Frontfocusing Sigma 18-35 1.8 on 7D

[Grummbeerbauer]

Hi all,

I currently have an EOS 7D with a number of lenses, mainly Canon mid-range stuff (24-105L IS, 70-200L F4 IS, 17-55 2.8 IS, 100L 2.8 IS Macro, 85 1.8 ) and two Sigmas (30 1.4 EX (i.e., the old one), and the 10-20 3.5-5.6 EX).
I had always wanted a short FL prime lens which ended me up with the Sigma 30 1.4, but its lousy AF (both regarding precision and consistency) and soft images below f2.0 make me use it only on rare occasions where I can live with a low number of keepers.

I had been eying the new 18-35 1.8 since it was announced, even more so when the first rave reviews came in. Prime like quality and aperture in a zoom, simply amazing. Getting what basically seems to be a "variable FL prime lens" I pulled the trigger when there was a good deal on Amazon recently (710€).

When it arrived, I was thrilled by the lens' haptics and feeling for quality, which is better than any of the Canon L lenses I own and just light years away from the Sigma 30 1.4 EX with its delicate "crinkle" finish and scratchy AF ring.

Focal-length-wise, while the 18-35 is definitely not an "always on lens", I found it to be very useful for a variety of situations.

I had no intentions to shoot any test charts, so I started testing it mainly indoors in mixed light with mainly static real-life subjects, which is my primary use case for such a lens. To my great disappointment, this particular copy on my particular body appears to exhibit a considerable frontfocus. So I printed an autofocus test chart (the one found at http://www.klein-bild.de/focus.htm ) and did some test shots at 35mm and varying subject distances. I always set focus back to MFD or infinity before autofocusing again and did three test shots at each distance and AF adjustment setting. For each subject distance, I concluded the series with a shot focused manually with live view for comparison.
The good news is that different from my 30 1.4 EX, autofocus was very consistent. But that's where the good news ended already.

It turned out that both at around 1 meter and at 0.5 meter subject distance, I needed to dial in a whooping +20 to get about the same level of sharpness that I got from MF (see screenshots below, the legend is in German, "Motivabstand" meaning "subject distance"). While I would not like the idea of having to do such a huge adjustment, I guess I could live with that.
However, at 0.35 meter subject distance, I got the best results at an adjustment setting somewhere between 0 to 10. That meant that at the +20 setting that I would need for the longer subject distances, I would be in the "barely acceptable" range of sharpness. Anyway, at that point it became pretty clear that even for a single focal length, the single adjustment value my body offers me would not suffice.
Being tired of test shots for the moment, I then did some additional research and found the "dot tune" focus adjustment method (check out this video: DotTune: Autofocus fine tuning in under 5 minutes ), which seemed to make a lot of sense and promises much faster adjustments and does not require taking hundreds of test shots.
I tried it out at 35mm and 1,75m subject distance (as recommended for dot tune, 50 times FL). I found the range where I got a stable focus confirm to be between -3 and +16, which would indicate a correction setting of +6 or +7.
The big surprise was at 18mm and 0,9m subject, the range of stable focus confirmation only started (!) at a correction value of +17.
I could visually confirm both correction values with actual test shots.
A quick test on my trusty 450D basically confirmed the results on the 7D: Considerable front focus at 35mm which gets better at shorter subject distance, and a plain massive front focus at 18mm.

tl;dr

I found that the Sigma's AF is consistent at a certain focal length and subject distance, but would require different correction values for different FL/SD combinations. This gives me some hope that I could fix it with the Sigma USB Dock and some patience, but I am just not sure if I am willing to accept this.

So what should I do?
Keep it and get the USB dock? I have a week before my "no questions asked" return period expires.
Or send it back for good?
Any comments or similar experiences welcome.

Regards

Grummbeerbauer

 

[neuroanatomist]

I don't fully believe the 'dot tune' method, I'd recommend FoCal or at least commercial tool like LensAlign. In particular, contrast detect AF on the 7D has been shown to lack precision, so as the basis got 'dot tune' I wouldn't rely on it.

Regardless, it's certainly possible for a lens to require a different AFMA at the MFD vs. a reasonable distance away (although 20 units is a big difference!). Then again, AF issues are Sigma's Achilles' heel. However, you might consider the Sigma Dock (which I'd assume is compatible with this lens, but I haven't checked). Unlike the AFMA on the body, with their dock you can set different AFMA values for different distances, right in the lens (a great innovation by Sigma, IMO).

 

[horshack]

I don't fully believe the 'dot tune' method, I'd recommend FoCal or at least commercial tool like LensAlign. In particular, contrast detect AF on the 7D has been shown to lack precision, so as the basis got 'dot tune' I wouldn't rely on it.

Contrast detect AF is not the basis of DotTune. As described in the video, either CDAF or manual focusing can be used, whichever method the photographer finds more convenient and/or accurate.

 

[Policar]

I have the same problem with mine on the 7D and Rebel XT.

Oddly, it's fine on the 5D Mark III.

 

[Grummbeerbauer]

I don't fully believe the 'dot tune' method, I'd recommend FoCal or at least commercial tool like LensAlign. In particular, contrast detect AF on the 7D has been shown to lack precision, so as the basis got 'dot tune' I wouldn't rely on it.

I am also not fully convinced of the dot tune method... after all, in my case I had a range 20 (from -3 to +16) in which I got a stable focus confirmation... that is half of the AF adjustment range! And that was at 35mm and f1.8, i.e., the narrowest possible DoF with that lens.

Regardless, it's certainly possible for a lens to require a different AFMA at the MFD vs. a reasonable distance away (although 20 units is a big difference!). Then again, AF issues are Sigma's Achilles' heel. However, you might consider the Sigma Dock (which I'd assume is compatible with this lens, but I haven't checked).

I am indeed about two clicks away from ordering the USB dock, because I would really want this lens to work for me. Seems that for this lens, the Sigma software allows adjustments at 18, 24, 28 and 35mm and for subject distances of 0.28m, 0.35m, 0.5m and infinity, see the screenshot in the DPReview here http://www.dpreview.com/articles/4406844237/sigma-usb-dock-quick-review/2.

 

[Grummbeerbauer]

In particular, contrast detect AF on the 7D has been shown to lack precision, so as the basis got 'dot tune' I wouldn't rely on it.

Contrast detect AF is not the basis of DotTune. As described in the video, either CDAF or manual focusing can be used, whichever method the photographer finds more convenient and/or accurate.

I used CDAF for initial focus and did some manual fine tuning. In any case, the precision was excellent.

 

[Grummbeerbauer]

I have the same problem with mine on the 7D and Rebel XT.

Oddly, it's fine on the 5D Mark III.

I presume that with "same problem" you are referring to the general need for different AF adjustments settings?

Interestingly, Roger Cicala at lensrentals.com had a four-part series of articles on autofocus precision, starting with http://www.lensrentals.com/blog/2012/07/autofocus-reality-part-1-center-point-single-shot-accuracy (links to the subsequent articles are at the bottom). Very interesting read in general. In articles 3a and 3b he also found out that newer Canon bodies (from 5DIII onward) if combined with newer lenses focus considerably more accurately than older bodies (including the 7D).
That sort of confirms my own experience:I have never been too satisfied with the AF precision of my 7d anyway, it has always been hit and miss (could of course also be my technique, but what technique can be messed up when setting up the camera on a tripod and point it at a non-moving high contrast target?).

I think I will have to do some shots at daylight this week to see if the type of lighting also plays a role. Although a prime lens like this is of little use if it doesn't AF well in low or artificial resp. mixed light.

 

[horshack]

I am also not fully convinced of the dot tune method... after all, in my case I had a range 20 (from -3 to +16) in which I got a stable focus confirmation... that is half of the AF adjustment range! And that was at 35mm and f1.8, i.e., the narrowest possible DoF with that lens.

The width of the confirmation range has no correlation to the accuracy of the VF confirmation since it's the midpoint of that range which is used for DotTune. The width of the range is intentionally padded by camera firmware to make it easier for those who rely on the VF confirmation for manual focusing - a narrower range would make it difficult if not impossible to quickly MF, particularly on lenses with short focus ring throws. Firmware applies padding in equal amounts to both sides of the tightest point of phase detection, thus the midpoint of that range is the apex of focus. On Canon bodies the amount of padding is the same for both AF and MF modes; on Nikon bodies the firmware pads the range by 2x when in MF to further assist those acquiring focus to the dot, which is why the video tells Nikon users to keep the lens in AF mode to assure that the confirmed range fits within the +/- 20 AF tuning range.

 

[Policar]

I have the same problem with mine on the 7D and Rebel XT.

Oddly, it's fine on the 5D Mark III.

I presume that with "same problem" you are referring to the general need for different AF adjustments settings?

Interestingly, Roger Cicala at lensrentals.com had a four-part series of articles on autofocus precision, starting with http://www.lensrentals.com/blog/2012/07/autofocus-reality-part-1-center-point-single-shot-accuracy (links to the subsequent articles are at the bottom). Very interesting read in general. In articles 3a and 3b he also found out that newer Canon bodies (from 5DIII onward) if combined with newer lenses focus considerably more accurately than older bodies (including the 7D).
That sort of confirms my own experience:I have never been too satisfied with the AF precision of my 7d anyway, it has always been hit and miss (could of course also be my technique, but what technique can be messed up when setting up the camera on a tripod and point it at a non-moving high contrast target?).

I think I will have to do some shots at daylight this week to see if the type of lighting also plays a role. Although a prime lens like this is of little use if it doesn't AF well in low or artificial resp. mixed light.

By same problem I just mean it doesn't focus accurately or consistently; I think it's front focus. (I use the lens primarily for video.)

Fwiw, the 11-16mm Tokina I own similarly is less accurate on the 7D than on other cameras I've used it on.

Of course the one FF zoom I own that has focus issues (17-40mm L) focuses great on everything but the 5D Mark III. So does the Sigma 50mm f1.4 focus poorly on the 5D. Figures.

 

[Grummbeerbauer]

The width of the confirmation range has no correlation to the accuracy of the VF confirmation since it's the midpoint of that range which is used for DotTune. The width of the range is intentionally padded by camera firmware to make it easier for those who rely on the VF confirmation for manual focusing - a narrower range would make it difficult if not impossible to quickly MF, particularly on lenses with short focus ring throws. Firmware applies padding in equal amounts to both sides of the tightest point of phase detection, thus the midpoint of that range is the apex of focus. On Canon bodies the amount of padding is the same for both AF and MF modes; on Nikon bodies the firmware pads the range by 2x when in MF to further assist those acquiring focus to the dot, which is why the video tells Nikon users to keep the lens in AF mode to assure that the confirmed range fits within the +/- 20 AF tuning range.

Interesting info, thanks for that.
It would be great if one could choose less padding for MF in some custom setting.
While I am spelling out wishes: It would be great if all lenses offered adjustment options like the new Sigma lenses and if it could be done on camera (maybe via USB connection to have the better UI of a desktop application).

And it would be even better if we could just set up camera and test target, type in the measures subject distance, and have the software trigger test shots (one done with CDAF for comparison), adjust AF settings in between, and figure out the optimal AF adjustment settings all by itself. 8)

Of course this can only be done by a camera system manufacturer, not for third party lenses... well, unless Sigma gets in contact with the MagicLantern guys to add their own Sigma auto-AF adjust plugin. If ML can more or less directly access the EF mount contacts to implement Sigma's protocol for doch-2-lens communication in software.

 

[Grummbeerbauer]

To finish the story of my copy of the 18-35: I asked a colleague (semi-pro photographer with lots of experience, not a pixel-peeping gearhead ) to bring his backup 60D for testing. We tried it on both bodies in the same light on different targets at different subject distances and this basically confirmed what I observed on my 7D and on my 450D: At 35mm it is OK at close subject distances, way of at longer subject distances, and off by a mile at 18mm.

BTW: He had his Canon 50 1.4 on the 60D instead of a lens cap, so I tried it on my 7D: Focus was spot on. So it is definitely not an iffy AF on my 7D body.

Since I am not willing to bet on being able to fix these issues with the Sigma dock & software (and spending several hours of my precious free time in the process), the lens is going back to Amazon tomorrow. Since I am not yet willing to totally give up on this lens model, I asked for a replacement. Maybe I have better luck this time.

 

[Pi]

To finish the story of my copy of the 18-35: I asked a colleague (semi-pro photographer with lots of experience, not a pixel-peeping gearhead ) to bring his backup 60D for testing. We tried it on both bodies in the same light on different targets at different subject distances and this basically confirmed what I observed on my 7D and on my 450D: At 35mm it is OK at close subject distances, way of at longer subject distances, and off by a mile at 18mm.

This is a proof that you have a genuine Sigma lens. I have experienced this with the few Sigma lenses I tried.

 

[kwirk]

Registered here to post in this thread. I have the exact same problems with my Sigma. I tried resetting my 7D, deleted all my settings, everything. This is my second copy of the Sigma already, the first copy had the same issues. I'm going to buy the dock this time and see if that fixes it. This lens is amazing when it works, so I don't want to give up on that. I already lost $60 shipping my first copy back, which I could have used to buy the dock in the first place, so I figure I might as well just do it. Might be useful if they update the firmware anyway. I thought I was the only person having this issue. Phew. I also tried it on my 500D and it definitely has the same problem.

 

[candc]

i bought one a couple weeks ago and have the same problems it was front focusing on the wide end at close distance by about 6 or 8 inches but if you adjust for that with in camera afma then its way off at longer distance. its dead nuts on with the 70d and live view but i don't like having to use it like that. i have a usb dock but it will take a lot of fiddling to get it anywhere close to where it should be out of the box. i bought mine from digital rev and exchanging it is turning out to be a hassle so i am hoping there will be a firmware update. until then i would not recommend buying one unless you want to use it in live view mode or want to spend a lot of time trying to get the af right.

 

[kwirk]

i bought one a couple weeks ago and have the same problems it was front focusing on the wide end at close distance by about 6 or 8 inches but if you adjust for that with in camera afma then its way off at longer distance. its dead nuts on with the 70d and live view but i don't like having to use it like that. i have a usb dock but it will take a lot of fiddling to get it anywhere close to where it should be out of the box. i bought mine from digital rev and exchanging it is turning out to be a hassle so i am hoping there will be a firmware update. until then i would not recommend buying one unless you want to use it in live view mode or want to spend a lot of time trying to get the af right.

So you have the USB dock? Does the lens work now?

 

[candc]

yes i have the dock and i have it to where when its on its on but it is still inconsistent especially when the light gets dim. i wouldn't go through all this bother if it wasn't such a great lens optically i will try and post some test shots, they are all in camera jpeg wide open center point focused.

 

[candc]

another

 

[candc]

another

 

[candc]

another

 

[candc]

and another