No need to get defensive. I was merely making the point I learned in chemistry class in college
I did not want to come accross as being defensive. Your questioning the validity of the claims was partly justified, and I gave the proof. (The issue qwith ISO 3200 is simple arithmetic.)
And pixel-peeping has that tendency as well - there are quite a few pixel-peeping articles which come to a similar conclusion
Pixel peeping alone means nothing. I am analyzing the
unadultered raw images, to exclude the effects of different raw processors, different settings, etc. I wonder if anyone cames to a different result
on this factual basis.
As for your own DR calcs, how would you qualify that "slight" edge of the 40D? Significant?
Generally, I would say the difference is far less important, than the difference cause by not perfectly exposing (ETTR), or not selecting the most suitable ISO, etc.
I'm personally mostly concerned about shadow noise in long (30-300 second) night-time exposures
This is a special consideration; cameras need to be tested specifically to this purpuse. I can't help much, but if you cam convince some people to make test shots (raw) with appropriate settings, I can evaluate the results.
Several people contributed to a comparison with 5sec blackframe shots at ISO 1600, between Nikon D300, Canon 1DMkIII, 350D, 20D and several 40Ds (I myself have a 20D and a 40D). The results are telling, but
1. that was only 5sec,
2. the 450D was not even on the market that time.
The test shots were made with lens cap (or better: body cap) on, viewfinder sealed. Only for a feeling, how the results were evaluated, see following manylayer files with captures, crops (all from the black frames).
Histograms, raw imagesCrops from ACR converted blackframes