The all-new EOS metering test

Test scene

This is an update to the post I made a while back on the relationship between metering, histograms and camera settings for my Canon EOS 20D. The update is for a repeat of that testing but for my new EOS 40D.

With the conclusions I drew about the 20D histograms, I realised I'd want to optimise the settings on the 40D as well. A few initial test shots also made me realise that the settings are substantially different on the 40D than the 20D and so I'd need to do a whole suite of testing before I was going to be happy using the camera in anger.

So what did I do? I effective repeated the testing I did before by shooting a whole lot of parameters on a fixed scene in constant light, all shooting in a batch. As before, a nice high contrast scene to push the highlights. I tested variations in contrast, saturation, WB and highlight tone priority (HTP). All RAW files converted in 4 different software packages: Canon's DPP, CaptureOne 4.0, Lightroom 2.0, Lightzone 3.4.

What were the results? Largely the same as before: contrast setting is the key factor, white balance acts largely as a blue-shift. However, the contrast settings have a very different effect in the 40D than the 20D - reflects, I presume, the updates Canon have done to their firmware with the newer processors.

In all cases the RAW files are identical. Same scene, same light, same metering, same basic data. What you'd hope from RAW. This also means that all testing is absolutely consistent.

I found that a Contrast zero (C-0) reflected the same range in histogram as in RAW: 0-100% show for the JPEG reflected 0-100% in RAW. This means the setting has been tweaked to offset the limited DR of JPEG to match the overall tone range of RAW. That's actually quite nice - it means out of the box performance is very consistent. C+4 is very strong. About 1-stop difference between the indicated over-exposure (flashing highlights) and actual RAW highlight point. Too much headroom for me. At C+2 there is about 1/3 to 1/2 stop difference, which is nice. This is also consistent with how I've set up the 20D. This level means I can push exposure until the review gives some flashing highlights and know I've nailed ETTR.

Obviously, negative contrast settings go the other way. Even at C-2, there is a tendency for the RAW highlights to saturate before the histogram does. Not good.

As for other settings, no real impact. I've been testing the HTP mode as well to see how much headroom it actually gives in RAW conversion. With DPP I get a lot. But then the entire dynamic range & tone curve is adjusted. With all the others I've checked this with 2 types of exposure. Firstly just to the limit, but no blown highlights in the RAW. I then check exposure differences.
Second with just over-exposed and using recovery tools. I then compare degree of recovery required to just clip highlights.
In both sets, it's about 1/3 stop difference. Recovery tools typically only have 1-point setting differences, EV compensation 1/3-stop offset difference. This is nowhere near the 1-stop claims. More investigation required here. Contrast parameter settings seem to have more influence in judged exposure than HTP.

My conclusions:

First-up, as before, setting in-camera are important even for RAW shooters when evaluating exposure with histograms.
Second up, my personal settings. I'll be using contrast +2 for normal shooting and contrast +4 for low light (I like to slightly under-expose high ISO settings).
Third, I'm not touching HTP until I've got a better handle on its effect and that means more testing.

I've also got a bunch of conclusions from this about JPEG shooting but that is a whole other post.