Okay to use 1/3 ISO increments?

[Jordan]

So I heard something somewhere that using the ISO values that aren't the standard 100-200-400 etc... can actually produce MORE noise than the "correct" values, even that one lower? For instance instead of using 250 or 320, you should use 200 or 400... because I hear even 400 has less noise than 250 or 320. I don't really understand how this can be true. I've read that it has something to do with the ISO values that aren't "real" are "simulated" and those algorithms actually end up producing more noise but I wanted to get some verification on this. I've searched the web and have found very wishy-washy results... does anyone know of this??


- Jordan


www.freshphotohawaii.com

[crosbyharbison]

I disable 1/3's because when I'm touching the iso I'm moving it a lot and its less clicks to get it where I need it to go (:

[Colin]

I disable 1/3's because when I'm touching the iso I'm moving it a lot and its less clicks to get it where I need it to go (:
<div style="CLEAR: both"]</div>



Good point.


I don't know if there's any reason why non-doubled ISO numbers would be any worse, but out of habit, I use the round numbers anyway.

[Ralph]

I've read this sometime ago (I think it's from this site) that ISO 250, 320 and the like are produced through interpolation and in camera algorithms (same as how p&amp;s cameras uses digital zoom). By using native ISOs, I find it much easier to move from ISO 100 to 1600 using a few clicks (same a crosbyharbison's comment).

[Jon Ruyle]

I think the problem is that because with some canon cameras, those "tweener" iso's are fake (in other words, ISO 125 may be the same to the hardware as 100, then hacked by software to look like 125) you might get more quantization error and lose a little dyamic range (either as lost highlight headroom or increased shadow noise, depending on which way the software adjusts).


Personally, I don't care about any of that. I just turn them off
because I don't like having to turn the dial so much to go from 100 to
3200.


You always get less photon noise when using a lower iso (assuming the meter reads the same, the light actually is the same, and you don't adjust the exposure) because you get more light, so I don't think the claim that ISO 400 has less noise than 250 is always true (read noise may be worse, but overall noise will not always be... in fact, will probably not be in most cases).


I think Daniel can expand/correct the above.

[kitaoka]

I find the 1/3 ISO adjustments very valuable in my work.

[Keith B]

The tweeners are essentially pushed an pulled from natives. 125 is 100 pushed a 1/3 which will increase noise appearance. 160 is 200 pulled a 1/3 which clips shadow detail reducing dynamic range.

[Daniel Browning]

Jon hit the nail on the head.

[Jordan]

Hmmm... that's what I was afraid of... looks like I should change my ISOs then!

[Jordan]

Hmm... sorry that's it's off topics, but what about "Highlight Tone Priority" it's "Disabled" but should it be on? What is it? Doesn't it just increase the dynamic range somewhat to reduce blowouts?

[wickerprints]

I always have HTP OFF. The way it works is it underexposes to give more highlight headroom at the expense of increased shadow noise when the tone curve is applied. It's basically the opposite of ETTR, where instead of trying to maximize S/N by pushing the histogram up to the right as far as it goes without clipping too much off the highlights, you are trying to prevent highlights from clipping by pushing the histogram to the left, which increases shadow noise.


From the following link: http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html



<span style="font-family: Helvetica Neue Bold,sans-serif; font-size: small;"]By the way,
underexposing at lower ISO is precisely what Canon cameras do
in the raw data when Highlight Tone Priority (HTP) is enabled; and what
Nikon cameras do when Active D-Lighting (ADL) is enabled.
Instead of using the ISO gain set by the user, the camera uses
a lower ISO (but exposes with the indicated aperture and shutter speed),
effectively underexposing the image; this provides more highlight
headroom. In post-processing, the image data can be brought back up
while preserving the highlights with a modified tone curve
in higher exposure zones. The place where image quality suffers
is in shadows at lower ISO, precisely as the above quantitative
model predicts.



For the really technically inclined, the above link is a very interesting read.

[Jordan]

Thanks! I'll stick with the way you explained it though ;) haha