85mm f/1.2L II - AF issues, and how many shoot mostly manual focus with this lens?

I'm just asking the following questions out of technical curiosity. I bought this lens earlier this month, and the first thing I noticed (after the bokeh that lives up to the legend) was how the through-the-lens phase-detect AF on my 6D really isn't precise enough to count on with this particular lens, especially noticeable when shooting at f/1.2 of course. The center AF point does alright, but the outer ones... forget about it. I know it's not the 6D's fault, because all the AF points work great when I'm using my 24-70 f/4, and it's not a matter of the extra depth of field simply including my intended focus point on the subject, I've tested it with a proper target and ruler setup (what I use for MFA), and the focus points just work more precisely with that lens.

What I'm wondering is: is it the lack of sharpness of the 85L wide open towards the edges of the frame what is resulting in the poor AF performance? DxO and photozone.de both note that the wide-open sharpness of this lens falls off dramatically away from the center, and is quite soft where the outermost AF points are. I'm assuming since the AF system has to look through the lens at f/1.2 when focusing, that lack of sharpness would impact AF precision, right? Or is the lack of sharpness just something we notice because we see the whole visible spectrum, and the lens is basically "sharp" at the one specific infrared wavelength the AF system uses? I'd be curious if anyone has a really good grasp on exactly what is going on that makes it so hard for my 6D to get precise focus with the farthest out AF points. I'm not doing focus and recompose, BTW, and everything I'm talking about here is valid for my tripod test shots as well has handheld--it's not an issue of the camera moving ever so slightly between the time of focus lock and the time of shutter release, I've ruled that out quite exhaustively.

When I say the AF is not precise, I mean exactly that. It's accurate, in that I've taken 100 test shots using my MFA ruler and target setup (proper tripod, remote, mirror lockup, etc), and I found that while there is a great lack of precision and repeatability, no micro adjustment is required--the shots miss in the front and back focus directions equally often, and the spread of the sample points is centered on the target distance just fine. The spread is just very wide. Only 22 of the 50 shots I took hit the correct focus distance dead on using the center AF point, and only 5 of 50 when I used the rightmost point. In all tests, I had the ruler scale near the center of the photo, I only moved the focus target. Because of the thin DoF at f/1.2, this situation translates into some pretty low keeper rates, particularly with the outer horizontal AF points.

So, I've been shooting using only manual focus with this lens. I have the Eg-S "super precision" focus screen in my 6D, which I absolutely LOVE. I mostly focus through the viewfinder, and I get a 90+% keeper rate this way in decent lighting. In lower light, I've been using live view mode at 10x to see the focus better. Letting the camera AF in live view mode also gets a decent keeper rate. I'm talking here about shooting with this lens at f/1.2 mostly, because that's how I use it. Stopped down to f/2.8 or higher, the lack of AF precision is still noticeable in my shots viewed at 1:1, but the expanded DoF means that the intended focal point on my subject is usually within the acceptable sharpness zone, in all but the most extreme outlier cases of AF miss or if I need to do 1:1 crops.

While I can't comment on the AF of my 6D's outer points, I will say that with my 50L I often shoot wide open and whether I am using AF or MF I take 2-3 shots in continuous mode.
The reason for this is I want to help eliminate my own user error when precision focus is critical.

I have test all of my lenses and used MFA to correct them, still I shoot 2-3 shots at a time as getting a 'keeper' is much much similar to delete 2 in post precessing than to miss something important.

I also bought the EG-S screen for my 6D and hardly ever use the Live View. I enjoy it.

I won't use none cross type AF points, period. It's one of the reasons I went with a 5DIII.
Also, I do not like the feel of the fly by wire manual focus of the 85LII.

I don't own one, but did you buy it new? Check out this review link. It should have fine AF.
http://wlcastleman.com​/equip/reviews/85mm/in​dex.htm

That is the nature of the 85. Half body wide open will be hard to nail on any body consistently.

It can be precise at times, but quite slow and not quick enough for micro movements. For full body with more DOF, generally not an issue.

artyH wrote in post #17687434
I don't own one, but did you buy it new? Check out this review link. It should have fine AF.
http://wlcastleman.com​/equip/reviews/85mm/in​dex.htm

I bought it from Canon through their refurbished store. They put up 60 of them at once, and when I asked the rep in online chat, she said they were "returned overstock". The lens came in all of it's original packaging, except that the outer retail box had been replaced with a Vixia camcorder brown box, probably for legal reasons (to not compete with retail units). Also, that link is for a study that only considered the center AF point, and the test data is from a 15-year-old 1-v film camera. I don't think it really covers my use case.

I have no reason to believe there is any fault in the lens or camera body. AF works as expected in view view, images at f/1.2 are ridiculously sharp, and center-AF-point only shooting works well in real-world situations, just not as well as other lenses when examined in my MFA test setup. The only thing I've found unusable in practice are those outermost horizontal AF points when shooting wide open. I'd need to take 10+ shots to guaranty a keeper, which is why I'm shooting manual with this lens.

As I noted above, I suspect this is because of how the sharpness falls off at the edges of the lens. Most of my lenses don't have this characteristic. My 40mm f/2.8 STM is the exact opposite, in my kit it's the best performer in terms of AF performance and keeper rate, >95% keepers no matter what AF point I use. Look at the DxO sharpness field maps below for the 40 STM as compared to the 85L II, it's a huge difference. My main question is: is this why I'm seeing the AF issues I'm seeing with the 85L and the outer points?

40 STM @ f/2.8

I finally Dumped my 85L and went back to the 85 1.8.. Focus issues. Its a PITA...85 1.8 never misses at 2.0........not so true with the 85L........frustrating lens........135L looks better and never misses. If I have room its on my camera

I find the 85L II to be a very consistent and accurate focusing lens on my 5D3. It's slow, but very accurate in my experience.

So, something else I should have mentioned: I'm shooting miniatures, toys, dolls at the moment, and it appears that the problem is greatly exaggerated with close subjects (I'm not actually at MFD, though, that's not the issue). You can try this test for yourself if you have an 85L:

1. Put camera in AI Servo
2. Select center AF point, and focus on a still subject at about 1 meter distance, observe for a few seconds.
3. Select farthest left or right AF point and do the same.

With the center AF point, I see and hear the lens making an adjustment about every second or so, which is normal considering I can't hold the camera perfectly still. But with the outer AF points the lens is adjusting focus almost continuously, both in front of and behind the actual target.

I do notice too that targets without lots of contrast exhibit this behavior more, so there is no doubt that the non-cross nature of the outer points is coming into play also.

absplastic wrote in post #17687782
So, something else I should have mentioned: I'm shooting miniatures, toys, dolls at the moment, and it appears that the problem is greatly exaggerated with close subjects (I'm not actually at MFD, though, that's not the issue). You can try this test for yourself if you have an 85L:

A little tangential, but the 85L II is a terrible lens for macro and near-macro work as the MFD is quite long and the maximum magnification is very low. Along that line, I don't understand how you are shooting miniatures with the lens and are not nearly at MFD.

Beyond that, I don't have experience with that kind of shooting. I use my very fast primes like the 85L for more distant subjects where I want to frame a large area but still want to blur a background.

Your signature lists 100/2.8L macro and 300/4L (IS?) which are both great choices for near-macro work.

JeffreyG wrote in post #17687967
A little tangential, but the 85L II is a terrible lens for macro and near-macro work as the MFD is quite long and the maximum magnification is very low. Along that line, I don't understand how you are shooting miniatures with the lens and are not nearly at MFD.

It's not macro work, exactly, though I am shooting quite close to MFD. What I meant was that I'm not working at exactly MFD, such that the focusing problem is due to getting too close. And the "miniatures" in this case are smaller than life size, but not like Lego mini figure kind of small. I have a collection of Dr. Who Daleks I'm using for one project, and they are about 13" tall, so I'm shooting scenes with those from 6 to 8 feet away. Another project I'm working on has doll subjects that are 22" to 25" tall.

JeffreyG wrote in post #17687967
Your signature lists 100/2.8L macro and 300/4L (IS?) which are both great choices for near-macro work.

Yes, the 100L is my go-to lens for small stuff normally, but with these larger toys, I'm finding it a bit too long at times; framing a scene with the Daleks in particular, I had to be like 12 feet away. Also, one particular project I have right now, which involves the 22" to 25" dolls, I especially wanted a softer look that the 100/2.8 just can't deliver, and the 85L can (example photo below). The 100L bokeh is decent, but it's just not on the same level as the 85L.

I've used my 300/4 IS for close-up work of insects and small animals, but the working distance is too great for what I'm using the 85 for right now. The 300/4 is indeed the IS version, and it's my wildlife lens that lives on my 60D.

Example of 22" doll, shot with the 85L II from a distance of about 1.2m, manual focus due to issues being discussed.

For that kind of shot, I'd be manual focusing via live view with almost any lens. I guess overall shooting near MFD with a fast lens and large aperture it going to be taxing on AF in general.

I've never used an 85L II. I do use a 6D and some primes: Σ15 FE, 28 1.8, 35 IS, 50 1.4, 85 1.8, 135L. USING THE CENTER POINT ONLY, my 6D is dead accurate with all of them, wide open. In incandescent light, it struggles a bit if the light-level is low, as in normal house lights at night. I have no complaints about the 6D, as long as I stick to the center point. That is OK with me.

My experience is completely different with the peripheral points. I find them right next useless. In really good light, with lots of contrast, and a non-moving target, they seem to work OK. My last attempt at peripheral points was a daytime, bright sunshine shoot of my grandson's baseball game. Hardware: 6D/100-400L, AI-Servo. Using the center point, I got nearly 100% keepers. Using any of the peripheral points yielded near 0% keepers. Somewhat in defense of the hardware, I was in portrait orientation and zoomed-in tighter for most of those shots. I could not make the peripheral points work.

I mentioned this in another post and another member suggested maybe it was not so much the peripheral points as the AI-Servo. I can't speak to that. I just went back the center point. My 6D is happy with that. So am I.

GeoKras1989 wrote in post #17688484
I've never used an 85L II. I do use a 6D and some primes: Σ15 FE, 28 1.8, 35 IS, 50 1.4, 85 1.8, 135L. USING THE CENTER POINT ONLY, my 6D is dead accurate with all of them, wide open. In incandescent light, it struggles a bit if the light-level is low, as in normal house lights at night. I have no complaints about the 6D, as long as I stick to the center point. That is OK with me.

My experience is completely different with the peripheral points. I find them right next useless. In really good light, with lots of contrast, and a non-moving target, they seem to work OK. My last attempt at peripheral points was a daytime, bright sunshine shoot of my grandson's baseball game. Hardware: 6D/100-400L, AI-Servo. Using the center point, I got nearly 100% keepers. Using any of the peripheral points yielded near 0% keepers. Somewhat in defense of the hardware, I was in portrait orientation and zoomed-in tighter for most of those shots. I could not make the peripheral points work.

I mentioned this in another post and another member suggested maybe it was not so much the peripheral points as the AI-Servo. I can't speak to that. I just went back the center point. My 6D is happy with that. So am I.

I'd agree. This is why I was forced to move to a 5D3. Pretty good success now.

JeffreyG wrote in post #17688449
For that kind of shot, I'd be manual focusing via live view with almost any lens. I guess overall shooting near MFD with a fast lens and large aperture it going to be taxing on AF in general.

Manual focus in live view does work quite well. For shots in outdoor daylight though, I find it hard sometimes to see the screen, even with the brightness turned up. I installed the Eg-S focusing screen in though, and it actually makes manual focus possible, especially for the more close up shots. Huge difference. I never had the precision focus screen in my 60D, because I used too many slow lenses with it. It hasn't been a problem for me when using my f/4 lens on the 6D either (the 24-70); the viewfinder did dim by about a stop, but I use this lens outdoors in good light nearly all of the time, and exclusively with AF. I often have a polarizer on this lens too, and even then it hasn't been an issue.