May 2, 2011

Let's Settle This: Do Canon DSLRs Perform Best at ISO Multiples of 160?

I'm confused. In the DSLR Guide's chapter on ISO noise, I originally stated that it was optimal to shoot at ISOs divisible by 160, because the camera's noise levels were lowest. Then I was told that this came at the expense of highlight headroom, so I recanted a bit. Now, as part of the Technicolor CineStyle picture setting release, the user manual mentions that one should shoot at "multiples of 160." I assume Technicolor knows what they're talking about, but the question remains.

As a response to Technicolor's recommendation, Andrew Schär conducted this test:

The noise levels are appreciably lower at the corresponding ISOs: 160, 320, 640, 1250, 1600, etc. However, this test doesn't show latitude or any sort of highlights. How much highlight loss are we talking? Is it always a best practice to shoot at ISOs divisible by 160, or is it situation-dependent (e.g., don't worry about using 160 ISOs if you want to retain highlight details)? What's the wisdom of the crowd on this?

Your Comment

23 Comments

I'd say, if you want to retain the highlights, use multiples of 160 (160, 320, 640, 1250, 2500) and underexpore the whole thing a bit. That's what Vincent Laforet suggests, too. Than overexposure it in the post, with the Technicolor picture style you should still have enough details in the darks.

May 2, 2011

0
Reply
Flu Popow

I haven't done any A/B tests to compare things side by side. However, I do have an idea about this.

Given that the Technicolor picture style adds a full stop or two of dynamic range, is more sensitive to noise, and Technicolor recommends using the 160 multiple ISO settings, it seems like a no brainer to stick with the 160 ISO multiples when using CineStyle.

As for using other picture styles such as Neutral or Faithful, I think the jury is still out on that one.

Having said all of that, I plan on using CineStyle and the ISO 160 multiples only from now on. I'd still be interested in hearing the results of these tests though!

May 2, 2011

-1
Reply
Nathan

"If you want cleaner shadows and your scene is not in danger of blowing out (or you don't care) then use the upper intermediates. If you want max dynamic range then use the whole numbers [100, 200, 400...]."
From cinema5D forum:
http://cinema5d.com/viewtopic.php?f=64&t=14020&start=0#p97724
which refers to some dpreview post.
Curiously, the technicolor setting refers specifically to dynamic range, which this guy says 160 multiples are not the best for - although less noisy.

May 2, 2011

1
Reply
Theo Dubeux

At the 2010 Canon Expo which I went to, I remember seeing a presentation given by Alex Buono, who is a DP for SNL. When this issue came up, he said that it was very complicated, but to use increments of 160 because the iso's in between were only simulated on the camera, and not a true representation of what the footage actually looks like.

May 2, 2011

0
Reply
David

DP Josh Silfen in his "Shootin' The Shot" blog addresses this same issue. He has some very interesting things to say on this subject. Here's a quote; "If you've spent any time researching Canon's HD DSLR cameras, you've probably come across discussions of which of the cameras' ISO settings to use and which to avoid. There seems to be a common misconception out there, held by even some very well-regarded experts that the cameras' "native ISO" or "true ISO" settings are the multiples of 160 (ISO 160, ISO, 320, ISO 640, etc), and that the rest of the ISO settings are produced digitally. This is not true. In fact the cameras' native ISO settings (that is, the settings that are derived from analog gain rather than digital exposure compensation) are the multiples of 100 (ISO 100, ISO 200, ISO 400, ISO 800, etc). However, that doesn't necessarily make them better."

Here's the URL to Josh's blog post. http://shootintheshot.joshsilfen.com/2010/05/13/canon-hd-dslr-native-iso/

May 2, 2011

1
Reply
Ken Wilson

That's the most convincing and comprehensive discussion of the topic I've seen yet; thanks for bringing it to my attention! I would say that unfortunately without Canon stepping in and telling us exactly how it is, this is all conjecture. But that post's treatment of the subject fills in a lot of factors that didn't quite add up for me as far as the received wisdom goes.

May 2, 2011

1
Reply
Luke

How does one get these ISO settings that are multiples of 160? On my Canon t2i it's only 100, 200, 400, 800, etc.

Do I need to install Magic Lantern or something? Last time I attempted that it was all wonky and I uninstalled it.

May 2, 2011

0
Reply

You need to change the settings in your menu. You have the option of having your ISO increments in 1 or 1/3 stops. I'm assuming your's is set to 1.

May 3, 2011

0
Reply
Joe

I recall my statement. It appears the t2i is only capable of 1 stop increment ISO settings.

May 3, 2011

1
Reply
Joe

install magic lantern

you'll get 160x ISO settings, plus histogram, waveform, zebras, peaking, focus help (10x zoom while recording), etc

May 3, 2011

0
Reply

oh, and I forgot to say: the build in the official page is old, I use this one:
http://groups.google.com/group/ml-devel/browse_thread/thread/850ec268bc8...

there's a newer one here, but I haven't tested it:
http://www.dvxuser.com/V6/showthread.php?218686-Magic-Lantern-for-550D-i...

May 3, 2011

-1
Reply

Thanks a lot! Hopefully this build will work better than the one I originally installed!

May 4, 2011

-2
Reply

I'll have these tests in mind the next time I shoot something WITH THE CAP ON

for anything else, I'll look at real world tests, like the ones I carried out:
http://www.vimeo.com/20486388
my conclusion was: in terms of noise it doesn't matter, the important factor is how much light reaches the sensor, and how much you push your footage in post (if you underexpose and then push in post, noise gets blown up)

now, the Technicolor people are surely much more knowledgeable than I am, so I'm ready to listen to any new evidence and revise my conclusions; but no tests with the cap on, please

May 3, 2011

0
Reply

Doesn't this issue just prove that using ISO a measurement is simply an anachronism to provide photographers a familiar terminology? How about starting to use terminology that makes sense for the way the technology actually works? Just call it gain, and let me put it at 0 or -3 db.

If it's more complicated than that, then show me the data. Professionals should be demanding more transparency from the camera manufacturers for how the image processing in their tools works.

May 3, 2011

1
Reply
Chris Larkee

That sure would make sense to me, but it is bound to piss off still photographers everywhere!

May 4, 2011

2
Reply
Max

"In the DSLR Guide’s chapter on ISO noise, I originally stated that it was optimal to shoot at ISOs divisible by 160, because the camera’s noise levels were lowest. Then I was told that this came at the expense of highlight headroom"

You're right indeed.

ISO values multiple of 160 deliver less noise in the shadows (that's why they're recommended) at the cost of less headroom on the highlights, (assuming it's well exposed). That could be summarized as a bit less of Dynamic Range.

The standard and real "native" ISO values (which are 200, 400, 800, etc) give a bit more Dynamic Range, but show a bit more noise at the shadows/dark areas.

All of above can be confirmed with some tests, especially at mid to high ISO levels.

Of course the tests must be done with ALO (Auto Light Optimizer) and HTP (Highlight Tone Priority) modes disabled!

Maybe we post some tests proving it, but it's an old already known fact (that applies to Photography and Video mode)

Cheers.

May 5, 2011

0
Reply

on a Mac, invert your monitor (Cmd-Ctrl-Option-8) to see the noise a little bit more clearly

May 5, 2011

0
Reply
Zach

Now I guess that's why stills at ISO 1600 on my 7D are hardly any noisier than stills at ISO 800.

I was really shocked by how much of a difference it makes from ISO 200 to ISO 320 in that video!

May 5, 2011

1
Reply
Heiko

I think the evidence is in the image. The 160 multiples look so much cleaner and in my experience have been cleaner in every situation. I always shoot 160 multiples. However, one should note that proper clean image will always be achieved by properly lighting the image.

May 5, 2011

1
Reply

From the snippets I've heard, I've been ignorantly going on the assumption that ISO 160 is ISO 200 digitally reduced by 1/3 stop. That would leave the highlights clipped 1/3 stop early and gobble up 1/3 stop of the noisiest shadows.

Has nobody empirically tested the 5D to find its native ISO?

May 6, 2011

0
Reply
Pippy

That's my understanding too. 200 is native, 160 is 200 reduced a third and 250 is 200 gained a third. And it goes the same upwards,

The easiest would be if Canon opened their mouths and outright told us all this so the guesswork is thrown out the window. But this is a company that puts more effort in stifling firmware-improvement by independents than pulling out the potential in their cameras. Would I pay for a firmware tailored for video? Gladly. Will they do it? I doubt any one of them have considered it. Grumble grumble millenium ham and shrimp!

May 14, 2011

0
Reply

In my own direct tests in manual and then enlarged to 200% and pushed in brightness and contrast using a graphics program and testing every iso setting using my Canon 50 D, I could find no noise at all at 100, 125, 160, and 200. I found barely perceived noise at 250, 320, 400, 500, and 800. I found perceived by very small noise at 1000, 1250, 1600, and 2000. I found noticible noise at 2500. I found objectionable noise at 3200, and.....drum roll.... 640. I didn't bother to test 6400, and 12,800.

June 21, 2011

-1
Reply
zarathu

This is my first time pay a visit at here and i am genuinely impressed to read everthing at one place.

August 19, 2014

-3
Reply

But isn't the live view in the lcd too dark when you set the iso to lets say 160?
How are you suppose to see what your shooting if it appears dark on the lcd?
I'm sure I'm missing something, so thanks for any clarification.

September 29, 2014

-1
Reply