Opticallimits

Very much do I like the lens tests that Photozone is making available to us, they are very helpful and easy to understand.

However, looking at some lens tests with a 20Mp Canon full frame camera, I have a problem with the MTF50 figures.

The used 21Mpixel 5D MKII to be seen in some pictures with the test lens mounted has 3,744 pixels vertically.

 

One would expect a maximum MTF50 of something like  3,744/sqrt2=2,647 LW/PH.

But as an example. the test  for a Canon 24-70mm F/2.8 II shows a figure of 3,732 LW/PH at F/4.0 in the centre, almost exactly the full amount of 3,744 pixels from the 5D MKII.

 

In case the resolution of the lens would be an extremely high 6,000 LW/PH (=125 LP/mm) combined with an ideal 21Mpixel sensor,  the MTF50 figure would become (1/3744^2 + 1/6000^2)^-1/2= 3,176 LW/PH, the MTF now already being reduced to 85% of the total amount of pixels.


​To achieve the figure of 3732 with a sensor of 3744 pixels, the lens should have a resolution of 46,700 LW/PH or 972 LP/mm, and that just with a sensor that has its MTF50 up to the full amount of pixels, which seems just as impossible because of Bayer demosaicing.

 

Am I overlooking something?

Could it be that this particular lens has been tested with another camera with (much) more pixels?

 

 

The sharpening applied probably accounts for the inflated numbers.

You are right, sharpening can blow up MTF50 figures quite a bit, but not to all the way up to the amount of pixels.

Sharpening can indeed improve MTF50 coniderably.

But it cannot improve that much that this figure lands on the amount of available pixels.