Improper autofocusing on 300D...

[jairy hunter]

I've been using the 135mm f2.0L lens with my digital rebel for a couple weeks.

Unfortunately although I love the low aperture, it seems to have trouble focusing on

the subject. This is compounded by the fact that I have to be so far away for portraits

due to telephoto quality--I can't tell whether the subject is in focus or not.

 

When the focus is good, the shots are brilliant, but I have to shoot a lot of exposures

to get a good one.

 

Is this something I have to live with?

 

My 70-200mm f4.0L focuses pretty well, even in lower light and when shooting

action shots, but again the same thing, occasionally the autofocus is just off the

subject. I expect this in action shots with a lot of players running around.

 

Thanks!

[steve_dunn2]

<p>Well, here are a couple of possibilities. One is the much-talked-about issue of a problem with the camera's AF system, in which it consistently (at least with a certain lens) focuses either in front of or behind the subject. I knew someone was going to jump on it so I mentioned it first. But really, option #2 is what you should be exploring before concluding you have a hardware problem.</p>

 

<p>Option #2: it's too harsh to call this one user error, but in essence, maybe the camera equipment itself is working as it was designed to work. This one encompasses a lot of possibilities; here are some, in no particular order.</p>

 

<ul>

<li><b>Depth of field</b>: You don't say what aperture you're using for your shooting, but a 135mm lens at f/2 (or even stopped down a bit) doesn't have a lot of depth of field; at a subject distance of 3m, for instance, your DOF is only a couple of centimeters on either side of the plane of focus (calculated using <a href="http://www.dofmaster.com/dofjs.html" target="_blank">this online DOF calculator</a>). Shooting at f/4 (e.g. with the 70-200 set to 135) roughly doubles this. This won't be the problem itself, because the camera's AF system sees the same shallow DOF you see*, but it can exacerbate other problems. It also means that the subject's entire face won't be in focus, unless they happen to have an extremely flat face.

<li><b>Changes in subject distance</b>: Is the camera on a solid tripod that doesn't move, or is it handheld (or on a monopod, which allows the camera to move)? Might the subject have moved slightly? Given the shallow DOF, you don't need much motion to throw the subject out of focus. This one isn't hard to test; put the camera on a tripod, sit an accomplice in a chair that has a back, and instruct them to keep their back firmly against the back of the chair. Take some test shots. If these shots are out of focus, you can rule out this possibility; if the shots are in focus, then maybe you're onto something.

<li><b>Size and position of the AF sensors</b>: In an ideal world, the AF sensors would be at exactly the same place in the frame as the little indicators in the viewfinder, and they'd cover exactly the same area as the little indicators in the viewfinder. But neither one is true of many cameras. It may be that the selected AF point did accurately lock onto something, but the something wasn't exactly what you intended (e.g. you pointed at the subject's eye, and it locked onto an eyebrow).

</ul>

 

<p>*: Note also that the camera does not necessarily focus exactly on the subject. The specs on the AF sensors are that the camera should focus within the DOF. So, technically, if the DOF extends 2 cm in front of the subject, the camera could focus 2 cm in front of the subject and still be operating correctly; never mind that the point you wanted in focus is now only just within DOF, and everything even slightly behind that point is getting a bit blurred. Now, usually the AF system will do better than this, and focus well within the DOF. But there's no guarantee of that.</p>

[andrew_certain]

At f2.0, if you have someone's face basically filling the frame, the DOF is going to be less

than an inch. In other words, the eye might be in focus, but the tip of the nose will not.

That's using dofmaster.com's .019mm CoC for the 300D. Now, if you expect the image to

be "sharp" when you view it at 100% resolution on your

monitor, you should probably use a CoC of 7.5 microns, which is the pixel size on the

300D sensor. That would give you a DOF of around a quarter inch. This is just due to the

laws of physics, and there is no way to take a close-up of a person at f2.0 and have both

the eyes and the tip of the nose in focus.

<p>

Even if you're willing to have the tip of the nose blurry, if you are shooting handheld, the

amount of camera movement between focusing and

exposure may be enough to throw off the focus from what you want. If you focus and

then recompose, <a

href="http://visual-vacations.com/Photography/focus-recompose_sucks.htm">you're

dead</a>.<p>

 

My experience with lenses faster than f2.8 on my 10D is that I have to indeed take a lot of

them to get ones that are focused the way I want them to. I don't think you're going to

find any equipment that helps you much here (other than maybe a tripod), as it's a matter

of optics, not technology.<p>

 

Andrew

[andrew_certain]

I should have just waited for Steve to answer. I wouldn't have spent the time, and you still

would have gotten the best explanation.... Good answer, Steve.

 

Andrew

[ericx]

Something that would be nice is a focus bracketing option. Working similar to

exposure bracketing, where it shoots at the AF spot, then slightly infront and then

slightly behind, to help fix AF problems, especially in low light.

 

What focus point setting are you using? (Just guessing here) but you may try to do

some tests and see which side of your DoF it's favoring, and then rock slightly

forward or backwards to bring the subject more into the middle of your range.

[jairy hunter]

Wow! Thanks for your quick and complete responses.

 

Actually the problem I am having is not so subtle. That is, a couple of cm here or

there are not an issue for me (ie the face in focus, nose slightly blurred).

 

And really, the focusing problem is not that bad or noticeable except to we

perfectionists.

 

I am completely and totally willing to accept the possibility of operator error (my

fault).

 

What is happening, specifically, is that when I compose a picture of a group of

people, say, three, on a front porch, I get a decent shot, but using autofocus, about

half the time with the 135mm (using P or automatic modes so I don't feel I have to

manage the aperture as completelu--handheld or monopod), the people are just

slightly out of focus--mainly observable at higher resolutions (say 200 pixels at 50%)

--but the front door, transom, and brick wall that the people are positioned in front

of --that's in full sharp focus.

 

Maybe I should try to take more control, but with 6 or 8 people in the shot, some

children, it's hard to imagine having to fiddle with the aperture settings too much

when the automatic modes and focus work so well. I guess I should say well enough.

 

Sample below (1/200s; f5.6; ISO 400); BTW keep in mind it has not been

photoshopped.

 

I think the most logical/understandable thinking for me is that the focusing zones on

the autofocus do not completely correspond to the zones on the viewfinder, which is

something I think i can correct for when shooting, if I know that to be true.

 

Thanks again for all the input and I'd appreciate any other thoughts you have on this.<div></div>

[steve_dunn2]

<p>Another possibility ... if you're taking pictures of groups of people, perhaps a 135mm lens is too long; even without the 1.6x crop factor, 135 is not a typical focal length for group shots. 135 is more typically used for tight head shots of a single person. Getting enough DOF for a group can be tough with a longer lens, and as you've noted, you're having to stand an awfully long way back.</p>

 

<p>If you need to keep using the 135 for group shots, manually select one focus point that you can be absolutely certain is on someone in the group (preferably someone who's around the middle of the group as far as front-to-back distance is concerned) without any chance of the background being involved, and stop down a fair bit to increase DOF.</p>

[andrew_certain]

The problem with autofocus is that it blindly (err... naively?) tries to maximize contrast

under the sensor. If there are several focus distances that have roughly the same contrast,

it will usually choose the nearer one. So far, so good. Unfortunately, if you have a

naturally low-contrast object (say, like a face or solid-color clothing) in front of a high-

contrast object (say, like a brick wall), the AF system sees that the best contrast position is

to focus on the wall. As Steve remarks, choosing an AF point (can you do that with a

300D?) that has no background nearby is a good option.

 

My guess is that the reason you see this on the 135 and not the 70-200 is that you are

using them for different subjects. If you used the 70-200 for the shot you posted, my

guess is you'd be likely to have the same problem.

 

Finally, Steve mentioned that you might try a different focal-length lens, but before you go

out and spend money on another lens, plug some numbers into a DOF calculator. As you

get closer to your subjects, the effect of lens length on DOF decreases. It's true that when

using a long lens, the DOF is shallower for a given object distance, but if you want to get

the same composition in the plane of focus with a short lens, you have to reduce the

object distance, which has the effect of reducing the DOF. The closer the focusing

distance, the more these two things cancel out and yield the same DOF.

 

See http://www.vanwalree.com/optics/dofderivation.html for some math.

 

Andrew

[phil vaughan - yorkshire u]

It's OK to allow the camera to choose your exposure settings, however you obviously aren't getting the results you expect by letting the camera choose the focus point. If you select the focus point and place it one of the childrens faces, then you will get a sharp (enough) picture, however using a small aperture and having the camera find a focus point, (where frankly it looks for the easy option - the high contrast door frame) will lead to frustration.

 

Remember the advice around here is at it's best when questions are specific, a 135mm f2 lens is a perfect portrait lens (for some users, therefore it's often recommended), however on a 300d it's a little long for a single head and shoulders, for a group it's just too long. For a traditional portrait look on a 300d I'd go for; group 22-35mm (depending on the size of the group), 3/4 portrait (or couple) 35-50mm, head and shoulders 50, 85 or 100mm. You must have been stood across the street to get them into an effective 200mm frame.

[gary_j._honis]

You may also want to test your camera to be sure that the problem with focus is not with the camera. In dis-assembling a Canon Digital Rebel I learned how critical distances within the camera are for precise focus. See the following page that explains how I tested the focus of my Rebel after modification and for the location of the camera's AutoFocus screw:

 

http://ghonis2.ho8.com/rebelmod9.html

 

If your camera is not focusing well from the tests, you should return it to Canon for repair. If out of warranty, you may want to consider making the Autofocus adjustment yourself. Doing the test would at least let you know how well your camera is focusing.

 

Gary Honis