Kit Lens Comparison; RF-S 18-45 vs EF-M 15-45, Charts, Imatest, Design

Sittatunga wrote:

BobKnDP wrote:

koenkooi wrote:

BobKnDP wrote:

What are the units on the horizontal axis of the MTF plots?

10 cycles/mm seems low, so I assume it's something else.

The horizontal axis is distance from the center of the sensor, in mm. So the rightmost section is the corner and the left most section is the center of the sensor.

OK.

In a former life, I was accustomed to seeing MTF plots as a function of spatial frequency.

What is plotted here? Obviously not MTF at a fixed spatial frequency, as that can only have a value of 1 at zero spatial frequency.

The black lines are radial and tangential contrast at 10lp/mm (coarse detail, more like the overall contrast of the image), the blue lines are 30lp/mm (fine detail for an A4 print from FF or an A6 print from APS-C). The distance apart of the lines is a measure of the astigmatism as it varies across the field of view. All at full aperture of the lens. Most manufacturers publish these theoretical charts in a similar format, but they're not all calculated in the same way.

You got in before I managed to edit my response.

I followed the links from "RLIght" to the Canon Japan site. I got the 10 and 30 cycles/mm thing from the legends. (Black is 10, blue is 30. Solid is sagittal, dashed is meridional. I used to see "meridional" as tangential.) A little Web searching suggests that this is a standard means of reporting lens MTF. (The sagittal direction is along the radius, so the grating lines would be perpendicular to that.)

30 cycle/mm seemed coarse to me, before I thought about it. Looking at it differently than from the point of view of a print: A Canon R5 has a pixel pitch of 4.39 microns. Considering the Bayer mosaic, I'd put the Nyquist sampling frequency at 57 cycles/mm. 30 cycles/mm makes sense in that context.

For a little more arithmetic, 30 cycles/mm on a full frame (36mm length) sensor corresponds to 3.64 cycles/mm if mapped to the full length (297mm) of an A4 sheet. Assuming 2 printer pixels per cycle, that's 180 printer pixels per inch. Sensible.