Severe focus issues with 40D + EF-S 17-85 IS

[antoine_clappier]

Hi all!

 

I have lately acquired a 40D with the EF-S 17-85 IS kit lens and am experiencing

severe focus issues.

 

After taking a few hundred photos today, I came home to discover that about 90%

where badly focused. My first thought: user error! Sadly, it does not seem to be

the case. Here is what I have found:

 

1- This is not a speed issue. The photos are not fully blurred. The focus

distance is simply wrong. Besides, the speed was largely in range for the focal

length/aperture/ISO used (even without taking into account the IS which was

turned on).

 

2- This is not a focus lag issue. I was not shooting Formula 1 racing cars but a

peaceful landscape!

 

3- The issue appears only for focal lengths between 27 and 35mm (35mm equivalent.)

 

4- Focus distance is always way too short. Most of my pictures were faraway

trees over a clean lawn taken at 27mm. Focus distance should have been infinity.

Photos show a focus distance of about 3 to 4m.

 

From there, I assumed another user error: bad selection of the AF point! Not so!

I went again outside and took a close look at what was going on:

 

1- First thing first, I activated only the central AF point. Aimed to a tree (I

love trees!) and shoot. Same problem: Tree blurred, the lawn in front of my

feet: crisp.

 

2- I then switched to tripod, LiveView and manual focus (BTW manual focus with

LiveView is a pleasure to use for landscapes). No problem: focus distance

perfect, picture so crisp it hurts!

 

3- Third test, LiveView, manual focus, then AF-ON: photo blurred, bad focus

distance. AF has actually ruined the focus previously done manually.

 

4- I will skip a few other tests, the conclusion is surprising: my 40D + EF-S

17-85 IS almost always focus at about 3m or less when using the low end of the

focal length. Prayers, threats will not change its behavior. My 40D thinks

infinity starts at 3m!

 

Questions:

- am I still doing something wrong (I just started using the 40D after all)?

- am I experiencing some kind of hardware failure?

- is there a known issue with the 40D AF (based on several forums it seems that

I may not be alone)?

 

Can someone on the forum with a 40D + EF-S 17-85 shoots at 17mm a few faraway

trees (or something else ;-) ) and see if he/she can reproduce the issue.

 

Sorry for the long post. I am highly interested to get your comments or tests,

 

Antoine

[rainer_t]

Sounds like a real problem to me ...

 

THis link ...

 

http://www.focustestchart.com/chart.html

 

... describes a formal and repeatable test that you could do.

If this test also supports your findings I guess it's time for your 40D and 17-85 to see a canon service. (Eventually it's just the adjustment of the secondary mirror.)

 

Can you test focus with any other lenses?

[jo7hs2]

Antoine,

 

Could you post a few samples?

[antoine_clappier]

Joshua,

 

I am fairly new to this forum. Is there a place where I can post full resolution photos? Or do I have to post on an other web site?

 

Thanks,

 

Antoine

[zigzag]

The normal thing is to post maybe a smaller overview photo and then post a '100% crop' of a detail area - i.e., with the full photo at 100% size crop out an area of interest and post that - so you don't need to post the full thing but people get to see the detail.

[PuppyDigs]

The first I'd do is try a couple other Canon lenses. If those are okay, you probably need to

have your 17-85 calibrated. If it's still blurry, your 40D needs adjustment.

[antoine_clappier]

Posted an example on Flickr:

 

 

Image developed in LightRoom 1.3 from RAW file.

 

Exposure is a bit unusual because this photo was taken for future HDRI treatment and was underexposed on purpose. I have corrected in part the underexposure in LightRoom.

 

Antoine

[antoine_clappier]

Another example:

 

 

Focus was on the house not on the iced lavender in the foreground. Same comment as previous post about the strange exposure!

 

Antoine

[jo7hs2]

Antoine,

 

Could you post one of your manual focus attempts, too, preferably with an autofocus attempt of the same scene? I can see a little out of focus areas on the one with the tree, but the loss of detail I see on the house one (in the trees) looks like edge sharpness loss to my untrained eyes viewing tiny images.

 

BTW, trees at that distance are probably not at infinity.

[brucecyr]

Antoine,

 

What Puppy Face said if you have any doubts. But I have doubts about your reports. For example in the last post the house just might be in focus and overall the exposure looks about right. What I mean is that it looks like a normal photo to me, although its hard to judge.

 

Several comments, like:

 

"Focus distance should have been infinity. Photos show a focus distance of about 3 to 4m."

 

make me wonder about your level of experience -- I'm not saying you don't know what you're doing, and you're obviously a very capable person, but I just wonder how much experience you have. For example, how can infer focus distance should have been infinity unless you set the focus point to infinity?

 

Also, the first photo seems to have been shot at f10, which would produce pretty good DOF on a cropped sensor body. Overall, I don't see any glaring problem with that shot either -- maybe what is evident to you, but not to us, is incongruity between your intent and your result!

 

I would suggest you try the linked focus test site above.

[davidhay]

Had the same problem with my Canon 10D. All lenses focused properly EXCEPT the 24-85 at 35 or wider and the 20-35 at all focal lengths. Serious front focus with both. Sent the camera and these two lenses back under warranty for calibration. Still using it five years later with no focus problems.

[antoine_clappier]

Many thanks for your useful comments.

 

By reading recent posts on this forum, I have found a very interesting link (Technical overview of Canon AF):

http://mkoehler.de/index.php?option=com_content&task=view&id=15&Itemid=9

 

 

The infos found at that page could explain some of my focus issues. As you noted, I was using a small aperture. This combined with the fact that the AF is calibrated for a circle of confusion of 0.02mm may explain that focus stops way short. On a 10mp APS-C censor, 0.02mm is 3.45 pixel wide (3888*0.02/(36/1.6)). This is a large number! If AF is satisfied with a 0.02 circle of confusion, the resulting photo will have pixels blurred on a width of 3.45 pixels (defect clearly visible at 1:1 resolution).

 

I have found a way to prove/disprove indirectly this theory. If the AF stops as soon as it finds a 0.02mm circle of confusion, the prior lens focus distance will have a measurable effect on the focus plan distance. A simple experiment just shows that:

 

- Set manually the focus distance to the minimum. Aim to an object at say 3m. Turn AF-ON. Note the focus distance.

 

- Set manually the focus distance to infinity. Aim to same object. Turn AF-ON. The focus distance is now noticeably longer.

Try this at home it is interesting!

 

 

I believe the previous experiment shows that the AF indeed stops at a given circle of confusion and does not go further. Then, the correct focus distance is between the two extremes of the Depth of Focus starting and ending with a circle of confusion of 0.02mm (probably the values determined by the previous experiment).

 

This said. I do not have the same issues with other Canon bodies (including the simple Canon G7 that I am still using). Also, why the problem suddenly stops happening after 35mm?

 

I have the feeling that there are two problems: it is possible that the 40D favors AF speed over accuracy (a much longer investigation is needed to prove that) and that I have a calibration issue with the lens/body that increases the AF problem.

 

Find below personal answers to your comments,

 

Antoine

 

Bruce:

how I have determined that the focus should have been infinity? Easy! The 40D features a very capable Live View that allows you to manually set the focus with almost perfect accuracy. From there it is easy to check what is the focus distance by reading the marking on the lens itself.

 

I know this is not a reliable method and gives only a ballpark number. But, in my case, manual focus gives a focus distance "around infinity" and auto-focus sets a focus distance around 3m. The difference is large enough to show there is indeed a problem.

 

About my experience: I am not a professional photographer but I believe I have a strong background in optics, computer graphics, etc. I have written several software in these fields and founded several companies. I am currently developing a Computational Photography software (HRDI, Synthetic Relighting, etc). I have bought the 40D to test my algorithms. In any case, I am here to learn. Your comments are welcomed!

 

Bruce, Joshua:

I agree with your comments about the photos. The problem is not glaring but that is just because I have used a fairly large DOF (aperture was respectively f10 and f9). The photos are not totally ruined and could be saved by using a fairly strong sharpening in post-process. Anyway, that does not change the fact that the camera has focused in the foreground instead of the background as wanted.

 

Rainer T:

Very interesting post. I will investigate this simple method for back/front focus test. I tried quickly with a simple book yesterday night and did not find an obvious issue. But I will try again with the provided chart.

 

Baby Face, Joshua:

I cannot do further tests today. The weather is absolutely awful today in the Loire Valley! I will continue on Monday if I can.

[marknagel]

Antoine,

 

Are you sure this isn't a DoF issue? Each picture has a focused area, it really depends what each focus point was and the aperture you were at.

 

m

[marknagel]

Never mind my above post, I looked at he Exif and see you were at f/9. You should have have more DoF than what I was thinking. Try another lens. I had two of these and one was awesome, one was terribly soft.

 

m

[brucecyr]

Antoine,

 

Interesting experiment. Basically you are showing that AF operation depends on its history. Whether your "first conformable circle of confusion" (FCCoC) hypothesis jibes with your data is another question.

 

Another unproven claim is that this is a general phenomenon. At this point, it could be (1) lens-specific, (2) camera body-specific, or even (3) specific-lens-copy-specific. Or it could be an unspecified interaction among these three and possibly other unspecified factors. N.b., David Hays' post would suggests, at least in his case, it was specific to his lens-X-body interaction.

 

I've reached the limit of my competence, but maybe others can help out. It would also be interesting to hear the result of your further investigations.

[brucecyr]

The other point I would like to suggest is that Live View is very much terra incognita. Its a fairly new capability and I don't trust it nearly as much as the old fashion technique of using a selected focal point in the viewfinder.

 

I would think your first step would be to replicate this experiment using the viewfinder and a selected focal point. Maybe others could also contribute their own trials because experimental work needs to be verified by repeatable trials from multiple, independent sources.

[jo7hs2]

I would tend to agree with Bruce C that live view may not be the best way to check focus. I've never found the LCDs to be all that useful, even zoomed in, to confirm focus in many environments, especially with regards to foliage. That's purely opinion, but I suspect that live view, despite being on cameras with *slightly* larger LCDs, wouldn't be all that much more helpful.

 

Maybe try putting your camera in A-DEP mode and seeing if the camera changes the way it autofocuses the scene. Read:

http://www.luminous-landscape.com/tutorials/dep.shtml

[antoine_clappier]

Bruce,

 

As I wrote in my today post, I do believe that it is probably due to a defect with the lens or body (which probably increase the "FCCoC").

 

In any case, it has forced me to look in more details how a modern AF works. The AF history phenomenon is interesting. It could be actually used to help the AF in some situations.

 

I will continue to post if I find an answer to my problems.

 

Thanks,

 

Antoine

[antoine_clappier]

Joshua,

 

I thought like you that Live View will not be precise enough for manual focus. I was surprised to find I was wrong.

 

The quality is very impressive. At full zoom, the Live View gives a 1:1 resolution. One censor pixel maps to one LCD pixel. The image is extremely sharp and is superior to the viewfinder (at least tothe 40D one!) for very accurate manual focus (on a tripod of course).

 

A surprising thing is the fact that the picture is much sharper during Live View than during review of the taken picture. I don't know how Canon achieves that.

 

I think anyway LiveView is more useful in a controlled environment (studio work).

 

 

Your A-DEP suggestion is interesting. I will try that too.

 

Thanks,

 

Antoine

[brucecyr]

Antoine,

 

In your detailed description of your procedures, its unclear whether you were using a tripod for the AF portion of your tests, i.e., the part where you had inconsistent focusing problems.

 

You do mention comparing AF with other modes, namely, tripod, Live View and manual -- which suggests that maybe you didn't use a tripod for the AF portion. OTOH, maybe you just failed to document this practice, which would be easy enough given the mass of detail you wrote.

 

While the lens does feature IS, it still seems that you should verify your results with a tripod to eliminate as many unmeasured sources of error as possible.

 

FWIW, I tried replicating your results in the local mall (too much wind for testing outside!) early this a.m. with a tripod, but I didn't see the apodeictic differences you observed. There may be something there, given the lack of total consistency I found, but I would need more time and less of a crowd of kibitzers. Unfortunately, that particular lens was recently sold, so when the post office in the mall opened, I packed and shipped it off!

 

I'm looking forward to an update on your explorations.

[antoine_clappier]

Bruce,

 

The sun was back on Monday morning and I had the possibility to do systematic tests on a tripod at various focal lengths. I will prepare a complete follow up post today with my findings. Let's just say for the moment that I think they prove that my AF is not working correctly for short focal lengths.

 

I will post under a new thread as the current one is getting a bit too long.

 

Thanks,

 

Antoine

[kay_bryan]

This is not an answer to your exact question, but I bought and returned TWO of the 40Ds before I finally got one that did not have focus issues. I am thinking that a lot of people that buy this camera will not realize that the camera even has issues unless they have used another digital slr camera and were used to taking sharp images.

 

I'm happy with the 40D now, but am extremely disappointed by the fact that I had to go through three cameras to find one that focused correctly. I suspect that there a lot of 40Ds out there with focus issues.

 

So there is a good chance that you may have one of those problem 40Ds. Good luck.

[jkantor]

Seems to me that most people are spending a lot of time trying to convince this guy he doesn't have a problem when he obviously does.

 

I don't have 40D, but I had exactly the same problem with a Sigma 15mm on my 20D.

 

Canon has very poor quality control and doesn't adequately test their new equipment when they introduce it.

[cmphoto]

I spent over an hour with Canon tech the other day. I have had some serious

problems with focus on my new 40D. I tried various "L" lens and even my

100mm macro. Very disappointing. We tried everything...sending it back for

a new one. I hope it doesn't take THREE to get a good one. In all fairness,

I have been shooting Canon's for over 33 years with no problem...lenses

or otherwise.