
SCART switcher
nunosousa/SCART-switcher
An old school SCART switcher. Its main purpose will be to automaticaly detect and switch the video output to the active input video source.

Programmable semiconductor curve tracer: STM32 firmware, two-layer KiCad PCB, SPICE Monte Carlo analysis, and a host-side SPICE-parameter extraction pipeline. Hardware in bring-up.
Curve tracer is an open-source printed circuit board by leedanie2, from the leedanie2/curve-tracer repository on GitHub. It is a 2-layer board of 135 × 92 mm with 88 parts on 99 nets, routed with 381 tracks and 31 vias.
It is built around the CC0603KRX7R9BB104, RT0603BRD0710KL and 0603B103K500NT, with the RT0805BRD071KL, RT0805BRD0720KL and RT0603BRE0723K2L among its other parts. Its parts include 25 resistors, 8 capacitors, 8 connectors, 6 diodes and 3 ICs. It belongs with the STM32 boards of this library.
KiCad project for the curve tracer PCB. Scope, dates and the non-negotiable
layout constraints live in curve_tracer_blueprint.md §14; the build
phases and date gates are §9. This file is the toolchain and the routing
policy.
MVP scope, unchanged from firmware/README.md: range 1 only, DC sweep, no
pulsed mode, no auto-ranging. One revision only — §14.4 lists what that
buys (test points everywhere, alternate-value footprints, headers over solder).
hardware/
├── README.md ← this file
├── DECISIONS.md ← where every value on the board came from: blueprint, sim or bench
├── H4_CARD.md ← printable bring-up card: numbered steps, settings, pass/fail, blanks to fill
├── CONTINGENCY.md ← if the board oscillates at bring-up: Part A goto steps, Part B the reasoning
├── ORDER.md ← everything to buy, by vendor (JLCPCB, LCSC, DigiKey), with stock, prices and what to verify
├── curve-tracer.kicad_pro
├── curve-tracer.kicad_sch ← H1, captured Oct 5, 2026
├── curve-tracer.kicad_pcb ← H2, routed Oct 6, 2026: 135 × 92 mm, two layers
├── fp-lib-table ← registers lib/ as footprint library `curve-tracer`
├── lib/curve-tracer.pretty/ ← two project footprints: J3/J4 range select, J7/J8 Nucleo socket
├── tools/check_topology.py ← asserts the §3 / §14.5 topology; run after every edit
├── tools/export_fab.py ← writes everything in fab/; run before every order
├── fab/ ← engineering BOM, Gerbers + drill, and the JLCPCB upload set in fab/jlc/
├── review/ ← layout renders and metrics, pass1/ and pass2/: the evidence for the write-up
├── analysis/ ← gitignored; kicad-happy analyzer runs
└── datasheets/ ← gitignored; re-sync with the lcsc skill's sync_datasheets_lcsc.pySymbols are stock KiCad 10. So are the footprints, except two. J3/J4 is a
stock 2×3 header with silkscreen added (RNG1–RNG3 per row, and "J3=J4
POS"). J7/J8 is the stock 2×19 socket with its pad numbering mirrored for a
board the Nucleo plugs into (DECISIONS.md L-2). Each symbol carries MPN, Manufacturer, LCSC and
Provenance fields. Provenance links to the symbol's row in DECISIONS.md.
The schematic is the source of truth and is edited by hand from here on. ERC proves it is legal; it does not prove it is the circuit §3 describes. Run both:
kicad-cli sch erc --severity-all hardware/curve-tracer.kicad_sch
python3 hardware/tools/check_topology.py --self-testcheck_topology.py exports the netlist through kicad-cli and asserts every
connection §3 depends on: each op-amp pin, the three gains and the ÷4, the
§14.5(5) tap position, limiter and clamp polarity (D3 included), the Nucleo
pin map, star ground, and test-point coverage. It also fixes the list of
parts on the DUT node (blueprint §3.3), so a part added there fails until its
current has been checked against range 3. --self-test also plants fourteen faults that
ERC passes, and fails if any goes uncaught. Change a value on purpose →
change the matching assertion in the same commit.
For H2: KiCad ships a STM32_Nucleo-64_Morpho project template
(KiCad.app/Contents/SharedSupport/template/). Its board places the two
2×19 morpho sockets at the correct spacing, with the mounting holes. Take
the CN7/CN10 geometry from it rather than measuring a Nucleo.
| EDA | KiCad 10.0.6 (schematic capture, layout, gerber/BOM/CPL export) |
| Review | kicad-happy v2.3.0 agent skills — ERC/DRC review, netlist and footprint checks, BOM and CPL sanity before the order goes out |
| Assembly | PCBA — assembled by the fab, express shipping (§14.3, §14.3a) |
| Routing | Analog-critical nets by hand. Everything else autorouted with Freerouting v2.5.0 (DECISIONS.md P-6) |
All five are installed on this machine: the first four as of Oct 5, 2026,
Freerouting on Oct 6. kicad-happy must be present before H1 (§9), and
it is.
| KiCad | 10.0.6, /Applications/KiCad/KiCad.app, CLI at /Applications/KiCad/KiCad.app/Contents/MacOS/kicad-cli. Installed from the official universal .dmg, not Homebrew — the cask writes to a root-owned path and wants a sudo password. Upgrades are therefore manual: fetch a new .dmg from kicad.org |
kicad-happy | v2.3.0 (aklofas/kicad-happy), cloned to ~/.claude/kicad-happy, with all 11 skills symlinked into ~/.claude/skills/ — kicad, spice, emc, datasheets, bom, digikey, mouser, lcsc, element14, jlcpcb, pcbway. Global, so available in every project. Upgrade with git pull in that clone — upstream documents /plugin update as unreliable, which is why this is a symlink install rather than a marketplace plugin |
| KiCad library tables | Default global sym-lib-table / fp-lib-table copied from the bundled template into ~/Library/Preferences/kicad/10.0/ on Oct 5, 2026. KiCad's first-run dialog normally does this; it had never been run here, so kicad-cli sch erc reported every symbol as an unknown library |
| Freerouting | v2.5.0, ~/.local/share/freerouting/freerouting-2.5.0.jar, from the GitHub release (freerouting/freerouting). Installed Oct 6, 2026; SHA-256 f6f51bb0…ffeb5de3c7 matches the release's published digest. Runs on Homebrew openjdk 25.0.2, which was already installed. Headless only: java -Djava.awt.headless=true -jar <jar> --gui.enabled=false -de board.dsn -dr board.rules -do board.ses. Even --version hangs once the GUI starts. Not a KiCad plugin: the board goes out as a SPECCTRA .dsn and comes back as a .ses (P-6 has the method). Upgrade by downloading the new release JAR and checking its digest |
ngspice | 47, Homebrew (/opt/homebrew/bin/ngspice), installed Oct 5, 2026 for the spice skill's subcircuit checks. LTspice stays the tool for sim/; the skill cannot drive the CrossOver bottle. Upgrade with brew upgrade ngspice |
kicad-happy needs Python 3.10+ and has no required dependencies (stdlib
only); this machine has 3.12.7. It does not need KiCad at runtime — it
parses saved .kicad_sch / .kicad_pcb files directly, and supports KiCad
5 through 10.
Optional, not set up: DIGIKEY_CLIENT_ID / DIGIKEY_CLIENT_SECRET,
MOUSER_SEARCH_API_KEY, ELEMENT14_API_KEY. Without them the sourcing skills
fall back to web search. LCSC needs no key, which is the one that matters if
the board goes to JLCPCB.
The kicad-happy skills are a review pass, not an authority. Every §14.5
constraint is a geometric claim about the finished board — trace length,
copper at a node, cap-to-pin distance, return-path separation — and a passing
DRC says nothing about any of them. Check them against the layout by
measuring, then have the skills review it. Neither substitutes for the other.
The analog-critical nets below are routed by hand and are not handed to an autorouter. The reason is not aesthetic. This circuit's stability is set by a parasitic — §3.1 puts stray capacitance at the op-amp inverting node at a pole of 2–3 MHz, against a ~3 MHz crossover, worth roughly 45° of phase margin — and an autorouter optimises for completion and length, with no model of node capacitance, return-current paths, or differential symmetry. It will produce a board that passes DRC and oscillates.
An autorouter is also the wrong tool for a one-revision board for a second reason: its output is not reviewable. §14.5's constraints have to be checked, and that requires routing someone chose.
Ordered by how much damage a bad route does.
| # | Net | Why it is hand-routed | Rule |
|---|---|---|---|
| 1 | FB_SENSE — R_sense tap → op-amp inverting input | The loop's only error signal, on a 7 kΩ node (R_f ‖ R_g). §14.5(1) | Shortest practical path. Never adjacent or parallel to BD139_C — the largest dV/dt and the full load current. Cross at 90° if a crossing is unavoidable |
| 2 | OPA_SWEEP_IN- — the inverting node itself, with R_f, R_g, C_f | This is the 2–3 MHz pole. §14.5(2) | Minimum copper: pads and the shortest links between them. No ground pour under this node, under R_f/R_g, or under the C_f pads. C_f pads sit immediately beside R_f — a footprint reached by a long trace adds stray capacitance to the node it exists to correct (§14.4) |
| 3 | BD139_B — op-amp output → R_B → BD139 base | R_B · C_jc is a ~13 MHz pole inside the loop (§3.1), and this net also carries the clamp-diode discharge, 21.9 mA peak (sim 09) | Short. Keep R_B and the 1N4148 B-E clamp physically at the transistor, not at the op-amp |
| 4 | BD139_E → R_sense → R_iso → LOAD | High-current path: 75.2 mA at the limiter, 107 mA into a hard short (§3.1) | Wide. Its return goes to the star point on dedicated copper — never shared with a sense return (§14.5(4)). R_sense and R_iso in series, in that order, with the §14.5(5) tap between them |
| 5 | SHUNT_HI / SHUNT_LO → difference-amp inputs | Worst-case CMRR is 74.4 dB from resistor tolerance alone (§13); asymmetric routing degrades it and cannot be trimmed back | Matched differential pair: equal length, equal width, routed together, same layer. Kelvin-connect at the shunt pads themselves |
| 6 | KELVIN_HI / KELVIN_LO → ÷4 divider → buffer | The four-wire measurement is what makes R_iso's uncorrected 1.1 V drop harmless (§3.1, §3.4). If this pair is compromised the stability fix is too | Matched pair from the DUT socket terminals. Sensed at the socket, not at the nearest convenient copper. Past the divider, VDIV is a 75 kΩ node (300k ‖ 100k): keep it short and away from BD139_C |
| 7 | BD139_C and every op-amp V+ | Decoupling placement, §14.5(3) | 100 nF within ~5 mm of the pin, measured along the trace, cap body to pin. Bulk 10 µF on the rail. The collector net is also the 0.89 W node — and the TO-126 tab is the collector, live at +15 V (§8): clearance to every adjacent net, no contact with a grounded enclosure |
| 8 | GND_STAR and all sense returns | §14.5(4) | One reference point. A shared milliohm at 107 mA is millivolts into a sense path whose range-3 full scale is 100 mV |
| 9 | DAC_SWEEP (PA6), DAC_GATE (PA4), ADC1_I, ADC2_V | Low-level analog into 12-bit converters; §3.5 wants the 10 nF at each ADC pin and BAT54S clamps to rails | Keep clear of BD139_C and BD139_E. 10 nF and clamps at the pin. The sweep DAC reaches PA6 from PA5 (DAC1_OUT2) through a shunt on the Nucleo, not on this board: keep CN10-11 unconnected here (§3.5) |
Nets 1, 2 and 5 are the three that cannot be fixed by rework if they are wrong. Route those first, while there is still freedom in the floorplan.
Digital, USB, the Nucleo header fan-out, LEDs, jumper selects: autoroute is
fine, then check it does not cut across nets 1–6 or break the star ground.
Done that way on Oct 6, 2026, with nets 1–9 locked first (DECISIONS.md
P-6, P-7). review/pass2/p2_split_crossings.png shows every net that
crosses the ground split, and its distance to the star tie.
hardware/tools/export_fab.py, run with KiCad's Python (it needs pcbnew: /Applications/KiCad/KiCad.app/Contents/Frameworks/Python.framework/Versions/3.9/bin/python3), regenerates everything in fab/
from the saved schematic and board. Re-run it after any change, after DRC.
It refuses to write the JLCPCB files if the BOM and CPL disagree.
| File | What it is for |
|---|---|
fab/jlc/curve-tracer_gerbers.zip | The quote page upload: Gerbers (Protel names) and Excellon PTH/NPTH, in JLCPCB's documented KiCad settings |
fab/jlc/curve-tracer_bom_jlc.csv | PCBA BOM upload. 33 lines, one per LCSC part, every one assembled by JLCPCB, through-hole parts and test loops included. No hand-solder lines |
fab/jlc/curve-tracer_cpl_jlc.csv | PCBA placement upload: 74 rows, PCBA parts only (35 through-hole: J1–J8, Q1 and 26 test loops), rotations corrected per DECISIONS.md P-10 |
fab/curve-tracer_bom.csv | Engineering BOM: every line including DNP. TP6 and TP28, bare pads, are not in it |
fab/gerbers/ | The same Gerbers unzipped, for review |
The consolidated order, with stock and prices re-checked on 2026-10-11, is
ORDER.md. This section keeps the reasons for each part.
For CONTINGENCY.md. None of these is on hand, and only D3 is on the PCBA BOM (spares). Stock was
checked at LCSC on 2026-10-07.
| Part | Value | MPN | LCSC | Stock | Qty |
|---|---|---|---|---|---|
| C_f sweep, C0G 0603 50 V ±0.1 pF | 1.5 pF | FH 0603CG1R5B500NT | C507055 | 1,900 | 100 (min) |
| 2.2 pF | FH 0603CG2R2B500NT | C313085 | 12,200 | 50 | |
| 2.7 pF | FH 0603CG2R7B500NT | C2836769 | 2,300 | 50 | |
| 3.3 pF | FH 0603CG3R3B500NT | C2836771 | 3,250 | 50 | |
| 3.9 pF | Murata GRM1885C1H3R9BA01D | C162229 | 1,035 | 10 | |
| 4.7 pF | FH 0603CG4R7B500NT | C313086 | 3,400 | 100 (min) | |
| R_iso up, axial 1 W 1% | 47 Ω | TyoHM RN1WS47ΩFT/BA1 | C385394 | 740 | 20 (min) |
| U1 fallback | OPA2196 | TI OPA2196IDR (cut tape) | C2057972 | 215 (2026-10-11) | 2 |
| D3 spares: refit after removing it for ranges 2–3, or replace a failed one | TVS 12 V standoff, SMA, unidirectional | MDD SMAJ12A | C113957 | 78,080 (JLC 126,199) | 20 (min) |
CONTINGENCY.md §4).
DECISIONS.md D3). Cathode (band) to DUT_D, which is D3's pad 1: the +y
pad, at y = 78 mm. A refit is checked with H4 step 13a.The order is event-driven: it goes out as soon as this list is clear —
target Oct 13, no later than Oct 20 (§9). Phase 1, the breadboard stability
gate, was dropped on Oct 8 as an accepted risk (§9). The oscillation test is
at H4, and CONTINGENCY.md is the response. Nothing on
the board improves by waiting, so do not hold a DRC-clean layout for a
calendar date. The Oct 26 abandon-the-PCB checkpoint is in §9.
Order with express shipping (§14.3a). The slack is all in front of the order; there is none between arrival and the Nov 16 freeze.
R_iso, R_B, R_sense (§14.4)C_f footprint placed beside R_f, unstuffed — value is fitted
after measuring on the real board (§14.4)DECISIONS.md L-1)ORDER.md ordered, and its Verify
yourself list clearedtools/check_topology.py --self-test passes on the frozen schematickicad-happy review pass on schematic and layoutR_f is 23.2 kΩ
(E96; 23.3 kΩ does not exist and has never been orderable)C_f footprint, the parallel alternates and
the contingency partsORDER.md)ERC checks pins against symbols, and DRC checks copper against copper.
Neither knows which way round the real part sits on its footprint. L-2, the
mirrored Nucleo sockets, passed both. For each part, check the one thing
listed on review/pass2/p2_top.png (crops: p2_orientation.png).
Re-render first if the board has changed. x runs right and y runs down,
from the top-left corner.
Then again in JLCPCB's placement preview, for the PCBA parts in that
list: C5, D1, D2, D3, D4, D5, D6, Q1, Q2, Q3, U1–U3, J1, J6, J7, J8. Q1 and J6
are the through-hole parts that fit their holes either way round: Q1
reversed swaps emitter and base, and J6 reversed faces its wire entries into
the board. The through-hole positions in the CPL are KiCad's footprint
origins (pin 1), not body centres, so the preview may draw them offset. The CPL rotations are community data
(DECISIONS.md P-10), not JLCPCB's. A rotation that is wrong there is the
PCBA version of L-2: every file checks out and the board does not. Fix any
rotation in the preview, then copy the fix into CORRECTIONS in
tools/export_fab.py.
Run it from H4_CARD.md, printed. It is a numbered card:
CONTINGENCY.md Part A;It replaces the checklist that stood here. Every item from that list is a
card step. Results go back as the filled-in sheet plus h4/*.csv, then
into docs/characterization.md.
Do not run a short test without the TO-126 heatsink fitted on Q1. On
the card that is prep step P4 and the precondition for Block 4. In a held
short Q1 dissipates 0.89 W. Bare, that is 129 °C at 40 °C ambient against a
150 °C limit: 21 °C of hand-calculated margin, on the part those tests
stress on purpose (DECISIONS.md P-8). The silkscreen where the heatsink
goes says the same.
| Qty | Value | Footprint | References |
|---|---|---|---|
| 4 | 100n X7R 50V | Capacitor_SMD:C_0603_1608Metric | C2, C6, C7, C8 |
| 4 | M3 | MountingHole:MountingHole_3.2mm_M3 | H1, H2, H3, H4 |
| 3 | OPA2197IDR | Package_SO:SOIC-8_3.9x4.9mm_P1.27mm | U1, U2, U3 |
| 2 | 10k 0.1% | Resistor_SMD:R_0603_1608Metric | R2, R10 |
| 2 | 10n X7R 50V | Capacitor_SMD:C_0603_1608Metric | C3, C4 |
| 2 | 1k 0.1% | Resistor_SMD:R_0805_2012Metric | R15, R17 |
| 2 | 1k 1% | Resistor_SMD:R_0603_1608Metric | R19, R22 |
| 2 | 20k 0.1% | Resistor_SMD:R_0805_2012Metric | R16, R18 |
| 2 | 23.2k 0.1% | Resistor_SMD:R_0603_1608Metric | R1, R9 |
| 2 | BAT54S | Package_TO_SOT_SMD:SOT-23 | D4, D5 |
| 2 | BZT52C12 12V | Diode_SMD:D_SOD-123 | D2, D6 |
| 2 | DNP (pull-down) | Resistor_SMD:R_0603_1608Metric | R23, R24 |
| 1 | +15V | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP7 |
| 1 | +3V3 | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP8 |
| 1 | 1 1% 1W | Resistor_SMD:R_2512_6332Metric | R12 |
| 1 | 10 1% | Resistor_SMD:R_1206_3216Metric | R5 |
| 1 | 10 1% (DNP) | Resistor_THT:R_Axial_DIN0309_L9.0mm_D3.2mm_P12.70mm_Horizontal | R6 |
| 1 | 100 0.1% | Resistor_SMD:R_0805_2012Metric | R13 |
| 1 | 100k 0.1% | Resistor_SMD:R_0603_1608Metric | R21 |
| 1 | 10k 0.1% | Resistor_SMD:R_0805_2012Metric | R14 |
| 1 | 10k 1% | Resistor_SMD:R_0603_1608Metric | R25 |
| 1 | 10u 50V | Capacitor_SMD:CP_Elec_5x5.4 | C5 |
| 1 | 15V BENCH | Connector_PinHeader_2.54mm:PinHeader_1x02_P2.54mm_Vertical | J2 |
| 1 | 15V IN | Connector_BarrelJack:BarrelJack_Horizontal | J1 |
| 1 | 1N4148W | Diode_SMD:D_SOD-123 | D1 |
| 1 | 1k 1% 1W | Resistor_SMD:R_2512_6332Metric | R11 |
| 1 | 22 1% 1W | Resistor_SMD:R_2512_6332Metric | R7 |
| 1 | 3.3p C0G (DNP) | Capacitor_SMD:C_0603_1608Metric | C1 |
| 1 | 300k 0.1% | Resistor_SMD:R_0603_1608Metric | R20 |
| 1 | 330 1% (DNP) | Resistor_THT:R_Axial_DIN0309_L9.0mm_D3.2mm_P12.70mm_Horizontal | R4 |
| 1 | 330 1% 1W | Resistor_SMD:R_2512_6332Metric | R3 |
| 1 | 47 1% 1W (DNP) | Resistor_THT:R_Axial_DIN0309_L9.0mm_D3.2mm_P12.70mm_Horizontal | R8 |
| 1 | ADC1_I | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP12 |
| 1 | ADC2_V | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP15 |
| 1 | BD139 | Package_TO_SOT_THT:TO-126-3_Vertical | Q1 |
| 1 | BD139_B | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP2 |
| 1 | BD139_E | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP3 |
| 1 | BUF_HI | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP25 |
| 1 | BUF_LO | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP26 |
| 1 | CN10 (morpho) | curve-tracer:PinSocket_2x19_P2.54mm_Vertical_NucleoCarrier | J8 |
| 1 | CN7 (morpho) | curve-tracer:PinSocket_2x19_P2.54mm_Vertical_NucleoCarrier | J7 |
| 1 | DAC_GATE | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP21 |
| 1 | DAC_SWEEP | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP20 |
| 1 | DA_OUT | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP22 |
| 1 | DUT G / D / S | TerminalBlock_Phoenix:TerminalBlock_Phoenix_MKDS-1,5-3_1x03_P5.00mm_Horizontal | J6 |
| 1 | DUT_D | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP19 |
| 1 | DUT_G | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP18 |
| 1 | FB_SENSE | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP4 |
| 1 | GATE_AMP | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP16 |
| 1 | GNDPWR | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP27 |
| 1 | GND_STAR | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP9 |
| 1 | HL2303 | Package_TO_SOT_SMD:SOT-23 | Q3 |
| 1 | IN- (pad) | TestPoint:TestPoint_Pad_D1.0mm | TP6 |
| 1 | KBUF_OUT | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP24 |
| 1 | KELVIN HI / LO | Connector_PinHeader_2.54mm:PinHeader_1x02_P2.54mm_Vertical | J5 |
| 1 | KELVIN_HI | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP13 |
| 1 | KELVIN_LO | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP14 |
| 1 | LOAD | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP5 |
| 1 | MMBT3904 | Package_TO_SOT_SMD:SOT-23 | Q2 |
| 1 | NetTie | NetTie:NetTie-2_SMD_Pad0.5mm | NT1 |
| 1 | NetTie — star point | NetTie:NetTie-2_SMD_Pad0.5mm | NT2 |
| 1 | OPA_SWEEP_OUT | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP1 |
| 1 | PTC 100mA hold | Fuse:Fuse_1206_3216Metric | F1 |
| 1 | PTC_OUT | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP17 |
| 1 | RANGE_FORCE | curve-tracer:PinHeader_2x03_P2.54mm_Vertical_RangeSelect | J3 |
| 1 | RANGE_SENSE | curve-tracer:PinHeader_2x03_P2.54mm_Vertical_RangeSelect | J4 |
| 1 | SHUNT_HI | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP10 |
| 1 | SHUNT_LO | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP11 |
| 1 | SMAJ12A | Diode_SMD:D_SMA | D3 |
| 1 | VDIV | TestPoint:TestPoint_Keystone_5000-5004_Miniature | TP23 |
| 1 | VIN (pad) | TestPoint:TestPoint_Pad_D1.0mm | TP28 |

nunosousa/SCART-switcher
An old school SCART switcher. Its main purpose will be to automaticaly detect and switch the video output to the active input video source.

mahmoud-elghamry/thermocouple
This directory is a self-contained KiCad 10 project. It does not share files with, and must not be confused with, the single-channel board in hardware/single-channel/.

mjm-12/transistor-hfe-measurement-circuit
Transistor hFE measurement circuit (KiCad, CERN-OHL-P-2.0)


dyknapp/-ion-optics-hv
Open Eurocard-HV-piezo-driver.kicadpro with KiCad 10.0.7 or a compatible later version. The project contains six schematic sheets and a four-layer PCB.

gallicchio/op-amp-pcb
The op-amp of Learning the Art of Electronics (Hayes and Horowitz, 2016), Lab 5L, built from six 2N3904 / 2N3906 transistors on a 100 x 100 mm board whose silkscreen is the schematic.


mkol-ME/A1-4RD
Local voice assistant with a character: Whisper and a 26B LLM on a reflashed datacenter GPU, a custom voice, a 3D-printed body and a KiCad PCB
It measures 135 × 92 mm and has 2 copper layers, 1.6 mm thick.
88 parts, built around the CC0603KRX7R9BB104, RT0603BRD0710KL, 0603B103K500NT and RT0805BRD071KL. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.
From the leedanie2/curve-tracer repository on GitHub, at the commit this page was made from: the board is hardware/curve-tracer.kicad_pcb and its schematic hardware/curve-tracer.kicad_sch. PCBKit opens the board in the browser, with nothing to install.
Its repository states no licence, so its author keeps all rights: ask leedanie2 before reusing the design.