>>4200870>They're directly related. In theory they should be, but PtP's graph clearly doesn't predict picrel. I don't know if the error is in PtP's testing, methodology, or presentation. It's likely just presentation, or perhaps you could say methodology + presentation. PtP's graph is based on a noise threshold, a line in the sand. The graph just tells you when the sensor crosses that line. It can't tell you what the noise looks like below that line...how fast the sensor is dropping below that line...so it's only a very rough indicator of performance at extreme ISO.
>Point is, it's not the pixel size.Pixel size is irrelevant over most of the ISOs photographers use. For a long time that led me to think/say that pixel size was always irrelevant to high ISO. But in digital night vision (SiOnyx, Starvis, etc) you quickly find that at extreme ISOs it becomes relevant. Camera ISOs are now crossing into that territory. So we're starting to see some differences, at the top extreme ISOs, that correlate with pixel size.
The complicating factor is that there are multiple sources of noise, and they scale differently. So whatever is correlated with pixel size at ISO 100k and above is invisible at, say, ISO 25k.
