Adom Fields
Public Made by Adomby adom
Adom Fields: where the current and the heat go on your board.
0.2 mm grid joins same-net copper 0.14 mm apart: a false neck sets VTILE's peak (133.6 A/mm2) and hides the real one
adom-fields 0.10.2 · adom/core 4.15.3 · board saved by KiCad 10.0.7 (format 20260206) · --pitch 0.2 (default)
What I did: adom-fields analyze on a 4-layer board where each power net has track-following zones on F.Cu and B.Cu. Two separate VTILE zones on F.Cu end close together. KiCad keeps their fills 0.142 mm apart (measured on the saved filled_polygons with shapely), and the net's two sides are joined only through other layers.
What happened: the solve treats the two fills as joined. VTILE's peak landed in the gap, as a neck that does not exist:
VTILE 1.5 A peakJ 133.6 at [117.9, 134.9] F.Cu
review.checks.ampacity: VTILE neck 0.32 mm at [117.9, 134.9], IPC 10C 1.05 A, solved rise 17.2 C
The J map shows a large share of VTILE's current running through that false shortcut. On the same board, two VSYS pieces were 0.081 mm and 0.291 mm apart (B.Cu and F.Cu) and were probably joined the same way, without setting VSYS's peak.
My reading: a cell counts as copper when its centre lies in copper, and neighbouring copper cells join. Same-net copper closer than about one cell then becomes one conductor. KiCad keeps no clearance between same-net zones, so this geometry is legal and easy to make. Because Fields reports one maximum per net, the false peak also hides the net's real one.
Expected: join copper only where the board joins it. For example, rasterize each filled polygon and track separately and connect cells across items only where the source geometry overlaps or touches. Or at least report "same-net copper closer than one cell was merged at (x, y)" as a finding, so the peak is not taken at face value.
Worked around by keeping separate same-net pieces on a layer at least 0.46 mm apart (more than two cells). I'll add the before and after peaks here when the new solve is done.