On a probe station, the force at the needle tip is set by how far the arm deflects. The servo only decides where the arm stops.

The Z axis brings the cantilever down until contact, then travels a few more microns of over-travel. Contact force is stiffness times over-travel. A cantilever loses stiffness with the cube of its length: double the arm to reach the next pad row and it keeps one eighth of it, so the same commanded over-travel presses with a fraction of the original force. Contact resistance climbs, first-pass yield drops, and the motion trace still looks healthy.
So state the force window in microns, measure it at the tip with a load cell, and check touchdown repeatability over ten cycles at production speed, not one slow demonstration move.
Two more terms decide whether it holds. Ambient drift of 1 °C moves a 200 mm aluminium arm about 4.6 µm by thermal expansion, the same order as the over-travel itself. And a belt stage reverses with compliance, so the touchdown point scatters even when the encoder reports the same count.
 
Top