The great EOS metering test

After the discussion that Paul Butzi has started on exposure decisions and a decent metering system I thought I'd do some tests on the EOS metering & histogram.

As the histogram is based on a JPEG conversion, would any of the shot parameters affect the histogram? Are any of these parameters fed back to the metering?

2nd question is the easy one to answer. No. As one would expect. As the metering and histogram are entirely separate functions, there's no link. Pretty much the point, I think, that Butzi's been getting at.

First question is a little trickier. My first set of tests involved only using the basic parameters - contrast, sharpness, tone etc. I took shots including a floor lamp which gave a small patch of brightness in a quite dark room. Shot below 1/50th (mains frequency) so lamp flicker didn't affect the results. I commented on one of the threads that only contrast above zero had an effect. closer inspection of the files showed that lower contrast only seems to affect the shadows, and then only marginally. Other setting had no effect.

This means that by applying contrast +2, you get a slightly more conservative histogram (by about 1/2 stop) which should mean getting better exposure in tricky light.

Then I thought about colour balance. Lots to play with there. Scene at the top was the target - looking across the street from a aside window. Bright sky (always tricky for to 20D), bright patch from the sunlit part of the white building. Very constant scene otherwise (tram rumbling by too far in the dark tones to worry about). Shots were taken in colour temp white balance mode, everything from min to max in about 1000K jumps. Still capturing RAW, though.

Turns out that while colour temperature acts like a blue shift, it doesn't seem to affect the reported luminance (histogram), nor the levels of red & green. This from observing the RGB histograms when converting. That might be because the brightest part of the scene isn't the sky. Did seem odd, though. The colour shift can confuse some RAW conversions, ACR was the only one of 4 (Lightzone, Capture One and DPP being the others) that seemed unfazed by the camera settings. Set everything to the same conversion settings and all shots came out the same. Using the white point adjusters had no noticeable effect.

From now on, I'm going to be using the high contrast setting (even in RAW) and pushing exposure up until it just clips on the histogram (few flashing highlights). Seems that will get the best "expose to right". I might have to do a little more experiment with a scene where a blue sky is the brightest part just to confirm the white point settings.

UPDATE 20/11/2008: I've redone this testing using an EOS 40D