First off and sorry to throw over your straw man: no-one in this subthread says or believes in digital grain
Once again: as a courtesy tell who you are linking to on Youtube
No it is not 100% fact and I posted at least one incorrect statement earlier and showed why it was incorrect.
I can guarantee that John knows a load more about this than either you or the author of your video.
Well, you've had time to reread and contemplate what I wrote. Do you still feel the same way? Remember, I did not claim that at very low framerates, the same ISO looks very different, with SNR linked to fps. I was talking about framerates higher than are standard now, like you might get with these ~240Hz-300Hz gaming monitors, where frames would be integrating over time to smooth out frame-to-frame noise variations. Hard to do even on a 60Hz monitor, no less 30Hz like my current desktop, which only gives me 72Hz max at 1920*1080 for some reason, even though it can do 4K/30Hz, at 1.67x the bandwidth. Of course, the persistence of a monitor also affects the smoothing of a pixel over time.
I would suggest performing your own experiments instead of blindly following "authoritative sources".
It's not "film grain", but like film grain, it does have clumps of correlated noise of various sizes, especially the chroma noise, which depends on both the sensor/ADC noise, and the demosaicing style, as well. In the future, I will try to use quotation marks on "grain" when it is not actual substrate of grain. Surely you've seen those globs of red noise and green noise in images shot in the shade with low exposure - they group into clusters of red-biased pixels and green-biased pixels.
Then it is absolute exposure changing; not just ISO setting.
Actually, the vast majority of cameras have the least "error" for recording electron counts at their highest analog gain. It is only the relatively low absolute exposures that make these smaller absolute errors larger, relative to the electron counts. If analog gain is used proportional to ISO up to 6400 on a camera, and you expose for 3200 at the ISO 1600, 3200, and 6400 settings, the 1600 will give the most noise, and 6400 the least; there is no extra "distortion" at 6400. If the camera is dual-conversion gain and switches at 1600/2000, as at least one camera does, IIRC, then the ISO 1600 setting will have even more noise or what you call "distortion", compared to the ISO 3200 and 6400 settings.
We're not seeing such problems with digital sensors, AFAIK. The only non-linearity that is usually seen is actually seen at base ISO on some cameras that don't clip away the non-linear response approaching full well capacity. My Canon 10D did this. The top 1/2 stop or so of RAW values at ISO 100 were non-linear, and compressed more than 1/2 stop of actual exposure range. Converters that did not recognize this gave the wrong colors when RAW conversions were pulling down highlights that went beyond what the JPEGs had.
There is no absolute need for a camera to choose ISO setting before exposure, and that only needs to be decided before any gain that goes along with an ISO setting, true, unless you have a dual-conversion-gain sensor, in which case either of two levels of capacitance must be decided before exposure begins. However, I think that most cameras actually decide the ISO setting (or have the user dictate it) before exposure. Either the user dictates it before metering even begins, or metering biases the auto-ISO setting.
--
There's no reason why the camera can't set voltage and conversion gain solely according to the exposure level at the sensor, given that it has an exposure reading, in the case of a mirrorless camera, from the sensor itself. There is really no need at all for the gain level to be linked to the ISO control. A positive side effect would be that people that weren't too good at guessing exposure (i.e, setting ISO) would get better results. It would be nice to have a manual option for gain for the real nerds, but for most of us it would be a real convenience.
Gee, you had time to think about it for three months and now you come back and all you seem to have done is resurrecting your old straw mannikins
Oh dear, here we go again.
www.rawdigger.com
