[Edit: when I first posted this, I thought you could turn off PDAF for native lenses with the a7RII. Turns out you can’t.; what I thought turned it off only turns off the display of the PDAF point. So this test is basically the same as the one in the next post. I’m leaving it here because there are external links to it, but I will probably remove it in a few months. Unless you’re an Internet historian,m it would probably be better if you just skipped to the next post.]
This is a continuation of a test of the following lenses on the Sony a7RII:
- Zeiss 85mm f/1.8 Batis.
- Zeiss 85mm f/1.4 Otus.
- Leica 90mm f/2 Apo Summicron-M ASPH.
- AF-S Nikkor 85mm f/1.4 G.
- Sony 90mm f/2.8 FE Macro.
The test starts here.
We’ve seen in posts over the past few days that the Batis has a moderate amount focus shift upon stopping down. Several have asked me if that will cause focusing errors when using autofocus. It’s hard to tell a priori, since the aperture of the Batis on an a7RII opens and closes during autofocus. Is it using phase-detect autofocus (PDAF) wide open and contrast-detect autofocus (CDAF) stopped down? Hard to say.
I lit a slanted-edge target with two Westcott LED panels, and set the color temperature to 5000K. The target has a medium-contrast Siemens star in the center, which I used for both manual and automatic focusing.
I mounted a Sony a7RII to the Cognisys computer-driven focusing rail. I set the controller up to use 192mm of travel and make 49 exposures 4mm apart. I set the assembly 8 feet from the on-axis target, mounted the lenses in turn, focused a git short of mid-rail wide open using manual focusing, and focused on the star using AF-S with expand flexible spot, PDAF display off, AF priority. When the Cognisys rail behaved, I exposed 49-shot series. I used f/1.8 and f/5.6. I used Jack Hogan’s Matlab program, MTF Mapper, and DCRAW to pick the horizontal edges and calculate the MTF50s for the raw color planes, imported the data into Excel, and plotted the results in cycles per picture height vs subject distance change.
Manual focus at f/1.8:
Autofocus at f/1.8:
This is an interesting an unexpected result. It appears the the autofocus system is optimizing the focus for the red plane, not for luminance, which derives mostly from the green plane. The red plane MTF50s obtained by autofocus appear to be nearly optimum, but the green and blue planes are distinctly sub-optimal. In looking at the MTF curves, we can see that if the AF got the red channel dead on, the blue and green channels would be about 800 cy/ph. They are a little better than that, but not much.
Manual focus at f/5.6:
The Cognisys rail had another mid-test hiccough, so don’t compare the horizontal scale of the f/5.6 curves to the f/1.8 ones.
Autofocus at f/5.6:
The red channel is once again the preferred AF channel, but this time the results are off what can be achieved with manual focusing, although not by a whole lot. The green and blue channels are less sharp than you’d get by focusing optimally on the red channel at f/5.6, so it looks like the camera is misfocusing on the far side of the red channel peak, which is the wrong side if you want the sharpest luminance.
Jim are the “manual focus” curves the curve for the manual focus you did by eye and set for mid rail, or for the best result of the focus stacking? If the latter it’s an interesting result, but doesn’t show you can do better in the field with MF. If the former, you might want to average a number of focus-by-eye results.
Also is the AF result a single pass of AF, or an average of number of exposures (of course if it’s super repeatable this might not matter)
And thanks again for all this great work!
cheers
David
The former.
Jim
Um, of course by “stacking” above I meant “bracketing”
Jim, does the a7RII have AF fine tuning? If so, are these lenses fine tuned?
The a7RII, since it has the PDAF sensors on the imaging chip itself, doesn’t have AF fine tuning for native lenses.
probaly CDAF sees that the total contrast (image from R+G+B) is peaking before G/B channels reach the maximum that’s it… not specifically paying attenion that it was R channel, just it so happens when CDAF is moving focus from one direction (for example from near focus to infinity)… one can assume that if CDAF will be moved from an opposite direction G+B will be the ones optimal… did you try to manually move the focusing plane slightly before and slightly after the focus before engaging CDAF to see if that is the case ?
It looks to me that it’s using CDAF at the max aperture (or close) and not adjusting for focus shift.
That might be why it’s even more heavily biased towards the red side, well past the point of best red focus even (judging by the blue and green channels), when you stopped down to f/5.6.
I think my prev. comment did not get in… did you try test AF first manually focusing in front of the target and then manually focusing after the target to see if AF actually lookes for total “R+G+B” contrast rather then for “R” plane contrast… approaching from one direction CDAF might indeed stop when R is maxed, but from a different direction when G+B is maxed.
You are right. Akismet flagged it as spam, but I have told it otherwise, and it probably won’t happen again. That’s the first false positive I’ve ever noticed from Akismet.
That’s a good idea about the focusing direction. I did not try to bias the focusing. It’s something to think about, but it would be labor-intensive to do that. I’m going to do a run with PDAF turned on as well.
Jim