Closed bug report

Thermal-via planner mistakes remote same-net pours for local heat-spreading copper

John Lauer · ·closed by John Lauer

AIFlow 0.1.70, source 706d95235c7e67f56b7bd4771fc051c4b19e4630; KiCad 10.0.5 / bridge 1.0.55 on Windows, Adom Bridge 2.1.172.

On the electronic-load run, describe the accepted local LOAD_POS diode regions on F.Cu and B.Cu, then run thermal-vias plan. It proposes 16 vias beside Q1 even though these copper regions are only around D1, elsewhere on the board. The exact output is Q1 LOAD_POS target 16 existing 0 planned 16. No proposed vias were landed.

The planner tests only net-wide presence of both layer names, not the candidate location. It also offers rings beyond the 3 mm radius used by its post-landing count. Expected: require local region coverage on both layers and use the same neighbourhood for planning and acceptance. Native filled connectivity and thermal checks must remain mandatory; planned polygons alone are not proof of heat flow.

A focused fix and regressions are in progress against the current immutable HEAD, covering remote/disjoint regions, via-edge coverage, malformed geometry, and plan/readback-count consistency. The external TO-220 heatsink design remains an independent unresolved engineering issue.

2 Replies

John Lauer ·

Fixed in source commit 1bc0e48412a8df82b48f451180f6090e02833d55, published publicly as 0.1.71 and installed/verified on this run. All 107 workspace tests pass; production build, component/widget/silkscreen and dashboard regression checks pass. Existing CLI command set is preserved.

The actual electronic-load plan drops from 27 proposed vias to 2, with all 16 remote Q1 proposals removed. Local polygon coverage and the same 3 mm neighbourhood are required for planning and shortfall counts. These remain necessary geometric checks, not native filled-connectivity proof. The disposable two-via and single-via shunt candidates each produced a ground connectivity break after native refill, and were rejected. The accepted live board is unchanged. No cooling qualification or finish pass is claimed.

John Lauer ·

Shipped in 0.1.71 (see the reply above); still in 0.1.94. Closing.

John Lauer closed this ·

Log in to reply.