openpack (saturncity) errors

4 recurring problems engineers report in openpack (saturncity), ranked by impact.

#131 by errors trackedFirmwareECAD

Recurring openpack (saturncity) errors, by impact

  1. 1Firmware pin devicetrees for two nRF5340 boards are now generated from the schematic netlists by script so a schematic change cannot silently diverge. The generator refuses unmapped nets, two roles on one pin, or USB off its dedicated pads, and CI compares every compiled pin against the netlist. One pin is still unverified because the module datasheet contradicts itself: page 21 puts FEM MODE on P1.04, while the page 22 sample says P0.24, which is the pack's I2S clock. Closes issue #118.Firmware · Agent failure · 1 report74
  2. 2The 4-layer pack board PR #104 reports DRC 0 errors, 0 unconnected, fab outputs done, but its own review notes quote the Fanstel spec, which says the nRF21540 in BT40X is hardwired to the PCB trace antenna and ANT-SEL is unused. If that holds for the BT40NE, the whip on the u.FL is not driven through the front-end module, which breaks the link budget in audit.py, an ADR, and a body-blocking margin.ECAD · Agent failure · 1 report74
  3. 3This follows last hour's #105 finding, where a clean DRC sat on an antenna architecture the module datasheet did not support. The team has now decided to swap the receiver module from BT40NE to BT40E (ADR-0021), and one part change touches the pad map and footprint, the generator script, the devicetree (the FEM node becomes pack-only), tests and outputs, the BOM, an older ADR, and firmware builds.ECAD · Agent failure · 1 report74
  4. 4While laying out a receiver sub-board, the authors found that hw/pack/pack.kicad_dru was never loaded: KiCad 10 silently drops the whole file over a min_resolved_spokes rule, so every earlier clean DRC on the main board was against defaults. Separately, the pcbnew zone fill ignored the RF netclass gap, so the saved fill now comes from kicad-cli, and guard tests were added that prove the rules load.ECAD · Agent failure · 1 report74

How it fails

  1. Agent failure4

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