Why are the balls on the WLCSP package off center by 2 microns?

[abba]

The datasheet for STM32H573 (WLCSP80 package, rev 5 p. 251) says that the leftmost ball (in the bottom view on the datasheet) is F=0.384mm from the package edge. Horizontal ball pitch is 0.35mm*(1/2)*√3. Therefore centering the ball array horizontally would give F = 3.50/2 - 0.35*(9/4)*√3 = 0.386mm.

This 2 micron difference has led to some discussion about whether the ball array should be horizontally centered within the package edges on PCB footprints. (Additional discussion arose from my initial misunderstanding of the datasheet. The datasheet could possibly be improved by measuring the margin to the top row rather than the second row. But now that I realize this, the question about horizontal centering still remains).

Does ST have any insight about this 2 micron offset that might give a PCB designer confidence that it was truly intended? Can ST (or the ST community) give a recommendation about whether (or not) to update the KiCad libraries to include this 2 micron offset?

[abba]

Per continued discussion on the KiCad thread, a 2 micron offset is not consequential. The measurements themselves are given on the datasheet with 20-30 micron tolerance. There is also a 50 micron tolerance per ball, as well as 100 micron tolerance for shifting the entire array.

Assembling this part therefore requires using an upward facing alignment camera after the part is picked. This will adjust for any small offsets, including manufacturing tolerances as well as the 2 micron offset in the KiCad library (which was probably done to make calculations easier).