Originally posted by southlander My quick'n'dirty method is to put two same strips of masking tape on the lens - one on the rotating focus ring and one on the fixed section of the barrel. Use a tripod so distance to subject remains perfectly static. Use an subject at desired test range. Focus with liveview. Check all looks ok with magnified liveview (and focus peaking if you have it). Mark a line across both pieces of tape. This is now the reference point for where the focus is spot-on. Now repeatedly focus using PDAF and observe the alignment of the marks on the tape. Work out whether an back focus or front focus is present (which can simply be gleaned from how the marks misalign). Start dialling in AF correction and retest until get fairly consistent results. Then move to some 10-15 metres distance from an object and retest. It may be necessary to compromise between two AF settings. Some of the hacking threads on the net suggests that lens firmware for the Pentax dslr system holds separate focus calibration data for two distances ('Near' and 'Far') which could provide a logical explanation of why the calibration may vary slightly between close and distant targets. In fact I think this selection between near and far for a given shot can be seen buried in the EXIF data of images.
What I found interesting from having a go with this method last weekend is the variability of individual PDAF attempts with my K-3. With an aligned mark on the lens, small variations in stopping positions upon multiple attempts are obvious and surprisingly frequent. Enough that when switching to magnified liveview, you can see that the acquired focus was not always optimal. The AF stopping point also looked to have some observable biases depending on whether you were starting at infinity or min. distance.
Liveview focus conversely was highly repeatable at where it stopped, but even then, just a touch more accuracy could be sometimes be extracted manually focussing using magnified viewing and focus peaking. Probably corresponding to the slop in the screw drive mechanism as the adjustment between liveview AF and what I judged to be optimal was no more than a 1mm or so of movement in the focus ring.
++++ This method obviously won't work for lenses that do not rotate the focus ring when autofocusing (I've just bought one such lens, so will need to use an alternate method for that lens). ++++
The bottom line take away I got from this exercise is that I should use liveview or even manually focus via focus peaking when I really really want critical focus (eg exposing at maximum apertures where shallow DOF won't hide small focus errors).
As an additional note, I got round to updating my K-3 firmware from v1.0 straight through to V1.21 last night. I hadn't experienced any of the mirror flop issues reported by others so had no pressing imperative to update. What I think I have seen though is some change to the AF correction required by the lens I happened to have on at the time. This might be due to factors not related to the firmware (eg different lighting - daylight vs artificial) but I will now need to evaluate this more carefully. I previously found I had quite a few lenses near the +10 mark, and the lens I was playing with last night suddenly looked to be much closer to nil adjustment. So maybe the firmware updates along the way have updated some of the PDAF algorithms (in a good way).
I will have a play about on the coming weekend in daylight and report back - middle of winter here so no free time at home on weekdays during good daylight hours.
Some of my lenses have appeared to need more than +10 correction so if the firmware updates have indeed improved AF performance, I may be able to avoid fiddling about in the debug menu. The firmware update to V1.21 last night was a precursor to see if this could be avoided. If will also be interesting to see if the variability of individual PDAF attempts is reduced.
I'm guessing this is a highly variable thing. Pretty much all my lenses that needed correction on the K5, work perfectly on the K3. Even the FA77 which was pretty far off. When I focus via live view, then refocus through the view finder it doesn't move at all. The only lens I haven't checked yet is my DA35, which also was pretty far off.