Now we see a fuller picture by looking at the colour channels as well as the luminance channels.

Again, larger (signal to noise in dB) figures are better and the best in each ISO setting are highlighted in blue and the worst in red.

The Nikon D50 stands head and shoulders above the rest here - no red numbers at all and quite a few blues. While the Olympus E-500 did very well in the luminance test, the colour channels bring it back down to earth with a bit of a bump. The Pentax *istDL doesn't show too well, especially compared to the Konica Minolta 5D and the Nikon D50, which both use the same Sony sensor. The Canon EOS 350D does surprisingly poorly as well, but we have even more DXO Analyzer data that may help to explain why the 350D appears to do better sample images than its test data indicates.

Interesting that these results are bearing out on the test bench what your closeups showed, the Nikon and Canon are streets ahead of the others.
Regarding the cameras using the same chip do you know who makes the processor for each of these? This, or the lens design, has to be the reason for the Nikon's superiority.
Tony

Hi Tony, the custom chips that do the image processing could be made by a number of companies. I only know for sure that Olympus uses a Texas Instruments custom chip.

For noise tests, the lens makes no direct contribution. Camera noise is unaffected by how sharp the lens is - it's entirely the product of exposure, sensor and image processing variables.

Ian

Ah, I was under the impression that the resolution of the lens did have an effect on the appearance of the noise, now know differently.

> Ah, I was under the impression that the resolution of the
> lens did have an effect on the appearance of the noise,
> now know differently.

Noise is usually measured against a featureless grey area. The amount of noise a sensor produces can affect resolution, but that's a different issue.

Ian

> Noise is usually measured against a featureless grey area.
> The amount of noise a sensor produces can affect
> resolution, but that's a different issue.

Hi Ian
Am I being thick? If noise can affect resolution are you not therefore better off with a high resolution lens resolving more lines per mm?
Tony

Noise erodes resolution - it's a one-way process.

The erosion of resolution by noise would, if you think about it, mean that finer resolution detail would be lost first, so I don't think a higher resolution lens would be of benefit.

Ian

Makes sense, thanks for the explanation Ian.

Hi Ian,
It appears that the Kodak sensor takes a big hit in the blue channel and doesn't do well in the red. I did read an article on the increasing use of flourecent lights for studio, because the provide more light in the blue channel, where the CCD & CMOS sensors need it the most. I wonder if your tests under different types of lighting would show any evidence of this in the noise world. While the studio flourecents are daylight balanced, I have been meaning to try it with house hold screw in versions for fun.
Bob

It's odd that the E-500's luminance noise is quite good and yet two of the three colour channels are quite bad.

I guess you can get good mono shots at higher ISOs with the E-500!

The problem with fluorescent lights is that their spectral response is not smooth. 3000K floods are used for the DXO Analyzer tests.

Ian