is 40mp too good for lenses?

It's non-reliable because without expensive specialized equipment it is, at the very least, non-repeatable. One can try, taking a dozen of shots, and see the results - they are conclusive only if a crude defect is present, and shots are taken with great care. More, some defects can't be diagnosed this way, while some can't be attributed.

A better way is to rely on aliasing, producing a "field", "rainbow" picture, as Jim explained. Results are fast and conclusive, no need in specialized equipment, and such test also allows to monitor lens and camera mount health - which is important for me in travels.

Rightly or wrongly, I find it useful in such matters to do edge detection in the image which can bring out the in-focus versus the out-of-focus parts and can show a "field" quite clearly.

For example: www.dpreview.com/forums/post/67137656

sometimes there are no other options because manufacturer does not make the body that you want with all other bells & whistles within you price range with less mp

[quote="@davidwien"]
If you have a 40 Mp sensor, you need lenses that are good enough to take advantage of it. Otherwise why have it?
😀
David

Good point. What lenses are "good enough to take advantage of [40 MP]" ?

I have a Panasonic m4/3 DC-G9 which be set 80 MP in pixel-shift mode. But the MTF curve for pixel-shift captures is pretty poor and the popular MTF50 occurs at a low spatial frequency relative to Nyquist. Which means that I don't need an Otus 55mm for that camera, or any of mine for that matter.

On the other hand, I can almost stick any old lens on SD9 with it's 9.12um pixel spacing ... 😀

On axis at optimal aperture, most decent lenses will be able to lay down detail a 3 um sensor can’t properly resolve.

Indeed, look at this Sony at f/2.8 giving well over MTF50 on-axis!

Don't know if that's "optimum aperture" for that f/1.4 lens and don't know the MTF curve for "a 3 um sensor" ...

Image by Roger Cicala www.lensrentals.com/blog/2019/03/sony-fe-135mm-f1-8-gm-early-mtf-results/

It is quite a specific statement. Ideally, sensors have a sampling frequency high enough that there is no aliasing. We are a long way from that.

I would like to see the MTF below .1 at the sampling frequency for a monochrome sensor. Lower for a Bayer CFA.

This is part of a photo taken with a7Rmk4, 3,76 micrometer pixels. The aperture was between 8 and 11, so the diffraction was in the play too.
What I see in the picture is aliasing if I understand how aliasing looks in pictures.
I'll bet aliasing was visible in picture taken with 3,1 micrometer pixel pitch sensor.
That means the lens outresolves the sensor?

The aliasing improvements in real world photography when switching from the 50 MP to the 100MP 33x44mm sensor was something I found important in my photography. But I still see aliasing with the GFX 100x and the Hasselblad X2D.

That's right. All test photos should be of like shrubbery and rolling hills, hand-held and JPEG with no EXIF to muddy matters.

agree if you are blind or have no commonsense or practical nous. 🤔😜

agree totally, went through this process setting up a recording studio for my daughter. high qulaity studio monitors and headphones are parramount.

Look, Aristotle!

It is easier to produce aliasing in a lab, but you don't need a lab to produce aliasing. As long as you have sufficient optical sharpness and camera/subject stability relative to pixel density, aliasing is always possible. What's so annoying about aliasing is that for any given sensor, the sharper the lens that you use and the more stable things are, which should be good things, the more likely you are to get visible aliasing.

Every Bayer capture that is moderately pixel-sharp has aliasing in the red and blue channels if the subject matter has aliasing potential; the raws are generally converted to "play dumb" with pixel-level color detail and hide the wild chromatic artifacts of color aliasing. This is even true with sensors with AA filters, because the point spread of the AA filter is small relative to the pixel spacing in the red and green channels. If it were big enough to greatly reduce aliasing in the red and blue channels, individual points of light could get ghosted into 4 distinct visible points in the implied luminance channel.

I think it is more productive to think in terms of the frequency of artifacts for any given usage and equipment, rather than "real world" vs "lab". How much can a lab improve over real world for stability, using a relatively wide lens and f/4 in sunlight at 1/1000? Why talk "real world vs lab" when one can easily tell the full story of "hand-held after too much caffeine vs tripod"?

I see no aliasing here.
I would expect aliasing to show up most clearly on a photo of a city scene, with plenty of slanted straight edges on buildings.

Don