Sat 27-Apr-13 08:34 AM | edited Sat 27-Apr-13 10:38 PM by Stagecoach
From the explanation of your concerns regarding focus accuracy repeatability and the tests undertaken I will make the following observations;
a) I estimate that in the test shots shown your camera to subject (focus point) distance is around 0.5 mtrs or less for which at 35 mm, f1.8 and 0.5 mtr would give you a 13mm DOF. (at 0.3 mtr distance the DOF would be down to 5 mm).
b) The minimum focus distance for the 35 mm lens is quoted as 0.3 mtr so if you were this close you are at its limit.
c) You mention in your explanation that between each shot you de-focused the lens so there would have been a possibility of slight movement.
d) Every electro/mechanical control system which in this case has sensed, adjusted and determined that the camera is focused and ready to shoot has what is called a 'deadband'. The deadband is an area that ranges either side of the 'ideal' position and its purpose is to stop hunting otherwise the focus motor would never stop.
e) Taking d) above (the deadband)into account there is going to be a difference, determined by deadband size, depending upon whether the final focus/ready position has been reached by travelling from a rear or near de-focused position.
So if we take all the above into account and the fact that you are testing at minimum focus distances and minimum DOF is this really a fair test or something that is simply showing the normal expected errors at the system limits.
You refer to taking pictures of people and the odd 'blurry' one but I suspect you would not be taking these with a 35mm at 0.5 mtr distance f1.8 where you are only going to get a DOF of 13mm.
Perhaps a test using similar distances and settings as when the problems occurred may be more beneficial to start with.