Focus problem when recomposing a shot

neonpollen

neonpollen

Messages
587
Name
Andy
Edit My Images
Yes
I recently took some shots shot of a racer friend on his bike and a lot are soft. I had the centre AF point selected and focussed on his eyes, recomposed the shot keeping the shutter button half pressed and took the picture. This is a crop (not 100%) of the resulting shot with my standard sharpening applied.

Recompose_Focus.jpg


A you can see his face is soft but his leathers at the elbow are sharp. When I was going through them later I thought that maybe this was a problem with the lens front focussing so I did lots of tests with all my lenses on a tripod covering a wide range of apertures and focal lengths. All the shots were sharp and with the sort of DoF I would have expected. I have a shot of the rider facing directly and again his face is soft but his leather are sharp where I recoposed the centre AF point to.

My camera is a 40D and the lens is a Ef24-70 F2.8l and I haven't changed any of the settings to alter the way the shutter works with AF so it shouldn't re-focus when I recompose the shot. I have tested again with recomposition and two shots out of a dozen or so are the same as above, it only happens when recomposing.

According to dofmaster.com my DoF for this shot should be around 1ft so if the lens was front focussing it must be doing it by quite some way, and why doesn't it do it on shots when I don't recompose.

Anyone experienced this before ? could the UV filter and the bright conditions have caused this ? I am hoping it's user error but I can think what I did any different on these shots than I do normally.

Confused of Lincolnshire ...
 
Only thing I can think is that you've got the camera set to continuous AF rather than single so when you recompose, the camera refocusses on whatever's under the selected AF point. Since the difference in focus distance is pretty small, you may not have noticed the AF working or seen any difference through the viewfinder.
 
When you perform a focus lock from a close distance and then recompose. The point of focus will change slightly because your camera will be at a slightly different distance too what it was when you locked the focus in. Maybe this could be the reason why. This is especially apparent when using large apertures.
 
Now, I may be about to make myself look stupid, but I think it could be that you had it on f/4. I think you would need a small aperture to get up to his face in focus, so maybe about f/8 - f/12.

Now this is how I understand f/ numbers. But it might not be correct, so hopefully someone will come along and either confirm or deny that.
 
Possibly what Jonno said, definitely not what Jo said but most likely what Nod said.

Jonno's theory can happen but at f/4 and looking at your shot I very much doubt that's what caused the problem this time.
 
Nod said:
Only thing I can think is that you've got the camera set to continuous AF rather than single so when you recompose, the camera refocusses on whatever's under the selected AF point. Since the difference in focus distance is pretty small, you may not have noticed the AF working or seen any difference through the viewfinder.
Click to expand...

I thought you had hit the nail on the head there but I've just been back and checked the EXIF info and the camera was set to One Shot AF rather than AI Servo.

Jonno_dudley said:
When you perform a focus lock from a close distance and then recompose. The point of focus will change slightly because your camera will be at a slightly different distance too what it was when you locked the focus in. Maybe this could be the reason why. This is especially apparent when using large apertures.
Click to expand...

That's a good point the distance to the subject will change and with the DoF being so small and the fact the sunject was close it might throw things out. Wouldn't that appear as back focussing though? or have I got that the wrong way around

Jo said:
Now, I may be about to make myself look stupid, but I think it could be that you had it on f/4. I think you would need a small aperture to get up to his face in focus, so maybe about f/8 - f/12.

Now this is how I understand f/ numbers. But it might not be correct, so hopefully someone will come along and either confirm or deny that.
Click to expand...

I do think the shallow DoF from an aperture of F4 is a factor here but other shots I took at F4 without recomposing are fine.
 
Jonno_dudley said:
When you perform a focus lock from a close distance and then recompose. The point of focus will change slightly because your camera will be at a slightly different distance too what it was when you locked the focus in. Maybe this could be the reason why. This is especially apparent when using large apertures.
Click to expand...

My initial thougts, untill I thought it through.
If this were the case then the area in focus would be further back that the riders eyes rather than in front.

My money would be on AI Servo being used - if it was a motorsport event this is the common AF mode to be in, and it is very easy to leave your AF in that mode by mistake when shooting after the races.
 
When you focus, lock and recompose, you can end up with your subject out of focus. I read it on a link from here somewhere. Basically, the plane of focus is in line with the plane of the sensor. So when you lock focus, it's accurate. As you rotate the camera, the plane of focus changes as you shift the plane of the sensor. This will end up, in the right circumstances with the focus point for your subject being behind your subject. It's hard to describe, but it'll show clearly in a little experiment. Tie a piece of string to a ruler, and anchor it to something. The anchor point is the camera, the ruler the plane of focus. Put an object at the centre of the ruler, then keeping the ruler perpendicular to the string, move the ruler left or right. The angle of the ruler will change as it forms the tangent to the circle (I wonder if I'm remembering my maths terms correctly :thinking:). This will put the plane of focus behind the subject. Although as has been pointed out the focus is on the elbow, this doesn't seem likely. I'm also not sure at what sorts of distances and apertures this effect will come into force.
 
Well this is how I understand F numbers. (please excuse the demo :LOL:)

fnumber.jpg


The 'target' symbolises the plain of focus using F Numbers. From the centre focus point the sharpness decreases heading outwards from all points (the red arrows symbolises this). You can see the F numbers marked along the way. If you select f/4 then from that point outwards the image starts to lose sharpness. From each F Number the same occurs.

fnumber2.jpg


Now I have put the same 'target' over your image. The white space around your image is a guess at the size of the entire image before you cropped. That target is over the eye, where you say you spot focused before recomposing.

fnumbe3.jpg


Now you have recomposed and pressed the shutter all the way. Look now where the points are for the different F Numbers and you will see why the face is OOF, where as the helmet and leathers aren't.

It's only a rough drawing and a guess at where the focus will start to decrease with different F Numbers. But I think you get the gist :shrug:

Again I could be completely wrong, but this is how I understand it. I hope someone with more knowledge can either confirm it or explain what it actually is. I'll learn from it too :).
 
I found this link that demonstrates what happenes to the focal plane when you recompose

Looking at this I would expect the focal point to be behind the subject but as he states in his article it's possible I may have moved backwards or forwards as well as refocussing.
 
Jo,

DoF affects the points to the front and back of the focus point, not the area surrounding the point if that is what you are describing here.
 
:bang: You know what? I actually knew that! :LOL: But for some reason I also thought this was the case :thinking: I think I need some sleep!
 
And just to confuse things more, here is the image full size showing the AF point positions. The center point isn't even close to the part of the image that is sharp

AF_Point.jpg


I am putting this all down to user error but I would like to know what I did so I can correct it :)
 
Maybe your subject moved back slightly during recomposition? He'd only have needed to lean to his right by a few inches to cause this problem.
 
Jo said:
:bang: You know what? I actually knew that! :LOL: But for some reason I also thought this was the case :thinking: I think I need some sleep!
Click to expand...

I think your diagrams are confusing that’s all Jo.

If I imagine your target as the sensor sees the image at the focal plane.... those target shapes given 3D depth as if looking into deep cones of hyperfocal distances, each new dof 's depth indicated by the f stop lines...Then imagine another cone going away from the back of the centre focus (the focal plane/the sensors surface) and each depth of field finishing at its corresponding rear fstop base.

I think that’s what your getting at anyways .. cool view point I thought, the stops would be much further apart I'm thinking.. . (y)

Edit.. if that made sense, hmmm.
 
neonpollen said:
I am putting this all down to user error but I would like to know what I did so I can correct it :)
Click to expand...

I do think the camera tilt has merit myself, only a foot to play with, easy enough for a few errors and the tilt to make 2/10 misses ...possibly... it is something I'm constantly on guard doing if composing at short range.
:)
 
There is a custom function on the canon 40D that lets the joystick control the which focus point is used (click the stick for the center point!)

This makes it easy to pick the best point to focus with once you are composed, and you don't need to focus/recompose, just pick the focus point on the bit you want in focus.

It maybe that the non-central points are a bit less good at picking up a moving target or if the light is bad, but under normal conditions they are great (and on something as upscale as a 40D they may all be the 'good' type I just cannot remember!)
 
bradleysmith said:
This ^^^

Take a look at this Jo

http://en.wikipedia.org/wiki/Depth_of_field
Click to expand...

I may be wrong but within the wiki article is this section which may indeed explain what is happening?

Camera movements and DOF

When the lens axis is perpendicular to the image plane, as is normally the case, the plane of focus (POF) is parallel to the image plane, and the DOF extends between parallel planes on either side of the POF. When the lens axis is not perpendicular to the image plane, the POF is no longer parallel to the image plane; the ability to rotate the POF is known as the Scheimpflug principle. Rotation of the POF is accomplished with camera movements (tilt, a rotation of the lens about a horizontal axis, or swing, a rotation about a vertical axis). Tilt and swing are available on most view cameras, and are also available with specific lenses on some small- and medium-format cameras.

When the POF is rotated, the near and far limits of DOF are no longer parallel; the DOF becomes wedge-shaped, with the apex of the wedge nearest the camera (Merklinger 1993, 31–32). With tilt, the height of the DOF increases with distance from the camera; with swing, the width of the DOF increases with distance.

In some cases, rotating the POF can better fit the DOF to the scene, and achieve the required sharpness at a smaller f-number. Alternatively, rotating the POF, in combination with a small f-number, can minimize the part of an image that is within the DOF.

:thinking:
 
neonpollen said:
And just to confuse things more, here is the image full size showing the AF point positions. The center point isn't even close to the part of the image that is sharp

AF_Point.jpg


I am putting this all down to user error but I would like to know what I did so I can correct it :)
Click to expand...
Now that we can see the whole composition the problem is all the more puzzling. This is a very classic focus/recompose scenario and usually it leads to back focus, not front focus. I cannot explain why you have the problem you have. I actually like this explanation for why you might get backfocusing in a situation like you had here....

http://visual-vacations.com/Photography/focus-recompose_sucks.htm

....but as you've got front focus it's a bit of a head scratcher.

As for the centre point not being anywhere near anything in focus, well you focused and recomposed, so there is no reason why it should be near anything particular - you focused and then you moved the point.

The best solution for the future, when photographing fairly close up with a shallow DOF is to minimise the degree of recomposing you do. You would have been better off choosing the top right diagonal focus point and placing this over his left eye. You would barely have had to recompose at all.

That's the correct solution to the usual problem, but your situation is not the usual case. I hardly dare raise the subject of front focusing from your lens but the only other thing that makes sense is that you inadvertently released focus lock for a moment, by accident, and then the camera re-acquired focus exactly where it was pointing just before you finished the shot - at his leathers.

BTW, it is usually not flattering to have people squinting straight into bright sunshine. Put the guy in shades (I mean sunglasses, but actual (open) shade would be good as well) or put the sun behind him, ideally with him in front of a nice big white wall that will bounce light back onto his face without completely blinding him.
 
Jo said:
Well this is how I understand F numbers. (please excuse the demo :LOL:)

fnumber.jpg


The 'target' symbolises the plain of focus using F Numbers. From the centre focus point the sharpness decreases heading outwards from all points (the red arrows symbolises this). You can see the F numbers marked along the way. If you select f/4 then from that point outwards the image starts to lose sharpness. From each F Number the same occurs.

fnumber2.jpg


Now I have put the same 'target' over your image. The white space around your image is a guess at the size of the entire image before you cropped. That target is over the eye, where you say you spot focused before recomposing.

fnumbe3.jpg


Now you have recomposed and pressed the shutter all the way. Look now where the points are for the different F Numbers and you will see why the face is OOF, where as the helmet and leathers aren't.

It's only a rough drawing and a guess at where the focus will start to decrease with different F Numbers. But I think you get the gist :shrug:

Again I could be completely wrong, but this is how I understand it. I hope someone with more knowledge can either confirm it or explain what it actually is. I'll learn from it too :).
Click to expand...

thats just plain wrong... its to do with distance FROM the camera, not distance in a flat vertical plane in parralel to the subject
 
this is a simple geometry question.. how far is everyhing away from a pont... the point being (approx) measured from the centre of the front of your lens

When you do the maths, you will realise that for close subjects, yo need to be more accurate and "off shell" errors are less forgiving.. the shell beingthe imaginary sphere with a radius of the focal point from the front of the lens

Throw into the mix off axis performance of the lens, and you will answer your question.

Throw in DOF - due to chosen f-stop and you may find your soloution
 
tdodd said:
Now that we can see the whole composition the problem is all the more puzzling. This is a very classic focus/recompose scenario and usually it leads to back focus, not front focus. I cannot explain why you have the problem you have. I actually like this explanation for why you might get backfocusing in a situation like you had here....

http://visual-vacations.com/Photography/focus-recompose_sucks.htm

....but as you've got front focus it's a bit of a head scratcher.

As for the centre point not being anywhere near anything in focus, well you focused and recomposed, so there is no reason why it should be near anything particular - you focused and then you moved the point.

The best solution for the future, when photographing fairly close up with a shallow DOF is to minimise the degree of recomposing you do. You would have been better off choosing the top right diagonal focus point and placing this over his left eye. You would barely have had to recompose at all.

That's the correct solution to the usual problem, but your situation is not the usual case. I hardly dare raise the subject of front focusing from your lens but the only other thing that makes sense is that you inadvertently released focus lock for a moment, by accident, and then the camera re-acquired focus exactly where it was pointing just before you finished the shot - at his leathers.

BTW, it is usually not flattering to have people squinting straight into bright sunshine. Put the guy in shades (I mean sunglasses, but actual (open) shade would be good as well) or put the sun behind him, ideally with him in front of a nice big white wall that will bounce light back onto his face without completely blinding him.
Click to expand...

My initial thoughts were that the lens was front focussing but I tested the focussing afterwards and it is fine. I tend to belive it's a mixture of factors, proximity of the subject, very shallow DoF and user error in me maybe not holding the shutter down properly. If nothing else it has alerted me to the issues in recomposing like this.

From now on I'll change the AF point to a more appropriate one, these were just quick grabs before he went back out on track again hence him squinting :) I was hoping to get his dad to move their white van and use that as a big reflector but we didn't have time.
 
It is possible to get front focus when you recompose. I've drawn up a little diagram to explain. I've stuck to two dimensions for simplicity!

3408897778_d87c215438_o.jpg


In this scenario, you focus the shot on the face (the green), and the you recompose to bring the centre focal point down (the red). When you do this, you alter the focal plane quite considerably, and the effect is very noticeable at larger apertures.

This will mean that the eyes that where crystal sharp when you focussed are now back focussed when you recompose, where as his arm (or gut in my case!) will now be nicely in focus even though they are in front of the original eyes.

So if you want to shoot with the lens wide open, it's always better to use one of the other AF points if at all possible to minimize the shift in the focal plane.
 
Furtim, assuming you have the camera at the level of the rider's face when you focus you have the shortest distance possible between your position and the subject. If you then tilt the camera downwards you will indeed send the plane of focus behind the eyes at the top (back focus for the eyes), and favour the closer areas further down.

However, if you look at the perspective here I think you will agree that it looks like shot was taken from a fairly low position. That means that the camera was not level with the eyes when focusing but was perhaps nearer to waist level or something. That meant that the distance between camera and eyes was longer than it would be with the first scenario I described. This is the same condition as that described in the link I posted earlier. With the camera tilted up for focusing, and then brought down to a horizontal(ish) angle you will actually be reducing the distance to the subject, but the focus will be set for a greater distance. Thus the camera will end up with back focus with respect to the intended point of sharp focus.

However, with all of that said, looking at the whole image again, I would say that the degree of recomposition is not excessive. I'd be surprised if the plane of focus moved more than an inch or two maximum. Andy said that the calculated DOF was about 1' and I really can't see the recomposing throwing things that far out of focus. At these reproduction sizes the DOF should actually appear even greater than 1'.

FWIW I took delivery of a 24-70/2.8L yesterday and set about calibrating the AF perfectly on my 1D3 and 50D. I was rather horrified to find that the degree of front/back focus varied significantlly depending upon the focal length. I am going to have to double check my results today but I am seriously alarmed. I ordered the lens in January and it only just came into stock. I need it for a wedding later this month. If it doesn't get sorted by the 18th I shall be very disappointed. Anyway, I wonder if Andy's lens is playing up a bit with respect to focus accuracy at different focal lengths.
 
If you are considering just a single central point at the origin of the focal plane, you are correct, but the camera sensor is not a 'point' as such.

The point (no pun intended), is you have to consider the whole focal plane - when you tilt a plane around that central point, as you move away from the centre, some points will be behind the original plane and some in front.

I'm not saying this is the reason for the OP's problem in this case, just saying you can get 'front focussing' effects from re-composing a shot.

Of course this is not an issue if you can keep the focal plan completely parallel to the subject at all times.
 
Starting to doubt my own sanity:wacky: on this one, but found someone else's explanation that puts it far better than mine. Better diagrams too!

dof2.png


Full article here...
 
You must log in or register to reply here.
Back
Top