RandThompson
Hey rand to avoid too many oversample discussions I merged your new question into the existing oversample topic, hope you agree this is the right place if not, I will rethink my decision.
As you can calculate oversampling from 8k to 4k = pixels become one. IN this case debayering isn’t needed in the regular way of how debayering works. You dont need to interpolate the information from neighboring pixels, instead the pixels are quantized into a super pixel.
In case you choose to scan the sensor for a smaller area like the s35 crop the pixel to pixel resolution is 6k.
(crop factor s35 over full readout =1.5 thus 8k/1.5=6k)
a 2X2 oversample mode in which the pixels are quantized instead of debayered should be 3k when shooting 6k s35.
In case you go opt for 4k resolution in s35 the sensor is first debayered into 6k and then as a 2nd pass downsampled to 4k, which is a quantization algorithm like how your post production nle’s treat downsampling, will still look amazing, but 2X2 on sensor quantization will yield a tiny edge over the other option. As you might understand this is all pixel peeping and when you do your own test you wont probably see a big difference, these are mostly theories and in practical use they yield about the same results.
Downsampling or oversampling are both great ways to lower noise values and increase color strenght. But 2X2 on sensor is theoractically the best way. I have yet to do intensive testing myself to find out how much these things matter in practical use. So expect to see some tests later on.