Dynamic Range (contd.)
The second part of the dynamic range story is of course the additional headroom available in RAW files. Experience has told us that there is typically around 1 EV (one stop) of extra information available at the highlight end in RAW files and that a negative digital exposure compensation when converting such files can recover detail lost to over-exposure. Much experimentation lead us to eventually use Adobe Camera RAW for conversion with the settings show below to retrieve the maximum dynamic range from our test shots.
It would appear that the EOS 5D's RAW files contain just over one stop of extra highlight detail, although interestingly this was pretty much unavailable in any Canon RAW converter because of their conservative approach to potentially incorrect data. At the shadow end it is possible to lift shadow detail however this also increases the visibility of noise. Another interesting artifact here was that using the default ACR conversion we lost a tiny bit of highlight range and about a stop of shadow range, it appears that ACR's default 'Medium' tone curve is a little more contrasty than the camera curve and also that ACR is doing less noise supression in shadows (hence it hit our lowest acceptable SNR earlier than in-camera JPEG).
- ACR Default: Exp. 0.0 EV, Shadows 5, Bright. 50, Contrast 25, Curve Medium (Default)
- ACR Best: Exp. -1.25 EV, Shadows 0, Bright. 50, Contrast 0, Curve Linear
The issue with highlight data above the normal 'clip' point is that the RAW converter can not be sure that it is 100% color correct, it appears as though rather than risk false color the Canon converters simply discard it. Adobe Camera RAW (and some other third party RAW converters) however do attempt to reproduce as much extra detail as possible and where a single channel is completely clipped they instead produce a gray of the correct luminance (seen below).
Real life example
Step wedge shots are great for analyzing the optimum response from a camera but nothing beats a real life example to get some idea what all this means. Below left you can see an example of an EOS 5D image with a large area of over-exposure. Using Adobe Camera RAW we produced two additional versions of this shot, the first with the 'optimum' -1.25 EV compensation and the second with a -2.00 EV compensation. While initially the -2.00 EV shot looks better a closer look shows that where ACR could not correctly calculate the color of a pixel it instead turns it gray (with the correct luminance), some of the brickwork has its correct color, other bricks are gray. The -1.25 EV shot shows no less detail it's just about optimum for the amount of additional information available in this RAW shot.
|Default conversion (original)||-1.25 EV, lower contrast||-2.00 EV, lower contrast|
Different RAW converters
Lastly it's worth comparing a couple of other RAW converters. In these cases we used -2.00 EV, despite the fact that it's more than the available headroom to explore how each converter deals with the lack of complete (accurate) data for some pixels. As you can see Adobe Camera RAW produces the most natural looking image with the best 'guess' for the value of pixels with inaccurate color data, RawShooter comes in second and Canons DPP takes the conservative approach, if it can't be sure of the pixel color then it just leaves it white / gray.
|Adobe Camera RAW, -2.00 EV||RawShooter, -2.00 EV||Canon DPP, -2.00 EV|