2

Assume the case of a Canon 6D Mark II. Its lowest ISO is regulary 100. If expanded, I can set it to L, which is ISO 50.

Until now, I was always under the impression the lower the ISO the cleaner the image, so whenever possible I've been using ISO 50. Now I came across this chart from photonstophotos.net:

chart

And that leaves me completely confused. According to this chart, I have a lower noise at ISO 100. ISO 50 seems to have a higher noise than ISO 300 too. Is that anywhere near correct? As long as I would not be clipping highlights, couldn't I just use ISO 160, 300 or 600 instead of lets say 50, 200 and 400 (depending on my needed exposure) and later drop down the exposure in post to get a cleaner image?

I've seen this question about ISO 50 but the two top answers are kinda contradicting:

Since your camera offers this 'expanded ISO" that provides for 80 ISO, you can assume that this ISO is sub-optimal, and could exhibit more noise or a loss of dynamic range than the 'native' ISO. [...] Some suggest Canon cameras are 'native' for ISO 100, and full stop ISO are best (100,200,400 etc.).

Vs.:

You can use it and it gives excellent dynamic-range and very low image noise but really barely any different from the ISO 100 setting.

The chart above would suggest the first answer is right, but that definitely doesn't hold true for the latter part of the quote, as ISO 160, 300 and 600 seem to be the best choices.

Can someone tell me if I interpret the chart correctly, and if ISO 160, later darkened in post, will really give me cleaner, less noisy images in the end than using ISO 50 (assuming that would be the perfect exposure) in the camera right away?

confetti
  • 881
  • 1
  • 8
  • 23

2 Answers2

3

Canon has been doing this with the +1/3 and -1/3 stop ISO settings since at least the original 5D back in late 2005. The last EOS DSLR that did not appear to do it was the APS-H EOS 1D Mark IIN introduced in mid-2005. The next 1-series cameras, the APS-H EOS 1D Mark III as well as the FF EOS 1Ds Mark III that were introduced in 2007 both demonstrated this as well.

There's an extended answer regarding how all of this works out in the accepted answer and the comments following it at:

Is it really better to shoot at full-stop ISOs?

Rather than copy/paste that entire answer here, I'm voting to close this question as a duplicate. Although the questions aren't exactly the same, the answer to both is.

The only additional information that might be relevant is about ISO 50. It is a "virtual" ISO that uses the sensor amplification set at ISO 100 and then "pulls" the exposure one full stop when the raw file is converted, just like ISO 160, ISO 320, ISO 640, etc. "pull" exposure by one-third (1/3) stop from the sensor amplified for ISO 200, ISO 400, ISO 800, etc. Even if you use a third party raw conversion application, the EXIF info attached to the raw file will let the app know to apply the exposure adjustments.

With ISO 50, the effect of the "pull" in development is to reduce the brightness of the entire picture, including the shadows where noise tends to be most noticeable, by one full stop. It also reduces the highlights by one full stop. So any areas that are right at the clipping point in the raw file (which is probably a stop or two brighter than what could fit into a jpeg with typical gamma and contrast curves applied) are also reduced by one full stop.

Michael C
  • 175,039
  • 10
  • 209
  • 561
  • I'm sorry, I haven't seen that other question. It does explain most of the questions I had though, but just to make sure I'm not missing anything: That means I can basically take a shot at ISO 640 (which will be overexposed by 2 2/3 stops but no clipped highlights), bring the exposure down later in post by 2 2/3 stops and have the same image but cleaner and with less noise than if I would've taken the same shot at ISO 100 to begin with? – confetti Apr 02 '19 at 19:23
  • 1
    @confetti Not exactly. If your 6DII is set to ISO 640, the sensor is actually being amplified at ISO 800, with an instruction to the raw developer to initially pull exposure 1/3 stop and call that '0'. You can then reduce it an additional 2 2/3 stops. If you use the exact same aperture and exposure time at ISO 640 that you use at ISO 100 and then reduce the ISO 640 exposure by 2 2/3 stops you'll have pretty much the same picture as long as you didn't blow the highlights when shooting at ISO 640. But that means nothing in your ISO 100 shot is within three stops of full brightness. – Michael C Apr 03 '19 at 07:31
  • 1
    There may be slightly less noticeable read noise in the ISO 640 shot after reduction by 2 2/3 stops. But you could get an even cleaner result by using ISO 100, using a three stops longer shutter time or a three stops wider aperture and then reducing the ISO 100 shot by three stops in post. This is because you would actually increase the amount of light recorded by the sensor in this case, and ultimately noise is about how much signal (i.e light) there is in ratio to the noise . – Michael C Apr 03 '19 at 07:36
  • Thank you for that last explanation, it's so obvious yet I didn't even consider that. Even though changing Av/Tv of course comes with changes other than just the light gathered too. – confetti Apr 03 '19 at 14:22
  • You can't change the light gathered without changing either the Tv or Av (unless you change the light itself). Changing the ISO only changes the analog amplification of the the signal from the same amount of light. – Michael C Apr 04 '19 at 06:10
2

While most of your questions are already answered by Michael or the question he linked to here are my two cents about your following question:

Can someone tell me if I interpret the chart correctly, and if ISO 160, later darkened in post, will really give me cleaner, less noisy images in the end than using ISO 50 (assuming that would be the perfect exposure) in the camera right away?

The chart shows the Read Noise in dependence of ISO. This was measured by taking a completely black picture. So as long as you only shoot with your lens cap attached, the answer is yes ;-)

Otherwise these measurements are not really relevant for practical purposes. There you use the Visual Noise. These are expressed in values from 0 to "unlimited". Values up to 0.8 mean near noiseless results, up to 2 for "low" noise, up to 3 for "medium" and everything above that for "clearly visible" noise.

I googled for tests with Visual Noise for your camera and found one at PDN. It's quite a read (with links explaining the methodology) but I think the following bullet point sums it up nicely:

The amount of observable noise is consistent in each viewing condition, at ISOs from ISO100 through ISO3200. At higher ISOs, the amount of noise increases.

So don't spent too much time considering if to use ISO 100 or ISO 160. The result will look the same most of the time anyway.

Lothar
  • 292
  • 2
  • 3
  • This is a great explanation and the article linked was super helpful to get more technical insights. It also addresses the chart directly and how the test was created, that's why I'm marking it as the accepted answer. I probably should've reworded my question and focused only on that part of the question to avoid the duplicate flag, but I guess it's fine how it is. – confetti Apr 03 '19 at 14:33