Inaccurate focus point problem 5DIII + 70-200 f2.8L II

[Adrian]

I am struggling a bit with focus accuracy using Canon 5DIII and 70-200 f2.8L Mk II IS (hand held).



This shot (which I am aware has too slow shutter speed but I used it here as it represents the issue well) was taken with the focus point toggled to be over the model's right eye (left as viewed here). This should be more accurate than using centre focus and then re-composing. The 5DIII has numerous focus options.

What has actually happened, if I look at the exif data, is that the real focus point is about 4cm diagonally away from the centre of her eye. I am certain that I focussed on her eye and subsequent test shots have shown a similar effect quite a lot of the time. I use glasses and prefer to have the diopter adjusted so that the viewfinder is optimised with my glasses on.

I am wondering if the inaccuracy in the focus point is caused by my glasses (perhaps creating a parallax error) or whether the focus points on the lens are not 100% accurate? Clearly the other issue is whether I am not holding the lens steady enough. In this case I was relaxed, IS was on, I was supporting camera and lens properly.

In this case the ISO was 320 (the photo was actually under a vine canopy and surprisingly dark: I should have pushed the ISO up a bit more. Neither flash nor tripod was available. Shutter speed was 200 which was on the low side at 200 and would have been better if I had pushed the ISO more. Hence the image is a little soft, but that is not the point here as I do have crisp images from later in the session. The issue is me missing the intended exact focus point which is super critical for portraits with shallow depth of field.

Thoughts?

Adrian

[Polar01]

Adrian: yet suppose that the shutter was fast enough, so to answer you questions 1) glasses No, 2) camera focus point No, 3) holding the camera Yes. I say this because you are trying to hand hold a heavy pig of a lens, focus on a small point from a distance (it you have been nice if you had supplied the distance). Just the pushing of the shutter release would have caused the camera to move guess what direction down and to the right, now 4cm or 1.57 inches is not that far if you shot at some distance a 200mm lens something as small as .5mm at your end could cause the 4mm movement at the subject's end. If as you state "The issue is me missing...shallow depth of field." then you should be locked down on a sturdy tripod, using a cable or remote release, locking the focus point or better still manual focus.

Cheers:

Allan

[Adrian]

Actually when you look a the exif, which shows all the focus points available and the actual one, the focus point is not down (as one might expect from a a shutter depression) but upwards. It has focussed on her eyebrow.

This particular issue has nothing to do with shutter speed: as I mentioned I do have other images that are sharp: this image was chosen because it was easy to explain the focus point I was aiming for versus the one I actually got.

Whilst I obviously agree that using a tripod may eliminate these types of issues, as I said in my post, a tripod was not available. Venues frequently do not permit tripods or there may be other reasons why one cannot be used: this was the case here.

This lens has Canon's latest generation IS and is generally fine for handholding. I can however replicate the shot even at a higher shutter speed (brought about with higher native ISO on Av mode) where the focus point achieved will be 4cm or so diagonally up and to the right from the intended point when I am positioned at a typical portrait distance. In other words, this is not an isolated incident: the picture given is just an example.

This was a portrait setting (under some time pressure) and I was around 5 or 6 metres away from the subject at a guess. The lens was at 200 according to the exif data. Autofocus on this lens is instant and silent. Back button focus was in use as I recall.

I am inclined to think that my eye / glasses orientation against the viewfinder has something to do with this. I suppose the only way I can be sure is a tripod test and also try live view (which I don't like at all for every day hand held shooting).

[rpcrowe]

Adrian,

I know how frustrating it can be when shooting with a lens that should provide exceptional sharpness and ending up with a less than a gnat sharp image.

A couple of thoughts. Athough I normally wear glasses, I prefer to view my shooting without glasses and adjusting the diopter to compensate for my vision. I tried using the Canon add-on which increases the distance between eye and viewfinder and found that my cropping was quite off. The same thing might happen when one is wearing glasses while viewing the image. You would not need to be very far off to result in a focus error. Additionally, it is somewhat difficult for me to see my viewfinder accurately while wearing my glasses.

Focusing a 5-6 meters, you would have a DOF or .1 meter to .15 meters with the maximum of .07 meters in front and to the rear of the point focused on when focused at 6 meters. That certainly doesn't leave any room for error. Even when a photographer is standing still as well as the subject being still, normal body sway might total .05 or .07 meters between subject and photographer. And when the focus point is a bit off the results could be an image that we will not accept.

If the IS on your 70-200 f2.8L Mk II IS is as efficient, or more efficient than the IS on my 70-200mm f/L IS lens, which I am sure it is, 1/200 second should absolutely be no problem. I consistently shoot as low as 1/60 second with confidence that my image will be sharp and can get a reasonable number of sharp images at 1/30 second.

I hold the camera with left hand under the lens and the right hand gripping the camera. I think that most photographers using long lenses shoot (or should shoot) in this manner. The left hand holding the lens "could" act as a fulcrum and thusly, the front of the lens "might" elevate when downward pressure is put on the shutter release. If a photographer grasped the camera with both hands while shooting with a long and heavy lens, there might be a increased chance for inaccurate framing/focus point.

Other than shooting at a slower shutter speed allowing a smaller f/stop (and greater DOF) or increasing the ISO to allow for a smaller f/stop, IMO the only solution might be a tripod or monopod...

[Polar01]

Adrian you have me at an disadvantage as I do not have any program that will show me the focus points in the exif. data, it may well be that, I do however I have never make use of or know how to activate to use. Do you get this with other lens, this movement or is this the only lens where this happens? I wonder as you say you can replicate this at higher shutter speeds. A large number of people tend to push the shutter too hard so the movement would be downwards, this does not seem to be the case. Now the 70-200mm f/2.8 is a large heavy lens and your supporting hand would be somewhere around the middle of the length of the barrel making a good pivot point now a slight downwards pressure on the shutter could cause a rotation to the right and upwards.
Now you say that you use a back button that has the function of activating the auto focus, I am going to assume that it locks the focus distance to the object, then you readjust your grip as you have to use a finger or thumb to activate the back button and press the shutter. As you have locked the focus distance it does not matter where the focus point is now located on the object as that is locked. So if there is any movement of that point it does not matter as it has not refocused.
Now something else if that back button is not set to lock the focus once you start to depress the shutter it will want to refocus before taking the image on a new spot.
If you were 5 to 6 meters @f/2.8 with 200mm then your DOF would have been in the range of 5 inches 2 in front of focus point and 3 behind it.
As you think it maybe your eye/glasses (I do not however I maybe very well wrong) are you glasses single focus or bi-focus.
If you are going to try a test I would suggest that you set the tripod so the camera is the same height as if your were holding it, use say your son as model, now take an image using shutter button to focus, then one with back button, then one with live view using zoom and manual focus to find focus point as other images. Now remove the camera from tripod and step into location of it, take 3 images as you would normally then repeat again but not using the back button to focus.
The first three should be dead on, the next ones should be interesting to see if there is any movement.
Just trying to cover all the possible bases as I see them to root out the problem.

Cheers:

Allan

[Adrian]

Thank you Allan and Richard. Food for thought there.

I use bifocals. I can adjust the eyepiece to compensate for not wearing glasses, but in practice I prefer to wear my glasses if I can. I shall have to experiment as suggested.

The other lenses that I mainly use are a 16-35 f 2.8L wide angle, and a 24-70 f2.8L "standard" lens. I have never had a focus issue using either lens.

I take quite a lot of staged portrait shots for work purposes: invariably indoors. For these I sometimes use a 85mm f1.2L, but mostly now the 70-200 f2.8. However, these are almost invariably tripod mounted and I tend to use live view or tethered for these on either my 5DIII or a 6D body that we use at work (we do a lot of video). No focus issues that are not clear operator error, and super sharp focus on the eyes is easy to achieve. However, these are always done with a central focus point.

I did some more tests tonight, hand held in the garden, using a static subject. In 4 out of 12 shots I ended up with focus shifted up and to the right: but only when I had toggled the focus to the left hand bank of focus points. If I used the centre panel (12 more shots) all were spot on, and the right hand panel I had two slight focus errors, but that could easily have been me moving as I was getting a bit tired of hand holding by the end!

I am a little bit stumped, but will try again tomorrow minus my glasses. (Or with a different operator).

I am shooting standing up, left hand supporting the lens from below, right hand operating the camera conventionally. I have quite steady hands usually and never normally have a problem.

The reason I thought a shutter speed of 200 could have been higher is the 200mm lens (so 1 over focal length pus allowance for a slightly moving subject) plus the model (my wife) blinks a lot when she is being photographed. So we have to use a countdown technique with her eyes closed, open on a count of three and I take the shot. This is a new thing and it had not occurred to me that just me talking to her might be part of my problem. However, it does not explain the static results above. It is very frustrating as I missed several shots that day, all of them by a very small amount. Thinking about it the depth of field plane between her eyebrow and the pupil of her eye can only be slight: the diagonal off point wold not matter at all if the actual focus point was on the same DOF plane as the intended focus point.

Adrian

[Polar01]

Adrian there is one more test, you setup the shoot, camera, lens, distance everything, this is going to be hand held, you shot some images than a different shooter so only difference is the shooter everything else the same. Couldn't hurt to see that would take out one more variable.

Cheers:

Allan

[dan marchant]

Test on a tripod,
Test the lens handheld on a different camera with IS on,
Test at faster shutter speed with IS off,
test using a different shooter using your camera/lens.

[dje]

Adrian I think you need to prove your camera's operation using a static subject and with the camera on a tripod.

I notice that the 5D MkIII has two options for single point (manual selection) focus - a "standard" option and one called Single Point Spot AF which apparently has a finer selection point. Which one are you using ?

Incidentally, when Googling for the User Guide, I came across this Canon document which goes into more detail on the 5D MkIII's AF system.

When you say you look at the Exif info for the focus point and it is not where you expect, are you referring to the use of DPP to display the location of the focus point ? If so, it must be borne in mind that this will only show you the actual location of the focus point used by the camera within the frame of the sensor at the time the focus is set. If there is any movement of the camera between when focus is set and when the shutter is operated, then it will not show you exactly what part of the subject was used for focus. There wouldn't have to be much movement of the camera for this shift to become apparent.

All the camera knows is which focus point was used. It has no way of knowing which part of the subject that focus point was aimed at before the shot was taken (ie when focus was set and locked in).

Dave

[Adrian]

These are good points Dave.

I a viewing the focus point exif in the Aperture info box which shows focus points, metering etc. Clearly you are right, it shows the grid point chosen not the exact point I focused on when the shutter button was depressed fully. I had not considered that. I can't tell from the exif exactly what focus option was set that day and it is two weeks ago so I don't remember either.

I will do some tests. It is clear though that tripod mounted and using live view, there is no focus drift at all.