>So you did not actually test the method I proposed yourself?
No - not yours. Many others though, including the commercial units I've reviewed.
>The whole idea of leaving *only* one dot visible when focusing >is to eliminate the typical errors coming from targets placed >at 45 degrees. Indeed, the rest of the target is here pure >white and cannot contribute to the phase detection of the >focusing system. This is easily confirmed by the total >inability of the camera to focus if the collimator is not on >the dot.
In my experience, cameras can sometimes assume focus, for a variety of reasons, when no normally obvious focus target is present. For example, in inconsistent lighting, the camera can sometimes read a faint shadow edge sufficiently well to attempt focus and allow the shutter to fire even when no obvious focus target is present on an ostensibly clean sheet of paper.
I'm not sure what you mean by the reference to a collimator. The hole in the top sheet which exposes the target is not a collimator - it's a mask. A collimator is particle modification device which is used to realign (collimate) a beam of particles (e.g., photons). A camera lens is a kind of collimator. Are you referring to the lens being used? If so, I'm missing your meaning.
An edge of the mask can also be a focus target as often as the target print itself. A flush or unflush far mask edge being grabbed hard (instead of the target because it's too small in the frame for the selected AF sensor) by the AF system, coupled with a lens that tends to focus slightly ahead, coupled with a camera AF system which focuses slightly ahead, may give a test result which looks spot on. There are all sorts of combinations of these three factors alone, which when coupled with 45 degree testing in particular, give photographers absolute fits of AF fine tune angst. To get a sense of what the camera is actually doing - not matter what testing metholdology is being used - the tester has to take a hundred photos or more just to chart the variability inherent in the particular camera and lens combination being tested.
I'm not trying to shoot down your efforts. I'm only injecting the important variables that I think should be fully accounted for in any AF focus testing system. The simplicity of your approach is genuinely attractive, but I don't see any way for it to be used, as-is, to produce accurate test results on a consistent basis. Then again, a 100-200 shot test series might produce enough variability data to make a judgement about AF fine tune. I just don't have the time to give your proposed test method a serious try, but I hope that someone else can.
What I'm suggesting though is that any testing at a 45 degree angle is difficult, at best, to do accurately. The angle in and of itself is a compromise solely for simplicity of test set up, but it materially challenges any camera's AF system in an inappropriate way. It's only ever been promoted by the producers of hobbyist AF testing kits because it's simple and small enough to fit into a manageable retail package. Angled test kits don't exist for any other reason. Parallel/plane target focus testing at various distances is generally a much more effectively accurate AF testing method, but it still requires more room and greater measurement accuracy to establish a parallel/plane relationship between camera sensor and target. It also generally produces more reliable results. Using your single target point by itself, in a parallel/plane test set up, would determine how effectively a cross-type sensor can accurately lock on to it. Maybe you've designed a target that is better than many of the others I've seen (noting too that some others still, have proven to be very good test targets no doubt). That last bit would be a real coup for you - seriously.
Once again, I'm not in any way criticising your goal of accurate AF testing and calibration, just any methodology which needlessly incorporates a 45 degree angle setup. To my way of thinking, the compromises and the test results are unacceptable.