1D Mark IV center point focus issue?

[SteveYoung]

I just recently upgraded to a Mark IV and am overall very happy, however I noticed what might be an issue with the center point AF and wanted to check in and see if this is normal or not. I've been doing a bunch of BIF shots on days when weather permits and I've noticed that even when I am very sure that I have the center point over the bird, focus will shift or drop and I will have to attempt to re-acquire. What it seems like is that the center point is much larger than represented in the viewfinder or possibly offset? I'm not sure. Once AF's acquired it hangs on pretty tenaciously but I've been getting frustrated with shots that should not have missed refocusing on near BG elements. This wasn't an issue with my 7d and if my shots were OOF with that body I could be sure it was user error. This only seems to happen with moving subjects that are relatively small in frame but larger than the center point. Should I be contacting CPS or is this just how the AF works on this body?

[arash_hazeghi]

Hi Steve,

The center point should not have a problem. You might have expansion turned on and that is why the focus is locking on the BG.

If the bird becomes too small in the frame (smaller than the center meter circle) it will be difficult to hold focus against a varied BG.

[Doug Schurman]

I have found that the center focus point must be quite a bit bigger than just the square. Sometimes I use it with long lens for identification of seabirds that are several hundred yards out. When the bird is really small in the frame (close to the size of the center box) it seems like the focus will get the water infront of the bird. sometimes it help to raise my focus point to accommodate this.

[SteveYoung]

Thanks for the response. I double checked and made sure expansion point was turned off, which it is. I haven't been able to replicate the problem on static subjects so it's probably just some slight difference that I haven't gotten used to from switching camera bodies. I just wanted to see if there was some known issue or quirk before I got freaked out and shipped my camera off to Irvine.

[arash_hazeghi]

If you look at the AF pattern for 1D4 you can see that the AF sensors are densely packed so, obviously each sensor cannot be much larger than the AF frame indicates. So that is not the issue.

125569

1D4 AF pattern

The main difference between the 1D4 and lesser bodies is the sensitivity of the AF sensor. This means that during tracking if the bird slips off the sensor even momentarily the camera will respond and lock on the BG. This maybe hard for the user to see or feel in the midst of mirror going up and down 10 times a second but it happens. Setting AI-servo tracking sensitivity to slowest helps mitigate this issue, but overall to get the best out of MK4 you need to track the bird dead centered rock-solid. Many of the issues with MK4 AF stem from imperfect tracking technique. It is like a sports' car, it responds to every input.

Tip: do not attempt to track birds that are smaller than the center meter circle, because even if the focus holds the crop will be huge (<10% of the frame) so the details will lack anyway and the output will be poor.

hope this helps

[SteveYoung]

So basically this is the situation here:

http://i.imgur.com/dYNT2b8.jpg

Granted this was at dusk and it was a fast moving eagle but I have a series of these with the center point right over the bird and they are all OOF.

In related news, this issue allowed me to discover ZoomBrowser, which I'd never used before, so that's cool.

[arash_hazeghi]

In this case it is quite obvious that the focus will lock on the BG. The AF will not see such a tiny target and even if it did lock, the bird would have no detail at all.

[SteveYoung]

The main difference between the 1D4 and lesser bodies is the sensitivity of the AF sensor. This means that during tracking if the bird slips off the sensor even momentarily the camera will respond and lock on the BG. This maybe hard for the user to see or feel in the midst of mirror going up and down 10 times a second but it happens. Setting AI-servo tracking sensitivity to slowest helps mitigate this issue, but overall to get the best out of MK4 you need to track the bird dead centered rock-solid. Many of the issues with MK4 AF stem from imperfect tracking technique. It is like a sports' car, it responds to every input.

Tip: do not attempt to track birds that are smaller than the center meter circle, because even if the focus holds the crop will be huge (<10% of the frame) so the details will lack anyway and the output will be poor.

hope this helps

Right, totally. I was noticing the problem when I wasn't actually releasing the shutter and was just trying to pan with the bird. I think what I'm gonna do is just use ZoomBrowser along with as much panning practice as the Pacific Northwest weather will allow and see if I notice a consistant problem with either my technique (probable) or with the equipment.

[SteveYoung]

In this case it is quite obvious that the focus will lock on the BG. The AF will not see such a tiny target and even if it did lock, the bird would have no detail at all.

Huh, ok. That's what I needed to know then. I realize that detail would be poor here, I was mostly tracking the eagle for practice and because it was hunting ducks, which is a pretty exciting thing to watch. Thanks so much for your responses, they were very helpful.

[arash_hazeghi]

Just watch for the size of the bird and the center circle and you should be good to go. Hope you get the eagle images you want :)

[Roger Clark]

Steve,

First of all the AF sensitive region is much larger than the AF rectangle, contrary to what was said above. Phase detect AF uses and out of focus part of the incoming light cone from the lens so sees a larger diffuse zone that is 2 or more times the size if the AF rectangle. Thus, adjacent AF rectangles sees an overlapping zone with other AF rectangles.

A simple way to determine the active AF region is to use a small bright light in a dark background with the light being smaller than an AF rectangle in the viewfinder. For example a distant street light, with no other lights in the field of view, or the moon in a clear dark sky. Move the lens out of focus and then activate AF at different distances from the AF rectangle and see how far you can be for AF to work. You may be surprised at how far the sensitivity is.

Now, with that knowledge, consider the following. The AF system will "choose" the brightest high contrast subject within the diffuse active AF zone. In the example image you posted, the brightest subject near the AF rectangle is the bright (snow?) patch below the bird. The AF system will most likely lock onto that, even if the AF rectangle was a full AF length away from the active AF sensor. It is very difficult to track a dark subject like your bird with brighter distracting elements in the foreground or background. It is best to wait for the subject to be closer (filling more of the frame), and/or does not have the distracting brighter elements nearby if you wanted to make an environment image.

Roger

[arash_hazeghi]

The above is incorrect. The area that sensors see is slightly larger but no more than half the spacing between to adjacent sensors. This is because a 1D AF sensor is not a simplified phase detector. The AF system has two layers of micro lenses, at the sub mirror and on the die surface which columnate light rays.

That's why you can have a spot AF which is even a tighter sampling area.

It is super easy to track a dark bird against a bright background, a dark bird against bright BG it is actually like a magnet for AF. AF is challenging when bird and BG have similar tones, but every photographer knows that.



Steve's problem is very simple, the bird is too small in the frame. nothing more.

[Roger Clark]

The above is incorrect. The area that sensors see is slightly larger but no more than half the spacing between to adjacent sensors. This is because a 1D AF sensor is not a simplified phase detector. The AF system has two layers of micro lenses, at the sub mirror and on the die surface which columnate light rays.


Simple: try the experiment I outlined. I have. I know the area my 1DIV AF sensors are sensitive to, and it is much larger than the AF rectangle.

Roger

[arash_hazeghi]

I don't think that experiment is a good way to test this. The dark enviroment will cause the AF sensor to switch to high gain, then the light source will saturate the pixels on the AF sensor causing erratic readings and AF oscillations...it may give a false lock and you can't tell if it is in focus or not...This is an extreme case when a slight bleeding will just saturate the sensor. I don't really see the connection with BIF photography.

[Roger Clark]

I don't think that experiment is a good way to test this. The dark enviroment will cause the AF sensor to switch to high gain, then the light source will saturate the pixels on the AF sensor causing erratic readings and AF oscillations...it may give a false lock and you can't tell if it is in focus or not...This is an extreme case when a slight bleeding will just saturate the sensor. I don't really see the connection with BIF photography.

A very simple thought experiment. If one tries the experiment, one will find quite accurate AF with the subject quite outside the AF rectangle. Try it too with the moon in a daytime blue sky. You get the same result. I have. And if it gives a false focus lock. what is to say a bright subject near an AF point when trying to focus on a bird will not cause the same thing to happen? It illustrates the problem of finding correct lock on subject with complex foregrounds or backgrounds. This is likely what is happening in the posted image. The point of the experiment is to show SENSITIVITY of the AF point outside the AF rectangle, not to say it provides accurate AF. In fact, as one moves away from the AF rectangle, AF accuracy falls. But that just makes the problem worse. It is really a simple test that can be done in many ways to show the sensitivity is larger than the AF rectangle.

Roger

[Roger Clark]

A further note. The size of diffuse AF region that is larger than the AF rectangle will vary by f/ratio, with faster f/ratio being larger and more diffuse. This is not your selected f/ratio but the f/ratio of the system, as AF occurs before the aperture is closed to your selected f/ratio. Thus, the size of the AF zone will be large with an f/2.8 lens, then will decrease and become less diffuse when you add a 1.4x TC, then decrease more and be less diffuse when a 2x TC is added. Same with an f/4 lens and adding TCs.

Roger

[arash_hazeghi]

I have never seen a camera lock focus on a subject fully outside the AF sensor area.

Stick a pole in front of a white plain BG and lock focus on it and then move the camera a bit, you will see that AF will start to hunt as soon as the pole exist the AF frame. very simple.

In AI-servo mode the focus may stay locked on the pole even when it is outside the frame for a few moments, but this has nothing to do with sensitivity of AF sensor. It is the AI servo custom functions that tells the camera to hold focus upon subject loss.

[SteveYoung]

There's some interesting points in this discussion. Out of curiosity, is there a significant difference in the AF point sensitivity between the 1D and 7D? As I mentioned in my original post, with my 7D I never noticed this issue with oof birds provided my rectangle was over the target. Would the smaller sensor play a role or could it be just an apparent difference based on the narrower field of view?

I should note, too, that a friend of mine that I often shoot with also recently upgraded to the mark IV from the 7D and noticed the same difference.

[Roger Clark]

I have never seen a camera lock focus on a subject fully outside the AF sensor area.

Stick a pole in front of a white plain BG and lock focus on it and then move the camera a bit, you will see that AF will start to hunt as soon as the pole exist the AF frame. very simple.

In AI-servo mode the focus may stay locked on the pole even when it is outside the frame for a few moments, but this has nothing to do with sensitivity of AF sensor. It is the AI servo custom functions that tells the camera to hold focus upon subject loss.

Arash,
Your scenario simply verifies what I have been saying. The AF will try and lock on to the brightest subject. So when you move the AF point off the pole it tries to lock on to the bright background. If there is any structure on that brigh background, AF will lock onto it.

I was out looking for comet panstarrs this evening and made some demonstration images that show AF locks onto a nearby brighter subject.

In the first image, AF locks onto the bright contrail that is about 1 AF rectangle away.
In the second image, AF locks onto the tree line, about 1 AF rectangle away. Dark edge on a bright bland field
In the 3rd image, AF locks onto the tree line, about 1 AF rectangle away. Bright edge on a dark bland field.

In all cases, the focus is spot on (300 f/2.8 lens with 1D Mark IV). However, increasing the distance much more and the AF accuracy drops rapidly.

Roger

[Roger Clark]

There's some interesting points in this discussion. Out of curiosity, is there a significant difference in the AF point sensitivity between the 1D and 7D? As I mentioned in my original post, with my 7D I never noticed this issue with oof birds provided my rectangle was over the target. Would the smaller sensor play a role or could it be just an apparent difference based on the narrower field of view?

I should note, too, that a friend of mine that I often shoot with also recently upgraded to the mark IV from the 7D and noticed the same difference.

Steve,
Do you have the AF sensitivity set differently on the two cameras? I have a 1DIV and 7D and find similar response of the larger AF area than the red AF rectangle.

Roger

[arash_hazeghi]

What I mentioned had nothing to do with the brightness of the BG, you can try it with a dark BG too. My point was that when the pole exists the AF frame , the AF does not see it and therefore loses focus.

I am not sure about your test, it is not clear where the camera had focused as everything is so far away and appears in focus any way. the last image was heavily underexposed but in the original scene there was enough light for the camera to focus on the trees. The MK4 can focus in even lower light conditions.

All I can say is that none of my Canon cameras lock on a subject outside of the AF frame, I'm pretty sure about it.

It would be very stupid of Canon to make an AF pattern like that if each AF point saw such a big area, it doesn't make sense to me,

[Roger Clark]

What I mentioned had nothing to do with the brightness of the BG, you can try it with a dark BG too. My point was that when the pole exists the AF frame , the AF does not see it and therefore loses focus.

I am not sure about your test, it is not clear where the camera had focused as everything is so far away and appears in focus any way. the last image was heavily underexposed but in the original scene there was enough light for the camera to focus on the trees. The MK4 can focus in even lower light conditions.

All I can say is that none of my Canon cameras lock on a subject outside of the AF frame, I'm pretty sure about it.

It would be very stupid of Canon to make an AF pattern like that if each AF point saw such a big area, it doesn't make sense to me,

This really shouldn't be this hard. Phase detect AF systems work on an out-of-focus portion of the light cone from the lens. By definition, it is out of focus, so there are no hard edges to the AF sensor. It is basic optics. Try doing a wikipedia or other web search and learn about phase detection. Note, specifically, the masks used to isolate the light from each side of the lens are not at the focal point, thus have no hard edge to the sensitivity of the AF region. The AF rectangles simply represents the peak response of the AF point, and not the limit of the response.

Regarding my examples, I moved the lens out of focus for each position and let the camera focus. You see the results.

Roger

[arash_hazeghi]

It is not hard but I think you should study this subject more, as you seem to confuse the AF module in a modern SLR camera with a simple general phase detector. It is not like that. As I pointed out the light from the lens is colmanted by the various lenses in the AF sub-system several times before it reaches the AF sensor. So it is not a simple light cone in the simplified model in your mind.

It is good to discuss basic concepts for the purpose of learning, but your problem is trying to explain complex engineering systems with oversimplified models and assumptions that are misleading and almost always to the contrary of field experience by every single experienced photographer on this site. But this has been discussed before many time so I see no need to bring it up again.

If your camera locks focus on subjects outside the focus frame you need to send it to Canon for repair because a healthy camera doesn't do that.


Good luck

[brian simpson]

Hi roger, I just tried experiment with my nikon D700 and yes you are correct, that is how my camera works also. All camera settings to single point center no expansion no lock on, Tree trunk in feild of bright white snow autofocus sensor just to the side not on tree and the tree snaps into focus ?

[Roger Clark]

It is not hard but I think you should study this subject more, as you seem to confuse the AF module in a modern SLR camera with a simple general phase detector. It is not like that. As I pointed out the light from the lens is colmanted by the various lenses in the AF sub-system several times before it reaches the AF sensor. So it is not a simple light cone in the simplified model in your mind.

It is good to discuss basic concepts for the purpose of learning, but your problem is trying to explain complex engineering systems with oversimplified models and assumptions that are misleading and almost always to the contrary of field experience by every single experienced photographer on this site. But this has been discussed before many time so I see no need to bring it up again.

If your camera locks focus on subjects outside the focus frame you need to send it to Canon for repair because a healthy camera doesn't do that.

Good luck

Arash,

See this demo from class CS178 at Stanford:
http://graphics.stanford.edu/courses/cs178/applets/autofocusPD.html
Be sure to play with the focus slider.

Not sure what colmanted means. If you mean collimated, yes, I understand the collimation. In order to do phase detect AF, one must isolate portions of the incoming light cone from each side of the lens. This can not be done in focus. It is a brilliant concept: fast focusing and focus tracking of moving subjects using out of focus information. Even when the subject is properly focused, the AF system is measuring light from the out of focus field. So whether a field stop is used outside of focus, or the colimating lenses are used as the field stop, as in the above example, both are in the out of focus field. In order to have sensitivity over a range of f/stops and focal distances, that means the optical response of the system will not have a hard edge to AF sensitivity. Period. I provided direct evidence of this effect. Every camera I have tested behaves the same from film DSLRs to the latest cameras.

Understanding these effects will help one to achieve accurate AF on one's desired subject in more conditions.

I'm done with this thread.

Roger

[John Chardine]

Steve- re. your pane 6 post. Without the techs of the image it's hard to say what is happening but the blurriness may not be due to focus but due to motion blur. There is clear evidence of this in the wooden structure to the left. It is quite possible that the bird has some motion blur as well. It looks to me like no part of the image is sharp- the BG does not look in focus either so the AF has not locked there.

pane 11- Roger is right here. Reference to page 19 of the white paper on the 1DIV clearly shows that the AF sensor area is much larger than the red rectangle in the viewfinder.

[Daniel Cadieux]

Hi all, what began as a very interesting thread has now veered off topic and pretty much run its' course and I am closing it. Some nerves were understandably touched on both sides but let's all take from what was in it originally and move on from there...

Thanks guys!