RCicala Offline Upload & Sell: Off
|
sebboh wrote:
i just read Roger's article and i have to say i don't follow the logic of his conclusions in a number of cases. perhaps he can chime in to tell me what i'm missing.
he says: "So why would the Zeiss 35mm f/2 fare worse on the camera, while it is probably the best lens of the three tested on the optical bench?"
but the results he shows from the optical bench indicate that the ZE 35/2 is the worst lens on the optical bench in the center and is noticeably worse (particularly at high frequency where it matters for the comparison) than the canon off axis on the optical bench as well. given these results on the optical bench the imatest results between the canon and the zeiss are exactly where i would expect them to be, not at odds with each other as Roger suggests. is there another comparison where he puts those two lenses on a canon camera and gets the opposite imatest results? because i don't see any contradiction between the optical bench and imatest with any lens except the sony. this makes me suspect that the difference between the sony and the other lenses comes down to the adapter. Roger says this is the best adapter from his many samples, but even the best adapter should have roughly twice the tilt on camera as a native lens because it has twice as many interface points. i would think the best way to test whether the difference comes from sensor topping issues or adapter issues would be to also test the canon and zeiss lenses on the optical bench with the adapter (as roger has done previously with other adapters).
i'll add that the while ray angle certainly can have a huge impact (as people have demonstrated with lenses that truly do have an exit pupil close to the sensor), the difference is ray angle between the ZE 35/2 at f/2.8 and the canon 35/2 IS at f/2.8 cannot be very large (based on geometry, i can't measure them myself).
finally, i'll add another way to test whether the adapter is the issue or the sensor coverings. using ultra wide angle slr lenses will minimize the effects of the sensor coverings and maximize the impact of the adapter. this is because ultrawide angle slr lenses have to be extremely telecentric and small differences in alignment have a much greater impact on ultrawides. if you see a greater discrepancy between the optical bench results and the imatest results with ultrawide angle glass then the source is the adapter....Show more →
I can't answer everything (because I don't know), but I do want to clarify I misstated the lens order in the article. In my first draft I compared just the Zeiss and Sony (the Zeiss is better off camera). Then I went back and added the Canon lens but didn't make the proper correction in the test. The Canon does outresolve both the Sony and the Zeiss off camera.
So off camera, the results were Canon > Zeiss > Sony. On camera it was Sony barely> Canon >> Zeiss
As to adapters - I'm looking at a 36 point horizontal and vertical resolution set, which I don't reproduce in the article. It's extremely sensitive to tilt - and Imatest measures tilt in addition. I ran multiple lenses on multiple adapters and took the results from the one with the least tilt. Given % change in resolution and astigmatism, the amount of tilt here is no more than the variation I'd see in multiple copies of a lens on one camera - around 0.04 of a degree.
Don't get me wrong - I don't advocate testing on adapters for just this reason. But in this case there's not much option I can think of.
|