Originally posted by nomadkng Human presses focus button, PDAF gives focus lock on a calibration chart. Human presses "Focus Review" which locks focus mechanism and aperture at wide open and transitions to LV. Human can then Zoom in LV to 100% and adjust for front or back focus in real time. Since I believe this actually shifts the sensor forward or backward you would see real time feedback at 100%. Human then selects "Focus Test" and PDAF cycles the lens to infinity then min and then attempts focus lock. Human then presses "Focus Review" and checks results in LV. Repeat as necessary until lens is calibrated.
This sounds like it could also work and wouldn't be that hard to implement.
I'm not quite sure though that an automatic process would be that complicated. For PDAF I assume the camera must keep track of the screw/HSM motor as PDAF is a predictive AF system (from what I understand). PDAF attempts to guestimate how much out of focus an image is and then sends the signal to the AF motors to operate for a certain amount of time or rotations to attempt to achieve focus. Once the motors finishes it checks again and then if any minute adjustments are needed and then (if needed) sends the signal to the motors again. That's why there is always one or two stages to PDAF focus. (Again I could be wrong here).
So if the camera is able to keep track of the position of the motor for focus then the process could work like this. Human sets camera to Focus Calibration mode. The camera focuses on an object using PDAF. The camera then switches to Live View. The camera then increases focus by a fixed amount and then decreases by a fixed amount constantly reviewing the contrast of the image. It then can determine where the contrast is the sharpest and then makes note of the position of the motors relative to where they were during the PDAF. If they are different then adjust the calibration and repeat until there is no difference in the position of the motors.
I wouldn't think this would be a complicated program especially for Pentax. I would think the programming required to shift the sensor for Astrophotography would be equally as complex.
Alternatively another method would be for it to simply step through the 20 levels of calibration one at a time and determine which works best. The camera would start at -10, use PDAF capture the image. Then switch to -9, use PDAF and again capture. Repeat until +10. Then review all of the images using the CDAF engine to see which one is sharpest and there you go. It is more of a trial and error approach but would also work with much simpler programming.