pcbcraft (tangency-labs) errors

8 recurring problems engineers report in pcbcraft (tangency-labs), ranked by impact.

#56 by errors trackedECAD

Recurring pcbcraft (tangency-labs) errors, by impact

  1. 1A symbol carries only a footprint name string and pins have free-text types; the proposal is one verified part record joining symbol, footprint, datasheet requirements and 3D body.ECAD · Agent failure · 1 report66
  2. 2Delete the highest-numbered component, save, reopen, and its ID could be handed to a new object, so anything still referring to it pointed at the wrong thing. The fix saves an allocation watermark.ECAD · Agent failure · 1 report66
  3. 3DRC skipped clearance pairs inside one physical copper island, so a short between two intended nets could pass. The repro: merge two nets, route a bridge, then delete the joining schematic wire and accept the split while the copper stays. The schematic says two nets, the copper says one, and DRC said nothing. The fix adds an explicit short finding.ECAD · Agent failure · 1 report66
  4. 4Per-footprint clearance and mask and paste overrides in KiCad library files were dropped at import, so DRC checked fine-pitch parts against the looser board-wide rule.ECAD · Agent failure · 1 report66
  5. 5Renaming a symbol pin meant removing it and retyping the full pin definition, which dropped its net attachments and no-connect flag, and swapping two pin numbers needed two edits with an invalid state in between.ECAD · Agent failure · 1 report66
  6. 6The starter library gives generic resistors 1k and caps 100nF, placement copies them in, and validation only needs a non-empty value, so an unspecified part looks chosen. The issue also says DNP and BOM exclusion have different consumers and must not be text conventions.ECAD · Agent failure · 1 report66
  7. 7Today a schematic deletion removes the physical footprint right away. The PR adds an accepted board baseline so footprints, positions and pad connections hold until a reviewed update, the way KiCad's Update PCB step works.ECAD · Agent failure · 1 report66
  8. 8Libraries disagree on pad numbering (1..n versus A/B or 1a/1b) and today the only fix is re-authoring the whole footprint by hand.ECAD · Agent failure · 1 report58

How it fails

  1. Agent failure8

Sourced from public community reports. Costs and delays on individual entries are modeled estimates, not measurements.