DXOs P-MPIX is NOT a Lens Sharpness Metric | FF in Crop is same as Crop Camera with same pixel pitch
Thank you to our sponsor https://www.lensprotogo.com for providing the D850, D500 and 24 1.4 Lens used in this test! This was not a paid promotion, they simply lent me the cameras and lens. If you need a fast rental, they are my go to!
For many years I have had questions about DXO Marks various camera tests and scores. The thing that bothered me the most was that they were publishing data and I was never really certain as to how they were coming up with some of their numbers, and I know many others feel the same way I do.
One of their most controversial metrics is the P-Megapixel or Perceptual Megapixel which for the life of me I always refer to incorrectly as the “perceived” megapixel score. It is a cool idea in that it is trying to give viewers an idea of the amount of resolution they would get in a final image when pairing a lens to different sensors. There are a huge number of problems with this idea, the biggest of which we still aren’t exactly sure as to what formulas they are using to come up with these numbers.
Heres the bigger problem, new camera owners, or those shopping for a new lens often use these scores to try to get an idea of how “sharp” a lens is. This is not what the P-Mpix score is, which is instead trying to to describe how many megapixels one would get when using the combination. I actually stumbled upon this accidentally when I was testing lens sharpness on a full frame D850 in crop mode compared with a D500. I chose those cameras because they have similar pixel pitch, a variable that impacts diffraction (and therefore image sharpness) and what I learned was, same lens, on similar sensor designs (including pixel pitch) equals similar sharpness.
While DXO doesn’t have D850 test scores that I could find, there were scores for the D810, a 36MP camera. The D810’s PMPIX Score was 23, compared to the D500s 13 (out of 21).
Crunching the Numbers we get:
23/36.3 = 63.3%
13/20.9 = 62.2%
So this leads me to believe they are taking their tests and applying a “lens penalty” to the overall resolution of the camera. This is whacky because it means if you start with a higher MP camera, the score will be higher, regardless of the actual lens and image performance.
Granted, the D810 is not the same as the D850 and there are likely some differences, but what we can gather from this, is that even when the crop areas are similar to the naked eye, their PMPix scores are vastly different.
This is why we cannot consider DXOs PMpix score as a lens sharpness metric, it should be viewed as a total image detail metric. Having it labeled as “sharpness” is very misleading to the general public, because without doing a similar test, there is no way to really know, and the perception is this PMPIX is a lens sharpness score. It isn’t.
I invite anyone interested to download my test images and compare them for yourself: https://www.dropbox.com/s/xs844ccv2wsisd2/Finals-D850-D500.zip?dl=0
There are a huge number of valuable and other interesting questions that can be asked about image sharpness, and Ill have at least 2 more videos on the topic coming soon!