Comparing FF vs. APS-C

[neuroanatomist]

I will probably reshoot just the ColorChecker with one camera, varying the shutter speed with ISO instead of using filters to control exposure.

Ok, so I tried this. I used the 5DII and the EF 100mm f/2.8<span style="color: #ff0000;"]L Macro IS, framing just theGretag Macbeth part of theColorChecker Passport. A constant f/8 was used, and shutter speed was varied in tandem with ISO, from 1/15 s at ISO 50 to 1/8000 s at ISO 25600. Exposure was set just short of clipping the white square in the upper right, although the lettering did clip (and I mean just short of clipping, since I was able to adjust the position of the two goosenecks for a finer adjustment than 1/3 of a stop). FWIW, the exposure was 2/3-stop higher than metered.


(click image for a 1600 pixel wide version)





Compared to the previous test, the effect of increasing ISO on red desaturation is much less. It's still there, though. However, the effect of increasing ISO on image noise is also much less than in the previous test, because in this case the exposures were different, which affects the relative contributions of read and shot noise. From a noise perspective, the previous test was better, IMO - the ND filters I used to permit constant exposure settings on the camera simulated 'real world' dim light by incrementally dimming the illumination. In this test, there was plenty of light (which also applies to real world scenarios where you require a fast shutter to stop action in relatively bright light). I do wonder how much of the desaturation in the previous test was due to the ND filters, and how much was due to exposure-related factors...

[HDNitehawk]

John


Was your observation of red saturation made visually? Looking at the samples of both tests closer it seems to me the de-saturation is across the board. It is more noticeable and pronounced on the red but this very well may be because how the human eye see's the colors.


Rick

[neuroanatomist]

Yes, visually - looking at the red square, but also looking at other colors with a subtantial red component, e.g. the pink and orange squares, which shift to lavender and yellow as the red is subtracted out. The green channel also desaturates, to a lesser degree. The blue channel seems relatively resistant to the effect. This may be a result of the demosaicing algorithms that Canon uses - DxOMark shows an example of color responsiveness of the Canon 500D vs. the Nikon D5000, where you can see that the Canon's spectral response curves have substantial overlap between the red and green channels, meaning the RAW processing do a fair bit of work to unmix the colors (which makes me wonder if it really is mainly an effect in the red channel, since that would also impact the green channel based on the overlap).