Network cabinet fan controller
Public Unreviewedby matt
Two independently controlled 12 V fans, Ethernet and Wi-Fi, and an RGB speed display in a custom printed puck. Routed development preview, not approved for manufacture.
16485bb
Local analysis adapters
The shared installed Adom tools and pinned Bridge runtimes were preserved. This project uses isolated source builds under tools/aiflow-local and tools/fields-local.
- Fields resolves KiCad numeric net IDs to their named nets and uses 18 µm plated-hole walls for this declared screening case.
- Fields validates explicit thermal-via requirements against the named physical pad and net. A tabless package no longer creates a fictitious net named
leadsfor that check. Missing/ambiguous pads and insufficient actual vias fail. 70 unit tests pass. - AIFlow compares reported via plating with the declared positive spec value (default 25 µm when absent), rather than requiring 25 µm for every project. Distributed plated-wall coupling, removed drill voids, current-density and temperature limits remain required. 36 analysis tests pass.
- The project-library adapter permits the explicitly selected complete local CAD library. Library provenance remains in libraries/provenance.json.
- The local 3D review command preserves background focus. It still requires the native canvas to paint before recording.
- tools/kicad9-cli implements actual native KiCad 9 zone refill before native DRC. It verifies UUID-to-net identity across refill/save/reload. Other CLI operations use the native remote adapter.
These repairs are local, not upstream releases. Tests of these adapters do not constitute physical qualification of the fan controller. Detailed test and build logs are retained in checks/.
The local regional-temperature correction samples each package's maximum over all its pad centres plus the solver's source-deposition radius. It retains the full declared loss, convergence tests and per-part rise limits. AIFlow accepts that verified package-region solve in place of its area-only pour estimate, retaining the original estimate in the result. It checks that the region covers every native pad; first-pad-only and stale reports cannot use this path. An absent tab means heat through all leads, as implemented by Fields, rather than a net named leads. Regression tests include a distant hot lead and refusal of incomplete regions.
The final project is saved by KiCad 10.0.6. Its current DRC uses that version’s native --refill-zones command. The probe checker now accepts an explicit F.Cu/B.Cu measurement layer and checks bodies and neighboring pads on that face; invalid layers fail. Four probe regression tests pass. The plan uses eleven underside test points and the front-side SW test point; probing occurs before fitting the carrier.
# Local analysis adapters
The shared installed Adom tools and pinned Bridge runtimes were preserved. This project uses isolated source builds under tools/aiflow-local and tools/fields-local.
- Fields resolves KiCad numeric net IDs to their named nets and uses 18 µm plated-hole walls for this declared screening case.
- Fields validates explicit thermal-via requirements against the named physical pad and net. A tabless package no longer creates a fictitious net named `leads` for that check. Missing/ambiguous pads and insufficient actual vias fail. 70 unit tests pass.
- AIFlow compares reported via plating with the declared positive spec value (default 25 µm when absent), rather than requiring 25 µm for every project. Distributed plated-wall coupling, removed drill voids, current-density and temperature limits remain required. 36 analysis tests pass.
- The project-library adapter permits the explicitly selected complete local CAD library. Library provenance remains in libraries/provenance.json.
- The local 3D review command preserves background focus. It still requires the native canvas to paint before recording.
- tools/kicad9-cli implements actual native KiCad 9 zone refill before native DRC. It verifies UUID-to-net identity across refill/save/reload. Other CLI operations use the native remote adapter.
These repairs are local, not upstream releases. Tests of these adapters do not constitute physical qualification of the fan controller. Detailed test and build logs are retained in checks/.
The local regional-temperature correction samples each package's maximum over all its pad centres plus the solver's source-deposition radius. It retains the full declared loss, convergence tests and per-part rise limits. AIFlow accepts that verified package-region solve in place of its area-only pour estimate, retaining the original estimate in the result. It checks that the region covers every native pad; first-pad-only and stale reports cannot use this path. An absent tab means heat through all leads, as implemented by Fields, rather than a net named `leads`. Regression tests include a distant hot lead and refusal of incomplete regions.
The final project is saved by KiCad 10.0.6. Its current DRC uses that version’s native --refill-zones command. The probe checker now accepts an explicit F.Cu/B.Cu measurement layer and checks bodies and neighboring pads on that face; invalid layers fail. Four probe regression tests pass. The plan uses eleven underside test points and the front-side SW test point; probing occurs before fitting the carrier.