Focus Accuracy: Center-point v. Side-point

[bmpress]

Hi Bob,

While reading user reviews of one of Tamron's lenses there were several comments saying that focus accuracy was off when using a left or right focus point.

Could this be a lens issue or one of camera accuracy? Can you please explain the dynamics of this issue.

Thanks,

Barry

[Bob Atkins]

I'm not sure what would be happening in that case.

The Canon AF system works by looking at the phase sensitive AF detectors in the camera, calculating the degree and direction of defocusing and then generating a signal that drives the lens to the predicted focus point.

I could understand if AF was off with all focus points - that would be a lens that was out of calibration. With many camera bodies the center AF zone is (or can be) higher precision, especially with fast lenses, but that doesn't mean that the other focus points aren't accurate. Focus accuracy for all zones is, at worst, within the DOF of the lens at he set focal length and aperture, at worst. Also some cameras have cross sensors in the center but only linear sensors elsewhere. However that should really only affect whether or not the camera could get a focus lock, not really the accuracy of focus once focus lock has been obtained.

With a properly aligned lens it's hard to see how the lens could focus properly with the center zones but not edge zones. The camera measures focus and generates the focus signal which it supplies to the lens. The focusing signal from the camera shouldn't depend on which AF zone is in use unless the camera has a problem.

An electronic problem with the lens is unlikely because the lens doesn't know which AF zone the camera used for focusing.

Could some sort of optical defect cause the phase sensitive AF detectors to generate an incorrect focus signal? I suppose it's possible, but Canon don't release any details of their AF detection system. In fact they don't even share their AF lens protocol with anyone. the 3rd party lens makers have to reverse engineer the system.

Assuming the effect actually exists, you'd need to try different lenses on the same camera body and use different camera bodies on a lens exhibiting the problem to try to track down exactly which component was causing the AF problem.

[Frank Kolwicz]

Under what kind of conditions were the focus errors found? General photography or frame-filling test targets? Real world subjects aren't as easy to focus on as flat walls.

I don't know if this is any help, but, I've found that for bird photography the area I need to have in focus, the head and especially the eye, are often similar to or smaller than the square defining any one of the focus sensors (and the sensor is actually bigger than the square). Under those conditions focus is unreliable.

What I think is likely to happen is that some part of the sensor picks up some part of or color contrast feature of the bird outside the small area that's important to me and so, under those conditions, I'm lucky to get 50% sharp images and the smaller the bird's head in the frame, the worse that gets. When I'm lucky enough to get frame-filling images, focus accuracy is not a concern and it goes up to around 90% or better no matter which sensor element is selected.

It's possible that the auxiliary sensor elements, with their different geometry, are even more unreliable when the area that needs to be sharp is relatively small. Of course, if you're photographing a planar subject that fills the frame and this happens, you probably have a system error.