Originally posted by Nicolas06 We can likely get even more over time until we get 100% of the light for each color instead of getting 50% of green and 25% of blue and 25% of red. Meaning with a perfect technology we could likely get 2-3EV more.
That's not that simple.
You could improve luminance noise by 1-2 stops, that's true. By having more than 50% of pixels detect green, and over a wider gammut. That's improved B&W images.
However, you can't improve color noise by more than maybe 1 stop. Because the human eye throws away much of the spectrum (tristimulus) and you have to do the same in the sensor (or apply low weights if you capture more than three colors in the first place).
Including progress in quantum efficiency, I do not expect low light performance to grow by more than 1-2 stops, 2-3 stops for B&W.
Originally posted by Rondec I don't know what is possible.
There is no physical limit for base iso or dynamic range. There is a physical limit for dynamic rangeat a given iso though.
For really low values of base iso (below iso 50), one probably needs digital integration and continuous readout on the chip. Something the new Sony Exmor RS stacked architecture looks promising for (not yet used for this though).