Predictive AF - 10D vs Mark II

[Rob Larsen]

Can anyone comment on differences in predictive AF between the 10D and Mark II? I know that the pro series cameras have better focus system performance tolerances and more focus points, but I'm looking for some comments via practical experience.

Lately I've been shooting a lot of track with my 10D and 70-200IS f2.8 with and without 1.4TC. Filling a focus point spot on the eyes of an approaching runner and using high shutter speeds is still turning out slightly OOF images more often than not.

I have done some measurebating and ensured the AF is accurate for stationary objects. This issue only crops up when using shallow DOF on subjects moving quickly toward the camera.

[defordphoto]

Having just made the move from the 10D to the MKII I can testify that the AI Servo tracking is much more accurate and results in significantly less OOFs. I use the 70-200L f/2.8 IS, the 100-400L IS, the 17-40L and the 400L f/5.6 with the MKII primarily and am extremely pleased with the results.

Some people are reporting some low-light AF issues, but I have yet to run into those. I've done some limited testing, but have been happy--so far--with those results too. All AF camera will have low-light issues of some sort. Some more than others.

We have a photographer here who primarily shoots model airplanes and has had some AF problems on cloudy days.

[Pekka]

10D tracking speed is slower than Mark II (Mark II has two CPU's for AF), the speed is also lens dependent:

See http://www.canon.com/camera-museum/tech/report/200304/200304.html (frame from http://www.canon.com/camera-museum/tech/report/f_index.html )

http://www.canon.com/camera-museum/tech/report/200304/img/zu5.gif

[Rob Larsen]

As I wait for my Mark II to be shipped, it is very comforting to hear first hand experience and see quantitative data on the AF improvements since that was my primary motivation for upgrading. Thank you for the comments.

The low light concerns are disturbing though... hopefully "low light" refers to "very low light" as I also shoot indoor sports. I wonder if that is related to some of the AF issues that have been cited on other forums?

[KennyG]

Rob, even with the pro bodies you have to allow time for AI Servo to settle. It is much faster on the 1D and 1D MK-II than the 10D (way faster), but you still have to get focus a second or so before you take the shot. You can see this happen with burst mode where the first shot is sometimes OOF and subsequent shots are right on target.

[Pekka]

The low light concerns are disturbing though... hopefully "low light" refers to "very low light" as I also shoot indoor sports. I wonder if that is related to some of the AF issues that have been cited on other forums?

Hi Rob,

The low light AF or Mark II is tested and explained thoroughly by Chuck Westfall in http://www.robgalbraith.com/ubbthreads/showflat.php?Cat=&Number=245591&page=12&view=colla psed&sb=5&o=&fpart=1

It seems the "issue" was that people expect in low light to have similar ultrafast AF they have in normal light. Why it is not so is explained well in paragraph:

Conclusions

Even under the most ideal shooting conditions such as the use of a fast wide-angle lens with a contrasty subject and the presence of a 550EX AF Assist beam with One-Shot AF and center-point focusing, it may take an EOS camera almost a full second to lock focus on the subject in extremely low light. This is especially true if the lens is starting from its closest focus setting and attempting to autofocus a distant subject. This is simple physics: it takes X amount of time vs. Y amount of light vs. Z amount of contrast for the AF sensor to accumulate enough information to function. That’s the way it is, and firmware updates will not change it. Therefore, if fast response is more important than accurate AF, consider the use of manual focus instead. If accurate AF is more important than fast response, have the patience to wait for the AF system to work. If you can’t get a lock in One-Shot AF even with an AF Assist beam, consider using manual focus to get "in the ballpark" and then let the AF take over to fine-tune if necessary.

I recently made also one observation from above explanation which I have noted to be accurate by some limited tests: in low light you'll get more accurate focus if camera is steady. If camera shakes the chance that focus point is off increases. This happens because AF sensor needs to gather light, and if light is not steady then the end information is gets is wrong - and when focus lock is based on wrong data you'll end up with front or backfocus. It all brings us down to the way old D30 needed to be shot in low light: if there is no clear focus area find a steady contrasty spot on subject plane, focus on that and recompose. These days people rely on camera to do it all, but there are times you'll need to show it the right direction no matter how advanced it is :)

Term "low light" is also very important. If you have 50/1.4, put it to Av, ISO 100 and 1.4, then find a level of light which gives you 2" exposure. It is surprising how bright this level of light is to adjusted human eye. Note that AF will work in less overall light too if there are even slight brighter contasts to detect, but AF will not be as accurate in most cases. To know or "sense" the light levels which are near the EV0 limit of 1D Mark II focus system helps you to troubleshoot AF errors and be aware why they may happen.

[John57]

My wife used our 1D Mk 2 and 70-200 2.8 today to photograph a charity fun run and I went along to see how the 10D tracked the runners in comparison on AI Servo. I had our 35-350 on the 10D. Both lenses were used at 200mm focal length and both cameras were set up roughly the same.

The results from the 1D Mk2 were far better. On the 10D on average only 1 in 4 images was what I would call crisply in focus with the others being ever so slightly out. On the 1D Mk 2 it was probably 9 out of 10 being OK. Static shots produced similar images - it was merely the tracking ones that really showed the differences.

Aside from the differences in lenses, etc. it was very clear from the results that the 1d Mk 2 is vastly superior in this area to the 10D..... but then if it wasn't I would be a bit disappointed in view of the cost!

[defordphoto]

Great followup, Pekka. That is also what I have found too. So far I have not run into problems focusing under any conditions with this camera.

[CyberDyneSystems]

..lol,. I just spent an afternoon tryong to use the 100-400mm IS on the 10D tracking flying Terns... I had gotten o used to the MkII's flawless AF.. I forgot that the 10D simply CAN'T DO THIS!

There is really no comparison.. The 10D could muster up a better showing withthe faster larger 70-200mm IS.. but still no comparison.. but with the 10D and 100-400mm.. I literally got ZERO shots.. whith the MkII I would expext to get about 90% perfect as far as the camera goes.. maybe more.

[MarkH]

My wife used our 1D Mk 2 and 70-200 2.8 today to photograph a charity fun run and I went along to see how the 10D tracked the runners in comparison on AI Servo. I had our 35-350 on the 10D. Both lenses were used at 200mm focal length and both cameras were set up roughly the same.

The results from the 1D Mk2 were far better. On the 10D on average only 1 in 4 images was what I would call crisply in focus with the others being ever so slightly out. On the 1D Mk 2 it was probably 9 out of 10 being OK. Static shots produced similar images - it was merely the tracking ones that really showed the differences.

Aside from the differences in lenses, etc. it was very clear from the results that the 1d Mk 2 is vastly superior in this area to the 10D..... but then if it wasn't I would be a bit disappointed in view of the cost!

I suspect the 1D MkII would be much better here, but this test shows nothing. Maybe if you had swapped lenses between the camera for a bunch of shots then you would have a good test.

Any f2.8L lens should be better at fast and accurate focusing than a f5.6 lens (even if it is a f5.6L)

[defordphoto]

John57:

I shot the 10D for a year from motorsports to soccer and never had any tracking problems with it. Granted it is not as fast as the MKII (duh!), but it is by no means a slouch. I shot 18,000 photos with it and 95% (or better) of them were nicely in focus using AI Servo. I got a cover and 4 two-page spreads using the 10D and it's an awesome camera. Yes, the MKII is much more camera (duh!), but that does make the 10D any less of a camera.

[John57]

I was not trying to rubbish the 10D... far from it - I love it but merely wrote what happened on one day; yesterday. I don't normally use the camera on AI servo and everything else I have done with this camera has been excellent.

The 70-200 lens is superior to the 35-350 and so it should be. As for changing lenses over to make it a better test thats fine - and a fair comment - but I accompanied my wife who was the official photographer at the event which meant I got the kit that was left over to play with. I didn't go out with the idea of 'testing' anything but thought it would be a good idea just to see what happened on the day.

[KBMphotography.com]

I echo all that RFM has said. I too moved form the 10D to 1Dmk2 recently after a year with the 10D. The 10D was slower - but otherwise fine when bolted to the 70-200 2.8 L IS.

This was the main reason I updated to the 1D.

[rebenstein]

Last night was the first time I got to check out the AI servo mode in action. I shot a basketball game indoor with relatively poor lighting (ISO - 800, f/2.8, 1/125). I came from the D60 so, I did not expect too much from the AI mode which was too slow to be effective. I shot in both one shot and AI to compare (practice games). The AI worked extremely well - much better than I had hoped for. When I got an initial focus on a player, I could track them as they came up court. There were far more keepers than with one shot.

[Mike H]

Last night was the first time I got to check out the AI servo mode in action ... I could track them as they came up court. There were far more keepers than with one shot.

Which camera were you using? Thanks.

Mike H

[rebenstein]

Which camera were you using? Thanks.

Mike H

1DMKII

[Canuck]

Is it not proof that you can get some real quality out of the 10D, being that Jim got published with it? It isn't a 1D M2, but look at the cost difference and also we are a year on in technology. That is a long time in the world of constantly evolving technology. It also isn't fair to comapre a 10D to even the original 1D. They are totally aimed at different markets. It is kinda like the 10D/300D comparison. It really can't be don fairly...again different markets. Again, the 300D isnt a 10D, and a 10D isn't a 1DM2, and so forth.

[Rob Larsen]

Rob, even with the pro bodies you have to allow time for AI Servo to settle. It is much faster on the 1D and 1D MK-II than the 10D (way faster), but you still have to get focus a second or so before you take the shot. You can see this happen with burst mode where the first shot is sometimes OOF and subsequent shots are right on target.

I got burned on that a few times in the beginning. Sometimes I get assignments like "capture the athletes in lanes 1, 2, 5, & 8." This is all supposed to occur during a 100 yard hurtle. Initially I tried to quickly pan to each lane -- only later to find the AI had not settled. Now when I get tough assignments like that I will try to group as best possible and crop to individual frames later. Prefocusing is something else I've grown to depend on in sprints.

Where I became frustrated enough to spring for the 1DMII was after tracking distance runners on the long stretch with Servo, fast shutter, f2.8, and focus point locked on face for several seconds/frames and still getting a high percentage of slightly OOF frames. (Note, Servo has worked well for me when tracking slower approaching subjects or using greater DOF.)

Here is one of the events I've found challenging to nail with a high success rate. A little deeper than average DOF helped on this shot.

http://www.roblarsen.com/galleries/people/main/hurtler640_std.jpg


It's great to hear everyone's encouraging words of experience that there is a difference in AF performance with the Mark II. Especially since I just got confirmation that the unit shipped out today! I can't wait for it to arrive. For some reason I find myself frequently needing to track it's slow progress from NY to CO... :D

[Rob Larsen]

The low light concerns are disturbing though... hopefully "low light" refers to "very low light" as I also shoot indoor sports. I wonder if that is related to some of the AF issues that have been cited on other forums?

Hi Rob,

The low light AF or Mark II is tested and explained thoroughly by Chuck Westfall in http://www.robgalbraith.com/ubbthreads/showflat.php?Cat=&Number=245591&page=12&view=colla psed&sb=5&o=&fpart=1

....


Pekka, thank you for the very informative followup. I feel better now about how "low light" is being defined"

The EV0 test is a great idea to help train your eye... I will check that out once the camera arrives. As for looking for a high contrast focal subject, years of shooting habit with a EOS 620 still have me doing that even when not needed with the 10D (or most modern AF I suspect). The 45 AF points in the 1DMII hopefully will break me of that. :)

[Pekka]

http://www.roblarsen.com/galleries/people/main/hurtler640_std.jpg



For images like this you could use the central AF point with AI Servo and take a short burst with 8fps - if you aim/start the focus slightly before shooting you'll surely get 100% in focus. If the right foot is up high may it may be better to use some upper focus point with AF point expansion on runners chest or face or you could just use slower focus tracking C.fn. I think you can get 4-5 runners easily in 15 sec time with Mark II :)

[slin100]

One thing to keep in mind about AI Servo on the 10D is that the first shot is not guaranteed to be in focus. The 10D operates in a release priority mode for the first shot, meaning that the camera prioritizes the release of the shutter over focus. Subsequent shots taken in continuous drive mode with the shutter button held down are taken under focus priority.

Details about the AI Servo are provided in this article (http://photonotes.org/other/ai-servo.html).

I don't have the reference but I think AI Servo on the 1D Mk II operates a little differently. Apart from AF being much faster, it also guarantees a much higher probability that the first shot will be in focus.

IOW, AI Servo on the 10D is a very different beast than on the 1D. For quick moving subjects, it would probably make sense to switch to continuous drive mode and take at least two shots in succession.