Import a PCB. Find out what you can actually simulate.
A KiCad PCB file gives us pad, net and part information. It does not by itself provide behavioral models for every chip. LabWired can import supported source files, resolve modeled parts, and test circuits whose required behavior is implemented. Parts without a supported model must stay visible in the coverage report, not silently disappear.
From PCB file to simulation evidence
- Import the source. Use the existing
labwired_importMCP tool with an explicitsource_kindand file content. The hosted expert importer also accepts asource_urlfor supported public HTTPS KiCad or Altium files, including GitHub blob/raw file links (import implementation). Import preserves source references and nets where the parser supports them; check the report for unresolved mappings. - Inspect model coverage. Distinguish mapped parts, unmodeled active devices and missing MCU/peripheral support. A parsed board is not yet an executable board.
- Run supported behavior. The hosted expert
labwired_runcan execute an import-producedanalog_simulationwith native ngspice for supported transient R/L/C networks and sources. This does not run unmodeled ICs. Supported MCU firmware checks are a separate verification path; nonlinear devices and vendor SPICE models require an appropriate native model, not the browser's linear adapter. - Verify with a negative control. State input stimuli, observe waveforms or firmware outputs, and confirm the test fails when a relevant component or connection changes.
- Report what remains unknown. A missing regulator, sensor or MCU model blocks whole-board verification even if a smaller subcircuit passed. KiCad design-rule checks are not a substitute for circuit-behavior tests.
Reproduce a transient import test, including negative controls
The hosted analog proof runbook
and test script
specify the import, labwired_run measurement, and refusal cases. The following
values are test expectations, not measured production results. For an ideal 5 V step,
the RC node should follow V(t) = 5 × (1 − exp(−t / RC)).
| Input and model | Expected result at 10 ms | What it verifies |
|---|---|---|
| 5 V step; 10 kΩ and 100 nF | RC node ≈ 4.999773 V | Transient charge response, baseline |
| 5 V step; 100 kΩ and 100 nF | RC node ≈ 3.160603 V | Resistor perturbation changes the waveform |
| 5 V step; 10 kΩ and 1 µF | RC node ≈ 3.160603 V | Capacitor perturbation changes the waveform |
| Divider without an independent source | Explicit missing-source refusal | Missing testbench cannot produce a false pass |
| AC-only source in a transient request | Explicit unsupported-analysis refusal | AC sweep must not be presented as transient data |
The fixture contract requires finite sampled probes, ordered time, deck hashes, disclosed model gaps, and a 2 mV numerical tolerance. It includes a separate Altium divider case and an unsupported-U1 case. A successful live workflow is needed before treating these expected values as deployed measurements. Even a passing transient proves only the modeled network, not a whole board.
ESP32 circuit simulation: check the module, not just the chip
Two original HeyPCB ESP32-S3 sensor boards exposed a package-mapping defect. The bare WROOM module and the DevKitC carrier use different physical pin orders. WROOM pad 1 is GND and pad 2 is a 3.3 V input. The DevKitC header starts with two 3.3 V pins. Substituting the carrier created a false power-to-ground short in the imported BME280 design.
The pinned ESP32-S3-WROOM-1-N16R8 acceptance record
documents the exact 41-pad module mapping and independent negative controls.
The LDR test also preserves the GL5528 photoresistor instead of substituting a fixed 220 Ω resistor.
Both original PCBs use source revision 57a2122a19ac1aa9c092e43fb5b123c461508752.
| Original PCB | Mapped / dropped / ignored | Mapped terminals checked | Net-pair checks |
|---|---|---|---|
| ESP32-S3 LDR, 22 parts | 17 / 4 / 1 | 35 | 595 |
| ESP32-S3 BME280, 24 parts | 19 / 4 / 1 | 44 | 946 |
Net-pair checks compare whether mapped terminals share a source net or remain separate. They do not test every component's behavior. Negative controls reject the wrong package, conflicting pin names and connected reserved PSRAM pads. A deliberately added ground-to-3.3 V short still fails validation.
Evidence snapshot: 12 November 2026, unmerged engineering branch. PR #3164 produces simulation manifests for both test fixtures. Neither board has a verified original-firmware browser run or whole-board electrical result. The AP2112K-3.3 regulator, USBLC6-2SC6 protection IC, BOOT switch and RESET switch remain explicit model gaps. The module remains non-runnable until firmware, memory, supply and reset acceptance pass.
Start with a supported browser circuit demonstration while these complete-board models remain unverified. Check current Asset Library acceptance evidence before selecting an imported board.
Open KiCad acceptance study: HeyPCB
An experimental LabWired acceptance change tested seven open PCB files from HeyPCB/hello-8051 (MIT) and HeyPCB/power-kit (CERN-OHL-S-2.0). All seven files imported in that test. None was verified as a complete, operating board. These are results from an unmerged engineering branch, not a production coverage claim.
| Board | Import test | Whole-board simulation gap |
|---|---|---|
| hello-8051 | 15 mapped, 7 dropped, 5 ignored | AT89S52 MCU, ST7920 LCD, SRAM and logic models |
| power-kit: dual-rail | 26 mapped, 3 dropped, 5 ignored | TPS563201 and AP63203 regulators; AO3401A |
| power-kit: PD trigger | 12 mapped, 3 dropped, 2 ignored | HUSB238 controller, MOSFET and selector |
| power-kit: boost-1s | 11 mapped, 1 dropped, 1 ignored | MT3608 switching behavior, SS34 diode |
| power-kit: power mux | 12 mapped, 1 dropped, 2 ignored | TPS2116 behavior |
| power-kit: USB eFuse | 12 mapped, 1 dropped, 2 ignored | TPS259251 protection behavior |
| power-kit: USB meter | 4 mapped, 1 dropped, 3 ignored | INA226 measurement behavior |
The counts are importer classifications from the pinned test corpus in PR #2934, not percentages of modeled functionality. Passing import does not prove copper equivalence, correct device behavior or physical hardware acceptance.
Independent source pad/net audit
A separate audit of the seven pinned KiCad source files counted 129 footprints and 575 physical pads. Of those pads, 510 have a net assignment and 65 have no assigned net. All 129 source footprints reconcile with the experimental import classification: 92 mapped, 17 dropped and 20 ignored. These are source-side counts; they do not prove that imported pins or nets match the source.
The KiCad
pad/net inventory checker records a SHA-256 digest of each source board's physical
pad number, component reference and assigned net. The source verification command below
re-fetches immutable upstream files, validates their Git blob hashes and checks the
recorded pad/net digests. Comparing those digests with a real labwired_import
output is still pending; imported net and pin completeness remain unmeasured.
Reproduce the source and coverage audit
The public HeyPCB acceptance manifest records each PCB's exact Git commit, upstream license and file blob SHA. A small source-integrity and verdict checker rejects a whole-board PASS when component models, net/pin completeness or control tests are missing. It independently checked all seven pinned PCB source blobs and both license files (nine files). This checks input provenance and evidence classification, not live simulator execution.
To check the immutable open-source files yourself, clone
LabWired Examples and run
python3 scripts/heypcb_evidence.py --verify-source. The 8051 source
is pinned under MIT; the power-kit source
is pinned under CERN-OHL-S-2.0.
Evidence stages are separate: fetched → parsed → imported → mapped → analog simulated → firmware verified → whole-board verified. Here, import passed for seven files; part mapping was partial, and the whole-board verdict remained unsupported for all seven. Net and pin completeness were not measured. The 239 indexed candidates in the separate, unmerged source catalog are not verified imports.
One circuit that the acceptance study could check
The boost-board import preserved a 15 kΩ / 2 kΩ feedback divider. A separate Arduino UNO observer tested that divider with an assumed ideal 5 V supply. Its positive control passed the expected firmware marker; swapping the resistor values caused the negative-control assertion to fail. This checks the divider and the observer, not the MT3608 converter or the original board's power behavior.
What would count as a simulated PCB?
A supported circuit needs resolved part identities and connections, power and device models, applicable MCU firmware when present, meaningful test inputs, and measured outputs. A whole-board verdict also needs to account for all required active components and relevant faults. If one is unknown, the result is partial or unsupported—not a successful whole-board simulation.
The browser's analog adapter currently supports linear circuits; complex nonlinear and vendor SPICE models need the native co-simulation path. Read the mixed-signal reference or request a feasibility review for your board. HeyPCB is an independent open-hardware publisher and does not endorse LabWired.