It's made to challenge MF cameras, and generally to be used at ISO 800 or lower.
[a href=\"index.php?act=findpost&pid=162888\"][{POST_SNAPBACK}][/a]
What a camera
is is what is relevant, not what some abstraction of what it is supposed to be is. Regardless of what it is "supposed to be", the ISO 1600 pixel read noise of the 1DSmk3 is anywhere from slightly less to about the same as the D3 (variations in samples), and that is with the 1DSmk3 possibly being really ISO 1900 to 2000, and having 74.4% more pixels, which has an effect of reducing the image noise to 75%, relative to the full image of the lower-res camera.
The D3 captures about 6700 photons at RAW saturation at ISO 1600; the 1DSmk3 about 3750. Even if both were truly ISO 1600 (in the sense that saturation occurs at the same exposure level), that means that in the area covered by each D3 pixel, the Canon collects about 3750*1.744 = 6540 photons, which is a tiny bit less than the D3 per unit of area. So, image shot noise is about the same, and if the the Canon is really 1.2 to 1.25x the stated ISO, as they usually are, the Canon has less total shot noise. This is the noise that runs all the way up into the midtones and highlights, the only significant noise in a well-exposed scene. The quality of shot noise does
not vary from camera to camera (as read noise can). It is a property of, and, in fact the very fabric of light. The photons captured is all that matters for shot noise, and this shot noise is a major determinant of DR for very conservative standards of DR (read noise dominates for more liberal standards).
Read noise is a little more complex. Read noise is not always random in 2 dimensions; it often has components that are 1-dimensional, either horizontal or vertical. In the case of these two cameras, banding noise is present but weak, mainly horizontal in the Canon and vertical in the D3 at high ISOs. 2-dimensional read noise in dominant in both of these cameras, in the specimens I've seen.
When is someone going to shoot the same scene, with the same lens, same distance, same Av and Tv values, complete with an out-of-focus color checker, including a pushed-to-ISO-100,000 or so shot, from both of these cameras, in addition to a well-exposed ISO 3200, and provide the RAWs so they can be processed *exactly* the same way in DCRAW, IRIS, or ImagesPlus?
This is the kind of thing that is needed to settle this issue of what the cameras are capturing once and for all, but no one wants to do something this straightforward and simple. Instead, speculation on images processed with slight-of-hand NR tricks is what people seem to like doing.