How to select ISO (tutorial)

[duydaniel]

I am sure some of you already know this but in case.
Here is the formula I discover on my Canon

Supposed you have ISO speed increment set to 1/3
Base ISO: 100, 200, 400, 800, 1600 ...

So the formula will be:
[x1] [no] [yes] [x2]

where x1, x2 are base ISO.
Example:

[100] [125] [160] [200][/b]
You should avoid the ISO 125
Preferred: 160

[800][1000][1250][1600]
You should avoid the ISO 1000
Preferred: 1250

The author here explained much better than I can:
http://shootintheshot.joshsilfen.com/2010/05/13/canon-hd-dslr-native-iso/

For your specific camera model, you can look here:
https://home.comcast.net/~NikonD70/Charts/RN_ADU.htm

Below is picture of iso 125 vs iso 1250
where if you follow the formula, the iso 125 fall in to the NO and 1250 into the YES
Any comments?

 

[Loswr]

Couple of clarifying points:

1. Despite the statements in the first linked article, DxOMark shows that when Canon dSLRs that show the behavior described are set to ISO 100, the real (empirically measured) ISO is 80, meaning ISO 160 is exactly one stop above the real base ISO. So, while the phenomenon is real, the linked author's explanation is wrong (and the one he debunks is actually closer to the truth!). ISO 100 is noisier than ISO 160 because ISO 160 is native analog amplification, and ISO 100 is digitally pushed from the base ISO 80. ISO 125 is even noisier because it's pushed further, and so on up the ISO range.

2. This doesn't apply to all Canon dSLRs - if you look at the 1D X plot from Bill Claff (second link in your post), it's smooth, not jagged (although the base ISO is still 80). This makes sense given that it's the same sensor used in the flagship HD-DSLR, the 1D C. The sensor on the 1D X / 1D C has better amplification circuitry than most other Canon sensors.

 

[Mt Spokane Photography]

So, how do you set your camera to only use those true ISO's? The only way I see is to manually select them, which might be fine if I'm just taking images at a fixed place with constant lighting.
I find it more important to get the shot than to fool with the ISO settings by referring to a table of values to decide if I should use ISO 12800 or whatever the best value is. Trying to remember to use 5120 instead of 6400 or 10240 instead of 12800, or 20480 instead of 25600 doesn't seem like time well spent, the noise at those ISO's is high in any event.

With my D800, the camera set ISO values like 7184, 8063, 9051, 10159, and 11,404. Now, those are weird.

 

[duydaniel]

I find it more important to get the shot than to fool with the ISO settings by referring to a table of values to decide if I should use ISO 12800 or whatever the best value is. Trying to remember to use 5120 instead of 6400 or 10240 instead of 12800, or 20480 instead of 25600 doesn't seem like time well spent, the noise at those ISO's is high in any event.

That is why I figured the formula started from base iso, you turn up 2 knots (2/3 stops given you set 1/3 iso increments) to obtain that optimal iso.

 

[duydaniel]

Couple of clarifying points:

1. Despite the statements in the first linked article, DxOMark shows that when Canon dSLRs that show the behavior described are set to ISO 100, the real (empirically measured) ISO is 80, meaning ISO 160 is exactly one stop above the real base ISO. So, while the phenomenon is real, the linked author's explanation is wrong (and the one he debunks is actually closer to the truth!). ISO 100 is noisier than ISO 160 because ISO 160 is native analog amplification, and ISO 100 is digitally pushed from the base ISO 80. ISO 125 is even noisier because it's pushed further, and so on up the ISO range.

I wonder if you have Canon's documentation on this?
Also, what do you think about the Dynamic range of iso 160 vs 100?

 

[Loswr]

I wonder if you have Canon's documentation on this?

Canon doesn't publish such documentation. But consider - we have an observation by the video crowd that multiples of ISO 160 are cleanest, and that's backed up by Bill Claff's measurements. There are two hypothesis to explain that observation, one that multiples of 160 are cleanest because they're multiples of base ISO (which was suggested to be ISO 160, but could easily be ISO 80), the other that the are cleanest because they're pulled from the next-highest real multiple of base ISO, i.e. 100, 200, etc. DxOMark's measurement data are reliable, and they measure the base ISO of the cameras in question as ISO 80, not ISO 100. That supports the first hypothesis, and means ISO 100 is, in fact, pushed from ISO 80.

Also, what do you think about the Dynamic range of iso 160 vs 100?

Based on Claff's data, DR is highest at the same multiples of ISO 160 that are least noisy - a win/win, IMO. As with noise, the 1D X has a smooth curve for DR vs. ISO, not jagged like other Canon cameras.

As Mt. Spokane stated, this concept is useful for a planned shot with setup time, less so for fast moving situations. In the latter case, Auto ISO is often a useful option, and even if you restrict the ISO setting to full stops (which I do, simply because it makes setting the ISO faster), Auto ISO will use the 1/3-stop increments. Personally, I'd take a little more noise in a shot with ISO 500 to missing the shot because I was trying to set ISO 640 instead. Fortunately, with the 1D X I don't have to worry about it, since the curve is smooth (and at all ISO values except 160, the 1D X has lower read noise than the 5DIII).

 

[Dylan777]

thanks for sharing....with 5D III, just keep it below 5000ISO(still do able @ 6400ISO) then you be just fine.

BTW...I don't see any dif. between 100ISO and 400ISO outdoor-daytime shooting.

 

[Loswr]

I don't see any dif. between 100ISO and 400ISO outdoor-daytime shooting.

Which makes perfect sense, because the data from Bill Claff, as well as the video shooters doing tests with the lens cap on, are showing only read noise. In bright light, photon shot noise is the dominant factor.

In fact, that brings up a good point about the above discussion. Knowing about the phenomenon described, and whether or not it affects your camera, is more important than the reason behind this phenomenon. But it's just as important to know when it is a factor, and when it really doesn't matter at all.

 

[cid]

thanks for sharing this info, I didn't know about this at all ???

I don't see any dif. between 100ISO and 400ISO outdoor-daytime shooting.

Which makes perfect sense, because the data from Bill Claff, as well as the video shooters doing tests with the lens cap on, are showing only read noise. In bright light, photon shot noise is the dominant factor.

In fact, that brings up a good point about the above discussion. Knowing about the phenomenon described, and whether or not it affects your camera, is more important than the reason behind this phenomenon. But it's just as important to know when it is a factor, and when it really doesn't matter at all.

maybe this info could be somehow useful for night sky shooters, or simply for shots where black is dominant color

 

[CanonLITA]

For your specific camera model, you can look here:
https://home.comcast.net/~NikonD70/Charts/RN_ADU.htm

Thanks for sharing. Anyone has some data/guess on the 6D? Should I assume it to be close to the 5D mark III (or mark II)?

 

[Old Shooter]

Very interesting! Thanks for sharing!

So if I read the 5DIII chart correctly, you actually have lower read noise at 160, 320, and 640 than you do at the "base" ISO of 100...

50 ISO is just fractionally better than 100 ISO, and is still "beaten" by 160, 320, and 640...

Interesting that when you get to ISO 2560, the "pull" of 3200, the graph suddenly becomes basically linear... Almost as if ISO 3200 is the last "true" ISO and everything from there on is "pushed"...

Jackson_Bill - looking at the 7D graph - if I had a 7D - ISO 160 would be my new "base"...

 

[chasinglight]

When looking at the "read noise in ADUs vs ISO" chart, why is it that lower resolution cameras such as the 5Dc and 1d3 have significantly lower read noise than the 7d? Does that have to do with fewer pixels = less noise? Even comparing the 5d3 and the 5dc, the 5dc apparently has much less read noise based on that chart.

 

[Meh]

Doesn't the jagged performance shown in
http://home.comcast.net/~NikonD70/Charts/RN_ADU.htm
mean that, for the 7D, the base is 100, 125 is pushed from 100, and 160 is pulled from 200?

What it means, as indicated in the notes to the chart, is that there may be a separate amplifier for intermediate ISO values. That separate amplifier (and where it is in the read-out process, if it's analog or digital, etc.) would appear to add more noise to the signal. A question we might ponder is why add a separate amplifier if it adds more noise... the answer, of course, is complexity and cost or possibly Canon engineers were just behind but the improvements are in the 1DX so we'll see that trickle down. The bottom line is that the noise performance is so good now that these differences are negligible except in low-light shooting as others have pointed out.

 

[DominoDude]

I'm one of those that mostly stick to multiples of ISO 160. Wanted to see if we didn't have more on that very subject in all of our topics here. It's so long ago that I started to use this method so I have forgotten where I initially got it from, but even when I look at my old shots from the 50D I see my habit of using ISO 320 and ISO 160. I do experience these shots as "cleaner" (generally speaking) in shadowy areas, or in areas that mostly consists of tones of one particular colour.

Backtracked bookmarks and searched a little earlier today to get more info on the "ISO mystery" with Canon, and tried to see if I could find something that reminded of the facts I first reacted upon.

Here's a few links I visited:
An article and example video showing noise patterns --> http://petapixel.com/2011/05/02/use-iso-numbers-that-are-multiples-of-160-when-shooting-dslr-video/
Truth about native ISO for Canon --> http://indigoverse.com/the-truth-about-native-iso-for-canon-dslrs/


Would be interesting to read more about why you select the ISO that you use most, and if you also stick to these odd multiples.

 

[siegsAR]

I've also came across this but couldn't remember where, all I remember there was a quick dismissal on the topic.

Good thing its being discussed again. Thank you, new thing to keep in mind and I've learned a thing today.