10D - AF Help needed!

[ingenus]

I 've had my 10D for about a week now, and the biggest issue I have is with focussing.
My feeling is after doing some tests that the camera/lens is front focussing. However, I'm not sure that I fully understand the AF FP system.
Some basic questions:
1) In a shot with objects at different distances, and with several FP points going red, which one(s) does the camera focus on? Can this not lead to the main subject being out of focus?
2) I seem to get better results when using the centre only FP, and at the moment am sticking to that. Is this the correct thing to do? Still, I feel as if I am missing out on one of the cameras main features.
Can someone give me an explanation of how the 7 point system works in practice because I feel I'm doing something wrong.
Thanks
Iain

[Roger_Cavanagh]

Iain,

From your description, I don't think you are doing anything wrong. I, and many other 10D'ers, use the centre focus point only. The algorithms that control which of the seven points is used seems pretty simplistic. Generally the closest object in the scene is chosen, which is not always what is required.

The 10D definitely does not have the greatest AF system in the world, but is not as bad as sometimes made out. It does, however, suffer from one significant design flaw: the size of the centre AF sensor is actually larger than the square etched on the focusing screen, so it is possible for the camera to select a high contrast edge that is outside the focus box. Consequently, while the camera is working perfectly, the user gets unexpected results and thinks there is a problem.

Check this article (http://www.rogercavanagh.com/helpinfo/23_focuscheck.htm) on my site. It describes a simple test that Canon Germany recommend to check that the AF is working OK. It doesn't tell you whether there is front focus or back focus, but that doesn't matter - if it fails, the camera still needs to go to Canon Service for recalibration.

Regards,

[ingenus]

Roger, thanks for the reply. Please see below results of a test I did a couple of days ago.

The results below show without a doubt (unless you tell me otherwise!) that the camera is definitely front-focussing by quite a big margin. The test was done with the card at 45 degrees to the camera - the right hand side to the front - as recommended, and the 2 lenses - 18-50 and 55-200 Sigma - were used wide open at f/5.6. I also stuck a couple of batteries in the shot, and the focus on them really proves the point I think.
The shots below are crops @ 35mm, 50mm, 55mm and 70mm. Obviously, the crop factor is higher as the focal length decreases. No flash was used, but tripod was.

This one was 35mm:
http://freespace.virgin.net/iain.mackenzie/35Auto.jpg

50mm:
http://freespace.virgin.net/iain.mackenzie/50Auto.jpg

55mm:
http://freespace.virgin.net/iain.mackenzie/55Auto.jpg

70mm:
http://freespace.virgin.net/iain.mackenzie/70Auto.jpg

One weird (or not as the case may be) thing was that I was unable to get the camera to auto-focus at any length greater than 75mm on the 55-200 lens. Manual focus produced good results, but auto just wouldn't focus, and I couldn't take a shot because of the flashing focus dot!

What do you think?

Iain

[Roger_Cavanagh]

Iain,

The problem with using angled targets is that there is no way to be certain exactly what the camera decided to focus on. This is partly related to the "over-sized AF sensor" issue I mentioned in my first post.

There are more sophisticated designs for testing BF or FF, but I would say: don't waste your time. It makes no difference which way the focus is off - front or back, it still has to go back for recalibration and it is of no benefit to the service technicians to know what the camera is doing. They just use a factory lens to adjust the camera back to factory specs.

The Canon Germany test I mentioned is much simpler: you take pictures of a nice, high contrast subject on a flat surface at 90 degrees to the camera. If you can get sharper images using manual focus than AF, something is wrong. That's all you need to know.

If you have several lenses, you should test each lens as well, in case lens calibration is off. If you get the problem on all lenses, chances are it's only the camera that needs adjustment. Otherwise it's a good idea to send both cameras and lenses for recalibration.

Not sure about the focus lock problems - I assume you weren't inside the minimum focusing distance. Does that lens change max aperture with focal length? Maybe not enough light at the longer length with a smaller aperture?

Regards,

[CyberDyneSystems]

Iain,

As Roger mentioned in his first reply,. your first order of business needs to be to set the 10D to use only the center focus point...

The 7 focus points are fine for certain tasks,. but they take some getting used to and in the meantime we all end up with out of focus subjects in our photos.

See if this helps,. and let us know.

[ingenus]

Roger,
The test you suggested works out OK with the AF in focus on all extremes of both the 18-50 and 55-200 wide open.
However, although clearly a good test of focussing on the main subject, I'm still not convinced that is a good test of front/back focussing. If you look at the shots I posted a link to earlier, the main subject is - the vertical line - in focus, but it is what happens to the DOF in front and behind that matters.
The sheet was placed at 45 degrees to the wall, and the batteries in the shot were equidistant to the focus point. From each shot you can see that the front battery is in focus whereas the back one is not.
I would have thought that these shots confirm that the camera is front-focussing by quite a long margin.
What do you think?
Iain

[Roger_Cavanagh]

Iain,

If your 10D passes the Canon Germany test then it is working OK. Front or back focus issues would have resulted in you getting better results with manual focus than AF in this test. It does NOT matter that you are shooting a vertical subject, FF or BF would still show up in blurred images.

Unfortunately, your pictures of the angled target only prove that the 10D does sometimes choose the wrong target on which to focus. It is one of the things you have to learn to manage with the 10D. You should use centre point AF. It's also worth changing custom function 4, so that focus is controlled by the * button instead of the shutter button.

Regards,

[ingenus]

Roger,
I used the centre FP in all shots posted earlier. I wasn't trying to make any point about using vertical subjects.
I think I am merely trying to say that surely the DOF should be the same front and back of the main subject in any shot. Apart from any other issues, that reduces the margin of error in focussing on the main subject, surely.
Iain

[Roger_Cavanagh]

Iain,

I'm not sure what point you're trying to make here - you'll have to elaborate.

DOF is distributed around the main plane of focus, though I think the ratio of front to back will change according to subject distance, but without digging out any reference books I can't remember exactly how it works.

Another point about the 10D (and all non-pro SLRs from Canon), the AF is designed to be accurate to within 100% of DOF at max aperture, so the sharpest focus can be in front or behind the desired subject and still be considered to (by Canon) to be working OK. Pro cameras like the 1D, 1Ds, 1D MkII and the film equivalents have AF that uses 1/3 DOF at max aperture as the standard, which is why there are not many complaints about focus from the owners of those cameras.

There are issues with the standards that Canon use because they are based on assumptions about size of enlargement and viewing distance and 35mm film. The "standard" enalargement is 10 x 8 inches with 35mm. When you examine a 10D image in Photoshop at 100% zoom on a 21" monitor, this is the equivalent of something like a 30 x 40 print viewed from only a few inches. DOF is not the same!

Returning to whether your camera works OK: if you cannot tell the difference between results using MF or AF, then your camera is working OK.

Regards,

[CyberDyneSystems]

I do NOT think that DOF is distributed evenly from the point of focus,. front to back... In fact I'm sure it is uneven... as the subject is moved further from a lens the depth of feild changes.. so the dof behind the focus point,. being further away from the distance in front of the focus point.. will therefore be "different"

I say different, because for the life of me I can remember whether you are supposed to have more DOF in front,. or less? I think I have it backwards in my head,... but I thought you would have more behind...

Anyway,. they are surely "different"

[KennyG]

Roger,
The test you suggested works out OK with the AF in focus on all extremes of both the 18-50 and 55-200 wide open.
However, although clearly a good test of focussing on the main subject, I'm still not convinced that is a good test of front/back focussing.

You can not use an angled target as a means of checking front to back focus. You can set up a simple test that is reasonably accurate, using items you should have to hand.

Print a high contrast target on a sheet of matte (it MUST be non-reflective) A4 paper. A checkerboard is fine. It should be printed at least two inches from the sides of the paper. Stick it to the side of a cereal box.

Get or make a ruler/scale around 12" (30cm). Set up the box on a table with window light falling on the 'target'. Prop up the ruler at an angle at the right side of the box so that you can read the markings (it should look like a ramp climbing away from you). Camera on tripod level with the target, good lighting, lens wide open, frame the target (making sure you can see all the ruler), confirm AF and take the shot.

You can then tell from the ruler (which was not in the AF area) if you have front back focus. This is a much simplified version of the test set-up that Canon use. The marking on the scale that is level with the front face of the serial box should be in focus.

Any of the tests using angled rulers. scales, batteries, etc. are just not valid.

[ingenus]

Roger and others,

I have looked back at an old Pentax manual lens with a depth of field indicator, and of course you are right the distance back and front is very different with much more back than front. For example at f/16 a subject at 3 feet, has a DOF of 2-8 feet, i.e. 1 foot in front and 5 behind.
This adds to my confusion with the 10D as my DOF is much greater in front than behind.

Kenny, it was not a simple ruler test, it was as described in http://www.photo.net/learn/focustest/
The battery thing was simply an aid to confirm the situation!

Iain

[KennyG]

Kenny, it was not a simple ruler test, it was as described in http://www.photo.net/learn/focustest/
The battery thing was simply an aid to confirm the situation!

Iain

It is still an invalid test no matter where it was published. I have seen the Canon test rig for real and the most important thing about the test is the target has to be in the same plane as the camera, not at an angle. The Canon engineer I know personally (he browses here from time-to-time) just shakes his head in disbelief about some of these so-called focus tests.

[ingenus]

Maybe I'm missing something here, but the angled 'ruler' seems like a perfectly good way of measuring focus and particularly DOF. By focussing on the central vertical line one can see at a glance where the depth of field is by looking at the lines on the ruler that are in focus. Surely it is merely a way of simulating similar subjects at varying distances.
The same effect can be got by photographing blades of grass.
Iain