EOS 5D
v
EOS 5D Mark II
PART 4 - Detailed pixel sniffing continued - Colour (18 January 2009)
Testing colour with the White Wall of Doom
I found out a while ago that the ultimate torture test for colour reproduction is absence of colour. For example whilst metamerism on colour inkjet prints may be fairly unnoticeable on normal colour prints, if you print in black and white using colour inks you are likely to find that a result which looks great under Tungsten light looks bright green in daylight. I never knew that Fuji Crystal Archive colour paper also displays slight metamerism until I printed an A2 sized black-and-white and put it on my office wall. It goes slightly magenta under office lighting but looks perfectly neutral in natural light. The colour prints which hang on either side look perfect under all lighting. They must have the same colour shift but in a colour print it is not noticeable.
So whilst Ken Rockwell has his Wall of Shame, my torture test for colour is the White Wall of Doom. The bad news is that the Mark II fares really badly.
The top image is from the Mark II and the bottom from the old 5D. My method was to choose the lens with the worst vignetting I have (the 24-105 f4L) and select f4 in aperture priority mode, ISO 100 and allow the camera to choose the shutter speed. One of the reasons the Mark II result looks worse is that it chose a shutter speed of 1/6 second compared with the 5D's 1/3 second, so in comparison it underexposed by a stop. I was slightly surprised to note that the metering response is different on the Mark II.
I can see all sorts of colours in the Mark II result, from green to reddish-brown. By contrast the 5D looks fairly neutral, with perhaps a hint of green as the corners drop away to blackness. Note that the RAW files were processed with no adjustment in DPP - no noise reduction, exposure compensation, saturation adjustment or anything else.
So that the old 5D doesn't have a (fair?) advantage due to its exposure meter doing a better job in these circumstances, I tried again with the 5D also shooting at 1/6 second f4:
Again the Mark II is the top photo. Again I see more colour variance in the new Mark II result, although with equalised exposures it is less obvious. By boosting saturation on the same images in Photoshop to 50, the difference is much more obvious:
As we are engaging in pixel sniffing to a heinous extent, I would say the 5D Mark II result (top) smells of strawberries at the right hand side and kiwi fruit in the middle, whereas the old Mark II has perhaps just a slight whiff of peach towards the right hand side. What does this tell us? Only that in extremis, the 5D has better colour than the 5D Mark II. Given that these tests are at ISO 100 and all I've done to highlight the issue is boost saturation a bit in the bottom pair of images, it is quite a disappointing result. In my view this is a much bigger issue than the "black dots" which caused a bit of a storm in a teacup before Canon fixed it last week. Black dots would certainly disappear when re-sizing a 21 megapixel image to 640x480, but colour problems are visible in all output at any size, even little web sizes.
In real life, like distortion which looks awful on pointless brick wall photos, it probably matters very little. But chalk up one victory for the old 5D!
Real life colour
Looking again at the resolution comparison, which was taken in very dull wintry light, I noticed that again the Mark II had underexposed, producing a very subdued looking image which looked less colourful than the 5D:
The original 5D image (bottom) looks much more colourful, especially the browns of the bricks, although this may be largely down to the better exposure which the 5D has chosen. The two images below illustrate two points. First, quality of light makes a lot more difference than camera. Secondly, in the sort of light you might actually want to take real pictures in (ie the good sort), I can't detect any difference in colour rendition between the 5D mark II (top) and 5D (bottom).
Part 5 - noise (and a return to detail)