Panopeeper wrote in post #7707433
This is not the question of plots but of taking and recording the measurements. It is quite time consuming; I got shots of a color checker sheet, and I pick the measurements on 26 patches, separate for the three raw channels.
Your efforts are certainly appreciated.
I did this for ISO 250, only as a demo. There is no point in it, because the camera does not have ISO gains for the 1/3 stop ISOs. 160, 320, 640 and 1250 are identical to 200, 400, 800 and 1600, respectively; the pixel values are simply multiplied by 0.8 (roughly 1/1.26, corresponding to 1/3 EV). Look at the fine histogram of an ISO 640 shot underneath: every fourth pixel value shows much higher "presence" than the others (the y axis is logarythmic, the small spike represents twice as many pixels with that value).
The full stop + 1/3 ISO steps are much worse: they are not from the next lower ISO step, as with the 40D, but from the higher steps. Correspondingly, the noise of ISO 250 is as much as with ISO 400.
The reason I mentioned it was that I'd seen (on another forum) some visible evidence that the noise performance of the 5D2 appeared to be better at 160, 320, 640, etc than it was on the corresponding 200, 400, 800, etc settings (essentially cleaner at 1/3 stop slower). I realize that's not exactly what you were measuring, but I was interested in some measured data that would go along with my observation. I haven't taken the time to visually test my own 5D2 for this yet. I'm still somewhat enjoying the clean performance in the 800-3200 range.
I'll look it over as well - thanks. I'll also study the chart you pasted in the thread. Lots of information there.