ISO 1600 v ISO 800

[earth titan]

Well I used to use ISO800 and then switched to ISO1600 on my modded 1000d with a Instantly improved my images. :-)

However I'm interested in others views regarding noise, exposure times, benefits v negatives etc. I've heard two schools of argument - one which suggests it doesn't increase noise, the other that it does.

With sufficient darks, will I remove noise?

Should I drop back to 800 and increase exposure?

Typed by me on my fone, using fumms... Excuse eny speling errurs.

[Ags]

In my limited experience, the higher ISO can work for faint stuff - whereas ISO400 might only give you 10 grey levels, ISO1600 might render the same object with 40 levels.

Yes a higher ISO does push up noise, but I find the noise in my 1100D similar between ISO800 and ISO1600... also, the higher ISO lets you shoot more subs, so that also helps control the noise.

[GlassWalker]

Trick is not to think of this like daytime imaging. In that case, for a given image "brightness" a higher ISO means shorter exposure time, all else being equal. Shorter exposure means more noise, but the benefit is shorter exposure.

For deep sky astro, you simply need sub exposure time to get the faint stuff out there. ISO setting does act as a gain, and it can help pull signal out of some of the low end noise, but at the same time, it can amplify some other noise types too. Plus a high ISO reduces dynamic range.

Darks can help with some types of fixed pattern noise, but not all. Best way to remove random noise elements is to have more exposures for stacking.

I've standardised on ISO1600 myself... for faint stuff at least, where I'm not in danger of running out of headroom.

[nmoushon]

Shooting at iso 1600 will drastically make your single sub 'pop' more than an iso800 sub of same exposure length. It is beneficial to shoot at iso1600 if you dont have guiding capabilities and cant do 5min subs. But my view on this is that iso800 has much less noise than 1600. Its get more prevalent in longer exposures (3-5min+) while in shorter exposures its not as bad. But the longer your exposures and the longer you imaging session is the noise just keeps increasing. From my experience that is. On brighter objects iso800 is better because you just dont need to shoot at 1600 and why introduce more noise if you really dont gain anything from it. LP plays a factor in this as well. But even though I dont guide with my set up I still only use iso800. Call me picky but Id rather take an extra night to get more subs than have to deal with noise in the post processing. Especially since my PS skills arent that good.

[GlassWalker]

nmoushon, when you say noise is lower at ISO800, is that for the same sub exposure duration as ISO1600? I've experimented before, and for the same exposure time at either ISO, the noise levels were practically identical once you're stretched them to the same output brightness. We're looking at optimal signal-to-noise ratio here, and the sub exposure time is more important there.

[nmoushon]

I did my own experiment with my 450D at iso 800 and iso 1600. I took 30x60sec exposures of M42 at iso800 and a separate set at iso1600, both with 20x60sec darks. I did two sets of stacks for both sets. One with the darks and one without the darks. Too see how the difference in noise was in the lights and how much darks decreased it by. I stack both in DSS with same setting and PS both. Just levels and curves and tried to get the stretching the same for both as best I could. Not super proficient in PS by the way so could be some slight skewing in the results but not much I think. This was a while ago and don't remember exact details but I remember that from my results I concluded that iso800 kept the noise down better than iso1600. And I've used iso800 ever since.

Note that this was just a noise test and not an overall signal test. The amount of data I got would not be sufficient enough for a test to compare noise AND signal. Obviously the iso1600 image was 'brighter' than the iso800 image but it was noiser. Also a disclaimer at the time I lived in the midwest where temperatures change a lot and had a good amount of LP. So that could also have effected the results.

Now since this brought back this memory I think this winter I will have to test it again. This time go for much much more total data and redo my experiment. Thanks for giving me another winter project

[earth titan]

So if I understand it correctly, the longer exposure time causes noise as does the higher ISO.

Therefore it's also almost like a 'dose' of noise. As a result, both speeds might generate the same effective noise to generate the same 'brightness' of DSO. Under UK skies then, where more subs reduce noise, a higher ISO makes sense with the kind of weather we get!

Typed by me on my fone, using fumms... Excuse eny speling errurs.

[Ags]

Perhaps I am misreading, but I would have thought the test would be 30 x 120sec @ISO800 vs 60 x 60sec @ISO1600 - double the ISO should double the number of exposures.

[Ags]

I have always wondered if the sensor gets hotter is the ISO is higher... anybody know?

[GlassWalker]

nmoushon, sounds like you did a deeper test than I did, as I was only comparing individual subs.

Ags, as far as I know the sensor heating would be the same for a given exposure duration regardless of the ISO, but I never tested it...

[nmoushon]

Tom - Noise, from my understanding - im no expert here comes from your camera heating up and adds noise/hot pixels to your image.(This is the main cause) Your camera heats up from you just using it. Long exposures and long use times heat it up more and faster. Outside temp pays into that as well. Which is why there are mods to cool you DSLR and there are cooled CCD cameras. These two thing add noise to your image. So long exposures don't really add noise to your image its the heat generated from the long exposures that does. But there is the signal-to-noise ratio that comes into play (which is really the driving factor in AP) and thats a whole other animal. But if you are looking ONLY at noise higher iso adds more noise to your image. But it adds more signal to your image as well. So you have to find the best ratio of signal-to-noise. Since I'm not from the UK I cant really comment on the weather and how that would effect it, if at all.

Agnes - When you said doubling the ISO should double the number of exposures is not right. Doubling the ISO would cut your number of exposures in half since your doubling the signal gained. So if I took 30x120sec iso800 subs I would only need to take 15x120sec iso1600 subs. From what you said at 30x120 iso800 vs 60x60 iso1600 is not quite right because you doubled the number of exposures but halfed the exposure time so you would actually get less signal with the iso 1600 than the 800. If you wanted to match the signal your numbers should be 30x120sec @iso800 and 15x120sec @iso1600 or 30x60sec @iso1600. And if I was testing for signal-to-noise ratio then I would have needed to that but I was testing just how iso effected noise gain on a single sub and a stack of subs and how much darks effect noise reduction. So I needed to shoot the exact same length in exposure to have a control and correctly compared test of iso.

[nmoushon]

Glass - I'm not sure how exact my experiment was back then or how controlled it was. I was still very new to AP at the time. I do need to do another, more controlled test, this winter under similar circumstances. Would like to know if the same results are multiplied, stayed the same, reduced, ect. with longer subs and longer total exposure time.

Agnes - You snuck that question in while I was typing lol. I don't have a solid answer on that but I don't think ISO effects the heat gain on a camera. Could be wrong but I don't think so. ISO and heat both contribute to the total noise but are separate-from my understanding.

[RikM]

I find ISO1600 better with my unmodded 1000D for faint stuff and I have tested it a few times.

For my last proper comparison, I took a bunch of subs of the Monkey head at ISO800 and ISO1600 and found that I didn't have to stretch the ISO1600 image as hard to get the brightness I wanted. By needing to stretch the ISO800 image further, it increased the noise. These were all 5 min subs but after processing there was no real difference between the sets and the processing was easier on the ISO1600 set, so I use that as default.

I only drop the ISO if shooting clusters, because stars are bright and it keeps the colour a bit better. Not much in it though.

I do have a fair bit of sky glow but no streetlights in direct line of sight. I am sure this is a factor in my results.

[nmoushon]

Rik brought up a good point I forgot to mention.

I've only ever shot bright objects because of my bad LP where I live. So I have not tried it on fainter objects. Guess I'll have to do 2 experiments now - one for bright and one for faint objects. Oh what fun this winter will be! Now just hope for clear skies and free weekend!

[Luke]

My own not very scientific experience with the Canon 450D is that ISO800 vs ISO1600 makes no real difference unless you are very much underexposing or overexposing, that the higher ISO is just a gain applied, and that although the 1600 looks more noisy, it is brighter, and when I brighten up the 800 to the same level as the 1600 when editing, it's just as noisy.

My approach is that the most important thing is the length of the sub, and then I pick the highest ISO that will give it a decent amount of exposure without overexposing it. As said, not highly scientific and would be really interested to see what others think and to see any tests!

[Ags]

Tom - Noise, from my understanding - im no expert here comes from your camera heating up and adds noise/hot pixels to your image.(This is the main cause) Your camera heats up from you just using it. Long exposures and long use times heat it up more and faster. Outside temp pays into that as well. Which is why there are mods to cool you DSLR and there are cooled CCD cameras. These two thing add noise to your image. So long exposures don't really add noise to your image its the heat generated from the long exposures that does. But there is the signal-to-noise ratio that comes into play (which is really the driving factor in AP) and thats a whole other animal. But if you are looking ONLY at noise higher iso adds more noise to your image. But it adds more signal to your image as well. So you have to find the best ratio of signal-to-noise. Since I'm not from the UK I cant really comment on the weather and how that would effect it, if at all.

Agnes - When you said doubling the ISO should double the number of exposures is not right. Doubling the ISO would cut your number of exposures in half since your doubling the signal gained. So if I took 30x120sec iso800 subs I would only need to take 15x120sec iso1600 subs. From what you said at 30x120 iso800 vs 60x60 iso1600 is not quite right because you doubled the number of exposures but halfed the exposure time so you would actually get less signal with the iso 1600 than the 800. If you wanted to match the signal your numbers should be 30x120sec @iso800 and 15x120sec @iso1600 or 30x60sec @iso1600. And if I was testing for signal-to-noise ratio then I would have needed to that but I was testing just how iso effected noise gain on a single sub and a stack of subs and how much darks effect noise reduction. So I needed to shoot the exact same length in exposure to have a control and correctly compared test of iso.

I think you misunderstand me. I am figuring you have a fixed amount of imaging time - let's say an hour. The question is not how long you are going to image for, but how you will fill that time - the total imaging time is a constant. So the effect of ISO is 30x120@iso800 vs 60x60@iso1600. What I am keeping constant is (1) total imaging time and (2) brightness of the individual subs.

[VigdisVZ]

I have always wondered if the sensor gets hotter is the ISO is higher... anybody know?

ISO is a form of post processing that cranks up the "volume" inside the camera after it has captured the information. Therefore it should "in theory" not affect sensor temperature. Or that is my understanding.

[jimmyjamjoejoe]

Don't canon state an "optimal" of 1600 on their newer cameras? I've always assumed lower iso = less noise, no matter what, but that word "optimal" made me think otherwise.

[Ags]

ISO is a form of post processing that cranks up the "volume" inside the camera after it has captured the information. Therefore it should "in theory" not affect sensor temperature. Or that is my understanding.

I believe on the contrary that ISO is gain applied on the chip - that it happens at the hardware level, not in post-processing.

[nmoushon]

I think you misunderstand me. I am figuring you have a fixed amount of imaging time - let's say an hour. The question is not how long you are going to image for, but how you will fill that time - the total imaging time is a constant. So the effect of ISO is 30x120@iso800 vs 60x60@iso1600. What I am keeping constant is (1) total imaging time and (2) brightness of the individual subs.

You're right I did misunderstand a bit. But what I said still applies.

This isnt completely accurate but without getting into a lot of math that I cant even do lets go with the common iso1600 is double the 'brightness' of iso800. What you are doing with what you stated is comparing individual subs at equal amount of signal gain. 1x120sec @iso800 = 1x60sec @iso1600. Which in AP the total exposure time trumps individual exposure time in regards to signal-to-noise. Exposure length determines the depth of your image. So restraining the test to 1 hr is not an accurate test. If you are trying to test for noise then they need to be equal subs at equal lengths (which are the constants) and your variable is then the iso.

But if you are trying to figure out how best to divide up that hour it comes down to your target. Using the same exposure and number on M42 will not be the best for M27 because the objects have different surface brightness. Since total time is 1 hour - M42 you could get all the faint faint nebula with 30 or 60 sec exposures while with M27 you would need minimum of 5min exposure (or more) to get the faint outer shell. 1hour of 5min on M27 would get the outer shell while 1hour of 1min would not. This is the example of how exposure length adds depth.

[Ags]

I think differently - there is a school of thought strongly in favor of ridiculously short subs, where the depth of the image is related only to total integration time, not individual sub length. I think the truth must lie somewhere between your position and the super-short sub school of thought.

For my imaging, I only get one or two hours to shoot a picture so the optimal use of that time is the question I want to answer. Of course you also need to take into account the brightness of the target, as you would for any photography.

[nmoushon]

I think differently - there is a school of thought strongly in favor of ridiculously short subs, where the depth of the image is related only to total integration time, not individual sub length. I think the truth must lie somewhere between your position and the super-short sub school of thought.

For my imaging, I only get one or two hours to shoot a picture so the optimal use of that time is the question I want to answer. Of course you also need to take into account the brightness of the target, as you would for any photography.

I don't think I've heard of the super-short sub theory. You mean like 30sec exposure short or even shorter?

I hear you. I don't get much time either and it very far between. Last time I was able to go out was 2 weeks ago. Time before that was 6months! Combo of weather, work and having to drive 1-2hrs to get out of Chicago LP and only getting Friday or Saturday night to shoot. So finding the best combo is high on my list of figuring out. When winter and long nights hit I will do some serious testing.

[Ags]

Have a look at what this guy says:

http://www.samirkharusi.net/sub-exposures.html

Last time I got out was three weeks ago when I shot the Cygnus region - I'd been promising myself that session since June. And that session was clouded out for the most part, and the 8 subs I salvaged were shot through thin cloud. Aaaaaargh!

[nmoushon]

I'll have to read that through. Thanks for the link.

[allcart]

Ooh an interesting topic for discussion. I have read Ags link and do understand what his theory is, and it makes sense. But I am also interested in nmoushon's thoughts about longer exposures giving more depth to the image.

I'll have to wait until he reads the link and composes a reply.