|
Yeah, Justin's image is pretty impressive! I think this is an issue I really want to spend some time testing thoroughly. The other issue I didn't mention was that dynamic range increases with stacking, but that can't compensate for saturated pixels because of reduced dynamic range in a sub. I also can't see any logic that says the signal will come out of the noise as you increase the gain (ISO) which stretches both signal and noise. The 'sweet spot' will vary with camera, but I really doubt I'd actually get more signal-to-noise from ISO1600 over ISO800 (or 400-640), as it logically doesn't make much sense, once there are fewer electrons to ADU points. But one day I'll prove to myself whether I'm right or totally wrong!
Craig Stark's article is only two years old, but the camera he used most in the test was a bit older (an EOS 450D), which as you say only pushed ISO to 1600. The question about Canon's treatment of darks is another one for testing too...
|