PHOTOOG Photography writings by Olivier Giroux

23Jun/102

DXO Lens Tests, Mixed Feelings

DXO just added lens tests! Already on day one there’s tons of data there, enough to spend hours sifting it.

The T-stop data is a particularly nice touch, this is showing the industry-standard 1/2 T-stop stolen:

Nobody was collecting this data on such a wide catalog of lenses – I think that is actually DXO’s real contribution here. You could make a similar argument with their detailed chromatic aberration data. The rest we’ve seen before.

The unfortunate side of this new site is that you have to personally dig into it because I think the summarized score number is pretty deeply flawed. Here I’ll pick on the first-generation 70-200VR Nikkor because I know it well. On the score chart it wins the #1 spot with the D3X body.

But with a little bit of digging you’ll find the big smudge at 200mm on 24x36mm sensors:

How could this lens win the top score, of all the lenses DXO has tested? One chart that turns up deep red everywhere like that should knock a lens down several notches. The issue there is that the main chart is just “peak score” which is aligned to that performance profile (see DXO’s explanation).

But they offer no other chart in visual form! This deeply flawed chart is the only one that’s visual. There is no equivalent to the DXO sensor chart.

My other pet peeve – a less important one – is that DXO also fell down a cosmic Internet rathole with this little chart…

“Not applicable”, what the!? Some of the best contemporary landscape art is being done at moderate telephoto lengths! Actually this chart only gets worse with other lenses. Like the top Canon-family performer, the 85mm f/1.8 USM, it’s apparently not applicable to anything except portraits.

Bottom line? Awesome new data. Crappy top level scores.

Also : DXO, it’s time to go test some (non-Sony) German lenses! ;^)


Comments (2) Trackbacks (0)
  1. T-stop testing is a nice touch, but other than that I thought the rest of it is quite confusing. How are you supposed to figure out the exact resolution looking at some fancy red/yellow/green screen?

    • The exact resolution figure isn’t as interesting as the uniformity of its distribution in my opinion.

      This said, it’s true we humans do poorly extracting data from the color of a chart. It’s better if you stick to curves.


Leave a comment

No trackbacks yet.