70-200 F4L IS - Slow focus

futureal33

futureal33

Messages
2,390
Name
Nick
Edit My Images
Yes
Im just wondering if this is normal?

I recently got a 70-200 F4L IS and it is painfully slow to focus on my 5DMk2. I was at the zoo yesterday and missed the shot on quite a few occasions because the lens was still focusing.
In one case, it took about 5 presses of the shutter button before it focussed - as it kept moving a few mm each time, but didnt "lock" focus until the 5th attempt.

Using a mixture of "One-Shot" and "Ai-Servo" modes

Ps - it is taking around 3-4 seconds to focus.
Pps - I tried alternating the 1.2m - infinity and 3m - infinity and it made next to no difference
 
Nope, what are you comparing it to? I hate AI-Servo personally, tends to hunt around unnecessarily regardless of lens I use. I think this is where 7d > 5dii.
 
Comparing it against all my other lenses.
This is my first L lens, and was expecting the UltraSonic motor to be a lot better, im just trying to work out if there's a problem as it is very very slow to focus.
As a a rough comparison, my Tamron 28-75mm 2.8 (which is supposed to be very slow focus according to reviews) is around 2-3x faster than it, from infinity to macro.

The 70-200 is taking about 5 seconds to go right from infinity to macro...
 
On the 5D2 the outer focus points are linear sensitive only, not cross sensitive. That means that some outer AF points are blind to vertical lines of contrast and others are blind to horizontal lines of contrast. That lowers your chances of a focus lock in some situations, where the centre point would probably nail it.

You need to make sure you aim at an area with adequate contrast. Fur is a notoriously poor target for an AF system. Make sure you position your AF point on a target with good contrast. Usually the nearest eye is clear favourite, but if that's in shadow, for example, you might need to locate an alternate spot.

Unless you need the tracking capabilities of AI Servo it would be better to stick to One Shot if you are having problems. Unlike AI Servo, which is re-assessing focus constantly, One Shot takes a fraction longer to gather and analyse focus data, which can gain you a lock in tricky situations where AI Servo may fail altogether.

With centre point AF and a slow 100-400 zoom lens my 5D2 can easily track BIF. I've used it at the zoo a few times and don't recall having problems with AF, or at least none that would come as a surprise, and usually solvable. The 70-200/4 should perform a little better.
 
Tim,

I tried for about 45 minutes to track BIF at the zoo this weekend. Using centre AF and Ai Servo mode. It was hopeless. In fact it was so hopeless that it didnt even find focus then lose it, it never found focus full stop! In the end I went to manual focus, which was actually slightly more successful but having never done it before, the results werent great!

By comparison, my 40D + 55-250mm IS lens was considerably better, easily tracked BIF and focussed much faster.

Im by no means a good BIF photographer - but in this case I think it was more down to the camera/lens than me as the guy behind the camera, as it wouldnt even find focus on where I was aiming.
Im also wondering if there's anything in the settings I could have messed up?

For isntance, AF selection point expansion - I tried this on both "enabled" and "disabled" - didnt seem to make any difference.

Any suggestions as to other possible settings that could be issue?
 
It is broken....

I has the F4 IS ands it is very quick to focus...

This was true on my 450D and 7D in server and one shot....
 
It does sound like your lens may have a problem, but there is one more thing to try in terms of technique....

When the subject is sufficiently far OOF that the camera can't determine in which direction to start focusing to lock on it will always rack the focus closer before reversing and trying to search farther away. If you don't engage the distance limiter, or even if you do, that can waste precious seconds while the camera tries to figure out what is going on. For an approaching bird, it may even be that with the focus heading in the wrong direction it may never catch up. One solution is to manually pre-focus at a distance a little further away, maybe even on the background if it is not too far away. That will make it easier for you to locate the bird in the viewfinder and give the camera an easier job too.

Another point not to be ignored - when using AI servo to track a moving subject you need to track for up to a second to let the camera get tuned in to the speed of the subject and predict where it will be by the time the shutter is released. This is not the kind of thing where you can just snap away. It is also of paramount importance to make sure your AF point stays where it should be. Ove the years, whenever I've analysed BIF shots that fell below my expectations I've discovered that with the AF point placed where I had it it would be a miracle for the subject to be focused correctly. Use DPP or Zoombrowser etc. to check on where your AF points are. You may be surprised.

Here's an example from the zoo with my 5D2 and 100-400 at 190mm, 1/1000, f/8, 400 ISO. Sharpening is at Lightroom defaults.

20100317_121429_2022_LR.jpg



Here's another, this time from a local park, with a faster bird coming head on, again with sharpening at Lightroom defaults. EXIF is 400mm, 1/800, f/5.6, 400 ISO....

20100215_102723_1504_LR.jpg



Last one - 5D2 and 100-400 at 400mm, 1/1600, f/5.6, 400 ISO, sharpening at Lightroom defaults....

20110325_104045_8032_LR.jpg



Certainly the camera can do it and the 70-200 certainly should be able to. If you're doing OK with the 40D and 55-250 then it seems your technique is OK. That doesn't leave many places to look for the problem.
 
Just to be completely clear, they were taken with Ai-Servo mode?

Thank you for the advice, I will go down to the canal tonight and hopefully try to catch some ducks as they come into land, similar to your shot using the technique you mention above.

Another thing - you mention that the camera needs to "learn" the speed of the subject etc.. that would suggest the AI Servo mode is intelligent? I thought that it simply kept up focus with whatever it was aimed on and then the shutter fired when you press the button down. From what you are saying it sounds like the camera makes a little alteration just prior to the shutter release which calculates the difference between time it takes for the shutter to fire, and the last moment you see through the viewfinder?

In the last photo, of the dog running towards you, can you just confirm if this is what you did.
Ai-Servo mode, focus on the dogs face as he is quite a distance to you. He begins running towards you, you keep the AF button pressed to track him, making minor camera position adjustments to keep his nose aligned with the central focus point...then at the desired distance you click the shutter?

Thanks for the help here, I will put this to test tonight
 
:agree: everything Tim said.

AF on my 70-200L 4 IS and 5D2 is very fast, like fractions of a second. Several seconds is crazy - even live view AF is quicker than that.

Servo is indeed smart. It uses all the focus points, and hidden ones (AF expansion) too, to get a fix on the subject and where it thinks it's going to be when the shutter opens. Some cameras are better at this than others, and it's not the 5D2's strongest suit ;)

In that sense, the lens is only as good as the camera - all the lens does is respond to instruction. To reduce the influence of the camera, try it on a static target and compare with your other lenses to try and isolate the problem.
 
Last edited:
AI Servo is "predictive". It assumes a steady speed of approach (or retreat) by the subject and needs time to work out what the speed is and to anticipate shutter lag and have the focus where it should be at the moment when the exposure takes place. A minimum of half a second is advised, even on a 1D camera, but a full second is probably the safer bet. If you want to track for five seconds then that's fine too, although offers no advantage.

Given the above, AI Servo does not work so well with random changes in the rate of approach or retreat of the subject. An approaching vehicle is perfect. A fluttering bird darting all over the place is a bit of a disaster.

With the dog, yes - I tracked her from a distance as she approached me and probably fired a short burst or a few individual frames as she got within a sensible striking distance. I would never try to snatch a shot such as the dog - you have to track the motion first before releasing the shutter. If you don't then I think success would be very much down to random chance.

Here is the AF point placement and camera settings for the shot. I don't know whether I had expansion points enabled. I certainly do use them but not always, and I can't spot any clues in the EXIF either way.

20110801_142723_.JPG


The tennis ball is key to the success of AF tracking with a shot like this. Without it there really is insufficient contrast for the camera to stand much chance of locking on to anything. I also try to pick my lighting angles so my subject is at least partially front lit. Shooting this into the shadow side would also be a fruitless exercise. Black dogs with no markings are not the easiest subjects to track at high speed.


One more point - The fewer AF points you can get away with the better. When you start adding expansion points into the equation, or going for automatic AF point selection (all points active), there is a lot more focus data being generated and requiring analysis and more decision making needed by the camera. The fewer choices you give the camera the more assured the result. I try to shoot with a single active AF point if I can. I progressively increase the number of points in use only if I'm having problem keeping one AF point where it belongs. There are other consideratoins too, like what else in the scene is competing with the subject for attention from the AF. When you're shooting BIF against blue skies it's hard to get it wrong, unless you have a huge bird and end up focused on a wingtip instead of the eye, but when you have a busy background, such as woodland, you really need to help the camera out by using the fewest number of AF points possible and placing it (them) bang on target.

If you are using AI Servo and all points active (a mode I practically never use, regardless of camera body) then remember that you must first acquire and track the subject with the centre point before the camera knows what the subject is and can pass off AF control to peripheral points. In this respect it works completely differently from One Shot with all points active.
 
Last edited:
Would I be right in saying that if you zoomed to 100% or there abouts on that photo, it wouldnt be sharp?

Because I have one I took of a vulture approaching, which looks fine at web size and even full screen size, but the moment you zoom in, you can see that it is OOF

The shutter speed was high enough that it should have been sharp (over 1/1000th IIRC) but its just missed focus I think - or couldnt keep up either way!
 
HoppyUK said:
AF on my 70-200L 4 IS and 5D2 is very fast, like fractions of a second.
Click to expand...

Mine too.


futureal33 said:
By comparison, my 40D + 55-250mm IS lens was considerably better, easily tracked BIF and focussed much faster.
Click to expand...

So sounds like your technique isn't that bad to cause such slow focusing! If it were me, I'd double check the contacts are clean and the lens seated in the mount properly (if you've not done it already), then check the focus speed on a high contrast static target (manual focus infinity, point at target a few meters away and see how long it takes to lock on). Then if that seems fine, I'd test servo mode on moving cars, because they are much bigger/nearer/high-contrast and the 5dii center point should handle them just fine.
 
Here is a 100% crop of the dog. You can see some parts look sharp - the tennis ball and some whiskers, while other bits look soft, probably due to the water creating an odd effect in her fur....

20110325_104045_8032_LR-2.jpg



For my purposes I'd call that sharp enough. The vertical whiskers look sharp, meaning the focus was good. The horizontal whiskers are not sharp. There are signs of motion. Whether that is vertical camera shake or vertical blur of the dog I don't know, but there is a vertical blur of some sort. For normal viewing sizes I think it is not a problem.

Viewing a 5D2 file at 100% is like viewing a print almost 5' across. If you printed at that size you surely would not try to study the picture from 12" away. You should be 6' or more from the print (or your screen). Then it should look sharp enough to get the job done. Pixel peeping is great for seeing just how sharp your pixels are. When trying to judge the quality of your PHOTOGRAPH it can set you on the road to ruin.

It is actually a bit more complex than that, but rules of thumb like 1/(focal length x crop factor) for minimum shutter speed and normal DOF values do not apply when you view at 100%. Even shooting at 1/1000, that does not eliminate shake or blur. It reduces it. At normal viewing magnifications it may reduce it to invisible proportions. Blown up to 100% it may not.
 
Last edited:
You must log in or register to reply here.
Back
Top