Sorry - professional astronomers use the theory all the time. And it works. Many millions of pounds would have been wasted if it didn't!

NigelM

[Zodiacal_Light]

The way I understand it is that to get the same image quality with shorter subs compared to longer subs, you need a longer total integration time with the shorter subs. So you can't really compare 120x1min subs with 4x30min subs. The total integration time for the 1min subs would have to be much longer to match the total integration time of the 30min subs.

[Daddystu]

1 minute subs just dont grab enough detail. I could take a million 1 minute subs and I wouldnt get anywhere near the image i'd get with 10x10 minute subs.

[Jessun]

A one minute sub at f3 is worth more than seven minutes at f8, so that argument, or any based on minutes alone, is not useful.

In collecting signal, an open shutter, electric or mechanical, is an open shutter, and the photons will start counting. With short exposures - relatively speaking for your particular setup - you are cutting the signal short each time, losing out on the dynamic range that the CCD offers. You can get around this by adding subs, say in groups of 10 and 10 etc, in software to reach a number closer to full well. These 'combo subs' can then be stacked in any ordinary fashion, and the signal will be there - just as strong. That elusive faint signal is not affected of how often you read out the CCD - apart from the seconds it takes to actually do this.

Sadly noise is part of the equation, so the above approach would be far from ideal. The signal to noise ratio is the holy grail here, so it's obviously better to push the CCD to its max between each readout.

The biggest factor is sky quality. This is not noise per se, just unwanted random light of whatever source. This can make read noise discussions somewhat academic, since the sky itself is the weakest link. In such case taking more subs is better for a given time to make the signal burn through and randomise the sky. For a messy sky, sub count up to 100 or beyond still shows improvement in my experience.

Most of us probably battle the sky rather than CCD characteristics, and find some sort of middle ground in terms of sub count / length.

So in the end it's simple. Go for as long an exposure your mount and CCD can handle given the sky conditions on the night. Then take as many of these as possible.

/Jesper

[ollypenrice]

Jesper has it.

I have a dark site. To exploit it I use long subs.

I'd like to stress that I'm not anti-theory, I'm just anti wrong theory.

Olly

[narrowbandpaul]

What a fascinating discussion

This is a perennially discussed topic amongst imagers it seems. With a physics background I thought I would add my view in here as well. Firstly in defence of science, not that it should need defending, the theory behind the SNR of an image does take in to a whole host of factors- things like read noise, shot noise, dark shot noise, dark fixed pattern noise, light fixed pattern noise, background noise, telescope efficiency, QE, filter transmission, elevation of target and seeing (I posted an article on this several years back). Some of these a easily measured, some are a pain in the ass. I have written down equations to get from signal on the CCD to magnitude using as much info as I could incorporate. The usual result is nonsense. Measuring the efficiency of the telescope must be a nightmare. So yes, science has covered all the bases in an equation but some terms are nothing better than guesses. All the theory is good but there are some things that can't be estimated easily and thus limits the usefulness of the analysis.

So to the topic at hand. Long or short? You absolutely need to image until you get above the read noise floor. 10 subs will have more read noise than just a few. How long it takes to do that depends on your background sky flux and the dark current of camera. More cooling from a dark sky will need longer subs. DSLR from the middle of London? You will beat the read noise in a minute maybe less. Once you are out of that regime you are then shot noise limited (assuming flats get used to removing fixed pattern noise). This is the best you can do. In this regime it is mathematically equivalent to take 10x10 or 100x1. All Astro images have a wide dynamic range, like M31 for example. The bright core might be in the shot noise regime, but the outer arms will still be read noise limited. Want to get more arm signal? Longer subs. Beat the read noise. This is at the heart of the issue I feel. What's the difference? From the dark sky site, the camera has had to integrate for many minutes to beat the read noise floor. There is lots of signal. But the dark sky has lead to there being low noise- a high SNR. From the city location, just a minute had been required so there is not a lot of signal, couple this to a higher background count.- a low SNR

So science has brought us the telescope and the camera and the filters and the computers. It has given us the equations to accurately assess an image or set up and work out its failings or limiting factors. It can't however know everything about the atmosphere. Or telescope and filter combination transmission. So only practical demonstrations can reveal the truth for any imager. It's a thing that will vary from site to site. Olly with his vast experience has demonstrated the need for longer subs. This is the experiment that proves the theory. We know that you need to beat the read noise floor, the advanced imagers know that you need to beat the noise floor.

The scientific process is a very rigorous one. Take a theory, make an experiment and see if it matches. If yes, hooray, win Nobel prize. If not amend theory. Arguing that the data itself is at fault is not scientific method. The experimenters are good. Instead of arguing about the data, why not investigate what needs amending. Why does 100x1min not have the SNR of 4x25mins. What have the theory or you (perhaps) missed. Is there a scenario when equal integration times having equal SNRs is true?

Hope that was coherent.

Paul

[narrowbandpaul]

This is an article by Richard Crisp- he began the Narrowband technique amongst amateurs. He is also a scientist. This is a mathematical view of the issue, graphs included!

hope it is of interest

Paul

http://www.narrowbandimaging.com/images/exposure_equivalence_crisp_04162008.pdf

[Earl]

IMO its all target specific.

There possibly could be an equation worked out which would work out the point of a set point diminishing return threshold for the equipment, vs the overall luminosity of the target.

[JamesF]

There are all sorts of mathematical assertions made about imaging and not all of them are correct. One much discussed at the moment is the Nyquvist theorem. Praticising imagers with the right kit and the right experience are satisfied that they are disproving it and the pictures they are posting in support of that idea convince me. In the end I'm more interested in the pictures than the theorem but this might not describe everybody.

Do you have any references for this, Olly? Sounds like it might be something interesting to read up on. I have to admit that I'd struggle to believe that the Nyquvist theorem is wrong, though I might be persuaded that it is perhaps being misapplied or perhaps just isn't an appropriate tool in the case of imaging.

James

[ollypenrice]

Do you have any references for this, Olly? Sounds like it might be something interesting to read up on. I have to admit that I'd struggle to believe that the Nyquvist theorem is wrong, though I might be persuaded that it is perhaps being misapplied or perhaps just isn't an appropriate tool in the case of imaging.

James

Tim's yer man for this one. I don't image at massively below the apparent seeing, though Yves' scope must be beyond the theoretical limit. I don't have a different point of reference, though, since my next step is to being massively undersampled. However Tim posted some images which convinced him (and most of us) that there is resolution to be had beyond what the seeing would seem to allow. We should not let this hijack the thread, though.

Olly

[nmoushon]

Wow havent been on in a couple day and this thread EXPLODED! lol It's a good thing as I love reading the back and forth as I'm trying to learn as much of the technical/math side of AP.

But one thing I did notice with the back and forth is that my original post was directed at specifically the camera being a DSLR but most the comments are using CCD as their source instead. Though CCD and DSLR have there similarities there are a lot of differences, mostly in practice, that play a big factor in the short or long subs. Though it definitely applies to both the practice of using the two is much different. Really appreciate Lewis posting is picture that he obtained with his DSLR during the summer.

[ollypenrice]

Wow havent been on in a couple day and this thread EXPLODED! lol It's a good thing as I love reading the back and forth as I'm trying to learn as much of the technical/math side of AP.

But one thing I did notice with the back and forth is that my original post was directed at specifically the camera being a DSLR but most the comments are using CCD as their source instead. Though CCD and DSLR have there similarities there are a lot of differences, mostly in practice, that play a big factor in the short or long subs. Though it definitely applies to both the practice of using the two is much different. Really appreciate Lewis posting is picture that he obtained with his DSLR during the summer.

The big difference is from heat build up. You need to cool DSLRs for very long exposure but even a freezer physio gel pack will help. Also summer and winter may give different answers to How long?

Olly

[StuartJPP]

As other people have stated, this topic has been discussed a zillion times before. What we could really do with is having some clear dark nights for a change and actually do a practical test. Practical results are something that I can always believe, theory is just that, a theory.

I certainly don't believe that 2 minute subs are good enough, just try imaging M33! Plus there wasn't even any mention of aperture or focal length or...

In the work I do, there are two types of people, those who are practically minded and those that are theoretically minded. The theoretically minded produce a lot of PowerPoint Slides discussing the pros and cons to the nth degree with no actual deliverable content. Then on the other hand there are the practically minded people...they produce stuff that may not be the most 100% optimised solution but it gets the job done more than well enough, earns the company a fortune and we then move on to the next product/project. I fall into the latter camp.

[narrowbandpaul]

Stuart,

Sounds like there is no place in this world for theorists. There wasn't any mention of the scope used as for a fair test of a camera and sub length that would need to be the same. You dismiss science out of hand without seeming to know how it works.

There have been many theorists who have made great contributions. Einstein was one. Without GR corrections to time intervals GPS would not exist. Period.

[StuartJPP]

Stuart,

Sounds like there is no place in this world for theorists. There wasn't any mention of the scope used as for a fair test of a camera and sub length that would need to be the same. You dismiss science out of hand without seeming to know how it works.

There have been many theorists who have made great contributions. Einstein was one. Without GR corrections to time intervals GPS would not exist. Period.

So where do I say that I dismiss Einstein out of hand? What I dismiss are quasi-experts regurgitating things they have read on a forum somewhere, if it can be backed it up with empirical evidence then I will believe it. This is astro imaging with an end result, not quite in the same league as GR.

For the amount of times that this topic crops up (here and other forums) with no actual answers except theories perhaps I am a bit sick and tired of reading some people's theories and prefer to listen to the people that have actually tested it and produced something (either good or bad).

To the OP, get out there and take your best shot at it. In time you will learn what works for you and your setup and what doesn't.