I see what you’re saying, but that assumes that you’re combining photon counts from adjacent pixels, rights? E.g., if you grouped every four pixels together and counted the totals as a single pixel it would be equivalent to the 10 mp sensor with the same theoretical SNR. In that case, sure, more data is always better. But if you’re not summing the pixels then although you’d have four times as many with the 40 mp, the full well capacity would have to 1/4 of the 10 mp. Maybe not an issue at low ISO, but it’s going to limit you as you push higher, no?
The well capacity would be enough to be 1/4 but the light falling on each pixel is 1/4 as well, so there is no problem.
Well yes, that was my point. You would have to reduce the FWC to 1/4 to maintain exposure. So, you increase the ISO two stops. Which would increase the noise at each pixel.
However, the thing I admittedly never thought about until reading this thread: If you then downsample the image in post to the same size as the 10 mp, do the pixels average out to give the same general level of noise as the 10 mp sensor? The more I think about it, at low ISO the answer has to be yes. But, if you’re really pushing the high ISO are the results so far off that it will screw with the final average? And I guess I don’t fully understand how the averaging would work considering the Bayer layer – statistically wouldn’t more noise mean more issues with the green channel, so averaging out the noise isn’t going to be so simple.
Sorry to ramble on mindlessly, it just got me thinking. I can’t believe I never thought of it that way, I was always of the camp of ‘so long as you have enough resolution for your uses, the bigger the pixels the better’.