fix: make every BOM Value unambiguous for the fab, and switch D_INV_ISO1 to SOT-323

[thePunderWoman]

Summary

PCBWay flagged D_INV_ISO1: "BAT54W" is SOT-323 from nearly every maker, but we had picked Yangjie's SOD-123 variant. A fab only sees Designator, Footprint, Quantity and Value in the BOM — never MF, MPN or LCSC — so a bare part name is ambiguous. This fixes the class of problem, not just one part.

• Every Value is now unambiguous: <label> | <MPN> | <manufacturer> | <package>, e.g. 100nF | CL05B104KO5NNNC | Samsung Electro-Mechanics | 0402. The label is dropped when it repeats the MPN, and commas/periods are stripped so the BOM CSV never needs quoting. BT1 keeps its RIGHT-ANGLE HORIZONTAL note. Applied to all 81 parts across every sheet and the PCB.
• scripts/bom_values.py builds, applies and checks these values (--apply / --check), and fails if the PCB copy of a part's Value, MF, MPN or footprint differs from the schematic. New footprints must be added to its package table on purpose; hand-sourced parts with no MPN (J_STICK1) are listed explicitly rather than silently blank.
• D_INV_ISO1 → Diodes BAT54W-7-F, SOT-323 (the common part, made by several vendors). Pin 1 anode / pin 2 not connected / pin 3 cathode, confirmed against Nexperia's numbered pinning table and the Diodes datasheet. Same BAT54 family electrically (30V, 200mA, 2µA leakage at 25V), so the SPICE model and scripts/spice/latch_sim.py still pass. PCB footprint replaced and rerouted.
• Layout diagrams now say AO3401A for Q_PWR1 (they were stale since it was swapped from DMG2305UX) and BAT54W-7-F / SOT-323 for D_INV_ISO1.
• Repo CLAUDE.md documents the Value rule and the script. Production files regenerated.

Notes

• Also includes PCB/snips_controller.kicad_pro: KiCad only reordered its used-designators list (one line, no functional change).
• scripts/jlc_match.py searches on Value, which now includes the MPN, so its Basic-library alternative search may behave oddly. That workflow is retired.
• The SOD-123 → SOT-323 change is a layout change on the inverter cluster; worth a glance at the routing around D_INV_ISO1.

Test plan

• python3 scripts/bom_values.py --check — 81 parts, all Values and PCB fields in sync (and --apply is idempotent)
• kicad-cli sch erc — 0 errors, 1 pre-existing benign warning (D_RGB1 lib_symbol_mismatch)
• kicad-cli sch export netlist — D_INV_ISO1 pin 1 → PWR_BTN_INV_BASE, pin 3 → PWR_BTN_SENSE, pin 2 no-connect
• kicad-cli pcb drc — 0 violations, 0 unconnected pads
• python3 scripts/spice/latch_sim.py — all checks pass
• Regenerated production BOM: all 40 lines use the new Value format, no commas/semicolons
• Confirm with PCBWay that the BOM now reads unambiguously

🤖 Generated with Claude Code