ISO speed...

Obviously the higher ISO speed, more light is absorbed and more grain, vice versa.

ONE SOURCE tells that this "native ISO" can make the sharpest (making it like a lens aperture sweet spot) out of an image, and it tells too that ISO 100 can be too soft while ISO 200-400 is ideal for outdoor lighting. THIS SOURCE (that I'm trying to look for again) just made me confused - do ISO speed affect grain ONLY or sharpness and grain?

Yah, I've read similar things. Plus how you're better off shooting at certain 1/3 increments based on how the camera is push/pulling the exposure to achieve the fractional iso. Frankly, my eyes just can't tell the difference. I just shoot at the lowest iso I can in order to get the aperture and shutter speed I'm after.

Just to be clear - higher ISO does not lead to more light being absorbed. That is a matter of aperture and shutter speed. ISO does control the 'gain'.

I just wanted to be clear for any new folks that may not be clear.

I wish I could answer your question & hopefully someone with greater knowledge will.

<My speculation>
I have heard similar claims - that native ISO can be preferable to lower ISOs. That makes some sense to me, as the less you do to the signal the better. I think noise/grain and sharpness are intertwined. But for me, I do not notice any degradation from using ISO 100 or 200. There is some threads here that discuss the more technical aspects that may interest you below. They center on pixel size, but some of the discussion may be useful.
https://photography-on-the.net/forum/showthre​ad.php?t=747749
https://photography-on-the.net/forum/showthre​ad.php?t=706255

Interesting.

I think of ISO as I did ASA on film. Difference is, digital has no grain, it's digital noise. Film was made with tiny particles (grains) that when exposed to light, would change (I always imagined them flipping). That is what accounted for the grain. Lower ASA (<200) was made from smaller grains of silver, so the change from one grain to the next was not noticeable. Whereas ASA's greater than 200 had larger bits of grain, and you could see them in the final image.

Now, on to your question. I think* the perceived 'increase in sharpness' is only perceived, and not true. Take a look at a photo at 100%, and then increase the sharpness slider all the way. It looks 'grainy'. I think it's the same theory. I'd rather have the same image with as little grain/noise as possible.

Pixel size on your sensor does play a role in both aspects, but I think it's more reliant on the glass in front of the sensor, and not the sensitivity (ISO) of the sensor that makes an image soft or sharp.

I'd like to see some tests regarding this.

ISO only affects the digital gain applied to the sensor (not sure at which point of the chain). When the light is low, signal to noise is lower, and when you amplify all of that you get to see the noise more than if your S/N was better. Noise doesn't make the image less sharp, but it makes it harder to see the sharpness behind all the crap, if you see what I mean.

Now, I can only speak about the 5D2 since that's the camera I researched before getting one, but I assume there's going to be similar work done on other cameras.
A number of people have graphed noise and DR against ISO, and found a few interesting things:
- noise levels at ISO 200 are exactly the same as at ISO 100.
- intermediate ISOs (in third stops: 125, 160, 250, 320, etc.) are derived from native ISOs (the 100, 200, etc. values) by taking the nearest native ISO and either pulling it up or pushing it down a third of a stop. As a result (*), the third-stops immediately above a full ISO are more noisy, and the third-stops immediately below are less so.
- since ISO 200 is the same as ISO 100, and since ISO 160 is cleaner than ISO 200, it stands to reason that ISO 160 is actually cleaner than ISO 100.
- in fact, even ISO 640 is as clean as ISO 100, which still kind of blows my mind.
- however, DR also suffers from higher ISO values. It turns out that you get the most DR out of ISO 100, with ISOs 320, 200, and 160 close behind in descending order.
- after ISO 1600, noise and DR have a linear relationship with ISO.

In conclusion:
- below ISO 800, you have no reason not to shoot at 160, 320, 640 instead of 100, 200, 400. In fact, unless you're going for the absolute least noise possible, you have no reason to shoot under 320.
- for some reason, ISO 1280, while way better than 1600, is worse than 800.
- at 1600 and higher, it's all linear.

HOWEVER, that applies to noise in absolute terms - not signal to noise. And that's where it gets fun: for the same scene, at ISO 100, you have twice as many photons as at ISO 200, since you're exposing twice as long. So while the noise at ISO 160 may be, what, 15% lower than at ISO 100, the signal to noise ratio shows a clear advantage for ISO 100 since the signal is, in fact, twice as strong.

So yeah, that's for the 5D2. It's all very interesting, but after all this I just choose to stick with the native ISOs. Maybe in certain fairly specific conditions you could squeeze a little extra juice from intermediate ISOs, but to be honest even at native 1600 on my 5D2 I don't feel noise is a real issue in practice, so why bother?


Sources:
http://home.comcast.ne​t/~NikonD70/Charts/RN_​ADU.htm
http://home.comcast.ne​t/~NikonD70/Charts/PDR​.htm
http://dokumentaryfoto​grafr.blogspot.fr …o-settings-and-noise.html
http://www.pages.drexe​l.edu …kIITest/5DMarkI​ITest.html
http://photocascadia.w​ordpress.com …d-mark-ii-iso-noise-test/
http://shootintheshot.​joshsilfen.com …canon-hd-dslr-native-iso/


(*) I say "as a result", but I don't actually understand why. I assume native ISO is gain applied at one part of the processing chain, and the third-stop gain adjustment is applied at some other point in a different fashion, because otherwise it'd just be a linear relation.

Ah yes, one point of interest many people forget to mention is that increasing ISO reduces your dynamic range, usually because in modern digital photography it's somewhat unobtrusive.

There actually exist two types of sensor designs when it comes to ISO handling, and the first kind you're already used to, native ISO starts at 100/200 and ramps from there. In this design, the ISO is fixed and can't be changed after the fact in a RAW processor and maximum attainable dynamic range is usually limited, but the slope at which you lose dynamic range and gain noise with ISO increase is relatively shallow.

The second type is typically used in film cameras like the RED and Alexa, where the native ISO is typically ISO800, can be changed after the fact, and the maximum DR at native ISO level is very high - a full 14 stops usually; the newly announced RED Dragon even claims 15+ stops DR! The downside to this design is that any change to ISO either way results in a linear (1:1 stop) decrease in dynamic range, and noise performance also, ISO1600 is exactly twice as grainy as ISO800.

It's quite impressive that the D800 has almost 14 stops DR at ISO100, but film cameras were already doing that at ISO800! It's not surprising that some photographers are actually buying Epics for photography use, although you have to pay closer attention to making sure the light of your scene gets as close to perfect exposure as possible at ISO800 if you want all the DR you can get, on a film set this is almost completely irrelevant.

A lot depends on the subject.

5D2 at 6400, hand held. No pp.

.

armis wrote in post #14869475
- intermediate ISOs (in third stops: 125, 160, 250, 320, etc.) are derived from native ISOs (the 100, 200, etc. values) by taking the nearest native ISO and either pulling it up or pushing it down a third of a stop. As a result (*), the third-stops immediately above a full ISO are more noisy, and the third-stops immediately below are less so.

I see this mentioned quite often. Isn't it simply the same as exposing to the right though?

It is understood that underexposing and pushing the processing in post increases noise, and overexposing and pulling in post can lessen noise, which sounds like exactly the same thing. The only difference being that the third stop ISOs under/overexpose then adjust the processing in camera?

If you habitually ETTR, is setting a stop one third under the native stop used going to have any effect, considering that you are overexposing anyway and pulling in post?

Under some circumstances, a higher ISO may result in lower noise. See here https://photography-on-the.net/forum/showthre​ad.php?t=730218 for a really, really exhaustive discussion and explanation - the posts by Daniel Browning are the critical ones for the explanation.

arkphotos wrote in post #14868879
Just to be clear - higher ISO does not lead to more light being absorbed. That is a matter of aperture and shutter speed. ISO does control the 'gain'.

I just wanted to be clear for any new folks that may not be clear.

I wish I could answer your question & hopefully someone with greater knowledge will.

<My speculation>
I have heard similar claims - that native ISO can be preferable to lower ISOs. That makes some sense to me, as the less you do to the signal the better. I think noise/grain and sharpness are intertwined. But for me, I do not notice any degradation from using ISO 100 or 200. There is some threads here that discuss the more technical aspects that may interest you below. They center on pixel size, but some of the discussion may be useful.
https://photography-on-the.net/forum/showthre​ad.php?t=747749
https://photography-on-the.net/forum/showthre​ad.php?t=706255

Ah well, "gain" seems a more legit word...

I might as well post test shots at different ISO speeds...

As pointed out already, high ISO grain in film is due to the fact that larger grains of silver are more sensitive to light, the same as how larger pixels are better at gathering more photons of light than small pixels.

In the case of digital, so-called 'grain' is actually DIGITAL CAMERA NOISE (I wish folks would quit calling it 'grain'!)...the base level of circuit noise has fewer photons of light to create a signal which is relatively weak in strength, compared to the signal created by lots of photons found in bright light. Post processing programs may have features like "Luminance noise" adjustments to reduce digital camera noise in some pixels vs. other pixels. But the pixels themselves are fixed in size, unlike photosensitive silver grains.

I have read in the past about full ISO settings (100, 200, 400, etc.) having fundamentally less noise than intermediate settings like 125 and 160. I really haven't bothered to confirm this (or not) simply because the intermediate values were necessitated by film emulsion compromises in design, in the playoff of sensitivity vs. grain size. So it is simply more convenient for me to have fewer choices to dial thru, as intermediate values are not really as necessary in digital as it was in film.

sandpiper wrote in post #14870173
I see this mentioned quite often. Isn't it simply the same as exposing to the right though?

It is understood that underexposing and pushing the processing in post increases noise, and overexposing and pulling in post can lessen noise, which sounds like exactly the same thing. The only difference being that the third stop ISOs under/overexpose then adjust the processing in camera?

If you habitually ETTR, is setting a stop one third under the native stop used going to have any effect, considering that you are overexposing anyway and pulling in post?

Yes if you are shooting RAW/ETTR then use fullstop ISO values. You will be doing the same as the camera will be doing, but with more control. If you are shooting in camera JPEG then using the two third stop ISO settings will give an effect close to ETTR.

Alan

KirkS518 wrote in post #14868902
I'd like to see some tests regarding this.

I wonder what subject fits this ISO test...

Wilt wrote in post #14872196
I have read in the past about full ISO settings (100, 200, 400, etc.) having fundamentally less noise than intermediate settings like 125 and 160. I really haven't bothered to confirm this (or not) simply because the intermediate values were necessitated by film emulsion compromises in design, in the playoff of sensitivity vs. grain size. So it is simply more convenient for me to have fewer choices to dial thru, as intermediate values are not really as necessary in digital as it was in film.

I also haven't tested the intermediate settlings like 125,160 etc,, I only use full increment settlings.

Apparently, the intermediate settling are software generated.

All shot wide open with my 50mm

ISO 100

5184 x 3456

EMBED PREVENTED, IMAGE TOO LARGE: http://img215.imagesha​ck.us/img215/8813/img9​294b.jpg Click here to see our image rules.

5184 x 3456

EMBED PREVENTED, IMAGE TOO LARGE: http://img35.imageshac​k.us/img35/7835/img929​5j.jpg Click here to see our image rules.

5184 x 3456

EMBED PREVENTED, IMAGE TOO LARGE: http://img19.imageshac​k.us/img19/2665/img929​6k.jpg Click here to see our image rules.

5184 x 3456

EMBED PREVENTED, IMAGE TOO LARGE: http://img267.imagesha​ck.us/img267/7362/img9​297p.jpg Click here to see our image rules.

5184 x 3456

EMBED PREVENTED, IMAGE TOO LARGE: http://img207.imagesha​ck.us/img207/4010/img9​298.jpg Click here to see our image rules.