I setup a tripod across my office pointing at a bookshelf. This was done at night, so light source didn't change and was a halogen bouncing off the ceiling. The cameras sit slightly different on the tripod, but the frames and crops are fairly well aligned. Camera settings were full manual (including focus) for all tests and timed shutter release was used. White Balance was Tungsten and Aperture was kept fixed at F/2.5 on the 50mm f/1.4 prime lens. I was about 10 feet away, so the expected depth of field is about 14 inches ... but if you pixel-peep, it's much smaller! Shutter Speeds were 1/25 second (ISO 400), 1/50 (800), 1/100 (1600), and 1/200 (3200). CF functions were defaults including CFII.3 on the 40D for no Highlight Tone Priority. To keep it simple, in-camera generated JPEG's are used; i.e. if one fiddled with RAW output, you could probably improve on this. Both cameras showed proper exposure in all shots, although note the XTi was slightly darker, especially at ISO 1600 - my guess is that really isn't quite 1600! ;-)
Some folks wondered if depth of field issues were making analysis harder, so I moved the halogen light to directly illuminate the bookshelf (yes, this makes for some strong shadows) and shot at same conditions as above, but at F/8.0 and went down to ISO 100. Shutter Speeds ended up being 2/5(100), 1/5(200), 1/10(400), 1/20(800), 1/40(1600), 1/80(3200) and both cameras oscillated between showing proper exposure and -1/3 stop. Depth of field increases to 45" ... but again, if you pixel-peep, it's less.
ISO 100 - 200 - 400 - 800 - 1600 - 3200 - Canon 40D with NO Noise Reduction - CFII.2=0
ISO 100 - 200 - 400 - 800 - 1600 - 3200 - Canon 40D with Noise Reduction - CFII.2=1
ISO 100 - 200 - 400 - 800 - 1600 - 3200 - Canon 40D with Photoshop 7-50-80-0 Noise Reduction
Update: I based my "one stop better" comparing against the
in-camera noise reduced 40D images. It has been correctly pointed out that's
not a fair comparison ... which is correct. I still think that there
is probably a third stop advantage to the 40D comparing to the
non-noise-reduced images. BTW, enabling in-camera noise reduction results
in a substantial decrease in burst depth - watch the
Canon 40D continuous shooting video.
For grins, I used Photoshop to apply a 7-50-80-0 noise reduction filter
to the non-noise-reduced images ... as can be seen, these are argueably
better than the in-camera noise reduction.
F2.5 Shots - shutter speeds of 1/25(400), 1/50(800), 1/100(1600), 1/200(3200)
Bytes Filename 4107440 40d-iso-400-nonr.jpg 3668818 40d-iso-400-nr.jpg 4506896 40d-iso-800-nonr.jpg 3993523 40d-iso-800-nr.jpg 4801448 40d-iso-1600-nonr.jpg 4073439 40d-iso-1600-nr.jpg 5827466 40d-iso-3200-nonr.jpg 4935976 40d-iso-3200-nr.jpg 4187685 xti-iso-400.jpg 4519588 xti-iso-800.jpg 4930530 xti-iso-1600.jpg
Bytes Filename 3684562 40d-iso-100-nonr.jpg 3394033 40d-iso-100-nr.jpg 3889114 40d-iso-200-nonr.jpg 3485773 40d-iso-200.nr.jpg 4244904 40d-iso-400-nonr.jpg 3812644 40d-iso-400-nr.jpg 4587469 40d-iso-800-nonr.jpg 4048430 40d-iso-800-nr.jpg 4834011 40d-iso-1600-nonr.jpg 4072288 40d-iso-1600-nr.jpg 5825260 40d-iso-3200-nonr.jpg 4937114 40d-iso-3200-nr.jpg 3829573 xti-iso-100.jpg 3989138 xti-iso-200.jpg 4308828 xti-iso-400.jpg 4582787 xti-iso-800.jpg 4929231 xti-iso-1600.jpg 5070794 xti-iso-1600-faster-aperture.jpg 5042598 xti-iso-1600-slower-shutter.jpg