Canon 5Dc native ISOs vs. push/pull ISOs test

Hey folks,

I recently read this thread which pointed me to this article. The author suggests that when using 1/3rd stop ISOs, the camera is pushing the lower native ISO setting in post and exposing for the 1/3rd setting you've chosen. Similarly, when using 2/3rd stop ISOs, the camera is pulling the higher native ISO setting and exposing the image for the chosen 2/3rd stop setting. Therefore 1/3rd stop ISOs would yield the worst S/N ratios and 2/3rd settings would produce the cleanest files. This only sounds logical knowing noise performance is the worst in underexposed images. So I decided to make a test of my own on my 5Dc. Strangely enough it still seems to not follow this logic and favor the native ISOs.

Here are the test shots. I added the full resolution file size for each test shot as it seems to represent very accurately the apparent amount of noise present between the shots. All done on 5Dc, Manual mode, body cap on, 1/200, Fine Large JPEG, imported in PS CS5, levels adjustment with white point on 70, then brightness/contrast adjustment 100/85 using the legacy algorithms. Saved in maximum quality, uploaded to DA which resized them to 900x600 resolution.

ISO-50 (L) (0.16 mb !)

ISO-500 (9.35 mb)

If you are going to shoot black frames to show noise differences, I would suggest you "equalize" the images (an option in Photoshop), it will enhance the effects to help make the differences stand out a bit better. Also regarding the ISO 50, people have to remember it is not a noise issue around using this "false" ISO level, it is the reduction in dynamic range that causes most of the discussions.

I don't even think about noise levels until I hit ISO 400 or higher personally.

TeamSpeed wrote in post #13943979
If you are going to shoot black frames to show noise differences, I would suggest you "equalize" the images (an option in Photoshop), it will enhance the effects to help make the differences stand out a bit better. Also regarding the ISO 50, people have to remember it is not a noise issue around using this "false" ISO level, it is the reduction in dynamic range that causes most of the discussions.

I don't even think about noise levels until I hit ISO 400 or higher personally.

I already tried equalizing them but all pictures become too noisy, even the ISO-100 sample, making them all more or less the same.

I know about ISO-50 and DR reduction, so I almost never use it (only when I want to get that extra stop of less DOF on a sunny day). Considering noise, from the samples above, ISO-50 to ISO-400 seem very clean with a considerable drop in performance at ISO-800 and ISO-1600 and, ISO-3200 producing very noisy pictures. These somewhat cover my observations in real life shooting, though ISO-3200 shots are still not bad at all if exposed correctly or ETTR. In my experience, having shot quite a bit with a 7D and a T2i, up to ISO-1600, the 5Dc is better in both noise handling and DR, with the 7D/T2i having a slight edge at ISO-3200.

One other thing that bothers me a bit in my 5Dc, compared to the 7D/T2i, display quality aside, is the lacking metering system.

Actually if you convert them to negatives, the noise stands out a lot better in the white background...

Yow - I thought I'd been magically transported to the "astronomy" section.

50 to 100 looks pretty clean, about what i'd expect.

MakisM1 wrote in post #13944210
Actually if you convert them to negatives, the noise stands out a lot better in the white background...

I tried that too but at least to me it seems noise is more apparent on black.

rick_reno wrote in post #13944482
Yow - I thought I'd been magically transported to the "astronomy" section.

50 to 100 looks pretty clean, about what i'd expect.

LOL
Don't forget these are pushed quite a bit to show noise ratio between the ISO settings, the original shots are very clean up to 1600 excluding ISO-1250 maybe.

Thanks for this test. Took me some time to get it.
So, ISO 800 and 1600 are very similar and 50 is not so bad on 5Dc.
I wonder how ISO50 would be with fast prime under bright light compare to ISO100.
I'll try it next time.

um... I'm more worried about real world tests. Take a picture of something colorful/real and lets see again, doubt you'll notice.

kf095 wrote in post #13944786
Thanks for this test. Took me some time to get it.
So, ISO 800 and 1600 are very similar and 50 is not so bad on 5Dc.
I wonder how ISO50 would be with fast prime under bright light compare to ISO100.
I'll try it next time.

You're welcome! In my experience, if exposed correctly, the 5D actually shines at ISO-1600. ISO-50 doesn't produce any noise whatsoever. However, as TeamSpeed pointed out, it does come with a slight decrease in dynamic range and you should keep an eye on your histogram so you don't blow out the highlights that matter. On the other hand the 5D exhibits so little noise up to ISO-400 that you don't need to worry about it at all at ISO-100.

Talley wrote in post #13944808
um... I'm more worried about real world tests. Take a picture of something colorful/real and lets see again, doubt you'll notice.

The story goes like this


I was a bit bored in front of the computer.

HTTP response: 403 | MIME changed to 'application/xml'

Byte size: ZERO | Content warning: NOT AN IMAGE

I'm not in the practice of shooting at ISO 50 with my 5DC, so I grabbed my camera and tried something:

It's overcast outside right now with some light clouds in the mix. So, I went to ISO 50 and tweaked my exposure until some of the clouds clipped then backed off by a click so no clipping.

Then, keeping the aperture and shutter speed the same, I went to ISO 100 and took the shot.

The good news, is that the ISO 100 didn't clip the highlights! The "white point" was pretty much the same between the two!

But, the shadows was a different story. The shadows and midtones were all pushed back, so that shadow clipping was a problem...

So, the question becomes...what good is ISO 50? I don't know, really. But it looks like Canon applies some kind of curve, so that a highlight on the edge stays on the edge but the rest is darkened in the ISO 50 results...not sure how that approach is really beneficial, since the ISO 100 shot didn't clip the highlights but got better shadow lighting. Except of course that pulling the exposure back would have "dampened" any noise, but then you have to ask yourself if the darker shadow areas really are worth it, since you lose clipped shadows and also will increase the noise if you do have to brighten the shadows in pp!

love the cartoons...

I really appreciate this test, and find the results very useful. Putting this kind of info to use is what helps us produce better images at the pixel level. Thanks!