Pass 134
2026-11-05 21:40 PST (2026-11-06T05:40:29Z UTC) · commit dfe4ee1 (dfe4ee170b19b5d51920f049ed8275a089125bcb)
ADOPTED
Shorts class attacked; verdict: it is live routes over foreign pads, not debris. A DRC-refereed deletion pass (scratch/drc_shorts_prune.py: KiCad proposes the track items of shorting_items, the contact graph vetoes own-net bridges, deletions applied per-net with DRC after each net and wholesale restore on any split) removed exactly 2 segments board-wide (CDR_DATP 1 + 1): KiCad DRC violations 420 -> 416, shorting_items 59 -> 57, unconnected 50 UNCHANGED. Everything else was vetoed as an own-net bridge or split on deletion (+3V3D batch splits +2, ULPI_NXT +1 – both restored and hand-listed): the shorting tracks are overwhelmingly fanout-lattice escapes that cross foreign pads while carrying their own net, concentrated at U5 (29 of 59: +3V3D stubs over U5 GND/CDR_CLKP/CDR_CLKN pads, CDR_DATP over U5.+3V3D pad, I2C tracks/vias over CDR pads) and the ULPI_D4/D6 x XO lattice (9). Deletion is the wrong tool for the rest; the fix is a coordinated rip-and-relay of the U5 fanout field (same locality as the CAP_CLK/CAP_DATA/CDR unconnected groups), then the ULPI/XO lattice. Via shorts (5) were not auto-deleted by design (a via on a foreign pad can be its net’s layer bridge) and are on the same hand list.
| Nets fully routed | not recorded |
| Pins failed | not recorded |
| Clearance violations | 142 |
| Unconnected nets | 35 |
| Audit verdict | FAIL |
Routes — all copper layers and each layer
F.Cu In1.Cu In2.Cu B.Cu
Top
Bottom
← Pass 133 · all passes · Pass 135 →