Baffling focus issue with Canon EF-S 17-55mm/2.8

I have experienced the exact same issues you are seeing on 3 copies of the 17-55 within the last month. Here is what I have observed. Note, despite how much I REALLY wanted to be able to use this lens for its great range and speed, I am throwing in the towel and calling it quits with this lens. Also note, I have 6 other Canon lenses that work fine with no problems on both my camera bodies.

- Camera bodies: 7DMK2 (mine), 7DMK2 (a friends), 30D (mine)
- Tests conducted on tripod with & w/o IS on. Tests also conducted in the field using both View Finder and Live View mode for 7DMK2.
- Tests on 7DMK2 using center AF, one shot and lenses MFA'd
- 17-55 lenses: 2 refurb units from Canon, one new unit from B&H

Results from close in subject distances (15' or less) are ALL in focus at all apertures and zoom ranges.

Results at long distance are all in focus at all apertures at 55 mm zoom setting.

Results at long distance are VERY inconsistent & mostly out of focus at all apertures and from 17-35mm zoom ranges when using the viewfinder on 7DMK2. Results under these same conditions when using Live View with 5x or 10x magnification were mostly sharp and in focus. This despite having MFA'd the lenses.

Tripod testing at distance while watching the distance scale on the lens showed consistent result under all conditions when using Live View with 5x or 10x magnification. I saw the same result when testing through the view finder at a zoom setting of 55mm. Using the view finder with remote and half pressing the release at zoom ranges between 17 to 35mm at distance under all aperture settings showed very inconsistent readings on the distance scale. This matches with the OOF image results I observed in the field.

It is clear to me that there is something wrong with the 3 lenses I have tried over the past month. I do not know if Canon has made a design change in the 17-55 lens' electronics that yields this result. In any case, it is a shame as I was looking forward to using this as my all purpose travel lens.

Gymbow wrote in post #17539782
I have experienced the exact same issues you are seeing on 3 copies of the 17-55 within the last month. Here is what I have observed. Note, despite how much I REALLY wanted to be able to use this lens for its great range and speed, I am throwing in the towel and calling it quits with this lens. Also note, I have 6 other Canon lenses that work fine with no problems on both my camera bodies.

- Camera bodies: 7DMK2 (mine), 7DMK2 (a friends), 30D (mine)
- Tests conducted on tripod with & w/o IS on. Tests also conducted in the field using both View Finder and Live View mode for 7DMK2.
- Tests on 7DMK2 using center AF, one shot and lenses MFA'd
- 17-55 lenses: 2 refurb units from Canon, one new unit from B&H

Results from close in subject distances (15' or less) are ALL in focus at all apertures and zoom ranges.

Results at long distance are all in focus at all apertures at 55 mm zoom setting.

Results at long distance are VERY inconsistent & mostly out of focus at all apertures and from 17-35mm zoom ranges when using the viewfinder on 7DMK2. Results under these same conditions when using Live View with 5x or 10x magnification were mostly sharp and in focus. This despite having MFA'd the lenses.

Tripod testing at distance while watching the distance scale on the lens showed consistent result under all conditions when using Live View with 5x or 10x magnification. I saw the same result when testing through the view finder at a zoom setting of 55mm. Using the view finder with remote and half pressing the release at zoom ranges between 17 to 35mm at distance under all aperture settings showed very inconsistent readings on the distance scale. This matches with the OOF image results I observed in the field.

It is clear to me that there is something wrong with the 3 lenses I have tried over the past month. I do not know if Canon has made a design change in the 17-55 lens' electronics that yields this result. In any case, it is a shame as I was looking forward to using this as my all purpose travel lens.

Thanks very much for your post. This adds useful info to the case.

I am going to try to find other copies of the lens locally to test their behavior.

BTW, how old is your lens?

Archibald wrote in post #17538959
The lens focuses fine at closer distances. It only fails at longer distances.

The reason I'm shooting at f/2.8 is to facilitate testing the lens. The focus failure also occurs at intermediate f stops, but it is not as obvious. I posted an example of normal shooting a few months ago (testing was not on my mind then) showing the focus failure, as Left Handed Brisket recalled.

DreDaze, are you suggesting that I should not shoot in AF but routinely shoot in manual focus and set it to the hyperfocal distance? Is that what you do?

Here is the picture previously posted. It was shot at f/5.6. The AF point was in the center of the pic, on the distant trees, but the sharpest area is on the trees in the foreground.

In this case (trees, lake and mountain scene), you used what appears to be an approx. correct hyperfocal focusing distance (about 2/3 of the distance between you and the most distant subject you want in focus is a good estimate).

The difference in sharpness, contrast and fine detail between the foreground and the distant mountain is due to atmospheric haze. It's no fault of the lens.

In this type of shot, a high quality, multi-coated UV filter might help by reducing the effects of atmospheric haze a little bit. See the blue in the shadow side of the mountain? A high quality, multi-coated circular polarizer may have helped too. If I were taking this shot, I'd have stopped down to f8, and probably would have made two or three shots... once without any filter and the others with one or more of the filters.

This image also could be improved to some extent in post processing (especially if it were originally shot in RAW). Some selective sharpening and contrast enhancements might do wonders for it.

Your first shots, of the two buildings across the lake, are more puzzling and concerning. For some reason the point of focus is quite different between the two cropped images. While you might see some loss of sharpness and IQ at such a cropped magnification, it should be the same in both images, assuming you focused on the same point. If that's the case, you are getting erratic focus for some reason.

7DII has something similar to 7D, a custom function that determines whether or not the shutter is released if focus has not yet been achieved. Might be worth trying that (though I thought it mainly applies in AI Servo, not so much in One Shot mode).

This type of shot is ideal for Live View focusing, too. Not sure about the 7DII with the new "Dual Pixel" focus.... But in old style Live View focusing the contrast detection type of focus, using the sensor itself, is often more accurate, although it's much slower than using the phase detection focus via the sensor array seen in the viewfinder.

amfoto1 wrote in post #17539865
In this case (trees, lake and mountain scene), you used what appears to be an approx. correct hyperfocal focusing distance (about 2/3 of the distance between you and the most distant subject you want in focus is a good estimate).

The difference in sharpness, contrast and fine detail between the foreground and the distant mountain is due to atmospheric haze. It's no fault of the lens.

In this type of shot, a high quality, multi-coated UV filter might help by reducing the effects of atmospheric haze a little bit. See the blue in the shadow side of the mountain? A high quality, multi-coated circular polarizer may have helped too. If I were taking this shot, I'd have stopped down to f8, and probably would have made two or three shots... once without any filter and the others with one or more of the filters.

This image also could be improved to some extent in post processing (especially if it were originally shot in RAW). Some selective sharpening and contrast enhancements might do wonders for it.

Your first shots, of the two buildings across the lake, are more puzzling and concerning. For some reason the point of focus is quite different between the two cropped images. While you might see some loss of sharpness and IQ at such a cropped magnification, it should be the same in both images, assuming you focused on the same point. If that's the case, you are getting erratic focus for some reason.

7DII has something similar to 7D, a custom function that determines whether or not the shutter is released if focus has not yet been achieved. Might be worth trying that (though I thought it mainly applies in AI Servo, not so much in One Shot mode).

This type of shot is ideal for Live View focusing, too. Not sure about the 7DII with the new "Dual Pixel" focus.... But in old style Live View focusing the contrast detection type of focus, using the sensor itself, is often more accurate, although it's much slower than using the phase detection focus via the sensor array seen in the viewfinder.

Speaking from my own findings (see my posting above), my tripod test subject was a bit less than 100 yards away. I tested with and w/o a B+W Pro uV filter and with setting of Focus priority for all cases on the 7DMK2. All images I took were preceded by a focus confirmation beep. The results I observed match with the OP shots of the hotel even though my subject distance was much less than his.

Compared to my other 7 lenses, the 17-55 mm images are horribly OOF in most cases in the zoom range of 17 to 35 mm (results at the lower zoom range are worse than at 35mm while results at 55mm are generally fine). There is nothing that can be done on the Raw image to fix this.

As I mentioned in my earlier post, the Images taken with magnified Live View are all sharp and look great (and performing MFA on the lenses did not help with the view finder images).

I contacted Canon tech support management today to relay my findings as I believe there is a systemic issue with recent copies of these lenses. I hope to hear back from their engineering department with their feedback.

Gymbow wrote in post #17539829
BTW, how old is your lens?

My lens was bought in 2007.

To the OP, per your request I just tried this at home with my lens. I took several pics at different apertures and different focal lengths checking the focus each time using live view zoomed in 10x. It hits focus each time without missing. There were no failures. Even hand held it would hit focus each time.

gjl711 wrote in post #17540133
To the OP, per your request I just tried this at home with my lens. I took several pics at different apertures and different focal lengths checking the focus each time using live view zoomed in 10x. It hits focus each time without missing. There were no failures. Even hand held it would hit focus each time.

I also get great results when using live view. It fails randomly, however, when in autofocus and focused on distant objects. It is most noticeable when shooting wide open and at intermediate focal lengths like 28mm or 35mm.

I would really appreciate it if you could verify that your copy of the lens works properly in AF on distant scenes (like landscapes).

Archibald wrote in post #17539001
Yes, you are right about the relative contrast, but I was using the central AF pt only. Also, I've done many tests with very high contrast subjects as targets and see the same problem. Ergo contrast of the subject is not a factor here.

The lens AF focuses OK about a third of the time, and always focuses OK in liveview in manual focus. So we can conclude that the glass is good.

Thanks for the suggestions.

What I'm really wondering is if anyone else has seen the same problem, or if I'm the only one in the world who has seen it.

I suggested shooting a wall as there will be a constant pattern to check for details. I had a Sigma 10-20 a few years back which would produce OOF shots, but not every time. This led me to think I was the problem. I couldn't rely on a sharp shot so I used to shoot my landscapes in portrait mode so the sky disguised the softness. ( it was soft on the left side)

I took my brick wall pics back to the shop where I bought it from. Sigma replaced the lens for me and the new one was fine. Since then I found out Sigma had a known issue with left hand side softness on that lens. Irregular soft images confused me as to where the problem was. Try a wall test to eliminate any similar issues, you never know.

h14nha wrote in post #17540411
I suggested shooting a wall as there will be a constant pattern to check for details. I had a Sigma 10-20 a few years back which would produce OOF shots, but not every time. This led me to think I was the problem. I couldn't rely on a sharp shot so I used to shoot my landscapes in portrait mode so the sky disguised the softness. ( it was soft on the left side)

I took my brick wall pics back to the shop where I bought it from. Sigma replaced the lens for me and the new one was fine. Since then I found out Sigma had a known issue with left hand side softness on that lens. Irregular soft images confused me as to where the problem was. Try a wall test to eliminate any similar issues, you never know.

OK, looking for a suitable wall now...

Archibald wrote in post #17540293
I also get great results when using live view. It fails randomly, however, when in autofocus and focused on distant objects. It is most noticeable when shooting wide open and at intermediate focal lengths like 28mm or 35mm.

I would really appreciate it if you could verify that your copy of the lens works properly in AF on distant scenes (like landscapes).

You misunderstand. I focused using the phase detect focus through the viewfinder just like normal. Then switch to live view to check the focus and see what it focused on. Each time there was no need to re-focus as it hit the target each time very consistently, it did not miss once. This was true whether focusing close or far at a larg aperture or small.

shooting a wall tests for various lens problems that are related to the optics. Bad glass, misaligned elements, or just poor sharpness in all corners.

IMO it won't do anything to help explain or provide a workaround for your focus problem.

edit: the other issue when shooting a big wall is that you would have to have the camera sensor exactly parallel to the wall, both up and down and left to right. It's much harder than you might expect, i know from experience doing copy shots.

amfoto1 wrote in post #17539865
In this case (trees, lake and mountain scene), you used what appears to be an approx. correct hyperfocal focusing distance (about 2/3 of the distance between you and the most distant subject you want in focus is a good estimate).

Uh-oh, the old (and wrong) urban legend at work again! The truth with an APS-C camera...let us assume Infinity = 100,000', and a viewer has 20/20 visual acuity to inspect an 8x12" print:

• at 17mm f/5.6, Hyperfocal = 25', DOF zone = 13' - 1329'
  
• at 30mm f/5.6, Hyperfocal = 79', DOF zone = 44' - 23310'
  
• at 100mm f/5.6, , Hyperfocal = 880', DOF zone = 440' - 100,000'
  

Gymbow wrote in post #17539940
Speaking from my own findings (see my posting above), my tripod test subject was a bit less than 100 yards away. I tested with and w/o a B+W Pro uV filter and with setting of Focus priority for all cases on the 7DMK2. All images I took were preceded by a focus confirmation beep. The results I observed match with the OP shots of the hotel even though my subject distance was much less than his.

Compared to my other 7 lenses, the 17-55 mm images are horribly OOF in most cases in the zoom range of 17 to 35 mm (results at the lower zoom range are worse than at 35mm while results at 55mm are generally fine). There is nothing that can be done on the Raw image to fix this.

Did you try to focus at Hyperfocal Distance, or did you simply focus on the subject of interest? Look at this next chart, it contains information about focusing at the theoretical Hyperfocal Distance with 17mm and 35mm, or focusing instead AT THE SUBJECT at 17mm and 35mm

Gymbow wrote in post #17539940
As I mentioned in my earlier post, the Images taken with magnified Live View are all sharp and look great (and performing MFA on the lenses did not help with the view finder images).

I contacted Canon tech support management today to relay my findings as I believe there is a systemic issue with recent copies of these lenses. I hope to hear back from their engineering department with their feedback.

This is what is odd, you looked at Live View and things appeared to be sharp, then discover the OOF when examining the image in the computer?!

Images as well as the scene are in focus when looking at Live View. The OOF issues only occur when taking images though the view finder under the conditions I described earlier in this thread when I was responding to the OP.