Forum Twitter Facebook Google + RSS

Author Topic: DxO needs a new m43 body for tests  (Read 2773 times)

Offline purethanol

  • Newbie
  • *
  • Posts: 1
DxO needs a new m43 body for tests
« on: January 24, 2013, 03:40:53 pm »
This a comment for this page on the website
I'm wondering what the results would be on a newer sensor with more resolution power. Like on an OMD? The GH2 is kind of outdated now.

Offline Sanpaku

  • Newbie
  • *
  • Posts: 2
Re: DxO needs a new m43 body for tests
« Reply #1 on: February 02, 2013, 10:00:40 pm »
The GH2 and/or GX1 are fine for m43 lens tests until a generation of higher resolution appears.  Why?

There are 5 components of DXO lens tests, and aside from the hard upper limit sensor Mpix places on resolution, sensor differences (in dynamic range, luminance noise, or chrominance noise) generally have negligible impact.

DXO found the Sony sensor Olympus cameras (EM-5, E-PM2, E-PL5) to have somewhat superior dynamic range and noise characteristics than other 16 Mpix m4/3 (GH2, GX1), but the pixel pitch, and hence the hardware upper limit on resolution, is the same.  The Olympus 16 Mpix cameras reportedly have weaker antialiasing filters than the the 16 Mpix Panasonics, but as I understand they still have them, so any difference in resolution tests would be smaller than the difference between the Nikon D800 and D800E (< 5%, and visible only upon pixel peeping).

Distortion is a large scale geometric defect rather than pixel level effect, and should be the same on any camera of the same mount.

Vignetting may differ on other camera mounts due to varying pixel well depth (incident light arriving at oblique angles may miss the well bottom), but 4/3 and m4/3 were designed from the start for digital sensors, with telecentric lenses and perpendicular incident light.

Transmission can be, and probably is, tested on an optical bench, with no camera at all.

Likewise, chromatic aberration can be measured in µm on an optical bench.  If it were measured in pixel widths, higher resolution would make CA more visible when pixel peeping.

As an exercise, look at lenses with multiple DXOmark tests on other mounts. You'll note the same general pattern:
- Camera resolution creates an upper limit on sharpness, but camera models with the same Mpix count differ trivially.
- Vignetting can vary lightly on non-telecentric lenses (but not 4/3 or m4/3)
- Distortion, Transmission, and Chromatic aberration are the same for a given lens regardless of mount.

So, might lens tests using an Oly 16 Mpix camera add 1 point to perceptual megapixels (P-Mpix) for the very sharpest m43 lenses (75, 60, 25).  Possibly, but in my opinion only if there's a shift from rounding down to rounding up.  It would have negligible impact on the relative scores of the lenses within the system or compared to others.