5D3: AFMA variations between different AF points

Alright, 100L seems to back it up, albeit not as dramatically as 35L.

This image is cropped from the left side of the frame - one picture is taken with the right point (which is sharp across the frame which also proves I am parallel to my target ) and one picture is taken with the left point, which is not even sharp in the area it was focused on.

Well, I'm way out of my return/exchange period since I got my bodies back in June, so I guess I'll send them in just for the hell of it and see what happens. I wish I could be there to talk and guide the technician as he recalibrates the AF points, just worried they might get lazy and just say "good enough" and send it back. Any more updates shot in daylight? This camera should/needs to kick ass in the dark, many people buy this for weddings.

Hi everyone

I've been searching to try and find out what this problem could be!

I've had the mkiii and 50 1.4 + 100 2.8 macro for a couple of months now. I hadnt been impressed with the sharpness consistancy i was getting from the 50 in my studio sessions and finally started looking into it the past couple of weeks...

I dont have the FoCal software so have just done basic tests, with a focus chart and a stationary object and from my results, came to the conclusion that the centre point on the 50 was good but the outer points front focus - when i tested the 100 focus is great for the centre and outer focus points, so that made me think it was the lens not camera for me.... now not sure what to do

Well for a prime the problem can usually be corrected with micro adjust, but if the center is in focus then a MA to fix the outer points will screw up the center point. It is possible that this is just the nature of these particular lenses, and an optic phenomenon that needs some new MA options to fix ALL points on a particular lens. I think it would be pretty easy to have an automated mapping of each particular lens with every single one of the focus points and give a corrected fix adjustment for each lens+camera setup. I don't think it would be too difficult to have each one of the many focus points programmable, and it could actually be possible with existing cameras with just a firmware upgrade. If this is not yet possible, I see that as one of the new features of future cameras ... where you can map every single point and on a zoom map even different zoom positions. Personally, I have yet to test in detail the focus of my lenses ... but with the 5 DIII I have to say I am extremely impressed with the auto focus so far. Maybe if I micro peak and test each single point I might get disappointed, so best is to avoid it

yeah it sucks! any shots with the outer points are scrapworthy to me maybe a replacement lens might be better?... hopefully!

Well I contacted my warranty centre and they replied with this...

"Yes, the centre focus point is strongest AF point with guarantee accuracy at f2.8, the other points around it are only guarantee accurate if you set F stop to 5.6"

guess I'll only be using the centre point then - so much for all those extra points! :/

Yeah, I did my own test and was pretty disappointed with the middle focus point on the very right hand side. The keeper rate was less than 50%. When using the center point, the keeper rate was 100%. This was with a DoF of about two inches. Color me extremely disappointed.

I tried some AFMA tweaks and the keeper rate on that outer focus point quickly fell to near zero. Sort of makes me mad that I waited to upgrade to the 5D3 because it had a "pro" AF system. Soooo much better than the 5D2 and 5Dc. Not.

dave_p wrote in post #15242830
Yeah, I did my own test and was pretty disappointed with the middle focus point on the very right hand side. The keeper rate was less than 50%. When using the center point, the keeper rate was 100%. This was with a DoF of about two inches. Color me extremely disappointed.

I tried some AFMA tweaks and the keeper rate on that outer focus point quickly fell to near zero. Sort of makes me mad that I waited to upgrade to the 5D3 because it had a "pro" AF system. Soooo much better than the 5D2 and 5Dc. Not.

To be fair the 5D and 5DII would have completely whiffed at those same outer AF point shots in bad light Cross type is better than non-cross type, even if you're still not getting 100% hit rate.

I recently got a 5DIII and battled through the same issues with outer AF points. I don't have FoCal but did run extensive tests and applied micro-adjustments. I initially had the impression that the right side AF points were weaker than the center and left, but I think I'm in a good spot now. This was most visible with the 35L at medium to long distances, almost undetectable in the 85L, and nonexistent in the 135L. I still have a bit of doubt which is ridiculous in a $3000 camera, but I'm now getting good results with outer AF points although not with 100% accuracy, so I'm leaning toward the prevailing theory that the outer points just aren't that dependable and this is normal behavior.

I think we're hitting against the limits of phase detect AF with AF points near the edge of the frame (problems caused by edge softness and angle of light, which is why the problems are more prevalent on wide angle lenses like the 35L) especially with large MP images where viewing at 100% zoom effectively enlarges the image to huge proportions, allowing us to scrutinize images in a way we never could before. It's tough-to-fix problems like this that make me want to chuck out all my DSLR stuff and go with a contrast detect system. Roll on mirrorless! Yes I know--there are still problems with that too, but at least you don't have to calibrate your lenses.

When you say "I think I'm in a good spot now", what do you mean? How did you "work around" this issue? I'm not attempting to be confrontational...I am genuinely interested to know what you've done or how you've made peace with it on your lenses where you've noticed the issue.

I literally had to throw out otherwise good shots from a shoot I did for an agency. This was with an 85 1.8 from about 10 feet away. And, as I said, my subsequent testing showed that the keeper rate was less than 50% with the outer most middle focus point (right hand side). So from a "real world" perspective, I might as well have a 5Dc or 5D2 because I no longer feel comfortable with anything but the center point on my 5D3. Certainly not trying to say you or anyone else isn't "real world", just trying to emphasize that when I need acceptable photos for my purposes, I can't trust the outer AF points on a 5D3. Because 50% is not acceptable. Not even close.

I've not done testing at middle apertures (say, 5.6ish to 11) such that my DoF is larger. Maybe the outer points are trustable once DoF increases enough.

dave_p wrote in post #15243070
When you say "I think I'm in a good spot now", what do you mean? How did you "work around" this issue? I'm not attempting to be confrontational...I am genuinely interested to know what you've done or how you've made peace with it on your lenses where you've noticed the issue.

I literally had to throw out otherwise good shots from a shoot I did for an agency. This was with an 85 1.8 from about 10 feet away. And, as I said, my subsequent testing showed that the keeper rate was less than 50% with the outer most middle focus point (right hand side). So from a "real world" perspective, I might as well have a 5Dc or 5D2 because I no longer feel comfortable with anything but the center point on my 5D3. Certainly not trying to say you or anyone else isn't "real world", just trying to emphasize that when I need acceptable photos for my purposes, I can't trust the outer AF points on a 5D3. Because 50% is not acceptable. Not even close.

I've not done testing at middle apertures (say, 5.6ish to 11) such that my DoF is larger. Maybe the outer points are trustable once DoF increases enough.

I had front-focusing issues. In the course of testing I concluded that the issue really only showed up to any noticeable extent with the 35L. It was ok at close distance but pretty bad at medium to long distance with the right outer points. I thought the 85L was spot on at first but the more I tested it the more I saw it not quite nail focus either using the right outer points--not nearly as badly as the 35L but it had some problems. Note that both lenses seemed ok at close distance, but at medium distances the misfocus was obvious. The 135L didn't seem to have any of those issues, which gave me hope that it wasn't a tilted AF sensor array or something awful requiring Canon service to intervene.

So I tested all my lenses and applied AFMA as needed, slowly chipping away at any softness I was seeing. I tested at close and medium distances, used the center and right center (upper center in portrait orientation) AF points. I also made a point to test in good light to avoid additional issues stemming from lack of light--my first round of tests were indoors at night which, although real world, doesn't do the camera any favors. The testing was a bit haphazard and by the seat of my pants--handheld, shooting a human target instead of a static, flat surface, etc.--but I eventually settled on +10 for the 35L and +4 for the 85L. This is what looked the best to me using the center and outer AF points.

I can't explain why I saw a discrepancy in right vs. center and left points at first but no longer do after AFMA. It still eats at me that the discrepancy has disappeared. I hate not knowing for sure what's going on and have no sensible theories to explain it. I went as far as going back to my original test setup indoors at night and still don't see the point-to-point discrepancy I saw before AFMA. Could be that my center point is now not as sharp as it was before, but it looks plenty good to me so I feel that that's not the answer.

Admittedly there were a lot of variables at play, which is why I only "think" I'm in a good spot. I'll know for sure after I've used it more particularly for jobs and have a better feel for it. Bottom line is I'm confident that I've got it calibrated to an acceptably good point but still need to work out any kinks with real shooting. I would use my 85L and 135L with total confidence even in portrait orientation using the outer cross-type AF points, but again the caveat is that I haven't really used the camera in anger. I'll be keeping a close eye on the 35L and won't use it at medium/long distance in portrait orientation with the outer AF points for critical shots until I feel more confident in it. That's the main area that still gives me pause. Otherwise I'm feeling pretty good with the understanding that I reserve the right to change my mind if results turn out poorly.

I don't know if any of that helped. I considered getting FoCal to do more scientific testing but decided not to since I was able to achieve good calibration after all. I still might though eventually, and I would recommend it as a valuable tool for anybody needing utmost AF performance and in need of a way to fully document their AF issues.

I'll say this: I have test shots with the 85L from about 10 feet away that are razor sharp at f/2 in portrait orientation, so you should be able to nail shots with the 85mm f/1.8 from that distance too. If you aren't getting the results you need with all the AFMA in the world then I'd lean toward a trip to Canon service.

Thanks for the detailed reply. I need to do some more testing, but I guess it's good to know that the camera is capable of focussing accurately with the outer points. I just wish my particular one was...I feel like I'm kind of cursed in the focus department. Oh well. Thanks again.

me too...

I'm unsure what to do in regards to their reply ""Yes, the centre focus point is strongest AF point with guarantee accuracy at f2.8, the other points around it are only guarantee accurate if you set F stop to 5.6"

so is it just tough luck that I cant use the outer points because they are "only guarantee accurate if you set F stop to 5.6" - I dont want to set to 5.6 - i want to get beautiful bokeh! Is there adjusting that can be done? should I send it somewhere?

Michelle BT - I would ask Canon to calibrate. They told me they could do electrical or hardware adjustments to calibrate the points that are way off (however, I exchanged my 5D3 that was particularly obnoxious). If you are within the warranty period, they should do it for free. Sending a candy bar or two in the box with a note that says pretty please works wonders.

DaveP - I feel much the same about my 5D3. If I am using 35L wide open or so, which I always am - I might as well just be shooting my 5D2. I bought the 5D3 and 35L combo to go together, and it has definitely not worked out like I thought it would.

I am thinking about trading my 5D3 for a 1D4. I really loved my 1D3, and sold it to fund the 5D3.

michelle blossom tree wrote in post #15251020
me too...

I'm unsure what to do in regards to their reply ""Yes, the centre focus point is strongest AF point with guarantee accuracy at f2.8, the other points around it are only guarantee accurate if you set F stop to 5.6"

so is it just tough luck that I cant use the outer points because they are "only guarantee accurate if you set F stop to 5.6" - I dont want to set to 5.6 - i want to get beautiful bokeh! Is there adjusting that can be done? should I send it somewhere?

While it might be true that the outer AF points are only guaranteed accurate to f/5.6, that doesn't mean that you should settle for a 5DIII that won't focus well with those points. After AFMA mine seems to be accurate enough to the naked eye, shooting wide open with a variety of large aperture primes. I've heard stories of other people having trouble with the outer points and having Canon fix it to perfection. Exchange if still within the exchange period or hassle Canon until they fix it. Be persistent and firm, but be polite. You should have a working camera sooner rather than later.

after reading this thread, maybe people have unrealistic expectations about their gear. Hell, I shoot a 5D2 and outer points the majority of the time, but when it's dark, center point and even manual focus when needed. I figure the 5D3 outer points can only be better than the 5D2 a certain number of EV's, so in a lot of low light situations, it's a no go.