Mark writes: >Nyquist for a full frame sensor is approx 16Mp -- thus "in theory" the 24Mp is somewhat "wasted" Mark, I'm curious to how you come to this conclusion, I have it another way My take on the Nyquist limit is the theoretical maximum info is 1/2 the frequency of what we're sampling (or rather, we need to sample at 2x the rate of information), but this is generally applied to a linear system such as audio. Let's say we have a device which is sampling is sampling at this frequency . . . . . . . . . . (sampling) and we have this input (data stream or whatever): . . . . . . . . . (high point) . . . . . . . . . . (low point) we can really only get this or this from the sampling: . . . . . . . . . . the high signal *OR* . . . . . . . . . . the low signal - it will only see one or the other to sample at 2x the frequency to get an accurate representation of the actual signal so if we were sampling at : ..................................... our data would be . . . . . . . . . . . . . . . . . . . which is accurate :) the problem with digital sensors is that the above only applies to data or image information that lies parallel to the vertical or the horizontal - so we'd have no problems with picture elements such as this: ---------- | | | but when we have this \ \ \ \ sampling will be inaccurate, which is what leads to moiré, which is a common occurrence in digital photos - in effect the sample rate we need is closer to 3x the information rate, or to put it another way, what we get from the sensor is 1/3 accuracy so in effect the limit to resolving power accuracy of a digital sensor of 24Mp in many scenes will be a mere 8Mp of course if we were photographing pure blocks of colour we could say the data was 100% accurate and the nyquist limit is irrelevant. karl
