PDA

View Full Version : 5D servo focus; expecting too much?


Phil Hawkins
June 12th, 2006, 11:37 PM
I'm new here, this is my first post. Referered by someone on Fred Miranda, so Hi Folks, I have a question about the 5D servo focus.

I shoot motocross racing, which is fast, fast fast. I shoot mostly for head-on motion to the riders. On Motocross tracks, they are coming at me either directly (90 degrees) or slightly to the left. On TT, (flat track oval course racing) they come at me in a circular motion, but head-on at bout 50 degrees off dead center.

I switched to Servo Focus due to the function that servo is supposed to measure the speed and direction of the subject and compensate for "perfect" focus (or as close to perfect as can be reasonably be expected.) I set for center focus-point which adds 6 additional focus points and my results are a bit disappointing.

Shooting usually at 2.8 to blurr background at 200 ISO and 1250 shutter speed.

full-frame image (slightly cropped)



http://yosemitefun.com/sharp_demo2.jpg



Now, cropped 100%


http://yosemitefun.com/sharp_demo.jpg

and, an image I consider to be prefectly focused; this is the result I'm looking for in all shots.

Full frame for reference:


http://yosemitefun.com/sharp_demo3.jpg

and 100% image


http://yosemitefun.com/sharp_demo4.jpg


I'm not getting the sharpness I got on this quad shot.

Why?

Phil

Daniel Harrison
June 13th, 2006, 12:34 AM
Firstly I would like to say that it is not an easy task to take oncomming photos :-) And I like your photos - regardelss of the 100% sharpness.

But here are the reasons I think your photo isn't as sharp as you would like.

You are shhoting at f/2.8 which means an extramly small DOF. You can see in the second photo that while the helmet is in focus the sholder is not.

You are using the middle focus points - and his head is no where near the middle- so obviously the camera will be tracking the bike (around the front area). Add to that you shallow DOF and you are sure to get a OOF head shot.

The sharp photo is not coming directly at you, so it is much easier for the camera to lock onto.

basically if you are going to shoot a f/2.8 and use the middle(and 6hidden) points you are not going to get the head in focus. I do not think you need to shoot at f/2.8. bump your ISO to 650-800 and stop down to f/5.6 and you shots will come out better, you probably won't notice much difference in the OOF background anyway. it will still be OOF

Otherwise try switching AF points - but I would still stop down.

keep up the great work
Daniel

Michael Tapes
June 13th, 2006, 05:02 AM
I concur..

Thomas Krueger
June 13th, 2006, 06:33 AM
You can for example calculate your depth of field with fcalc (http://tangentsoft.net/fcalc/) to get a rough estimate which parameters you need to get a sufficient range of DOF (depth of field).

Alexandru Petrescu
June 13th, 2006, 06:47 AM
Recently my 5D servo disappointed me on a bad day until I realized it was because of the circular polarizer filter I'd put on the 50/1.4. I hope you used no filter.

Phil Hawkins
June 13th, 2006, 08:15 AM
Firstly I would like to say that it is not an easy task to take oncomming photos :-) And I like your photos - regardelss of the 100% sharpness.

But here are the reasons I think your photo isn't as sharp as you would like.

You are shhoting at f/2.8 which means an extramly small DOF. You can see in the second photo that while the helmet is in focus the sholder is not.

You are using the middle focus points - and his head is no where near the middle- so obviously the camera will be tracking the bike (around the front area). Add to that you shallow DOF and you are sure to get a OOF head shot.

The sharp photo is not coming directly at you, so it is much easier for the camera to lock onto.

basically if you are going to shoot a f/2.8 and use the middle(and 6hidden) points you are not going to get the head in focus. I do not think you need to shoot at f/2.8. bump your ISO to 650-800 and stop down to f/5.6 and you shots will come out better, you probably won't notice much difference in the OOF background anyway. it will still be OOF

Otherwise try switching AF points - but I would still stop down.

keep up the great work
Daniel

Will do; but the 1250 shutter is very necessary due to the incredibly fast speeds at which these riders go. I'll see if I am freezing the action at 650 to 800 shutter.

Thanks!

Phil

Alexandru Petrescu
June 13th, 2006, 09:50 AM
Not using the center focus point on the headshot also means that the meter is fooled. With the 5D the meter can only be linked to the center point even if other point focuses. In this contrasty light all metering modes can be easily fooled.

Not an absolute necessary to use high shutter speeds. Very nice background effects can be obtained with 45-to-125 speeds if "panning" - camera exactly follows the moving subject. Or zoom out-in (from tele to wide) on a static but smiling subject with 1/30s and you get a very nice action photo. Add a 2nd curtain flash to it and it's almost a guaranteed success.

Phil Hawkins
June 13th, 2006, 01:02 PM
Thanks, petrescu

I am using AI Servo which uses the center point by default, and adds 6 focus points directly above and below the centerpoint. This, in theory according to the manual, "ensures" that enough image sampling can occur to render a sharp focus. Such is not the case.

Phil

Daniel Harrison
June 13th, 2006, 03:24 PM
Hi Phil,
You were shooting at ISO 200, bump it up (640-800) and you will have no problem with shutter speed. I wouldn't lower it.

Happy Shootng
Daniel

Alexandru Petrescu
June 13th, 2006, 03:27 PM
Ok, I'm wondering whether you're too fast deciding the 5d servo af is too slow. With a subject coming straight at you too slow focus and 2.8 would mean the sharp image is behind the point where you focused. I can't see on the pictures any area behind the subject that is sharper than other areas.

I hope the C.Fn-17 Custom Function is activated on the body (without it the Servo won't use all the invisible focus points).

I'm wondering whether the lens you used (which one?) has 2.8 as its max aperture because it's known most lenses at their max apertures don't deliver their sharpest pictures, should be stopped down. On my 70-200/4 the 4 gives only average sharpness; the 50 at 1.4 is so-and-so, starts good at 2.8. The 135/2 is sharper at 2.8 than at 2.0. The invisible points are used only if the lens is 2.8 or below (at 4 they don't work). So you'd need a very very good lens. All zooms start at 2.8, and I don't think you used a prime. Which makes me think you may have used a lens with max 2.8. So what lens did you use?

I wouldn't avoid using an iso1600 or 3200 for higher shutter speed, because in these daylight conditions the underexposed areas will be very small, I believe, maybe none at all. The noise at 1600-3200 comes mainly in the underexposed areas.

I think I've seen some tests with vehicles coming straight to 5d photographer at vehicular speeds (like 70kmph) and focus was following ok.

That's why I doubt 5d can't keep.

Daniel Harrison
June 13th, 2006, 04:20 PM
Thanks, petrescu

I am using AI Servo which uses the center point by default, and adds 6 focus points directly above and below the centerpoint. This, in theory according to the manual, "ensures" that enough image sampling can occur to render a sharp focus. Such is not the case.

Phil

Phil,
I think you are missing the point. The camera is focussing on the BIKE not the head. The middle AF point is NO WHERE NEAR the head. You are using f/2.8 which is a small depth of field or in other words only a very small amount of the image will be sharp. If you want the camera to focus on the head you are going to have to use a different AF point or use f/5.6 or higher. I don't think there is anything wrong with the camera, it's just that you are not fully understanding what it is doing.

My reference to 640-800 was NOT the shutter speed, but you ISO. Raising your ISO will give you the ability to stop down your aperture (f/5.6) and keep your high shutter speed.

If I am still not making sense let me know. But I bet that the front of the bike is in focus?

when you figure it out you will have to post us some more shots, they are quite good!!!

keep it up

Daniel

Lee Jay Fingersh
June 13th, 2006, 07:26 PM
My reference to 640-800 was NOT the shutter speed, but you ISO. Raising your ISO will give you the ability to stop down your aperture (f/5.6) and keep your high shutter speed.
Daniel

Absolutely. This was taken at f9 (EXIF says f6.3 because of stacked TCs) and ISO 800 on the 20D. It managed 1/2000th at those settings. Actually this needed a slower shutter speed for more prop blur (had it on 800 by accident) but it still shows what you can do with cameras that have good high-ISO performance when combined with slow lenses:

http://photos.imageevent.com/sipphoto/events/aamairshow2005/20D10988.jpg

This was again shot at 400mm f9 but this time at ISO 400 and 1/800th on the 20D. This plane is traveling about 150mph and was perhaps 200 feet away when I took it. Note that even at f9 I got pretty good subject isolation and the focus is dead on.

http://photos.imageevent.com/sipphoto/samplepictures/Spit%20reduced.jpg

I find that, when I have out-of-focus AI servo shots on my 20D or 5D that the main cause is the dummy at the controls. When I keep that center focus point on the target, both cameras basically nail the focus every time. R/C aircraft are small, erratic and traveling pretty fast so they aren't easy targets either.

Lee Jay

Daniel Harrison
June 13th, 2006, 08:20 PM
Thanks for the examples Lee Jay, good to see you here! I have seen you around on my travles at dpreview. Welcome :-)

Phil Hawkins
June 14th, 2006, 11:21 AM
Ok, I'm wondering whether you're too fast deciding the 5d servo af is too slow. With a subject coming straight at you too slow focus and 2.8 would mean the sharp image is behind the point where you focused. I can't see on the pictures any area behind the subject that is sharper than other areas.

I hope the C.Fn-17 Custom Function is activated on the body (without it the Servo won't use all the invisible focus points).

I'm wondering whether the lens you used (which one?) has 2.8 as its max aperture because it's known most lenses at their max apertures don't deliver their sharpest pictures, should be stopped down. On my 70-200/4 the 4 gives only average sharpness; the 50 at 1.4 is so-and-so, starts good at 2.8. The 135/2 is sharper at 2.8 than at 2.0. The invisible points are used only if the lens is 2.8 or below (at 4 they don't work). So you'd need a very very good lens. All zooms start at 2.8, and I don't think you used a prime. Which makes me think you may have used a lens with max 2.8. So what lens did you use?

I wouldn't avoid using an iso1600 or 3200 for higher shutter speed, because in these daylight conditions the underexposed areas will be very small, I believe, maybe none at all. The noise at 1600-3200 comes mainly in the underexposed areas.

I think I've seen some tests with vehicles coming straight to 5d photographer at vehicular speeds (like 70kmph) and focus was following ok.

That's why I doubt 5d can't keep.

I'm using the 70-200 f/2.8 IS "L" lens. It's very strange, the lens gives me very sharp images when I'm panning right-to-left on a consistent distance plane, but when coming straight-on sharpness drops precipitously.

I will try using 5.6 or 6.0 aperture next time and see if thta doesn't help.

Phil

Phil Hawkins
June 14th, 2006, 11:24 AM
Absolutely. This was taken at f9 (EXIF says f6.3 because of stacked TCs) and ISO 800 on the 20D. It managed 1/2000th at those settings. Actually this needed a slower shutter speed for more prop blur (had it on 800 by accident) but it still shows what you can do with cameras that have good high-ISO performance when combined with slow lenses:

http://photos.imageevent.com/sipphoto/events/aamairshow2005/20D10988.jpg

This was again shot at 400mm f9 but this time at ISO 400 and 1/800th on the 20D. This plane is traveling about 150mph and was perhaps 200 feet away when I took it. Note that even at f9 I got pretty good subject isolation and the focus is dead on.

http://photos.imageevent.com/sipphoto/samplepictures/Spit%20reduced.jpg

I find that, when I have out-of-focus AI servo shots on my 20D or 5D that the main cause is the dummy at the controls. When I keep that center focus point on the target, both cameras basically nail the focus every time. R/C aircraft are small, erratic and traveling pretty fast so they aren't easy targets either.

Lee Jay

This is EXACTLY what I was looking for. GREAT shots!! I will stop down and increase ISO. Seems like that is the ticket. Going down there tonight, shooting some more, will report in.

Thanks again!!

Phil

Lee Jay Fingersh
June 14th, 2006, 11:52 AM
This is EXACTLY what I was looking for. GREAT shots!! I will stop down and increase ISO. Seems like that is the ticket. Going down there tonight, shooting some more, will report in.

Thanks again!!

Phil

Sure, but the main thing I was trying to point out is that the most important thing is to keep that active focus point accurately on the target. You also have to let AI servo work for at least a half second of so before you start shooting. If the focus point drifts off the target during that time, it will take AI servo a while to reacquire the target's velocity and accelearation.

Stopping down will help too because it will increase DOF over the wide-open condition and thereby leave more room for error.

Lee Jay

Gary Jean
June 14th, 2006, 12:20 PM
I find that sharpness improves using AI Servo with the 2nd and following shots. The 1st shot is often not well focused. You may be doing this already, but fire off a continuous series of shots.

I'm sure you already know that your 70-200 is plenty sharp wide open, but the bikes may be too close to get adequate DOF at f/2.8.

I may be missing the boat here, but I don't like to use Cf 17-1 to get expanded focus points. I am never sure which point has the focus. Shooting flying birds, I try to put that center point right on the eye. Only if I think I'll have plenty of DOF, I'll also try all focus points.

I hope I am not in the minority by having lots of throwaways when I'm shooting fast moving action. I don't fault the equipment. In my case it is lack of operator skill.

Kirk Darling
June 24th, 2006, 01:33 PM
I find that sharpness improves using AI Servo with the 2nd and following shots. The 1st shot is often not well focused. You may be doing this already, but fire off a continuous series of shots.

This is a known operational factor of AI Servo that Lee Jay was talking about, fully discussed in Canon's pamphlet (Mr Westfall's pamphlet, I think) "Getting the Most Out of Your EOS-1 Class Camera."

You have to place the focusing point on the subject and half depress the shutter release for about half a second before the system "acquires" the subject and begins tracking. If you just immediately fully depress the shutter release, it's still going to take half a second to acquire the subject, and the first shot or two in that half second will be out of focus.

Phil Hawkins
July 6th, 2006, 10:05 AM
This is a known operational factor of AI Servo that Lee Jay was talking about, fully discussed in Canon's pamphlet (Mr Westfall's pamphlet, I think) "Getting the Most Out of Your EOS-1 Class Camera."

You have to place the focusing point on the subject and half depress the shutter release for about half a second before the system "acquires" the subject and begins tracking. If you just immediately fully depress the shutter release, it's still going to take half a second to acquire the subject, and the first shot or two in that half second will be out of focus.

Thanks, Kirk, yes the 5D manual explains this concept, and I always depress for a second or two then fire. That's what has me so perplexed about the lack of sharpness. I tried the 5.6 aperture and saw an improvement, but in using the 2.8 I was trying to blur my background. Shooting motocross and TT racing is notoriously difficult on some tracks as the backgrounds are sometimes very bad (outhouses, cars, ad banners, people, etc.) Guess I just have to live with the limitations...

Phil

dhphoto
July 7th, 2006, 02:13 AM
Just to say, I've been forcing myself to use AI Servo in conjunction with the centre focus point and CF4,3 on my 20D and 5D and the number of OOF shots has dropped dramatically.

It takes a while to become confident in it but it certainly seems worthwhile

Roger Lambert
August 19th, 2006, 11:29 AM
In this thread, it was concluded that it was better to close down a fast lens to, say, f/5 to allow a deeper DOF so the AF could acquire better focus on a moving target. Do I understand that correctly? Also, a larger DOF results in more of the subject being in full focus, but the thread is about AF focusing, yes?)

But, all Canon cameras using Canon lenses of f/ 2.8 or better, automatically open up to f/ 2.8 to acquire AF focus, and then close down for the instant of the exposure to the requested f-stop.

This is additional light at f/ 2.8 or better puts the camera into an increased-accuracy focusing mode. But it also narrows the DOF, which may promote focusing error rate, especially with moving targets?

So, the recommendations in this thread to increase the aperture number up to , say, f5 will not actually have an effect on the DOF at the time of AF focusing of the camera, if the user is employing a lens that is f/ 2.8 or faster?

Would it actually be better to use a slower lens that never gets down to f/ 2.8 when trying to photograph faster moving objects?

Asher Kelman
August 19th, 2006, 11:53 AM
Roger,

There two functions:

1. Acquiring focus: the wide aperture, f 2.8. helps the lens to focus by gathering more light for the process. If the amount of light becomes sufficient for the particular contrast and detail of the subject, then the wide aperture only helps the quality of the focus.

2: The DOF here depends on the 5.6 f stop and the f 2.8 can be forgotten.

Asher

Roger Lambert
August 19th, 2006, 12:50 PM
Yes, I understand that difference, Asher.

Let's say you have a situation where the DOF at f/2.8, or 1.4, is quite narrow, and the camera is trying to acquire focus. The focus point is looking for a line of either vertical, horizontal, (or both) contrast within a very narrow field of depth at that aperture.

If it finds such a line, it will focus more accurately at f/ 2.8 than it would at a smaller aperture. But at f/ 2.8 it has less subject to choose from. May it not, in fact, latch onto the best thing it can, which might be a worse choice than if the DOF were broader?

Just because the camera acquires focus does not mean that it is acquiring focus where we want it to acquire focus.

Or is this this effect only likely to come into play, if at all, at close distances, and not at all, at anything past 50 feet or so?

On the other hand, I'm in the midst of a migraine - I probably shouldn't be posting at all! :D :(

Ray West
August 19th, 2006, 01:27 PM
Like Asher said, but what else goes on? Lets do some crude sums - nothing precise, but just to try and get a feel of what is happening.

say the subject is moving at 60mph across your field of view, at a distance of, say 100 ft.

60mph = 88 ft per second. If you have a lens of about 58mm focal length, and a sensor of say 20mm wide, then in one second you would get a motion blur of about one full sensor width. For a 1/500th second exposure, you would get a blur of 1/500 sensor width, or for an average sort of pixel array of 4000 pixels wide, it would be an 8 pixel blur. There is a delay, when you press the button, before the actual image is captured - the mirror goes up, the lens is stopped down, both mechanical movements which take finite time - how many pixels? During that time, there is no af.

So, you pan with the object, but do you? do you follow through, like shooting clays/skeet. or do you stop when the button is pressed? With a complex highly patterned subject, the af point will skip around, I bet. There will be some combinations of pattern/movement that will not work. A bit like some of the pro bbc wildlife video I've seen - sea otters, with image stabilization. The is locked onto the wave pattern, not the otter bobbing up and down. Interesting to see how they corrected that in post. - they didn't...


To practice the follow through, rig up a simple pendulum or something - it costs nothing other than time.

And I didn't even mention the vibration caused by the mirror/shutter slapping about, and other camera movements.

Best wishes,

Ray