The QFN Wizard generates wrong pad sizes.
It seems that the wizard always draws rectangular pads (if “Oval” is off) and that the value for “Fillet” is added to the length of the pads.
I exaggerated this in the screenshot below, by making square pads (0.2 * 0.2mm) with a large “fillet” of 1mm. This generates pads with a length of 1.2mm, while the pads are square when I set “fillet” to 0.
We do not use the wizard for the official lib as it does not provide any increase in maintainability and is quite frankly bad in general (Unless the recent update truly fixed everything but your screenshot does not look like the required degrees of freedom have been added.)
I already had my suspisions that:
a). The official libraries were script generated.
b). The Footprint wizards were out of date (probably from V4, before fillets).
It would be nice to have working Footprint Wizards. They work quite nice for generating footprints with lots of pads.
For what I know the rounded corners are usefull for better release of solder paste from stencils, and the wizards do not seem to support it.
Maybe it should be filed as a bugreport or on the wishlisht to improve the footprint wizards with the capabilities of the V5 footprints, and use the same Python scripts as the official libs. Using different scripts for the same footprints seems very silly.
I’m a total python noob thouh.
it was me who brought the topic up in the EEVblog Kicad forum, many thanks to paulvdh for taking it to the Kicad forum here. I checked the patch and while it fixes the problem of incorrect footprints, it applies the new “offset” parameter in a totally un-intuitive way: the pads are moved outwards by offset/2 instead of just offset. This makes it unnecessarily convoluted to compute the correct offset if you try to match the land pattern of a chip from the vendor datasheet. I think this should get fixed before it goes into a release.