In the previous article about the best ISO setting for astrophotography, I’ve explained what ISO is, how it works in digital cameras and how to determine the best ISO to use for your DSLR.
However, theoretical discussions aside, the question is of course if we can even see the difference in reality? During my stay in Namibia I did a few ISO800 vs ISO400 tests, so I’ll share the results in this post. Please note that these are just examples to show the differences between the two ISO settings. At that time I didn’t have the level of understanding of the concept of ISO that I have now, or I would probably have done some different tests and for my Nikon DSLRs go all the way down to ISO100. However, if you are anything like me and it feels very counterintuitive to you to be using real low ISO settings for astrophotography, it will be quite interesting to see some examples that show the differences. Hopefully this will be another step that will help convince you to really look for the best ISO setting for your camera, even if this is a (very) low ISO setting.
Nikon D7000: ISO 800 vs ISO 400
When I was on my astrotrip to Namibia I did do a few shots on the globular cluster Omega Centauri. I took a few exposures at ISO 800 with 150 sec exposure time and a few shots at ISO 400 with 300sec exposure time. As you can tell from this, I was still ‘stuck’ in thinking in terms of the ‘exposure triangle’ and doubled the exposure time when I halved the ISO. But remember; the exposure triangle is a flawed concept and we really should also be considering the ISO as a setting on it’s own. Unfortunately I didn’t do that on this object, so let’s just look at the differences;
That is quite convincing. ISO400 is basically outperforming the ISO800 exposures in all areas.
But those are just numbers. Can we see the difference when we look at the individual frames side by side?
Please note that all the images below are stretched using the ScreenTransferFunction function in PI.
Hmmm, I can not really see a difference here. (Apart from the focus, which is not due to the ISO setting of course) But this is a difficult area of the picture to do a comparison on, let’s zoom in a bit further and look at a faint galaxy in the background;
The noise difference is hard do judge visually I think, but what becomes strikingly apparent is the brightness and definition of the faint stars. They are much brighter in the ISO400 picture. Intuitively this might still be strange, but considering the fact that the read noise for the D7000 is pretty much consistent right from ISO100 it completely makes sense. Basically you are capturing twice the amount of light, so sure the faint stuff becomes more apparent and brighter!
As said it would be better to look at the differences for exposures of the same length. So let’s look at a few exposure tests I did on the Helix nebula; ISO 800 with 600sec exposure, ISO 400 with 600sec exposure and ISO 800 with 300sec exposure.
First let’s look at the numbers:
Wow, look at the Noise, that’s a huge difference! And looking at the SNRWeight we can tell that we didn’t loose as much signal so we gained quite a bit of SNR by using ISO400 instead of ISO800, even if we compare it to the ISO800 with the same exposure time. The star support difference between 300sec and 600sec is obvious, the loss of 111 stars going from ISO800 to ISO400 might be statistical and not significant as it is only 2.23%.
Now let’s see if we can see the SNR difference in the images visually as well;
Hmmm, again the difference is not that obvious visually.
Maybe if we zoom in a bit more;
Well, the difference between the 300sec and 600sec can be clearly seen. However, the difference between the ISO800 and ISO400 in the single image is hard to tell.
It is clear that in order to get the faint signal, exposure is key!
Nikon D600: ISO 800 vs ISO 400
For my image of the Sagittarius Triplet I did a quick test for my 2nd and 3rd runs and upped the exposure time from 300sec to 600sec while dropping the ISO from 800 to 400. It would have been better to test both ISO’s at 600sec, but I didn’t know yet that at the time 😉
So let’s work with what we’ve got and see if we can spot the differences;
We see the same picture again; Noise is lower while SNR is improving as we lower ISO. This is clearly indicating that we are in the area where upstream read noise is swamping the downstream read noise, because otherwise we would have seen the reverse (SNR improving as ISO is higher).
So most likely we will have to go even lower for the optimal ISO for the D600.
Again, visually it is hard to see the difference. Only if we zoom in real close we start to see the difference. But remember, since we are in the ISO-less zone for the D600 at these ISO’s, we are really comparing exposure lengths and not so much the ISO settings.
Conclusion on ISO800 vs ISO400 comparison
For both the D7000 and the D600 it is clear that we are better of with using ISO400, even though it might not always be that clear visually. The numbers don’t lie, and we have to remember that we are comparing single subs here. The difference will become stronger when we start stacking, especially the dynamic range difference in terms of picking up the faintest signals.
Next time I’ll do a test on ISO100 and ISO200 versus ISO400, to see if I have to go all the way down to ISO100 on the D7000.
Interesting article! Keep up the good work!
Nice article. However, the tests *might* be inconclusive as you did not account for the thermal noise. This will definitely play a major role in long exposure according to many astrophotographers. If we take the thermal noise into account, I think we will have different optimum ISO’s at different exposure times?!?
This might be counter intuitive; but ISO has no impact on the thermal noise, other than the possible amp glow.
ISO is just amplifying the thermal noise that is already there, just as it amplifies the signal. So there is no impact on SNR.
Thermal noise is dependant on the exposure time, where longer exposures means more build up of the thermal noise. (and of course the temperature of the sensor)
Interesting, but a bit as expected of course. Unfortunately, the longer I expose the more bright airplanes I will get in a single frame 🙂 It’s a catch 22.
Thanks for all the great information and explanations. In your comparison of ISO values (Helix nebula) you conclude that “exposure is key”. Did you actually mean that integration time is the key? From your ovn explanations ISO does not increase sensor sensitivity. So 90 frames at 60 sec. should capture just as many photons as 30 frames at 180 sec.
Because of the above I miss a comparison (of noise values in particular) where two ISO 800, 300 sec. frames are stacked. Then we would have the same integration for all the comparisons.
I tried to do that math myself and got 3,58. If that is correct then stacking two such frames gives a better result than one ISO 400, 600 sec. exposure.
If I’m wrong, please explain why and why.
Excellent questions!
I did mean “exposure is key” in this article, since I was solely focussing on single exposure comparisons since you will only increase those results when you start stacking the same number of frames.
Whether stacking is 100% equivalent to adding exposure time (like you stated with the example of 90 frames of 60sec = 30 frames at 180 sec) is a whole different discussion and other factors will come into play like guiding ability and consistency, the time you have to spent on target, clouds passing by etc. etc.
As a rule of thumb I always try to get 25+ frames, and will increase exposure of single frames as much as possible based on the available time. This will make sure you make good use of the noise cancelling benefits the stacking process gives you.
Hope this helps and answers your questions
Hi and thanks for your swift reply. So in theory at least I’m right, all other things being equal. Final SNR after stacking will be the same for the two examples I gave above. (I’ve also verified that from other sites). The benefit of doing short frames is that for each frame that can’t be used for any reason I loose less integration time. And it requires less from the mount, guiding etc. Downside is that very faint signals may not appear in all frames and may be considered as noise by the developing software. But I suppose that is dependant of the quality of software.
I think I”ll make an effort to calculate an optimum frame time based on my local sky conditions. But I don’t know how to do that just yet. Work for a rainy day.
For now I think I’ll continue using short frames and a high number of them. Until somebody convinces me otherwise.
You mention all the right things to consider. I do want to emphasize the fact that you’ll need at least a few photons from faint signal in order to capture it in your photo, so longer exposure teams will be appropriate depending on the target.
I’ve seen a formula in the past that should calculate the best exposure time, also based on local sky conditions (light pollution), I’ll try to find it again and post it here 🙂
It’s always better to stack more images to reduce the noise. But for the exposure time it’s all depends from in what area you live. If you are in the red or white zone shorter your exposures due to the light pollution. In darker places the longest exposures are better.
Good job! Thanks from. SPAIN