Noise - ISO vs shutter speed

I'm trying to get a feel for noise in images. Although there are different sources of noise, if I'm correct, the big issue is signal-to-noise ratio. Scenario one - I am shooting a dark scent at 100ISO, and need 30 seconds of exposure. My concern is that with the shutter open for such a long period, it will result in significant noise in the image because of the length of shutter time. I could then change ISO to 200, and use 15 seconds, or 400 and use 7 or 8 seconds, etc. The shorter I go in shutter speed, the higher the ISO. At some point I will cross over to a point where noise is induced more from the higher ISO setting than the length of exposure.

My question is, at what balance point between ISO and shutter speed will I achieve the best (lowest) noise point? I know this will vary from camera to camera (sensor to sensor), but does anyone have any rules of thumb on this topic??

Just trying to glean from those more experienced in low light work, and start a conversation . . .

if you DSLR has long exposure noise reduction its not an issue.

the camera take a second "dark" exposure to determine the noise generated by the sensor heating up and then subtracts it from the original image.

IMO experience at 30s exposures at ISO100, i have yet to notice noise that is detrimental to my images, if i can even see it at all.

The more light you can let hit the sensor the cleaner your image is going to be. This is the rational behind ETTR, as it allows the maximum amount of light on to the sensor, without blowing wanted highlight detail. Yes you have to shoot RAW, but by using non expanded ISO's at the Base, 100, 200, 400 etc and ETTRing you will maximise the SNR and get the cleanest image possible.

For situations where you have to use very long exposures, that is not a problem, as others have said the Long Exposure NR works really well, and is applied to the RAW data too. The only problem is that it makes another "exposure" of the same length, so for a 30s exposure you also have to add the 30s dark frame so you are limited to one shot every minute.

Alan

Having a longer shutter speed won't increase noise, in fact, if you can pull it off and get a good exposure it will decrease noise!

tonylong wrote in post #15756192
Having a longer shutter speed won't increase noise

But it does. Each and every pixel has a small, but non-zero, probability of recording a false signal. The longer your exposure, the higher the probability of a pixel creating a false signal and the more pixels with false signals (noise) will appear in your image.

If exposure times didn't matter then Canon's Long Exposure Noise Reduction wouldn't need to take a dark frame with the same expsosure time.

It's two different types of noise. Long exposure noise is a fixed pattern noise that is specific to a sensor and sensitive to temperature (photodiode leakage currents, wikipedia calls it - although I'm not sure exactly about the physics behind it); high ISO noise is the result of the amplification of the random pattern read noise.

The former can be essentially removed with dark frame substraction; the latter can be removed by stacking and noise averaging (or, you know, using lower ISOs). Astrophotographers will tend to do both.

As such they're not really equivalent in terms of striking the perfect balance. I guess there's still an equivalence in terms of signal-to-noise, but honestly I'm not sure where it is. IMO the question is really driven by the specificities of your shot: if you need a long-exposure, low-ISO noise you only really need to bother with a dark frame or ten. If you're taking shots of moving objects, you can't really avoid ramping up the ISO to the max. If you have a choice, then I guess it depends on which noise-removal technique you intend to use, though I'd argue that strong long-exposure noise is easier to remove than strong ISO noise (but weak LE noise is probably harder to remove than just using a denoising plugin/slider in your preferred post-production software).

hollis_f wrote in post #15756514
But it does. Each and every pixel has a small, but non-zero, probability of recording a false signal. The longer your exposure, the higher the probability of a pixel creating a false signal and the more pixels with false signals (noise) will appear in your image.

If exposure times didn't matter then Canon's Long Exposure Noise Reduction wouldn't need to take a dark frame with the same expsosure time.

Ah, yeah, I think it was too close to my bed time, I wasn't thinking "long exposure" noise and I should not say much when it gets later at night!

I have seen someone experiment with the number of exposures vs. exposure time (don't remember the site now), but I don't think I've seen the test you are talking about. But really you have your camera, why don't you do your own experiment? I also agree that 'hot pixel' noise is easy to remove in PP which would argue that you should just use the lowest ISO and longest time, but I have not had the best results doing this in night exposures.

If I remember correctly the conclusions that I have seen about taking 1 long exposure vs several shorter ones concluded that about 2min exposures were optimal - at least that is the ballpark figure I (think I) remember. Ideally you'd let the sensor cool down between exposures as well, but in reality once you get beyond a few minutes, even the moon can move significantly, so I just try to get the entire exposure done as quickly as possible and get as much light onto the sensor as possible. From night exposures I have taken, my gut feel is that 2-5 minutes is optimal - i.e set the ISO for that time frame - and then stack exposures if you can from there. I could be way of base, but this is what I have gravitated to with the 5DII.

Just as with HDR, several exposures can also help you choose how much exposure to apply to different parts of the image. For this reason alone I usually try to split up exposures of more than several minutes, so my estimate about could be biased my this.

I have also stacked short (1/15s) exposure shots taken at high ISO of a static subject (dark and needed a reasonable shutter speed). I put the camera in burst and took 6 shots in a burst, remove any with blur and stack the rest. It helped considerably. I had to be careful to do NR in the right order - I think I applied a bit pre-stack, but most post-stack. i know not really the question though.

So a long reply that doesn't answer your question, very useful of me

An interesting question though and one I would have thought would have had more of a general consensus of what to do. I might actually test this myself this weekend so I have more of a plan for my next night exposure, rather than winging it to a great extent.

Thanks for all the great replies everyone. My purpose was to generate just such a discussion, since technique is as much of a talent in advanced digital photography as is composure and all the other artistic elements. Keep the comments coming, since it is an area where we can all learn a bit from everyone else . . .

OK, I finally got around to my first (maybe my last) test.
First was ISO100 @ 2min and all other multiples of ISO. ISO100 was clearly the best. So certainly for shots up to a few minutes, you are best off with the lowest ISO.

Now for longer exposures….

ISO400 @ 165s (left) vs. ISO100 @ 662s (right). These are with long exposure NR off. This was at room temp (~70F). The image was increased in exposure by 2 stops in PP to show the shadows, where we would expect to see most noise. Obviously no NR or sharpening.

The ISO100 has less random noise, but the hot(ish) pixel color noise is more evident. I'm not sure what stacking the ISO400 shots would do, but the color noise on the ISO100 shot is not that easy to remove (no chance in ACR). They both have very hot pixels which are much easier to remove.

I think I’d prefer the ISO400 version in this case

Now with NR on.

ISO100 looks pretty good, better than ISO400. So Long Exposure NR helps with a lot of the more subtle color noise, not just the super-hot pixels.

easy,

high iso speeds produces noise.
slow shutter speeds give more time to the camera to produce noise.

however, one stop of higher ISO speed produces more noise than one stop slower of shutter speed

Thanks for the testing! Great lessons here.

ejenner wrote in post #15799040
Now for longer exposures….

ISO400 @ 165s (left) vs. ISO100 @ 662s (right). These are with long exposure NR off. This was at room temp (~70F).

Room temperature won't be that important. Much more of a factor - did you give the sensor a chance to cool down between exposures? If you did the ISO100 experiment immediately after the ISO400 experiment then the ISO100 test would have been done with a sensor that had already been cooking for 165s.

Not sure how long it takes for a sensor to warm up or cool down enough to make a difference - a couple of minutes both ways would be my guess from what I've read (not much) in the video fora.

hollis_f wrote in post #15799542
Room temperature won't be that important. Much more of a factor - did you give the sensor a chance to cool down between exposures? If you did the ISO100 experiment immediately after the ISO400 experiment then the ISO100 test would have been done with a sensor that had already been cooking for 165s.

Not sure how long it takes for a sensor to warm up or cool down enough to make a difference - a couple of minutes both ways would be my guess from what I've read (not much) in the video fora.

I did consider it (I guess it is part of the whole point) which is why I did the long 11min exposure after the shorter one rather than the other way around. I did leave a few minutes between them and I left 10-15 mins between the without Long Exposure NR and with.

Actually thinking about it I have Magic Lantern which actually gives you the sensor relative temperature, so I could have used that to check - but didn't.

In the previous experiment I went from ISO6400 to ISO100 at 2 mins without a break and the ISO100 was still clearly better, so I figured if the sensor was heating up then it wasn't the major factor in a 1-2min exposure.

I had noticed better long exposures on my T1i when out in the cold (say 10F) than in the house, so I figured I'd mention that it was done in a 'reasonable' temp environment.