[Serafin]

I am currently gathering information on lenses, particularly Nikon FX, Nikon Z and Sony FE, to help me decide what system to invest in. Opticallimits has been a great resource.

As for the resolution tests, there are some things I don't quite understand. First of all, the interpretation of the MTF figures seems to have changed between the older tests for Nikon FX lenses and the newer tests for Nikon Z and Sony FE lenses. For example, in the older tests a value above 2920 lw/ph was considered "very good". Today, only values above 3750 are considered "very good".

Why is that? Is it because todays cameras have a higher resolution than those 10 years ago and therefore demand higher resolution lenses?

Also, it says in the tests that "MTF results are not directly comparable across the different systems". What does that mean if I compare for example a Nikon AF-S 50mm f/1.8 G to a Sony FE 50mm 1.8? The Nikon goes up to 4000 lw/ph whereas the Sony reaches almost 5000 at f4. Does that mean the Sony is sharper at f4 or not?

[Rover]

Just compare the bar heights related to the scale. The Poor / Fair / Good / VG / Excellent rankings are there for this reason. Bear in mind however that what is usually called "Good" around here means "actually rather soft but manageable in a pinch" (Klaus, correct me if I'm wrong?) - hence anything below "very good" levels is something to be wary of. And if a lens is just "good" in the center at some aperture (like wide open) it means it's either pretty bad or - if you want to put it charitably - a "character" lens like the old Canon 50/1.2L.

[fotorrhoe]

The MTF values only apply for a combination of lens and sensor. If you use a higher resolution sensor, the values increase for the same lens, at least in the center.
Therefore it is recommended to compare lenses only for the same sensor. With some limitations, even when used in different systems.
That is the reason why Klaus has this comment on his website, not to compare different sensors/systems but rely on his relative judgement.
And yes, good means acceptable without too many computational corrections, not more. A really good lens to todays standard should be at least very good in the center while we accept good at the borders.

[Serafin]

If you use a higher resolution sensor, the values increase for the same lens.

Ok, that pretty much answers both my questions.

So, in other words, these tests cannot precisely tell me if the sharpness of a particular Sony FE lens on a Sony a7iii is higher than that of a Nikon G lens on a D750. Because the Sony FE lenses have been measured with a Sony a7r which is 36 MP as opposed to 24 MP for the D750. So if the Sony lens had been measured with a Sony a7iii the test results would be more comparable, since the a7iii also has 24 MP.

Of course I can still draw some information from the test results of the Sony and the Nikon lenses, such as which lens has more sharpness falloff towards the corners.

BTW, are these MTF measurements based on unprocessed, unsharpened RAW files?

[Klaus]

The MTFs are processed using RAW-converted JPEGs with standardized sharpening.

It doesn't make sense to publish "pure" RAW MTFs because every real-life image out there is sharpened. There isn't a lot of difference anyway because unsharp masking has (mostly) a linear response. We don't use AI sharpening where this is different.