Originally posted by normhead "I looked into it (in my equivalency white paper) and it is a myth, an urban legend among photographers. The opposite is true. If you compare, e.g., the APS-C corners of an 31mm or 35mm at f/2.8 and the full frame corners of a 50mm at f/4, the latter wins by a significant margin (in absolute LW/PH resolution, not fall off from the center, to be clear here). Or any other equivalent combination."
Does that assume equivalent Mp?
The answer is true regardless of MP for reasonable numbers of pixels (i.e. not comparing 10x as many pixels). Photozone has a bunch of lenses tested on Nikon at ~12MP on both APS-C and FF that shows this is true at the same resolution.
Originally posted by normhead How useful is an analysis if it only looks at one element of a camera system at the exclusion of others and doesn't include factors such as sensor sensitivity, internal processing and Mp? Are they any use at all? And last, how is it that the K-5 sensor comes out rated above so many FF sensors, if it can't achieve the IQ of any of it's FF cousins.
Inquiring minds want to know.
Without knowing what an equivalent picture is, you cannot determine which sensor is better. Not that anyone should care what sensor is better - they should care what picture is better, and these 'equivalency' equations show that even a relatively poor sensor on FF is better than a relatively better sensor on APS-C. If you're trying to compare between sensor sizes, you need to know equivalency to interpret the dxomark, etc, tests. Keep in mind, too, that measuring the sensor does not (or should not) measure the lens... and larger sensors demand far less from their sensors for an equivalent IQ.
So a FF with a larger k5 sensor (in other words, the D800), would score 'off the charts' on the DxOMark test (which it did), but that still wouldn't account for the improvement in resolution likely from a same-cost-class lens (so the FF should be better still for many people).
FYI the k-5 sensor rates slighly better than the d7000 but I believe it's within the margin of error listed for the test. I don't argue that the k-5 is better.. but it's not that much better.