Higher megapixel sensor -> less resolution

Great Bustard wrote:

buellom wrote:

Great Bustard wrote:

...then I'd not put a lot of weight on the conclusion. For example, if the particular A7R5 that they used had a mount that was ever so slightly out of alignment, or the sensor were ever so slightly out of alignment, that would *easily* account for the difference.

And, of course, it could be an error made by those testing the camera, where one possibility is that the A7R5 was not aligned perfectly square with the test chart (which, I hope, is far more likely than a defective A7R5).

I mean, anytime a test has an anomalous result, it's best to question the test before deciding a theory that works for every other camera is suddenly wrong.

True, I can't rule out any errors in the test. But their results look quite consistent.

However, the website Optical Limits / photozone.de came to a similar result. They tested the 16 STM on the R and the R5. Here the resolution values for the R (30 MP) are better than for the R5 (45 MP) at the extreme borders, though, not by much.

Canon RF 16mm f/2.8 STM - Review / Test Report - Analysis (photozone.de)

So we have a second test. I'm not aware of any other test that has an eye on this issue.

Couple of things:

• The differences are *easily* small enough to be measurement error.
• In red, OL states the following: "Please note that we apply a MILD amount of sharpening during the RAW conversion, and sharpening is more receptive to cleaner edges on lower megapixel images, so, besides other reasons, the MTF numbers obtained with the EOS R are NOT comparable to the R5 variant." That right there says it all.

The MTF graphs at the bottom, of MTF50 vs f-number don't look like the shape that one expects. Lens testing at these kinds of resolutions is crazy hard. So, I think that there is every. Also, they say 'In the test lab on the Sony Alpha 7R V, the FE 20-70 mm F4 G shows no visible darkening of the edges. With all focal lengths and apertures, it is only 0.3 f-stops, even at the maximum. The laboratory tests on the Sony Alpha 7R III and 7 III show that a digital correction has a strong impact here, because the maximum edge darkening of 0.9 and even 1.1 f-stops is significantly stronger. But this also shows that Sony has increased the correction of the edge darkening more and more with the generations.'. If their lab tests are including digital correction for distortion then all bets are off in terms of lens testing. Who knows what the correction algorithm is doing.

And I'm also not aware of any test that actually shows the opposite.

Well, this test, for one. Theory says that the lower the resolution of the lens, the less difference the pixel count makes -- in other words, lenses with soft corners. The question, that you bring up, then, is why the larger pixel sensors appear to do a little worse, rather than a little better. I think the answer, barring hardware defects, chart alignment issues, and measurement errors, is in the italics I quoted above.

It's not common to compare the results of a lens on different resolution cameras. And I think the issue will be only visible with not that great wide angle lenses.

Well, the test I linked, to the super sharp Sigma 85 / 1.4A, shows way higher sharpness for the larger pixel count. But that's 'cause the lens is sharp right to the corners.

It's enough for me to reconsider my wish for a high resolution (80MP+) camera. At least I want to understand first the IQ compromises related to the high resolution sensor...

I think what you would find is that the situation you describe only comes up when the corners suck to begin with, and any differences would likely not be visible at all. Still, it is worth putting to the test. If anyone has an R6 and R5 and has nothing better to do, they could post such a comparison.

...of course these are dual pixel sensors, so the R6 is a 40MP camera (48 for the Mk II) and the R5 is a 90MP camera. If it's actually the case that small pixels are destroying edge resolution the Canon should be thinking very hard about DPAF.