135L autofocus performance - have I got a lemon?

[PeteH]

I've recently bought well-used examples of the 135L and 85LII and need to decide within the next week or so whether to keep them. I'm not convinced about the AF of the 135 and I'm wondering whether what I'm seeing is typical.

I'm using a 7D2. Other lenses include a bunch of EF-S zooms, Sigma 50 Art and the old 70-200 OS (pre-Sport), and the EF 85/1.8.

I've microadjusted the 135 to +8, which gives me a reasonable hit rate, but only reasonable. Focusing on my own car's number plate at a range of 10 m or so while it's sat on the drive, I'd say I'm getting maybe 60% spot on, 20% near miss that I'd consider acceptable (the sort of thing you could cover up with a bit of deconvolution sharpening and nobody normal would notice in casual sharing), and 20% miss. Misses seem to be randomly either substantially in front (like a metre or so) or a similar distance behind. (This is using centre or near-centre focus points, the hitrate seems to drop further with points closer to the edge.)

I'm talking here about focusing wide open. I'm using exclusively AI Servo and back-button focus as I do with all my other lenses. I never focus-and-recompose, I always click the shutter with AF active. The results described above would be using either single-point AF or single-plus-four expansion, I don't see much difference either way. I've also tried spot AF with the 135 and it maybe gives slightly more reliable results, but only slightly. (I only occasionally use spot AF in the normal run of things as typically I find tracking more reliable with single or single-plus-four).

I'm reasonably confident that the misses I'm seeing are not the result of user error - there is clearly nothing anywhere near the active focus point that is in focus, but I can generally find a band of sharp focus either in front of or behind the intended subject. (I have found that 1/250 is a little too slow with the 135L so I've got auto ISO set for minimum 1/500.)

Of these two lenses I currently have on trial I was expecting the 85L to be the finicky diva but actually I've found it not too bad. I certainly don't get 100% hitrate with the 85L wide open, and the tracking performance obviously isn't great, but I'd expected that. What I wasn't expecting was to find the 135L not really much more reliable at f2 than the 85L is at f1.2.

I'm also finding the 135L slower to focus than I expected. My go-to for capturing action would be the 85/1.8 which I've used to photograph sprinting dogs and suchlike with a decent hitrate, even wide open. That also goes for the Sigma 70-200 OS (which as I said above is the non-Sport version). The 135 isn't bad - obviously it's much faster than the 85L - but it's certainly nowhere near as quick as the 85/1.8 is, and it also seems slower than the Sigma 70-200 (though now I write that I haven't compared them side by side). I haven't tried it properly yet with anything moving especially fast but it just doesn't seem very quick.

Both the 85L and the 135L give much more reliable focus in Dual Pixel live view than through the viewfinder unless I'm trying to track action - no big surprises there.

Does that sound typical of your experience with the 135L, or is the one I've got on its last legs? Optically it's absolutely fine - the glass looks clean and clear - but the focus ring is fairly well-worn. I think it has seen a fair bit of use, though it doesn't look to have been knocked about too much.

Any thoughts greatly appreciated - thanks!

 

[Sporgon]

I wouldn’t use the 135L on AI Servo for static subjects, especially at f/2. I think that if I did with mine I’d probably get the results you have. Certainly try it in One Shot to find if it’s accurate or not. Also beware of setting an inaccurate AFMA.

 

[PCM-madison]

I've used my 135mm F2 L on a 7DM2 for sports/action quite a lot and get a very high hit rate. Your experience does not match mine.

 

[Del Paso]

Same thoughts here.
AFMA adjustments can be really tricky, so, this is what I'd check first, using a tripod, One Shot AF AND live view's 10 X loupe function for checking, after having focused thru the OVF.
If then you don't get a high rate of well-focused shots: return the lens to the seller!

 

[YuengLinger]

Yes, I remember the days of DSLR's! So many settings on the checklist.

But the ef 135 was outstanding--though I had to keep my shutter speeds way up!

Good advice about not using AI Servo if you know the subject will remain static, but it sounds, OP, as if you know how to AFMA.

On the EOS R, the advice about AF and static subjects is the opposite, in my experience: Just leave Servo AF on at all times. But on the dSLR it definitely reduced my hit rate for portraits. By the way, my ef 135 is still excellent, though I look forward to learning whether the R5 IBIS helps with handling it. (I actually got a monopod JUST for the ef 135. That's how much I cherish that lens.)

That lens is loved by so many. I hope you get it sorted and let us know what you decided to do!

 

[PeteH]

Thanks everybody for answers so far.

I've AFMA'd all my other lenses to my own satisfaction, and I think the 135's focus scatter is now centred about the right place. Given that I seem to be seeing approximately equal amounts of front- and back-focused misses I don't think there's anything else I can do there. I am wondering whether there's quite a strong distance dependence with AFMA on this particular lens (specifically, needing higher positive values at longer focus distances), which is something that is obviously possible but that I haven't personally encountered before.

It's compared in particular to the Sigma 50A and the Canon 85/1.8 that the 135L seems erratic to me. This is the first L lens I've used other than a short hire of the 70-200IS (II). Obviously I'm aware that you can potentially get shallower DOF with the 135/f2 than with say the 85/1.8, and that my technique would be tested more than I'm used to, which is why I'm here asking you kind people for advice.

I forgot to mention that I also have an old Sigma 30/1.4, which I adore in many ways but which has a bit of a tendency to just randomly focus wherever it feels like. This 135 isn't as bad as that, but it reminds me more of it than I was expecting. As I mentioned it does seem like it's absolutely dead on about half the time, it's just that it does also miss completely quite a lot.

On the subject of Servo AF it doesn't currently seem like I miss much more with a moving target (my kids on a see-saw or running across the frame or whatever) than with a stationary one. As I mentioned I haven't yet stress-tested it with anything coming fast towards / away from the camera.

I'll try it with single-point and see if it's any more consistent - that's an interesting suggestion, thank you. It could be that always using Servo is the sort of thing that I could get away with when using more forgiving lenses but not with this one.

Maybe worth mentioning that I absolutely adore the IQ when I do get a well-focused image, it's glorious. So I'm keen to work out what's going on here.

 

[Mt Spokane Photography]

To test a suspected problem lens, do not hand hold it. Remove as many variables as possible. Put it on a tripod, defocus before each shot and take 10 or so. Try at least two or even three distances because that does make a difference, some defective lenses only act near MFD for example.

A lens can definitely have excessive friction or binding or a number of issues that could throw autofocus off. I found my 135mm L to be very accurate and consistent with the cameras I used at the time, the 5D MK II for example.

If there is any doubt, return it. A new one is not that expensive and there will be sales before Christmas.

One of the modules in FoCal is autofocus consistency where it runs the test automatically but you must still set the distances. It gives a reliable report because it removes as many human errors as practical.

 

[PeteH]

Update for anybody who's interested. Thanks again for your replies and interest!

I returned the suspect 135L and have now obtained another, which cost me a bit more but is in rather better condition, and my word the AF on this second lens is completely different to that first one I had.

Autofocus is now fast and secure, rather than ponderous and a bit random. I think that first one really must have been on its last legs, it was decidedly leisurely about getting up to go and look for focus, and it often got lost on the way.

I was initially worried because this new 135L needed an all-time AFMA record for my camera of +12 (I actually had it set to +13 for a while). Having got that sorted out though it's been comparatively plain sailing. I'm now getting the kinds of results I'd have expected from an older L prime, which is to say, it's quick and accurate (albeit not completely infallible). The random misses I was getting with the first copy have all but vanished this time round; it hits reliably enough in Servo that I'm happy enough going back to using that mode exclusively, including wide open under tricky conditions (eg backlit and/or small focus target and/or low contrast and/or close-range moving target).

This lens gives the kinds of results I'd hope for in the test I described above, where when I shoot at a number plate at medium range wide open in Servo, almost every time I get focus either bang on or near enough for me not to worry about it too much (granted YMMV here). Very loosely and non-quantitatively I'd characterise the AF scatter now as one-and-a-bit-times the DOF wide open - it's not perfect every time at a pixel-peeping level, but it's almost never OOF in a way that a normal person would notice - whereas with the first lens I had the AF scatter was more like 3-4x the wide-open DOF, and obvious misses were frequent.

A couple of additional observations. I'm pretty sure it's not quite as robust in terms of action tracking as my 85/1.8, given a subject at equivalent size in the frame. I may be able to tune AF case settings to improve the results there a bit but when I really push it (think kids coming through the frame on a bike at head-and-shoulder size) it seems to start falling behind the subject a little before the 85/1.8 does.

It's also a little fussier about AF target contrast than my Art 50/1.4 or EF 85/1.8, and it reaches its limit a little earlier in terms of finding focus on tricky targets (eg a backlit head). This may simply be because it's a longish unstabilised prime - remember I'm using a 7D so crop factor plays a part here. Hand shake is very obvious through the viewfinder at this length and I don't find it hard to imagine the AF system finding it marginally more of a struggle to lock onto something that is dancing about in my shaky human hands.

DOF is noticeably shallower than I'm used to and I'm probably going to default to single-point AF rather than single-plus-four which is my normal go-to. Spot AF is even better as long as there isn't too much subject motion. I'm thinking I'll reconfigure the custom AF settings on my 7D to allow one-touch access to spot AF - currently I have one back button set for auto tracking which I use occasionally under circumstances where I think it'll be able to cope adequately (there aren't very many) and maybe I'll get rid of this.

Anyway there you go. Thanks again for your input, and hopefully this little anecdote is helpful or at least interesting to somebody.

 

[YuengLinger]

Great, PeteH! What subjects? Have you tried it with lake/pond water in the scene? Magic with OOF ripples and reflections.

 

[PeteH]

Great, PeteH! What subjects? Have you tried it with lake/pond water in the scene? Magic with OOF ripples and reflections.

Mostly my family, immediate and extended (I have two kids aged 3 and 6). Very occasionally other people or things, when people ask me to or the opportunity arises. Potentially the odd competition if I can find the time. No lakes or ponds, but I had it out by the sea at the weekend and while I haven't uploaded yet it does seem to render OOF sea-glint pleasingly (and the way it flares shooting against the sun looks exceptionally pretty too).

I'm "the guy with the camera" for various social and work purposes but I don't shoot professionally. I used to keep a Flickr gallery but it's now rather out-of-date: click me . I'm at the stage of life where I photograph a lot of children blowing out candles on cakes, or at least I did before... well, you know. But for obvious privacy reasons I don't tend to post anything like that anywhere public.

 

[Mt Spokane Photography]

Update for anybody who's interested. Thanks again for your replies and interest!

I returned the suspect 135L and have now obtained another, which cost me a bit more but is in rather better condition, and my word the AF on this second lens is completely different to that first one I had.

Autofocus is now fast and secure, rather than ponderous and a bit random. I think that first one really must have been on its last legs, it was decidedly leisurely about getting up to go and look for focus, and it often got lost on the way.

I was initially worried because this new 135L needed an all-time AFMA record for my camera of +12 (I actually had it set to +13 for a while). Having got that sorted out though it's been comparatively plain sailing. I'm now getting the kinds of results I'd have expected from an older L prime, which is to say, it's quick and accurate (albeit not completely infallible). The random misses I was getting with the first copy have all but vanished this time round; it hits reliably enough in Servo that I'm happy enough going back to using that mode exclusively, including wide open under tricky conditions (eg backlit and/or small focus target and/or low contrast and/or close-range moving target).

This lens gives the kinds of results I'd hope for in the test I described above, where when I shoot at a number plate at medium range wide open in Servo, almost every time I get focus either bang on or near enough for me not to worry about it too much (granted YMMV here). Very loosely and non-quantitatively I'd characterise the AF scatter now as one-and-a-bit-times the DOF wide open - it's not perfect every time at a pixel-peeping level, but it's almost never OOF in a way that a normal person would notice - whereas with the first lens I had the AF scatter was more like 3-4x the wide-open DOF, and obvious misses were frequent.

A couple of additional observations. I'm pretty sure it's not quite as robust in terms of action tracking as my 85/1.8, given a subject at equivalent size in the frame. I may be able to tune AF case settings to improve the results there a bit but when I really push it (think kids coming through the frame on a bike at head-and-shoulder size) it seems to start falling behind the subject a little before the 85/1.8 does.

It's also a little fussier about AF target contrast than my Art 50/1.4 or EF 85/1.8, and it reaches its limit a little earlier in terms of finding focus on tricky targets (eg a backlit head). This may simply be because it's a longish unstabilised prime - remember I'm using a 7D so crop factor plays a part here. Hand shake is very obvious through the viewfinder at this length and I don't find it hard to imagine the AF system finding it marginally more of a struggle to lock onto something that is dancing about in my shaky human hands.

DOF is noticeably shallower than I'm used to and I'm probably going to default to single-point AF rather than single-plus-four which is my normal go-to. Spot AF is even better as long as there isn't too much subject motion. I'm thinking I'll reconfigure the custom AF settings on my 7D to allow one-touch access to spot AF - currently I have one back button set for auto tracking which I use occasionally under circumstances where I think it'll be able to cope adequately (there aren't very many) and maybe I'll get rid of this.

Anyway there you go. Thanks again for your input, and hopefully this little anecdote is helpful or at least interesting to somebody.

Sounds much better. Set the focus limiter to match your focus range. That makes a big difference in time to find the subject. The lens does not need to search macro distances if you are shooting at over 0.5m or it doesn't search distant objects when you are at less than 0.5m. When in full range, it can take a while if its looking at the wrong distance when you start to AF.

 

[Kiton]

Glad to hear you returned it for another and are happy. I have had mine since about a year after they launched it and it is an amazing lens, not that you have a good one, you will love it no doubt!