I dissected a sample raw file from imaging-resource.com at ISO100, no NR, with IRIS.
It appears that the non-illuminated pixels are included in the file.
Found some numbers.
The Black Level offset is changed to 2048 (used to be 1024)
The read noise Sigma is around 4,7
The histogram maxes out with a barrier at 13580.
BL subtracted: Saturation at ADU 11532
Dynamic Range: 11,26 stops.
Dynamic Range normalized to 8 MP: 11,85 stops
Note that this doesn\'t tell anything about photon collection efficiency, as the system gain is unknown.
Neither does it tell anything about SNR.
wow sigma 4.7 is crazy low!!
Then again they did bump BC and actually wow this is still a lot lower than the Nikon D3X and D90.... so canon still can\'t deliver great ISO100 shadows?? Or nikon did something that tricked DxO??
Only small, lame consolation is it makes me not feel as bad about my 5D2 ISO100. This is better though since it is essentially the same DR as the 5D2 but, by all early reports, with less x-hatching. So it may well deliver, in some ways a better ISO100-200. Of course SNR will still be worse and some other stuff, but interesting if 5D2 had had this technology in the sensor it would probably do 11.5 and 12.2, well that is not wildly more only an extra 1/3 stop DR, but with less banding maybe 1/2-2/3 more usable?
anyway, it looks to be 1/2 stop better DR at low ISO than the 40D/50D and about the same as the 5D2. Early guess is at least 2/3 stop worse SNR than the 5D2 though, but maybe up to 2/3 stop better SNR than the 40D/50D (not based on serious calcs, just some peeping at IR images).
here is the data I got using IRIS for 5D2,50D,40D,20D(something seems off with the 20D data maybe I did somthing wrong with the 12bits or I had a bad copy):
Sep 01, 2009 at 06:09 PM
Previous versions of skibum5's message #7482541 « 7D noise vs 50D numbers »
