1D IV Tracking Experiment

[clk_walker636320]

Only had this body for a week so be kind. I have shot a couple of BB games with so so results. User error, not camera. But I was interested in the ability of the camera to track. So I setup a simple test which is not necessarily how the camera is to be used. However the tracking system should not care. The test:

Indoors using a resolution chart, AV mode, custom WB, ISO6400, AI Servo, tracking priority with manual AF point and 45 expansion points, 3FPS, 24-70mm f/2.8L @ 70mm f/2.8. I selected the furtherest AF point to the left and acquired the target and starting shooting and panning.

I would have thought that the AF point being used would remain on the target as I panned. The last AF point should have been the furtherest AF point to the right. Needlessly to say the tracking failed. I have not been able to get the camera to do this no matter what settings I use. Fast/slow no expansion points or 45.

So should I expect teh camera to do this or not? the panning was relatively slow.--
A bird in the viewfinder is worth...

 

[Scott Larson]

Indoors using a resolution chart, AV mode, custom WB, ISO6400, AI Servo, tracking priority with manual AF point and 45 expansion points, 3FPS, 24-70mm f/2.8L @ 70mm f/2.8. I selected the furtherest AF point to the left and acquired the target and starting shooting and panning.

A resolution chart meaning a chart with lots of lines? If so, the AF system will naturally be confused by that in 45 point expansion mode since it can focus on anything in the chart as it fills the viewfinder. You can't expect the AF to remain locked on one part of a large target when it finds an acceptable target under multiple focus points. It won't know which one you want.

 

[clk_walker636320]

Indoors using a resolution chart, AV mode, custom WB, ISO6400, AI Servo, tracking priority with manual AF point and 45 expansion points, 3FPS, 24-70mm f/2.8L @ 70mm f/2.8. I selected the furtherest AF point to the left and acquired the target and starting shooting and panning.

A resolution chart meaning a chart with lots of lines? If so, the AF system will naturally be confused by that in 45 point expansion mode since it can focus on anything in the chart as it fills the viewfinder. You can't expect the AF to remain locked on one part of a large target when it finds an acceptable target under multiple focus points. It won't know which one you want.

The chart occupies a small percentage of the FOV. Any AF point on the chart is counted as being in focus. I am concerned about the AF point jumping to the background. The chart is suspended so there are no near objects.
--
A bird in the viewfinder is worth...

 

[liquidstone]

Sorry to ask, English is not my native language.

Very interested to understand what you're saying.

Only had this body for a week so be kind. I have shot a couple of BB games with so so results. User error, not camera. But I was interested in the ability of the camera to track. So I setup a simple test which is not necessarily how the camera is to be used. However the tracking system should not care. The test:

Indoors using a resolution chart, AV mode, custom WB, ISO6400, AI Servo, tracking priority with manual AF point and 45 expansion points, 3FPS, 24-70mm f/2.8L @ 70mm f/2.8. I selected the furtherest AF point to the left and acquired the target and starting shooting and panning.

I would have thought that the AF point being used would remain on the target as I panned. The last AF point should have been the furtherest AF point to the right. Needlessly to say the tracking failed. I have not been able to get the camera to do this no matter what settings I use. Fast/slow no expansion points or 45.

So should I expect teh camera to do this or not? the panning was relatively slow.--
A bird in the viewfinder is worth...

--
Romy



PHILIPPINE WILD BIRDS (Over 260 species captured in habitat, and counting.)
Photo Gallery - http://www.pbase.com/liquidstone
HD Bird Videos - http://www.youtube.com/user/romyocon

 

[Scott Larson]

The chart occupies a small percentage of the FOV.

How small?

Any AF point on the chart is counted as being in focus. I am concerned about the AF point jumping to the background.

Everything on the chart is about the same distance from the camera so the AF will consider it to be the same subject. However as a subject moves, the AF is constantly losing the subject and reacquiring it as you pan with it. You can see how quickly and easily this happens in 45 point expansion mode.

If the AF loses the subject (meaning there is no target in the selected focus point near to where it last focused successfully) for a period of time then the "sensitivity" setting will tell the camera how long to wait before looking for something else to focus on. If you can't get the focus point back on your target, it will find something else to focus on, most likely the background. The challenge in photography is keeping that focus point on your subject so the camera won't want to focus on other things.

There is also a mode misleadingly called "Main Point Priority" which will force the camera to focus on whatever target it finds closest to the camera in all active focus points. This mode is intended for tracking isolated subjects that would make the AF want to lock on the background but it's a complete disaster for basketball as you can imagine.

 

[Scott Larson]

Sorry to ask, English is not my native language.

I think the poster means "farthest".

 

[liquidstone]

Thanks, Scott.

I've no further questions.

Sorry to ask, English is not my native language.

I think the poster means "farthest".

--
Romy



PHILIPPINE WILD BIRDS (Over 260 species captured in habitat, and counting.)
Photo Gallery - http://www.pbase.com/liquidstone
HD Bird Videos - http://www.youtube.com/user/romyocon

 

[clk_walker636320]

English is my native language but what do I know!! OK more details. The test setup is in the house with the distance about 10 feet. This experiment is equivalent to a small subject moving in an arc of a circle about the camera since the distance does not change. The AF point red box is about 1/5 the height of the resolution chart. I was panning very slowly to give the tracker amble opportunity to track correctly. Track has been typically lost when the subject gets about in the middle if the FOV or the camera is aimed directly at the subject.

This would seem to me to be a simple test and based on the writeup in the Canon white paper and other material describing the tracker performance using expansion points it should be easy.

Yes keeping the AF point on the subject should be easy and not very challenging if there are no interfering objects or background. But this experiment is a simple test of the ability of the tracker to handoff to another AF point to maintain track.
--
A bird in the viewfinder is worth...

 

[stoddard]

Can't remember what review or where I saw it, but the reviewer could not get the camera to hand off the focus point like you are trying to do. He was testing both the Mark IV and a Nikon D3. The D3 would show the focus point moving. Don't know if this helps you or not.

 

[lol101]

It would be easier to undestand if you posted a picture of your setup (maybe even a sequence of shots demonstrating the problem).

It's an interesting test and I can't quite undestand why it would fail.

What happens if you start with center AF point?
--
http://lol-photoblog.blogspot.com/

 

[clk_walker636320]

It would be easier to undestand if you posted a picture of your setup (maybe even a sequence of shots demonstrating the problem).

It's an interesting test and I can't quite undestand why it would fail.

What happens if you start with center AF point?
--
http://lol-photoblog.blogspot.com/

I will be glad to. Can you tell me how to get the AF points to show on the image? I use zoombrowser to view the AF points but not sure how to get them to embed. Thanks.
--
A bird in the viewfinder is worth...

 

[lol101]

It would be easier to undestand if you posted a picture of your setup (maybe even a sequence of shots demonstrating the problem).

It's an interesting test and I can't quite undestand why it would fail.

What happens if you start with center AF point?
--
http://lol-photoblog.blogspot.com/

I will be glad to. Can you tell me how to get the AF points to show on the image? I use zoombrowser to view the AF points but not sure how to get them to embed. Thanks.
--
A bird in the viewfinder is worth...

Don't know anyhting about ZoomBrowser but I just get screen captures when I need to show this kind of tests (I am using PicPick, it's free for personal use).

--
http://lol-photoblog.blogspot.com/

 

[riknash]

It would be easier to undestand if you posted a picture of your setup (maybe even a sequence of shots demonstrating the problem).

It's an interesting test and I can't quite undestand why it would fail.

What happens if you start with center AF point?
--
http://lol-photoblog.blogspot.com/

I will be glad to. Can you tell me how to get the AF points to show on the image? I use zoombrowser to view the AF points but not sure how to get them to embed. Thanks.
--
A bird in the viewfinder is worth...

Don't know anyhting about ZoomBrowser but I just get screen captures when I need to show this kind of tests (I am using PicPick, it's free for personal use).

In Windows, you can simply capture the active window to the clipboard by depressing Ctrl+Alt+PrntScrn together. Open up Photoshop and paste into a new layer, or use Microsoft Paint and paste into it.

 

[clk_walker636320]

It would be easier to undestand if you posted a picture of your setup (maybe even a sequence of shots demonstrating the problem).

It's an interesting test and I can't quite undestand why it would fail.

What happens if you start with center AF point?
--
http://lol-photoblog.blogspot.com/

I will be glad to. Can you tell me how to get the AF points to show on the image? I use zoombrowser to view the AF points but not sure how to get them to embed. Thanks.
--
A bird in the viewfinder is worth...

Don't know anyhting about ZoomBrowser but I just get screen captures when I need to show this kind of tests (I am using PicPick, it's free for personal use).

--
http://lol-photoblog.blogspot.com/

OK the following is sequence of screen shots that chsne only when the AF point changes. There was a a total of 37 shots at 3fps. Notice that tracking starts at the extreme left AF point and does move more to the center as I pan. The AF point gradually moves to the left of the target and finally goes off and acquires the background.

--
A bird in the viewfinder is worth...

 

[riknash]

Do you think that all that light in the background might interfere with the AF determining the distance of the low contrast focus chart and consequently which AF point to choose? The focus chart looks really dark to me.

 

[Schwany]

Looks correct. I'm pretty sure that you can't get all 45 points to work once you select one. Only a user specified number of AF points around the selected point will work.

How about putting some light on the chart? I usually try to shoot with light on the subject. Shooting at the shadowed side of a subject is generally a waste of time.

 

[clk_walker636320]

The camera is initially tracking as I pan. But the selected AF point (the one used for updating the tracker) appears to gradually move off the subject. With a manually selected AF point and 45 point expansion the tracker should be able to use any of the 45 AF points to assist in tracking. But the key is, if I understand the Canon literature correctly, is that the selected AF point (the one used to update teh tracker) will move and track the subject as the subject moves across the FOV. This test was an attempt to demostarte this capability. To me it would be no different than acquiring the subject at one extreme of the FOV and have the tracker track the subject as it moves across the FOV without the camera following the subject.
--
A bird in the viewfinder is worth...

 

[Schwany]

Yeah I get it, but actually didn't think it worked that way. Don't have time to look into it. I need to get out and see how bad my panning has gotten after a month of not using my camera. ;-)

Without light on the subject, don't you think this test is doomed to failure regardless of how many points are active?

The camera is initially tracking as I pan. But the selected AF point (the one used for updating the tracker) appears to gradually move off the subject. With a manually selected AF point and 45 point expansion the tracker should be able to use any of the 45 AF points to assist in tracking. But the key is, if I understand the Canon literature correctly, is that the selected AF point (the one used to update teh tracker) will move and track the subject as the subject moves across the FOV. This test was an attempt to demostarte this capability. To me it would be no different than acquiring the subject at one extreme of the FOV and have the tracker track the subject as it moves across the FOV without the camera following the subject.
--
A bird in the viewfinder is worth...

 

[Scott Larson]

Looks correct. I'm pretty sure that you can't get all 45 points to work once you select one. Only a user specified number of AF points around the selected point will work.

The 1D Mark IV has a "45 point expansion" mode which really will turn all 45 points on and will show in real time which point the AF system has switched to.

 

[Scott Larson]

OK the following is sequence of screen shots that chsne only when the AF point changes. There was a a total of 37 shots at 3fps. Notice that tracking starts at the extreme left AF point and does move more to the center as I pan. The AF point gradually moves to the left of the target and finally goes off and acquires the background.

It's locking on whatever point has the most contrast. Your test is a great example of that because your target is in front of a brighter background -- the worst case scenario. Additionally 45 point expansion is the worst thing to use in this situation.

Remember that the focus points are very simple sensors. They only see gray contrast, not objects. If a neighboring focus point has a more solid phase match on something than the one under your subject and whatever it has matched is not much farther away from your subject, it will switch to that point and leave your subject behind. The exact logic is something only Canon knows (maybe) but it's clear that at times it will favor contrast over tracking.

I see this all the time. Whenever a basketball player moves in front of the backlit courtside advertisements, I know the AF will want to switch from the player's face to the bright, high contrast lettering in the ads. Whenever a defender runs up to a ball handler, I know the AF will want to switch to the lettering on the back of the defender's jersey. Once you see these things coming, you can usually steer your selected focus point away from these tempting targets.

And that's why I only use one focus point. As you can see from Canon's white paper, the focus points in the Mark IV are gigantic already so I haven't found any need to use expansion points. I think it causes a great number of tracking problems that Rob Galbraith described in his Mark IV AF report in which he always used full point expansion around selected focus points.