ISO 160 vs. 100

[Quasimodo]

I think it was in the Jeff Cable review (reviews are pouring out now) of the 7DII where he made claims that ISO 160 on Canon cameras are better than ISO 100. I find this strange. Guided by earlier discussions on this forum I was led to believe that ISO 100 is the lowest (most clean) natural setting for Canon cameras, and shooting at the optional ISO 50 would clipping the highlights and affect the DR.

However, I am a simple guy, and if I am told by credible sources that 160 is the magic number for shots where you want a maximum clean image (where conditions allow you), that is then what I will shoot at

Any thoughts/insights?

 

[Maui5150]

You can seer here:

http://vimeo.com/23082874

Or here:

http://indigoverse.com/the-truth-about-native-iso-for-canon-dslrs/

I did similar tests and found the same at the lower end.

My understanding that the 1/3rd stops are pushed and the 1/3rd stops pulled. the ISO 160, 320, 640 seem to be slightly underexposed, but a lot less noise. This may mean you do trade some DR.

For the video example, I would have liked to see them have done 2/3rd jumps i.e. 100, 160, 200, 320, 400, 640.

No question from what I have seen and what the video shows 125 is more noisy than 160, 250 more noisy than 320, and 500 more noisy than 640.

Whether you see 160 as less noise? that remains to be seen, but to my eyes (and some of the charts posted in another thread, 160 looks a little less noisy than 100

 

[Coldhands]

Maui5150 has it right.

ISO 160 generally appears to exhibit less noise that ISO 100 or 125 (same applies for multiples thereof) because it is actually a "pulled" ISO 200 exposure. Because the whole image is being darkened, this tends to hide the noise. The opposite is true of ISO 125, as it is a "pushed" ISO 100 exposure.

The downside of using the pulled ISOs is a lower dynamic range, although the difference is likely to be academic.

Also worth noting: you may have heard that ISO 160 is the "native" sensitivity of Canon sensors, but this is a misconception due to misinterpretation of the results described above.

 

[Maui5150]

Not sure I have it right:

Also have seen this.

http://photocascadia.wordpress.com/2011/05/13/canon-eos-5d-mark-ii-iso-noise-test/

In the tests like I posted, the noise seems to be called pattern read noise so more an artifact that appears when doing this kind of lens cap on test.

I think when all is said and done, if you shoot 100, 200, etc you get the most information in your image.
If you shoot 160, 320, etc, you are shooting at 200, 400, etc, but then stopped down 20% and this appears to "clean" some of the pattern noise.

The test I haven't seen and what might be interesting to try is looking at long exposures at night and sensor noise from the sensor heating up. i.e. shooting 640 for 3 seconds versus for 400 for 5 seconds

 

[2n10]

I liked the bloggers conclusions. ETTR and then ITTR to get you best results all without blowing out your highlights of course.

Isn't that what most folks advocate when you first learn how to use a digital camera?

 

[Marsu42]

I think when all is said and done, if you shoot 100, 200, etc you get the most information in your image.
If you shoot 160, 320, etc, you are shooting at 200, 400, etc, but then stopped down 20% and this appears to "clean" some of the pattern noise.

+1, after shooting at iso160-multiples for some time I came to the conclusion that it isn't worth the hassle... and you cannot just put all sensors in one basket, a 18mp crop sensor on the old 7d1 will behave in a very different way than the latest clean, ff 6d

For other in-depth analysis by the Magic Lantern devs, read this thread "Do have iso160-multiples have more dr & less noise?": http://www.magiclantern.fm/forum/index.php?topic=9867.0

 

[Loswr]

Any thoughts/insights?

You have a 1D X, so I wouldn't worry about it. The 1D X uses a different analog amplifier setup, note the 'jagged' plot of noise vs. ISO for the 7D and 5DIII (with troughs at 160 multiples), vs. the 'smooth' (and overall lower) plot for the 1D X.

 

[sedwards]

very interesting chart Neuro. do you know if there is a similar one for the 70D and 6D ?

 

[drmikeinpdx]

If you want to know more about the work of William Claff, you can follow this link:

http://home.comcast.net/~nikond70/index.htm

He has a number of charts that apply to Canon sensors, mixed in with a lot of Nikon data.

 

[Coldhands]

Any thoughts/insights?

You have a 1D X, so I wouldn't worry about it. The 1D X uses a different analog amplifier setup, note the 'jagged' plot of noise vs. ISO for the 7D and 5DIII (with troughs at 160 multiples), vs. the 'smooth' (and overall lower) plot for the 1D X.

Very interesting. Is this because the 1D X applies the gain at an analogue stage for each ISO, instead (I presume) of for multiples of ISO 100 only, followed by digital push/pull on other sensors? If so that is surprising, as I would have thought such behavior is a product of the sensor architecture shared by all 500 nm process DSLR sensors.

 

[Loswr]

Any thoughts/insights?

You have a 1D X, so I wouldn't worry about it. The 1D X uses a different analog amplifier setup, note the 'jagged' plot of noise vs. ISO for the 7D and 5DIII (with troughs at 160 multiples), vs. the 'smooth' (and overall lower) plot for the 1D X.

Very interesting. Is this because the 1D X applies the gain at an analogue stage for each ISO, instead (I presume) of for multiples of ISO 100 only, followed by digital push/pull on other sensors? If so that is surprising, as I would have thought such behavior is a product of the sensor architecture shared by all 500 nm process DSLR sensors.

Others know more about this than me, but I believe the signal amplifiers are off-die (also accounting for 'amp glow'), and the 1D X has separate analog amplifiers for the 'tweener' ISOs whereas other Canon bodies do only full-stop analog amplification with digital push/pull for the 'tweener' settings.

 

[Maui5150]

Any thoughts/insights?

You have a 1D X, so I wouldn't worry about it. The 1D X uses a different analog amplifier setup, note the 'jagged' plot of noise vs. ISO for the 7D and 5DIII (with troughs at 160 multiples), vs. the 'smooth' (and overall lower) plot for the 1D X.

Is the 1Dx native 200 ISO? From the charts it looks like 200 is slightly better

 

[Loswr]

I'm pretty sure the native base ISO is 100 (or slightly less, according to DxO) for the 1D X.

 

[Quasimodo]

While I have a bit of difficulty of understanding all of the kind answers here (technological ineptitude), I am to understand (Neuro) that on my 1Dx Iso 100 is the optimal setting?! And I guess that the charts used here are based on empirical evidence, hence it would be impossible to say beforehand what the correct value for the 7DII would be, except than low?

 

[Loswr]

While I have a bit of difficulty of understanding all of the kind answers here (technological ineptitude), I am to understand (Neuro) that on my 1Dx Iso 100 is the optimal setting?! And I guess that the charts used here are based on empirical evidence, hence it would be impossible to say beforehand what the correct value for the 7DII would be, except than low?

On the 1D X there's really not much difference at any ISO from 100 - 400. No way to know for sure about the 7DII, but I'd guess it'll be more like the 5DIII/7D.

 

[Quasimodo]

While I have a bit of difficulty of understanding all of the kind answers here (technological ineptitude), I am to understand (Neuro) that on my 1Dx Iso 100 is the optimal setting?! And I guess that the charts used here are based on empirical evidence, hence it would be impossible to say beforehand what the correct value for the 7DII would be, except than low?

On the 1D X there's really not much difference at any ISO from 100 - 400. No way to know for sure about the 7DII, but I'd guess it'll be more like the 5DIII/7D.

Thanks

 

[jrista]

In my experience, the differences in noise for the full and +1/3rd (push) stops are not really enough to get worried about for current Canon generation cameras. I have noticed that the 2/3rd (-1/3rd or pull) stops are a bit noisier, however it's still usually not enough to worry about.


Older 18mp APS-C parts had more problems with noise, and the 2/3rd pull stops were pretty noisy. I don't know about older FF parts. So long as your using a current model, however, I simply don't worry about it. Canon read noise is high, and it's high no matter what (at lower ISOs)...a third stop change in high read noise isn't going to change things much. At higher ISOs, the shift with third stops is less significant, and so doesn't matter at all.

 

[Don Haines]

In my experience, the differences in noise for the full and +1/3rd (push) stops are not really enough to get worried about for current Canon generation cameras. I have noticed that the 2/3rd (-1/3rd or pull) stops are a bit noisier, however it's still usually not enough to worry about.


Older 18mp APS-C parts had more problems with noise, and the 2/3rd pull stops were pretty noisy. I don't know about older FF parts. So long as your using a current model, however, I simply don't worry about it. Canon read noise is high, and it's high no matter what (at lower ISOs)...a third stop change in high read noise isn't going to change things much. At higher ISOs, the shift with third stops is less significant, and so doesn't matter at all.

For both the 5D2 and the 60D you could easily detect the differences in noise with the "dark frame" test, but with actual photos there wasn't any noticeable difference between adjacent 1/3 stops...

 

[jrista]

In my experience, the differences in noise for the full and +1/3rd (push) stops are not really enough to get worried about for current Canon generation cameras. I have noticed that the 2/3rd (-1/3rd or pull) stops are a bit noisier, however it's still usually not enough to worry about.


Older 18mp APS-C parts had more problems with noise, and the 2/3rd pull stops were pretty noisy. I don't know about older FF parts. So long as your using a current model, however, I simply don't worry about it. Canon read noise is high, and it's high no matter what (at lower ISOs)...a third stop change in high read noise isn't going to change things much. At higher ISOs, the shift with third stops is less significant, and so doesn't matter at all.

For both the 5D2 and the 60D you could easily detect the differences in noise with the "dark frame" test, but with actual photos there wasn't any noticeable difference between adjacent 1/3 stops...

I see more noise in normal shots with the 2/3rd stop settings...no dark frame required.

 

[bret]

I did some test images on my 5D2 by taking a black frame with the shutter cap on at each ISO and then increasing them by the same amount to more easily see the noise. I always shoot at increments of 160.