Use of high ISOs necessary?

[pwnage101]

I have taken two shots with my Nikon D80 + 50mm lens. Each at f6.3 and 1/40s, but one at ISO 400 and the other at ISO 3200. The ISO 3200 shot was correctly exposed, and the ISO 400 shot was increased by 3 stops via ACR. For the purpose of this thread I will not label the following images, but there is full exif data available if you feel an urge



Maybe it will become immediately obvious, maybe not. Personally, I find it hard to tell the difference. I guess my question is this: why should I even bother with high ISO's if it just clipps highlights? I am actually quite disappointed after doing this test as I think back on certain occasions when I accidentally overexposed during high-ISO night events. Why not just shoot between 100-800 and just fix in ACR accordingly?

However, this may not be a completely fair test. On DXOMark, the SNR 18% graph of the Nikon D80 has a fairly harsh downward curve as compared to other cameras. This evidence is very much in accordance with my test, but shows that my philosophy does not hold with other cameras, especially those shown to have a linear db ISO relationship in their SNR 18% graphs. After all, this is just one test and two images.

[peterb533]

Hi Troy,

On my monitor the high ISO image looks more convincing in that the blacks are blacker and there is a greater tonal range. I guess that you have had to use the fill light slider to bring up the mid-tones and the result looks compressed. I'd vote for the high ISO shot despite the slight increase in noise.

Peter

[pwnage101]

Blacks I can always alter. It's the noise I'm referring to (which you already addressed). Close inspection does reveal a slight decrease in dynamic range.

[revi]

I get the impression that the first image has slightly less noise and is slightly smoother and softer than the 2nd. (Yes, i did check the ISO values, won't tell them here to avoid spoiling the fun for others).

That said, iirc, the camera does exactly what you did when changing ISO value (i.e. amplify the signal). However, in-camera amplification is done before digitising the signal, which means you will have more bits for the shadows, and lose range in the highlights (am I clear here?). I'd expect that if you did the same test with 100 and 3200 ISO, you'd see posterisation/banding in the shadow regions: scaling up 5 EV (going from '100' to '3200 ISO' in PP) you basically drop 5 bits of information, going from 12 to 7 bits (or 14 -> 9).

Another factor is noise: not sure how to express this, but I'd expect a difference in noise 'quality' between in-camera amplification and PP amplification. In-camera is analogue, so continuous values, which means that visible noise has a minimal amplitude of 1 bit in the final image. PP amplification is digital, so if you amplify 4 times, the weakest visible noise signal will have at least 2 bits difference, making it appear harsher.

So I think it's still best to use the lowest possible ISO if you need the darker parts of the image, and then you'll have to accept a few blown highlights (which you'd have anyway after PP...).

Hope this is clear,

Regards,

Remco

[Colin Southern]

Hi Troy (and others),

A few notes you might find helpful ...

- Increasing the ISO does increase amplifier gain prior to A/D conversion, but only for whole ISO amounts -- with odd amounts there is also a component of digital manipulation.

- Increasing the ISO won't cause blown highlights per sec, but it does decrease the dynamic range. Essentially the net result is the noise floor moves up, but not in a linear way) which isn't an issue with a properly exposed reflective scene that only requires around 4 to 6 stops (camera will typically have around 11 or 12 at base ISO, and around 8 at very high ISOs).

- The net result in under-exposing and correcting in PP is that the highlights are far enough away from the noise floor to escape relatively unscathed, but you'll see significantly more noise in shadow areas (as you can see here).

Hope this helps

[pwnage101]

I just read in another thread that HI.x ISOs (like my D80's 3200) are just digitally manipulated from the preceding ISO step. Doesn't this mean that I have no real reason to go higher than 1600 unless I shoot jpeg?

[Colin Southern]

I just read in another thread that HI.x ISOs (like my D80's 3200) are just digitally manipulated from the preceding ISO step. Doesn't this mean that I have no real reason to go higher than 1600 unless I shoot jpeg?

Hi Troy,

Although it's a "digital ISO", it still allows the same shutterspeed / aperture / metering etc as though it was a "real ISO", so it can be convenient.