# AF sharpness problems (part 2)

A couple of weeks ago I promised to report on an apparent problem concerning inaccurate focussing when an off-centre AF sensor is selected. I have now completed my first tests on this and can reveal the results...

The tests indicate that there may well be a real and genuine problem but before drawing that conclusion it is necessary to warn you that this column contains more mathematics than would normally be found in What Digital Camera!

In the case of the dSLR that was the first to undergo MTF testing using a variety of different focussing areas, some of the results obtained were as follows; 0.345, 0.333, 0.306, 0.344, 0.336, 0.338, 0.360 and 0.358 cycles-per-pixel. Clearly, one of the results (the third) differs significantly from the rest.

Although the result is different that doesn't mean it is "wrong". To illustrate this point consider flipping a perfectly-balanced coin: you have a 50:50 chance of throwing a head each time and that remains the same even if you have already thrown five heads in a row. But that doesn't mean that you can't keep throwing heads indefinitely; it simply means the chances of doing so get progressively less and less. In other words, an out-of-the-usual result could be nothing more than a statistical variation.

In this case, however, the MTF values are not isolated values: the eight figures given are actually just one line from a matrix of results that records the MTF figure for each AF area at every aperture of the lens used. The full matrix looks like this (with the problem area's values highlighted in yellow):

Matrix of MTF results of a given lens-and-camera combination for a variety of AF sensor areas at all aperture settings.

As would be expected for a focussing problem, the deviation of the problem area's MTF value from the mean of the other values generally decreases as the aperture is closed. The data do not follow this pattern exactly but from f/11 onwards the MTF of the problem area is within an acceptable range of the mean (defined by two standard deviations) meaning that the lens is effectively sharp across all AF areas when stopped down to f/11 and beyond.

In short, the likelihood is that this particular camera and lens combination may exhibit poor sharpness when the top-left AF sensor area is used. I am deliberately not naming the camera and lens used because I need to do a lot more checking into this effect but if anybody else has observed this effect then I would love to hear about it.