5DmkII ISO performance test

[Rentham]

Originally Posted by sk66

So you could be better off using ISO 800 instead of ISO 3200, underexposing by 2 stops, and fixing exposure during RAW conversion. The post adjustment would generate no more noise (possibly less) and you would have ~ 2 stops more dynamic range at ISO 800. The last "beneficial" ISO for a 5D II is 1600 where the A/D converter hits the noise floor.

That quote from this thread piqued my interest. I did a quick and dirty test to see what the real world difference would look like.

Settings for both shots are the same except for ISO - 800 and 3200. Menu option for High ISO noise reduction is turned off. Shot in RAW. Only RAW adjustment is pulling up exposure slider on the ISO 800 shot exactly 2 stops.





Available 1200px wide on Flickr so you can get a better feel for the noise in each.

[OsmosisStudios]

Umm, I dont know what sk66 is on: its almost universally frowned upon to increase exposure in RAW with the sliders because it makes any grain that's there stand out like it was on fire. Thats why people advocate shooting a higher ISO and overexposing (a bit) and then pulling in post. ETTR and all that.

[sk66]

What's your opinion of the results?
At this size (as viewed on Flickr) the second image, which is the ISO800 pushed 2 stops in post, appears smoother and with more dynamic range than the ISO 3200 image. At least to me.

[sk66]

Originally Posted by OsmosisStudios

Umm, I dont know what sk66 is on: its almost universally frowned upon to increase exposure in RAW with the sliders because it makes any grain that's there stand out like it was on fire. Thats why people advocate shooting a higher ISO and overexposing (a bit) and then pulling in post. ETTR and all that.

While it is true than many recommend "exposing to the right" etc etc...
Once you've hit the noise floor for your A/D converter (generally ISO 800-1600) you are getting more noise than signal from the A/D converter and loosing dynamic range.

ISO is something of a fallacy with digital....it has no affect on the amount of light captured. A higher ISO resulting in a faster SS actually reduces the amount of light captured and reduces the signal to noise ratio making images noisier.

It's not the "ISO" which makes images noisy, it's the lack of light captured.

[OsmosisStudios]

Originally Posted by sk66

What's your opinion of the results?
At this size (as viewed on Flickr) the second image, which is the ISO800 pushed 2 stops in post, appears smoother and with more dynamic range than the ISO 3200 image. At least to me.

I see the opposite. Better contrast, colours, and deeper shadows in the 3200 image. I can clearly see the grain in the 800 image that's been pushed: not so on the 3200 (on the images posted in the forum.

Looking at them in Flickr, The 800 image just lacks that "oomph", though the 3200 image would be almost unuseable, though in this case because of smearing of details.

[Rentham]

Just to be clear, and for those who don't feel like looking at the EXIF, the top image is ISO 3200. The bottom is ISO 800 pushed two stops in Lightroom.

Personally, I see better color and contrast in the ISO 800 image. The 800 image has less overall noise, but is clearly worse in certain spots - such as the lighter area immediately to the left of the glue stick. Could some RAW adjustments to the 3200 image make it better? Maybe it's a wash.

Neither shot is great, but I don't find either to be unusable. In a situation where you feel compelled to go ISO 3200 or take the shot underexposed by 2 stops, you wouldn't expect amazing fidelity in any case.

[sk66]

The only benefit of pushing is dynamic range...The differences can be much more dramatic with even higher ISO's. The relative noise levels in the pushed image will be dependent upon the RAW converter used and they keep getting better. So as you update your software your results from pushing can improve where the A/D converter will always be what it is.

I would say this is a fair test of what I was stating. But I would also venture that further adjustments using the tone curve sliders etc would show even more difference. (Also 14bit RAW helps a little).

The main benefit of this is in situations where the required ISO to get adequate SS is unusable (lesser cameras/ crop sensors) hit the signal floor earlier. It's also why 800 is usually the limit used by most celestial photographers.

[inkista]

Actually, I preferred the 3200 shot. The pushed 800 shot had what looked like banding issues to me.

Curious. Did you try properly exposing an iso 800 and then overexposing a 3200 shot and pulling it down two stops and comparing?

[Rentham]

No I'll leave that in your capable hands

[inkista]

[grin]. No thanks. It would cut into my play time.