Fixture: antmicro-microphone-board
A PDM microphone breakout: a Knowles SPH0644LM4H-1 microphone routed to a zero-insertion-force connector, with supply filtering and ESD protection.
Attribution
Upstream: antmicro/pdm-microphone-board at commit 20813e2b, retrieved 2026-07-28.
Copyright (c) 2024-2025 Antmicro, licensed Apache-2.0. The upstream license text is preserved verbatim in LICENSE.
Design files under tree/hardware/ are copied unmodified. Repository assets, documentation PDFs, CI configuration, and the local .kicad_prl state file were not copied. No design file was altered in any way.
Why this board
It is a real board that was designed to be manufactured, by people who do this professionally, and it is small enough to reason about completely. Twelve components, one sheet, three signal nets, three power nets. The upstream project is Apache-2.0, which is the same license this repository uses, so vendoring carries no license friction at all.
It is deliberately not impressive. A fixture’s job is to make agent behavior legible, and a board an author can hold entirely in their head is what makes an unexpected diff obviously unexpected.
What it establishes about real fixtures
Running verification on it at baseline, with kicad-cli 10.0.6:
| Check | Errors | Warnings | Detail |
|---|---|---|---|
| ERC | 0 | 44 | 32 lib_symbol_issues, 12 footprint_link_issues |
| DRC | 0 | 13 | 13 lib_footprint_issues, 0 unconnected, 0 parity |
Every one of those 57 warnings is library resolution: the design references upstream antmicro-footprints and antmicro-symbols libraries that are not vendored here. Nothing electrical is outstanding, and the two checks that matter most for an edit task, unconnected items and schematic-to-board parity, are both zero.
This is the normal condition of a real design in a bare checkout, and it is why tasks on this fixture assert erc_no_new_violations rather than erc_clean. See FIXTURES.md for the rule and the reasoning.
Scale, and what it does to a surgicality bound
The schematic is 8,489 lines; the board is 35,688. For comparison, a synthetic fixture is a few hundred lines total.
That difference is not cosmetic. AC-3.7’s 5 percent diff bound was calibrated against a small hand-written schematic, where 5 percent is roughly fourteen lines and therefore a real constraint. Five percent of this schematic is 424 lines, which is enough to rewrite entire subsystems while still passing. Tasks here use a bound near 1 percent, and the general rule now stated in FIXTURES.md is that a surgicality bound is calibrated per task against the file it constrains, not inherited as a constant.
Verifying this fixture
node scripts/hash-fixture.mjs --check fixtures/antmicro-microphone-boardRegenerating the baseline reports
Only when the reference kicad-cli version changes, and the new counts belong in fixture.json in the same commit:
cd fixtures/antmicro-microphone-board/tree/hardware
kicad-cli sch erc --format json -o ../../baseline/erc.json microphone-board.kicad_sch
kicad-cli pcb drc --format json -o ../../baseline/drc.json microphone-board.kicad_pcb