5D Mark III focus points off center

[Dick]

I noticed that the focus points seen in the view finder are slightly off center. Photos on the display show the used focus points slightly in different positions in comparison to the intended spots (~½ focus point width to the left). Is it the same for any of you? I checked two 5D3 bodies at a local store and noticed that the focus points were all over the place on them too.

I brought my 5D3 body to warranty repair and they told me there is nothing wrong. When I showed them what I meant, they noticed it too. Now I have yet another repair scheduled, but I would like to know if anyone else has had the problem and if it was possible to fix it. I would not really like to get unnecessary repairing done on the body. The body doesn't get any better if the problem still exists after additional repairs.

My thoughts:
1) Focusing screen not correctly in place?
2) Focus point display off center?
3) The issue has been there from the beginning. Now it bugs me as the smaller focus point doesn't hit targets.

What do you think?

 

[Skulker]

Interesting observation, how have you determined that the smaller focus points don't hit the target? Have you done some tests? I'd like to check my 5d3.

I know the focus points shown are never supposed to be actual focus area, but just an indication. I would expect them to be pretty close. Are you finding you can't focus on what you want to get spot on?

 

[Dick]

Interesting observation, how have you determined that the smaller focus points don't hit the target? Have you done some tests? I'd like to check my 5d3.

Haven't really done any real tests. The ~½ focus point width difference, however, is enough to place the small focus point off target.

I know the focus points shown are never supposed to be actual focus area, but just an indication. I would expect them to be pretty close. Are you finding you can't focus on what you want to get spot on?

I haven't really noticed any problems in photos that I could say are because of this issue. If I switch focus points on in PP or on the 5D3 LCD, they light up in somewhat "wrong" places. Taking pictures using live view on the other hand places the focus points exactly as intended.

 

[SiliconVoid]

You do realize that the AF point itself is not (1) pixel.. It is an array of many pixels in that given area, however during autofocus the camera can determine best focus based on the contrast in an area of just a few pixels depending on the actual subject/scene.

Basically what I mean is lets say under each focus point there is a grid of 100 pixels - and when focus is achieved under the selected focus point it happens to be on the right side of the 10x10 grid of pixels - then attempting to place a marker that represents the square/dot seen in the viewfinder precisely over the actual focus point in an image it is going to be off to the right of the super-imposed point in the viewfinder.

 

[Dick]

You do realize that the AF point itself is not (1) pixel.. It is an array of many pixels in that given area, however during autofocus the camera can determine best focus based on the contrast in an area of just a few pixels depending on the actual subject/scene.

Basically what I mean is lets say under each focus point there is a grid of 100 pixels - and when focus is achieved under the selected focus point it happens to be on the right side of the 10x10 grid of pixels - then attempting to place a marker that represents the square/dot seen in the viewfinder precisely over the actual focus point in an image it is going to be off to the right of the super-imposed point in the viewfinder.

Well, let me give you another example. If I place the small focus point on someone's eye ball using the LCD & then check the viewfinder and see the point on his/her nose, which one am I focusing on?

I'd say that the LCD is more likely to be accurate. Therefore if I use the viewfinder to focus, I will not place the focus point on the target I was aiming for.

 

[iSy]

I have 2 5D3's and I have just tested this on both of mine.
I focused on a metal pole, right in the centre on both cameras and when reviewing the images one camera is bang on in the centre and the other shows the red square slightly to the left.
Both images appear to be focused though.
I took a couple of test shots with my 70-200 2.8 on both bodies locking on to small text and they both seem ok although one does show to the left when reviewing the image with highlighted AF point on the display.
I did notice that when I put the 50 1.4 on and tried some focusing i had a few misses on the one body although some were ok but focusing hand held that close at 1.4 is tricky at the best of times.
I would be curious to see how you get on as I appear to have this "strangeness" on one of my bodies.
Cheers
Simon

 

[iSy]

Update - tired again with the 50 1.4 and focus on some keys on my MacBook Pro. Focusing at the closest distance on the ?/ key i focused is from the right at a low angle around 30degrees and nailed the focus on the ? and when reviewing its deffo in focus on the ? although the red af point shows it moved slightly down to the / symbol.
So from what I can see it seems the view finder AF point is accurate and when reviewing the images with AF point displayed its slightly off on one of the bodies which I guess isn't a real issue if the images are focusing ok.
Thoughts?

 

[Dick]

Update - tired again with the 50 1.4 and focus on some keys on my MacBook Pro. Focusing at the closest distance on the ?/ key i focused is from the right at a low angle around 30degrees and nailed the focus on the ? and when reviewing its deffo in focus on the ? although the red af point shows it moved slightly down to the / symbol.
So from what I can see it seems the view finder AF point is accurate and when reviewing the images with AF point displayed its slightly off on one of the bodies which I guess isn't a real issue if the images are focusing ok.
Thoughts?

Interesting! In the end all that really matters is that the photos have no flaws because of the issue. It is annoying though to constantly know that something is off.

 

[neuroanatomist]

Update - tired again with the 50 1.4 and focus on some keys on my MacBook Pro. Focusing at the closest distance on the ?/ key i focused is from the right at a low angle around 30degrees and nailed the focus on the ? and when reviewing its deffo in focus on the ? although the red af point shows it moved slightly down to the / symbol.
So from what I can see it seems the view finder AF point is accurate and when reviewing the images with AF point displayed its slightly off on one of the bodies which I guess isn't a real issue if the images are focusing ok.
Thoughts?

Not so sure this is a valid 'test'... The actual AF point (on the AF sensor) is larger than the little box that represents it in the VF. The AF system doesn't know a ? from a /, nor does it know which you want to focus on - if both are under the AF point (even if a little outside that box), it'll lock onto the highest contrast feature at the right orientation. I even have a suspicion (anecdotal) that if there are two similar high-contrast features at opposite sides of the (real) AF point and at different distances, the system will actually switch from one to the other on successive shots.

It does seem reasonable with assembly tolerances that the VF display array may be slightly misaligned with the AF sensor. FWIW, my 1D X is perfectly aligned.

 

[iSy]

Update - tired again with the 50 1.4 and focus on some keys on my MacBook Pro. Focusing at the closest distance on the ?/ key i focused is from the right at a low angle around 30degrees and nailed the focus on the ? and when reviewing its deffo in focus on the ? although the red af point shows it moved slightly down to the / symbol.
So from what I can see it seems the view finder AF point is accurate and when reviewing the images with AF point displayed its slightly off on one of the bodies which I guess isn't a real issue if the images are focusing ok.
Thoughts?

Not so sure this is a valid 'test'... The actual AF point (on the AF sensor) is larger than the little box that represents it in the VF. The AF system doesn't know a ? from a /, nor does it know which you want to focus on - if both are under the AF point (even if a little outside that box), it'll lock onto the highest contrast feature at the right orientation. I even have a suspicion (anecdotal) that if there are two similar high-contrast features at opposite sides of the (real) AF point and at different distances, the system will actually switch from one to the other on successive shots.

It does seem reasonable with assembly tolerances that the VF display array may be slightly misaligned with the AF sensor. FWIW, my 1D X is perfectly aligned.

I know what you mean, and I agree its not a very accurate way of testing it I was just saying it doesn't appear to affect the overall image which in the end is all that matters. I will continue to play with it and see if there is an issue but for now it appears to function fine even though I agree, when something isn't quite perfect even when it works it can bug you crazy lol

 

[East Wind Photography]

Ok you are worrying about a problem that doesn't exist. First off phase detect uses a separate sensor near the focus screen. Contrast detect uses the actual sensor to perform the AF. In live view the mirror is locked up and therefor PD sensors are out of the picture.

They don't have to be in alignment because both are not used at the same time.

 

[neuroanatomist]

Ok you are worrying about a problem that doesn't exist. First off phase detect uses a separate sensor near the focus screen. Contrast detect uses the actual sensor to perform the AF. In live view the mirror is locked up and therefor PD sensors are out of the picture.

They don't have to be in alignment because both are not used at the same time.

I don't think the worry was about phase vs. contrast detect. Rather, as you look through the VF at the focus screen (transmissive LCD) representation of an AF point, if that point (the whole array, really) is sufficiently misaligned with the AF sensor, that is a problem. For example, when framing a tight portrait an AF point (the real one on the AF sensor) pretty much covers an eye - if I put the VF representation of that point on the eye, but misalignment results in the selected AF point in the AF sensor being over the ear, the thin DoF of a fast prime shot at a wide aperture means the subject's eyes are OOF, and that's a missed/throwaway shot.

 

[East Wind Photography]

Yes but it sounds like he is comparing the viewfinder square with what is displayed on the LCD screen on the back. They are two different things not linked together...at least not on the 5D3.

The red AF point markers on the rear LCD are not tied to the point in the viewfinder...though they attempt to show you which AF point was activated during the shot, they are not calibrated to the PD sensor.

Ok you are worrying about a problem that doesn't exist. First off phase detect uses a separate sensor near the focus screen. Contrast detect uses the actual sensor to perform the AF. In live view the mirror is locked up and therefor PD sensors are out of the picture.

They don't have to be in alignment because both are not used at the same time.

I don't think the worry was about phase vs. contrast detect. Rather, as you look through the VF at the focus screen (transmissive LCD) representation of an AF point, if that point (the whole array, really) is sufficiently misaligned with the AF sensor, that is a problem. For example, when framing a tight portrait an AF point (the real one on the AF sensor) pretty much covers an eye - if I put the VF representation of that point on the eye, but misalignment results in the selected AF point in the AF sensor being over the ear, the thin DoF of a fast prime shot at a wide aperture means the subject's eyes are OOF, and that's a missed/throwaway shot.

 

[Dick]

Yes but it sounds like he is comparing the viewfinder square with what is displayed on the LCD screen on the back. They are two different things not linked together...at least not on the 5D3.

The red AF point markers on the rear LCD are not tied to the point in the viewfinder...though they attempt to show you which AF point was activated during the shot, they are not calibrated to the PD sensor.

So if I mount the body on a tripod, live view and viewfinder focus on different spots even when both for example use the middle AF point? Is that what you are saying, or is one of them just indicating a false AF point position? The middle points are not on the same targets.

The sensor remains in the same spot, framing is therefore the same. I doubt that these two modes have separate AF systems with different AF point locations.

 

[sturdiva]

Yes but it sounds like he is comparing the viewfinder square with what is displayed on the LCD screen on the back. They are two different things not linked together...at least not on the 5D3.

The red AF point markers on the rear LCD are not tied to the point in the viewfinder...though they attempt to show you which AF point was activated during the shot, they are not calibrated to the PD sensor.

So if I mount the body on a tripod, live view and viewfinder focus on different spots even when both for example use the middle AF point? Is that what you are saying, or is one of them just indicating a false AF point position? The middle points are not on the same targets.

The sensor remains in the same spot, framing is therefore the same. I doubt that these two modes have separate AF systems with different AF point locations.

Someone correct me if I'm wrong, but I believe that liveview focusing doesn't use auto focus points at all. The autofocus points are used only for phase detection autofocus, while liveview focusing does contrast detection by actually reading the data off of the sensor itself, rather than making use of the autofocus array.

 

[Mt Spokane Photography]

You do realize that the AF point itself is not (1) pixel.. It is an array of many pixels in that given area, however during autofocus the camera can determine best focus based on the contrast in an area of just a few pixels depending on the actual subject/scene.

Basically what I mean is lets say under each focus point there is a grid of 100 pixels - and when focus is achieved under the selected focus point it happens to be on the right side of the 10x10 grid of pixels - then attempting to place a marker that represents the square/dot seen in the viewfinder precisely over the actual focus point in an image it is going to be off to the right of the super-imposed point in the viewfinder.

Well, let me give you another example. If I place the small focus point on someone's eye ball using the LCD & then check the viewfinder and see the point on his/her nose, which one am I focusing on?

I'd say that the LCD is more likely to be accurate. Therefore if I use the viewfinder to focus, I will not place the focus point on the target I was aiming for.

Liveview Quick Autofocus is likely to be less accurate. It takes the image off the sensor, but the AF points displayed are electronically generated and have nothing to do physically with the AF sensor.
I'd expect the viewfinder to match the AF sensor, if it doesn't, Canon can adjust it, its likely not seated correctly.

In any event, its not likely to show up in real life because the actual AF areas are larger than what you see.

 

[neuroanatomist]

Yes but it sounds like he is comparing the viewfinder square with what is displayed on the LCD screen on the back. They are two different things not linked together...at least not on the 5D3.

The red AF point markers on the rear LCD are not tied to the point in the viewfinder...though they attempt to show you which AF point was activated during the shot, they are not calibrated to the PD sensor.

Sounds like he's comparing the selected AF point with the red marker displayed during image playback (with that function enabled). While they're not 'calibrated' if the center AF point was used, that red marker will be dead center in the image. If the shot was taken with a feature under the center AF point, and that feature isn't under the red box on review, there's a misalignment.

Note that it should be tested properly - on a tripod, etc., since if the camera is moved between half and full shutter button press (unintentional focus/recompose) the camera doesn't know that (unless it's a Hasselblad) and that will appear as misalignment.

I doubt that these two modes have separate AF systems with different AF point locations.

Your doubt is misplaced. Phase detect AF (viewfinder) uses a dedicated AF sensor, whereas contrast detect AF (live view) uses the image sensor itself to determine focus (which takes much longer).

 

[Dick]

Thank you guys for explaining! I guess I can just cancel the scheduled repair then.

 

[Skulker]

Thank you guys for explaining! I guess I can just cancel the scheduled repair then.

I've been away since my initial reply, work getting in the way again.

As others have said the focus in live view is completely independent of "normal" focus, they are different systems.

One point that has not been mentioned is that when focusing using live view you can move the focus area! Also the size of the focus area changes! You can move the focus point with the joy stick and zoom in with the zoom button. Sometimes in live view if you can't focus at one zoom setting then you can with another.

Live view focus is normally more accurate and much slower. (Although the 5D3 has really good focus so there is not so much difference.)

Hope this helps.

 

[Martin]

I had exactly the same problem and it was real. After three service visits they exchanged the mirror box. Atfer that repair the AF points' misplacement was corrected however since then I struggle with AF system. Since almost a year from purchase the camera is being serviced all the time without any good solution. My guaranty is almost over and the camera is like piece of crap. I can only suggest you to push the store to exchange it as probably any service will not fix that. I checked 3 canon services and every time if one thing was corrected the other one was not working properly. Right now center and left points are backfocusing and right points are focusing in opposite direction. One year of frustration without taking photos. Unfortunately IMHO if you think your AF points are misplaced-you are probably right. In my camera when I focused on the edge of any target it focused far away. And it was obvious that it happened only in one edge of focus point. It is normal that the real focus point has more area than its layout in viewfinder however not only in one direction. As I said-they fix this but camera probably lost all factory tolerances and now AF is just useless.