I have a question - does using ISOs that are not full stop increments degrade image quality?

i.e. ISO 500, 1250, etc. - do they look any worse than the ISOs that fall on the powers of two (100,200,400,800, etc.)

Thanks
-Glenn

It doesn't degrade IQ per se, but rather it's the in camera treatment of the RAW data that may increase/decrease noise and decrease dynamic range. If the base ISO is 100, then the camera shoots in 1 stop sensitivities at 200, 400, 800, 1600. Everything else is either pulled or pushed to simulate an ISO level of i.e. 160, 125, 640...

Which Mark 2? You have at least three to choose from.

Degrade image quality? You'd probably never see it, not even on the 5D, where the fractional ISOs are not exactly real. I don't know whether the 5D2 fractional ISOs are more real or not, but you'd still not likely be able to detect any difference in image quality unless you use rigorous tests with well calibrated equipment and superior testing procedure and analysis. I don't know about the other Mark II cameras.

Sorry about that - it's a 1D Mark II.

I had just noticed shooting at ISO 1250 that the results weren't quite what I expected. It's likely me, not the camera - but I wanted to ask just to make sure.

Thanks

I try to stay away from the fractional ISO. Maybe its just superstition but I prefer 1600 to 1250, etc.

I recently read a post from some1 who showed charts of various 1/3rd ISO levels VS noise levels for the 20D/30D cameras. You should find one for the mkII series....
Basically it showed that ISO 320, 640 had less noise levels than non-fractional ISO levels (200, 400). And I believe it to be true for xxD series...

TVRguy wrote:
I have a question - does using ISOs that are not full stop increments degrade image quality?

i.e. ISO 500, 1250, etc. - do they look any worse than the ISOs that fall on the powers of two (100,200,400,800, etc.)



For all Canon consumer-level digital SLRs, the hardware can only amplify signals at 1-stop intervals. ISO 100, 200, 400,... etc. All intermediate ISO values are done through software extrapolation, by shifting the RAW values forwards or backwards. So ISO 125 would be just ISO 100 RAW values pushed up 1/3 stop. ISO 160 would be ISO 200 RAW values pushed down 1/3 stop. This is true for all Digital Rebels and 20D/30D/40D series cameras. Intermediate ISO values are essentially worthless if you shoot RAW. You might as well shoot in 1-stop intervals and digitally adjust exposure during RAW conversion.

On most Nikon digital SLRs, all regular and intermediate ISO values are done through hardware amplification only. The amplifiers will actually process the incoming signals for ISO 200, 250, 320, 400..., etc. So there's no disadvantage to choosing intermediate ISO values at all.

On higher-end Canon models like the 5D, 1D, and 1Ds series, amplification is done through hardware as well. But it's done in two stages. The first stage amplifiers will bump up the signal to the nearest 1-stop interval (100, 200, 400, 800, etc). If you shot with intermediate ISO values, then the signal would then be passed through a 2nd stage amplifier to reach the intermediate values. Amplifiers are one of the biggest sources of read noise in sensors. The use of a second amplifier introduces quite a bit more noise than a single-stage amplifier would've done (like Nikon's). And so intermediate ISO values are still noisier than you would've expected.

In fact, the noise added from the 2nd stage amplifier is such that the 5D/1D/1Ds series may as well have done the intermediate ISO values through software extrapolation. In terms of signal-to-noise ratio, there's no practical difference. But at least they won't clip the highlights 1/3 stop early --- something that happens when shooting the consumer models at ISO 160, 320, 640, 1250.

This is exactly what I was thinking of. And it might explain why on some occasions I've seen results that really did look "pushed". I will try to avoid the intermediate ISO values and stick to the powers of 2.

Thanks all!!

-Glenn

OK - so next question - is there a way on my 1D Mark II that I disable the intermediate values? I don't see it as a custom function or personal function.

Thanks

TVRguy wrote:
OK - so next question - is there a way on my 1D Mark II that I disable the intermediate values? I don't see it as a custom function or personal function.

Thanks

No, you just have to not dial them in if you don't want them.. Since it is already difficult to change ISO on this camera, I would expect it not to be much more of a hardship

Gene

Glad the 40d has this Cfn...I disabled intermediate ISOs as soon as I turned on the camera for the 1st time (and the highlight priority thingy which cuts down fps).

GeneO wrote:
TVRguy wrote:
OK - so next question - is there a way on my 1D Mark II that I disable the intermediate values? I don't see it as a custom function or personal function.

Thanks

No, you just have to not dial them in if you don't want them.. Since it is already difficult to change ISO on this camera, I would expect it not to be much more of a hardship

Gene


Umm there is a CF that alows you to choose either full stop or 1/3 stop ISO values - default is full stop.

I was reading the ones to avoid re ISO 160, ISO 320, ISO 640, ISO 1250 and the ISO 250, ISO 500 and ISO 1000 are ok to use. That was for 5D, so don't know if it holds for 1D II.

You might be interested in a current thread at NSN on the same topic:

http://www.naturescapes.net/phpBB3/viewtopic.php?f=2&t=139621

LCollector wrote:
I recently read a post from some1 who showed charts of various 1/3rd ISO levels VS noise levels for the 20D/30D cameras. You should find one for the mkII series....
Basically it showed that ISO 320, 640 had less noise levels than non-fractional ISO levels (200, 400). And I believe it to be true for xxD series...

Beware that less noise in raw levels and less noise relative to a fixed signal in photons are two different things when the ISO is varied, since ISO changes the number of photons/raw level. The important number is the noise relative to the input signal, so there is a little math to do before concluding whether the noise at a given ISO is better or worse than the noise at another ISO.

ejmartin, I was referring to this.

http://forums.dpreview.com/forums/read.asp?forum=1019&message=19721647

LCollector wrote:
ejmartin, I was referring to this.

http://forums.dpreview.com/forums/read.asp?forum=1019&message=19721647

I thought you might be. It might look as though the noise is less at ISO 160. However, since all that ISO 160 consists of on the 30D is a digital manipulation of ISO 200, the read noise is the same as that at ISO 200 when referred to the number of photons it corresponds to. All that has happened is that the noise in raw levels at ISO 200 has been divided by a factor of 1.25, because the way that ISO 160 is generated is to divide all the raw values of an ISO 200 exposure by 200/160=1.25. The same noise represented by different numbers.