Mark III update from Rob Galbraith

real interesting

I wonder, do the film bodies (like the 1N or 1N-HS which I notice I still see people using) have similar focus issues?

Just a question or a thought: Why would a big compony put their name on the spell, especially in such a completive market. He mentioned that they have removed their technicians that tested with him –“In the wake of its publication, Canon USA abruptly ended our EOS-1D Mark III testing collaboration.” One never knows what politics are behind the scenes. I'm not a professional photographer, but I do work with people and if some one starts their report on how the Canon people are going about doing their busyness and then start to talk about his test results, I'm just wondering. I don't have any doubt that his test are professionally done, I'm just wondering what his big motive behind all of this is.

Unlike the vast majority of people who feel compelled to discuss the “AF issue” with 1D3 and/or 1Ds3 I actually use both cameras so lemme address the issue from my experience. I’ll concentrate on 1D3 in AI Servo AF with a longish lens (300/2.8 L IS, early model) simply because I haven’t done any high-fps “action” shooting with my 1Ds3.

First of all, I don’t believe that RG is completely wrong, the AF is not 100% consistent, but I do sense more than an ounce of exaggeration, perhaps even user error, on his part. Yes, the sub-mirror assembly had issues in some models (both my 1D3 bodies come from the “affected” S/N range) and the software could be better off the bat, but the issues were not nearly as serious, frequent or widespread as reported on his site or hyped (literally) elsewhere on the ‘net. Secondly, in my experience, the AF performance of 1D3 is much better than that of 1D2n (I’m in total disagreement with RG on this one) not to mention that 1D2n shots at 8.5fps only (a bit less than that in my experience, closer to 7.5-8) whereas the 1D3 gets the full 10 fps.

Canon loyally warns that AI Servo AF works best for subjects that approach/retreat from the camera at a constant rate (that’s why there are few known complaints from the Formula 1 photogs…) and that the first image in the series may be OOF but if you get the subject in focus with the center AF point, subsequent images will be OK. Indeed, each time I nail the center focus point on the first shot, I get a perfect series at 10 fps from the first (or second – see above) image in the series, even with over 50 images of a dog running at/away from me. Now, the dog is pretty much monochromatic, uniformly black - runners (favorite test subjects of RG) on the other hand are not uniformly colored, for instance the skin color differs from the uniform colors and then there is the starting number tag, which often is another color/pattern altogether. It is almost impossible to hold a focus point (esp. with a long lens) in the same (exactly the same) spot on a fast moving, bouncing, smallish target and the camera’s AF point (or points) can bounce to a different part of the subject (or a different runner in a bunch…) making it easy for the predictive AF software to re-qualify it as a different target, reassess its speed and direction and react accordingly. In my experience, one “bounce” like that causes the subsequent images to vary widely in focus. Yep, sometimes I get a really bad series but if I can nail the focus on a running dog, I surely can follow a runner. Or can’t I..? So, I chalk most OOF series as user (i.e. my) error.

The 1D3 can be set up differently in terms of and it is my belief, that many reported AF issues stem from the incorrect setup (plain vanilla out of the box) or from a total misunderstanding of the applicable CFn settings. Canon has not done a great job explaining this, BTW. Also, some lenses (most notably the 70-200/2.8 L, the only lens mentioned explicitly by Canon in the manual) can focus properly only with the center AF point if used with an extender yet I see lots of “expansion points” used for pretty much all lenses. FWIW – here are my 1D3 settings: All NR functions off; Neutral picture style if the light is bright; Fn. III-2 (Servo tracking sensitivity) set to slow; Fn. III-3 (Servo focus 1st/2nd image priority) set to 0; Fn. III-4 set to 1 (continuous AF track priority); Fn III-5 (focus search) zero; Fn. III-8 set to 1 (one expansion point.)

I do agree with RG that the AF performance varies considerably in different lighting conditions, esp. with changing contrast and direction of light but that shouldn’t surprise anyone who has taken more than five pictures in his lifetime: light meters fail in many situations so do AF sensors. Where the both - 1D3 and 1Ds3 – cameras fall flat on their faces is AF performance in backlit, low contrast conditions: 40D is an order of magnitude better in this respect!

In terms of AF performance I get a much larger percentage of keepers from 1D3 (and 1Ds3) than from any AF camera I have ever used, including 1D2n. The big question is whether the 1D3’s AF performance can be made more consistent with a subject bouncing in and out the focus tracking “scope” in varied intervals and with varied speeds. I hope so; it would be nice to have a camera that will automagically focus and expose “en punto” each and every time at high fps shooting. Realistically, I’d expect some incremental fixes and a better AF performance in the backlit situations (especially on 1Ds3) from the next round of firmware upgrades. I also expect Canon to extend the warranty on 1D3 and 1Ds3 in the US to 3 years. In the realm of “wishful thinking” I’d like Canon to replace all existing 1D3 (and 1Ds3) cameras with a hardware-fixed/redesigned upgrade to remove any lingering doubts and put this issue to rest. Chances of that happening? Snowballs in hell…

Michael,

Do you see problems with chest-focus, when hands pump in and out of that area? I don't understand why RG didn't focus on the faces (I'm not a sport shooter, just hobbyist).

At any rate, Canon's response has even me convinced that I'll need to look at Noink for when my 1Ds dies. That's sad. Canon must now fight to regain trust.

Lad:

focus on the faces
The head is small; the chest is bigger so it is easier to focus. With faster lenses (f/2.8-4) DOF is thin (esp. when the magnification increases) but should easily cover the entire silhouette and that’s the idea here.

look at Nikon
Well, Nikon has nothing to offer (perhaps “yet”, but still) to the shooters who are likely to experience AF issues with the 1D3, so this is a moot point.

I haven’t lost “faith” in Canon (1Ds3 is the best DSLR camera ever made, period) but I’m puzzled by their handling of that debacle. From the US perspective, it is a major PR disaster with no happy end in sight (i.e. Canon didn’t even attempt to spin it in public - bad, bad, very bad thing in the US.) I see two reasons for that: lawyers and “the Japanese art of covering shit with silence.” The former is fairly easy: any admission of faults in 1D3/1Ds3, other than the tepid "press relases", would have triggered a slew of lawsuits (after all this is our national pastime…), the Uniform Commercial Code violations, etc. The latter is the famous “losing face” thingy (I’m oversimplifying here): the Canon management was afraid to be disgraced in the eyes of their peers if they admitted any wrongdoing. Perhaps they woul dbe disgraced (or even fired...), but a full and open account of troubles, a recall and a fix would have preempted any lawsuits and we would all live much happier lives. Screw the lawyers and the lawmakers, eh?

>>One workaround suggested is to stop down to F5.6 if the light allows<<

But, the lens focuses wide open regardless...

I have owned (and still own) and used a Mkl, Mkll and Mklll in a sporting career spanning more than 30 years. I am not in a position to doubt Rob Galbraith's findings with the bodies he has used but, in my experience, the Mklll is by far the best of the three. I still use the Mkll as back-up from time to time but the focussing accuracy is far better on the Mklll - in my opinion.

It's worth noting that others have tested and compared the Mkll and Mklll and come to a completely opposite conclusion to Mr Galbraith.

I seldom shoot sports so I usually use one-shot focus mode; rarely AI. I cannot get AI on my 1D MkIII to work well, but one-shot works pretty much 100% of the time for me. Find a contrasty area, position an appropriate focus spot on it; wait for the beep; shoot.

Is anyone else finding this OK as I am?

Aren't we expecting too much from a camera autofocus? For all its faults it is still miles better than manual focus that is for sure!