Focus accuracy on new Tokina 11-20. Expected or defect?

Question

I seem to have a focus issue with my new Tokina 11-20. It's mounted on a Canon 1000D(XS)

The camera is on a tripod.

The lens is wide open F/2.8, at 11 mm. I manually focused on the T on the box using liveview for the "manual" shot.

Then I manually focused the lens to infinity and let the camera refocus using the center cross point, centered on the T. I also did a shot manually focusing to shortest distance and let the AF focus again; the result was identical to the previous one(still off).

I called the shop, they're willing to give me my money back, but they're out of stock. Basically I can't get a replacement any time soon (no competitors either).

So I guess my question boils down to this; Is this enough to be classified as a defect? Or is another lens just as likely to show the same inaccuracy(do I just deal with it?)?

The focus also seems to be very persistent, it "nails" it everytime, only 3 cm behind the subject. Is it possible to calibrate it?

Note: no focus microadjust in the Canon 1000D

Answer 1

It is within the realm of possibility that the lens is back focusing. To confirm this and measure exactly to what degree, though, the target and distance scales need to be properly aligned with the camera.

There are a couple of reasons why good alignment is needed to get a valid AF test measurement:

• If the camera is aimed at an angle to a target such as you have created then the distance from the camera to the "T" on the target is not precisely the same as the distance to the "0" point on the scale beneath it and to the side.
• Your AF system will attempt to focus on the area of greatest contrast within the entire active area of the AF point(s) selected. The area of sensitivity for most AF points is quite a bit larger than the indicator square that you see in the viewfinder. Modern AF systems often have overlapping areas of sensitivity between adjacent points. If the camera is not perpendicular to the target then the camera may be trying to focus on an area closer or further from the camera than the spot at which you think you are aiming it. For more, please see Which offers better results: FoCal or LensAlign Pro? and Are of cross-type focus points more accurate, or just faster?

If it turns out the lens is consistently missing in one direction with the 1000D, which lacks AF Micro Adjustment capability, you have a couple of options:

• Exchange the lens and hope the manufacturing tolerances of the next one matches up better with the manufacturing tolerances of your camera.
• Send both the lens and your camera to an authorized Tokina service center and let them adjust the lens to the camera.

Answer 2

The phase detection sensors used for viewfinder autofocus have limited resolution. Ultra wide angle lenses make everything appear very small and thus subjects are frequently beyond what the sensor can see clearly enough to accurately judge focus. Additionally these kinds of lenses provide a very large depth of field, even at f/2.8. The phase detection sensor just can't tell the difference between in focus and slightly out of focus anymore.

With my camera (a Nikon D7000) and my copy of the Tokina 11-20 at 11mm, I can see evidence of this resolution problem when using manual focus too. The camera offers a little indicator in the viewfinder whether or not the picture is focused. Once it indicates an in focus picture I can twist the focus ring a fair amount further without the camera telling me that things have gone out of focus.
A simple test setup similar to yours with the subject at 37cm distance (from sensor plane) revealed that the camera thinks the target is in focus anywhere between 0.3 to 0.4 meters focus distance as indicated on the lens distance scale.

It is possible that your camera was programmed to focus at the furthest point that is still measured as in focus.

Answer 3

With this lens mounted on a D5500 I have found that AF accuracy is a problem using the viewfinder ie. focusing using the AF sensor. In this case it front focuses and the severity of the eror is less at the 11mm end of the zoom range and is worst at the 29mm end. Almost the same results were found using a Nikon D3300 camera.

If the lens is manually focused using the in focus indicator in the viewfinder the result is the same and the amount of slop in the indicator from lit to unlit whilst rotating the focus ring is huge. If the lens is focused using live view the result is accurate and this is easy achieve with the touch screen on the D5500 being suitable to "pull focus". The focus results that I obtained were similar before and after the lens was returned to the Tokina agent for calibration without the camera. ie. when set up to factory specs it was still front focusing but not on "live view" which uses the main sensor to achieve focus.

Conclusion: This is a design issue for Tokina to address as it will not operate properly in viewfinder focus mode on two current models of Nikon DX cameras.