Stargazing What are the limitations of astrophotography and how can they be overcome?

[Drakkith]

I'm afraid that's wrong, too. Increasing exposure time does not increase SNR! What increases SNR is obtaining a larger number of independent observations of your desired signal -- i.e. short sub-exposures -- so you're not integrating much of the noise in each sub-exposure. When you average N sub-exposures, the SNR increases as \sqrt{N}.

Increasing the exposure times increases the signal gained. In the case of light pollution it also increases that, but either way the end result is the same. If your sky is 10,000 counts/min and your object is 1,000 counts/min, then doubling exposure times doubles both the noise and the signal. (I think because light pollution isn't normal noise, it is unwanted signal that increases linearly.) However, because the skyglow and the signal add together on your chip, I THINK you can subtract it with long enough exposures or more exposures. The readout noise, shot noise, dark noise, etc, all increases at the square root of the exposure time, not linearly, so longer exposures and more of them will definitely increase image quality.

Consider this: After 1 minute the pixels that get both the signal and the skyglow are at about 11,000 with the background at 10,000. (With the standard noise from dark current, readout, ETC) Up the exposure to 5 minutes and you will get on average a value of 55,000 for the object and 50,000 for the background. So your SNR between the object and the background is the same, but the difference is now 5,000 between the object and the background, giving you much greater range.

Averaging multiple images keeps the range the same, but it cuts back on shot noise, which is the random nature of the photons from the signal object. So you want to get BOTH exposure time and multiple images.

If I could, I'd build a rig that could take a million 1/4000th second sub-exposures, and it'd peer right through the densest pea-soup light pollution on Earth.

After combining the images the resulting image would be of similar quality of a hundred 4 second exposures or 10 forty second exposures. Worse actually, as at that low of an exposure time your images would be overwhelmingly dominated by noise. While you can subtract most of it you would never be able to get all of it.

 

[chroot]

Increasing the exposure times increases the signal gained. In the case of light pollution it also increases that, but either way the end result is the same. If your sky is 10,000 counts/min and your object is 1,000 counts/min, then doubling exposure times doubles both the noise and the signal.

Right -- if you double the exposure time, you double both the signal and the noise, so your SNR stays the same and you gain nothing. Sure, you get a brighter image, but you're no more able to separate the signal from the noise, and that's what astrophotography is all about. At some point you'll saturate your sensor, and that'll be it. You're toast.

Correct me if I'm wrong, but I think you agree that http://en.wikipedia.org/wiki/Shift-and-add" , but the targets must be sufficiently bright.

However, because the skyglow and the signal add together on your chip, I THINK you can subtract it with long enough exposures or more exposures. The readout noise, shot noise, dark noise, etc, all increases at the square root of the exposure time, not linearly, so longer exposures and more of them will definitely increase image quality.

Some of these sources of noise are fixed (the same for any exposure of any length), like readout noise. Others, like dark current, are actually linear with exposure time.

Consider this: After 1 minute the pixels that get both the signal and the skyglow are at about 11,000 with the background at 10,000. (With the standard noise from dark current, readout, ETC) Up the exposure to 5 minutes and you will get on average a value of 55,000 for the object and 50,000 for the background. So your SNR between the object and the background is the same, but the difference is now 5,000 between the object and the background, giving you much greater range.

Again, the ratio of your signal to your noise remains constant, so you've not gained much. You've spread your signal over a greater number of codes, so you've reduced the quantization noise, but it comes at a tremendous cost. Those long exposures will put a much greater burden on your mount (which has tracking error) and your seeing conditions (which change constantly). Very few amateur telescopes have tracking error that is acceptable for even one minute unguided exposures.

So, here's the bottom line: you definitely want to stack exposures to increase SNR and dynamic range. This method is used to great effect by professional astronomers all over the world. You want your exposures to be short enough to eliminate any visible tracking or seeing error, but not so short that they become dominated by the readout or quantization noise of your sensor. Somewhere in between, there's a sweet spot that will produce the optimal output. For me, with my equipment, that sweet spot is in the 10-30 second range. I hope that my photographs speak for themselves, and demonstrate the advantages of the method.

- Warren

 

[Drakkith]

Correct me if I'm wrong, but I think you agree that http://en.wikipedia.org/wiki/Shift-and-add" produces better SNR (and therefore better final photos) than individual long exposures.

To my knowledge it does not until your total exposure time with the stacked images exceeds the longer exposure images.

Some of these sources of noise are fixed (the same for any exposure of any length), like readout noise. Others, like dark current, are actually linear with exposure time.

Yes, and one of the reasons you need to get long exposures or stack many images is because of shot noise correct? I thought that because of the randomness of shot noise a long exposure will end up just like multiple short exposures will if the total exposure times are the same.

Again, the ratio of your signal to your noise remains constant, so you've not gained much. You've spread your signal over a greater number of codes, so you've reduced the quantization noise, but it comes at a tremendous cost.

How have you not gained much? Your shot noise has dropped thanks to the extra signal. I'm just wondering how effective it is vs the increased background noise.

Those long exposures will put a much greater burden on your mount (which has tracking error) and your seeing conditions (which change constantly). Very few amateur telescopes have tracking error that is acceptable for even one minute unguided exposures.

Sure, but let's leave that for another discussion.

So, here's the bottom line: you definitely want to stack exposures to increase SNR and dynamic range. This method is used to great effect by professional astronomers all over the world. You want your exposures to be short enough to eliminate any visible tracking or seeing error, but not so short that they become dominated by the readout or quantization noise of your sensor. Somewhere in between, there's a sweet spot that will produce the optimal output. For me, with my equipment, that sweet spot is in the 10-30 second range. I hope that my photographs speak for themselves, and demonstrate the advantages of the method.

I pretty much agree with you, however assuming you have good tracking and guiding, 5-10 minute exposures seem to be just as good as stacking ones of lesser exposure time. Heck, you're still going to stack the 5-10 min ones anyways. The advantage of shorter exposures compared to long ones (5-10 min exposures compared to 1 hour long) are that you can get rid of any frames that have satellites go through them, had a mount error, etc. That and it keeps saturation from occurring.

My only question is will getting many exposures of something help to counteract light pollution or not? I think so, but I'm not sure.

 

[Drakkith]

Excellent photo by the way Chroot!

 

[Andy Resnick]

<snip>

So, for many reasons... stack, stack, stack, my friends. You still need to nail your focus, but stacking can solve literally every other problem you're likely to encounter.

- Warren

I understand- that makes a lot of sense. I also found a site discussing stacking, and they discussed the rule of diminishing returns- for example, if my SNR goes up by (say) 10 by stacking 50 frames, I'd need to stack 500 (or something like that) to get up to 20, 5000 to get an SNR of 30, etc.

My main limitation is the lack of a tracking mount- my exposures have to be short enough to prevent motion blur using a high camera gain. This is also ok, because like any other noise, it's statistical and washes out with stacking.