fix(fpc): default pin count to 12 so a bare fpc renders (#786)

[zkasuran]

Summary

• default fpc's pin count to 12 so the bare fpc name renders instead of throwing a raw Zod parse error
• add a regression test that asserts bare fpc produces the 12-pin FPC-05F-12PH20 layout, with an SVG snapshot

Closes #786

Motivation

fp.string("fpc").circuitJson() threw a ZodError on num_pins because the schema had no default. Every other pin-count footprint defaults its count (bga 64, quad 64, ssop 8, dip 6, lga 14, son 8, sot223 4, the two-pin parts 2), so fpc was the only one that failed on a bare name.

The fpc dimensional defaults are already built around a 12-pin part. The existing fpc12 test reproduces FPC-05F-12PH20 from the schema defaults alone, so 12 is the count the rest of the schema was tuned for.

Implementation

One line in src/fn/fpc.ts:

num_pins: z.coerce.number().int().min(2).default(12),

The default only applies when no count is given, so fpc12, fpc30_staggered and every explicit form stay unchanged. Bare fpc now resolves to the same geometry as the explicit fpc12 reference.

On KiCad parity

The part this footprint models, FPC-05F-12PH20, is not in the KiCad cache. The nearest 12-pin 0.5mm entry there is a different manufacturer part, Hirose FH12-12S-0.5SH, whose mounting pads sit at ±4.65 (ours at ±4.44) with a different body origin. A parity run against it gives a 41% courtyard difference. Matching it would distort the FPC-05F-12PH20 geometry the repo already asserts. So this test follows the existing fpc test style: a hardcoded copper-pattern check against the reference part plus an SVG snapshot, not a kicad-parity test.

Validation

• bun test: 522 pass, 0 fail
• bun run build: success
• biome format: clean
• new test: bare fpc renders 14 pads matching the explicit fpc12 (FPC-05F-12PH20) reference, plus a new SVG snapshot

AI disclosure

AI assistance (Claude, Anthropic) was used in developing this change. The design, review and verification were done by the author. Verified locally before submitting: bun test (522 pass, 0 fail), bun run build and biome format.

[techmannih]

kicad test?

[zkasuran]

Added one, though it needs a caveat because fpc does not map cleanly onto the usual kicad-parity check.

KiCad has no single canonical FPC footprint. Every part in Connector_FFC-FPC.pretty is a specific vendor connector (Hirose, Amphenol, TE ...) with its own housing, mounting-pad geometry and courtyard. The footprinter default reproduces the FPC-05F-12PH20 land pattern, which is not in the KiCad library at all. What every 0.5mm 12-pin FFC part does share is the contact land pattern, so that is what I pinned.

tests/kicad-parity/fpc_kicad_parity.test.ts compares the bare fpc default against Hirose FH12-12S-0.5SH:

• 12 contacts, matching KiCad in count, 0.5mm pitch, 0.3mm pad width and X positions (-2.75mm to 2.75mm).
• an overlay snapshot of footprinter against the KiCad reference.

It deliberately does not assert courtyard parity, unlike the sibling tests. Footprinter draws a pad-bounding-box courtyard (11.38 x 4.95mm); the KiCad courtyard encloses the connector body (12.12 x 7.9mm on the Hirose part). Measured courtyard IoU against the four nearest KiCad 12-pin 0.5mm parts (Hirose FH12-12S-0.5SH, Amphenol F32Q/F32R-1A7x1-11012, TE 1-1734839-2) is 47 to 67% (diff 34 to 53%), all of it the housing, which footprinter does not model and which differs by vendor. The other parity tests assert well under 5%, so a courtyard bound here would either fail or be a made-up number that certifies nothing.

Verified locally: bun test tests/kicad-parity/fpc_kicad_parity.test.ts passes (30 assertions) and biome format is clean on src and tests.

If you would rather point it at a different reference part or record the courtyard as a documented non-gating number, happy to adjust.