Rounded corners on custom pad shapes

[Evan Shultz]

Please see https://github.com/KiCad/kicad-footprints/issues/658.

IPC 7351C will be recommending rounded rectangle pads. (Apparently this was desirable from the first release of the document but ECAD software didn't commonly support this pad shape so oblong was the recommendation.) While we can use rounded corners with rectangular KiCad pads, custom pad shapes do not allow rounded corners.

This prevents all pads from having rounded corners and makes a choice between not adhering the IPC's latest recommendations or having pads both with and without rounded corners (if only custom pad shapes aren't rounded). The example given in the link above is for a QFN (currently Package_DFN_QFN:QFN-28_4x4mm_P0.5mm) with funky corner pads that are best represented with custom pad shapes.

The librarians would like to have rounded corners available on custom pads so that the benefits of rounded corners are available to all pads, and all pads in KiCad can be compliant with the best practices according the IPC 7351C.

[Andrzej Wolski (awolski)]

Custom pad shapes can cave rounded corners, you just have to set polygon's outline thickness to a nonzero value.
Unfortunately, that thickness will be added to polygon's size, which makes creating precise pad shapes cumbersome.

[Evan Shultz (evan-shultz)]

Got it. Thank you!

While that does work, it's a much more difficult hurdle than being able to directly draw the outline with 0 width and apply a corner radius. And the radius cannot be easily changed later (essentially the custom shape must be redrawn).

While drawing this with a script would take out a lot of the challenge, and we are scripting more and more of the KiCad library, it would still be easier for both us librarians and users if custom shapes could have rounded corners applied directly so my wish still stands.

[Seth Hillbrand (sethh)]

This will require a file format change, so we can look at it during 6

[Rene Poeschl (poeschlr)]

I don't think it really requires file format changes. Polygon pads can have a line width which will add a radius to it.
The only thing necessary would be an abstraction layer in the graphical interface that automatically moves the polygon points inwards by the required radius and assigns a line width of double the radius.

[Seth Hillbrand (sethh)]

I'm pretty sure that that solution will only round the convex corners. Is that the desired limitation for pads?

[jean-pierre charras (jp-charras)]

Yes, I confirm it works only for convex polygons, and can modify the initial shape (I mean setting a radius to a given value and then to 0 does not give the initial shape in every case).

[Seth Hillbrand (sethh)]

@Rene- Do you know what the IPC recommendation is for non-rectangle pads? I don't have access to a copy of version C. Absent strong driver, I think it is safest to code this as a file format change for the pad but there may be a reason to facilitate the approach you suggest (with the convex limitation).

Do you (or anyone else) have thoughts/arguments for/against?

[Evan Shultz (evan-shultz)]

@Seth
IPC 7351C isn't released yet, but you can find some info at
http://www.ocipcdc.org/archive/What_is_New_in_IPC-7351C_03_11_2015.pdf (page 9+) and
https://www.ipc.org/committee/drafts/1-13_d_7351CGoals.pdf (last page).

KiCad has a limitation right now, so file format change or not it would be helpful to incorporate this feature. Perhaps waiting for 5.1 and making a robust and inclusive solution is best?

[jean-pierre charras (jp-charras)]

Using a custom shape could be a temporary option.
However this is not a good solution.

Custom shapes are here to build complex shapes.
Therefore it is not possible to automatically create thermal reliefs for such shapes.

This can be annoying for SMD pads.

[Rene Poeschl (poeschlr)]

The main reason to use polygon pads for us right now is that some qfn footprints just need a chamfered pad at the corner pads. These are pads where the current workaround of using line thicknesses works.

Something like this footprint: https://github.com/KiCad/kicad-footprints/pull/834

[Rene Poeschl (poeschlr)]

I assume this mailing list conversation is a response to this: https://<email address hidden>/msg31906.html

[Rene Poeschl (poeschlr)]

The current implementation does not fix all we needed. I have documented requested improvements in: https://bugs.launchpad.net/kicad/+bug/1824614

[KiCad Janitor (kicad-janitor)]

KiCad bug tracker has moved to Gitlab. This report is now available here: https://gitlab.com/kicad/code/kicad/-/issues/2158