Nature Photographers Network
This is the preserved archive of the original Nature Photographers Network. The community continues at naturephotographers.network.

Canon 5D II noise better at ISO 320 than ISO 100?

Discussion

4 replies
Posted by Sean Bagshaw on May 13, 2011 at 1:14 PM
Intrigued by various blogs and forum chatter regarding optimal ISO settings for minimal noise with different cameras I decided to conduct my own test with my Canon 5D Mark II. General logic and previous literature would have us think that the cleanest images can be obtained with an ISO setting of 100. However I found that with the 5D MKII ISO 320 produces less noise than ISO 100 and ISO 640 is better than ISO 125. Nothing is free, so does less noise at higher ISO settings come at the cost of something else, such as dynamic range?

You can read my entire article and see my test images on the Photo Cascadia blog.

I'd love to hear some other thoughts on this matter or responses to my findings. I'm sure many of you are far more educated on this than I am.
Posted by Phil Pogledich on May 13, 2011 at 2:19 PM
Very interesting theory, and your write-up provides evidence that you may be onto something. I'll have to experiment a bit myself and take into account your caveats about the potential for reduced dynamic range at ISO 320 and similar multiples of 160. Thanks for posting this.

Now for a silly question . . . you don't mention it in your write-up, but I assume that you turned in-camera noise reduction off for the testing?
Posted by Sean Bagshaw on May 13, 2011 at 3:19 PM
Phil - Yep, in camera noise reduction was off. If you search the web there is quite a bit of talk on various forums related to this and several other photographers who have conducted similar tests. QT Luong even did some testing with the 5D MKII, but his measurements were a bit techy and mathematical for me. I wanted to produce some test images that I could examine up close myself.
Posted by Paul DeCesare on May 13, 2011 at 6:27 PM
This is a very interesing topic, because in the cinema world they have clearly proven that shooting 1080p video at 24fps on the 5D Mark II looks best when ISO is 160, 320, 640 or 1250. People who do both stills and video (like Vincent Laforet and Philip Bloom) have stressed that it only applies to video, not stills, but I think it's clear that they say that out of default (surely ISO 100 is best, right??), I don't believe I've seen the same testing done by video folks specifically on still images. Why would it be different, video is basically a boatload of still images, correct?

I think you're on to something Sean, there must be something in the physics. All I know is I shoot my 5D2 at the odd-stop ISO settings all the time and am never unsatisfied! That 5D2 is just an image quality marvel, ain't it?

(Slightly OT for those shooting video, most DSLR cinematographers don't shoot higher that 1250 on the 5D2 simply because the post work for video noise reduction is a lot more difficult to achieve, that's why I stopped at 1250 when I wrote the ranges. You could certainly shoot as high as it goes for artistic effect, graininess didn't hurt the folks who did the Blair Witch Project, lol.)
Posted by Luis Argerich on May 13, 2011 at 10:16 PM
This has been studied in deep detail.

The only ISOs that make sense on a Canon body if you shoot RAW are the native ISOs: 100,200,400,800,1600 and 3200.
ISOs 50,6400 and all the intermediate ISOs are software emulations from the base ISOs, if you shoot JPG those can be useful, if you shoot RAW you lose dynamic range and win nothing.

The "hot" ISO on some blogs and forums is ISO160, the reason why ISO160 has less noise than ISO100 is fairly simple. When you dial 160 the camera takes the shot at ISO200 and then pulls back 1/3 of a stop. So the camera is exposing to the right for you and noise seems to be less.

If you want minimal noise the recipe is simple: ETTR & ITTR
First ETTR (expose to the right) expose as much time as you can. Photons defeat noise.
Once you can't expose more then ISO to the right (ITTR) increase the ISO if the histogram still has room.