Skip to the content
2 layers144 × 150 mmDigifant2-pruefstand, the printed circuit board in 3D

Automated test bench for the VW Digifant-2 ECU: Raspberry Pi Pico signal simulator and capture board (KiCad + firmware)

Size
144 × 150 mm
Copper layers
2
Thickness
1.6 mm
Parts
78
Nets
70
Pads
490
Tracks, vias
752, 78
Pours
4
Routed
Yes
KiCad
KiCad 10.0 (file 20260206)
Created
10 Sep 2026
Updated
30 Oct 2026

Overview

Digifant2-pruefstand is an open-source printed circuit board by designer2k2, from the designer2k2/digifant2-pruefstand repository on GitHub, published under the MIT licence. It is a 2-layer board of 144 × 150 mm with 78 parts on 70 nets, routed with 752 tracks and 78 vias.

It is built around the TS5A3159ADBV, AD9833BRM and SK6812, with the SS54 among its other parts. Its parts include 23 resistors, 10 capacitors, 8 ICs, 4 transistors and 4 connectors. It belongs with the RP2040 boards of this library.

Main components

4

README

README.md

Digifant-2 Prüfstand

CI Board

Hardware for a fully-automated test bench for the VW Golf 1 Cabrio Digifant-2 ECU: signal simulator + capture front-end, driven by a Raspberry Pi Pico 2, with a high-side load switch so the bench can cold-boot the ECU and cut power on overcurrent. Scope here is the Signal-Simulator + Erfassungseinheit only — the EPROM emulator is a separate future board.

The schematic is generated from Python with circuit-synth; the PCB is placed and autorouted headless (pcbnew + Freerouting).

firmware/OPERATION.md — how to operate the bench: display and buttons, PC control and scripting, reading the measurements.

BRINGUP.md — step-by-step first power-on checklist, tools needed, and the open questions to settle on the bench.

DESIGN.md — what every block does, part-selection rationale, the GPIO / I²C / VW-harness maps. USE_CASES.md traces every real bench use case through the schematic to concrete numbers (voltage/current/timing headroom) — a requirements/test matrix for the hardware, and where the AD9833 decoupling-cap gap turned up before it got fixed.

Layout

pathwhat
digifant2_pruefstand.pythe circuit — one @circuit function, flat single sheet
gen_custom_symbols.pybuilds kicad-symbols/Custom_Digifant2.kicad_sym (INA226, AD9833, 2N7002, DPAK P-FET, VW connector — pinouts verified against datasheets)
gen_project.pypython3 gen_project.py <name> → full KiCad project + netlist + lib tables, runs ERC/DRC via kicad-cli, auto-applies no-connect flags
place_and_route.pypcbnew functional-zone placement → Specctra DSN → Freerouting → SES import (run inside the KiCad flatpak; java runs outside it)
add_nc_flags.pystamps no_connect flags from an ERC report (circuit-synth has no NC API)
fix_layout.pypost-route layout surgery (real netclasses, GND/Kelvin re-routing) — see the header comment for the 2-phase, 2-process invocation it needs
reroute_missing.pyre-exports a Specctra DSN from a partially-ripped-up board so Freerouting only has to fill the reopened ratsnest
firmware/Pico firmware (Pico SDK, C): USB serial command protocol, build/flash instructions, what's implemented vs. stubbed
host/bench.pyPC-side client for the firmware protocol: interactive prompt, one-shot CLI, or importable Bench class; host/examples/ has a scripted rpm sweep
board/the current KiCad project (144×150 mm, 2-layer, 4× M3 corner holes, isolated OLED + USB clearance zones)
board/bom_output/bom.htmlinteractive BOM (InteractiveHtmlBom) — click a part to highlight it on the board, no KiCad needed
BOM.mdMouser sourcing: part numbers for every line item, with confidence notes — 6 of 30 need manual confirmation, see the file
mouser_cart.csvonly the Mouser-verified part numbers, ready to paste into Mouser's bulk order tool
JLCPCB_ASSEMBLY.mdhybrid plan: JLCPCB machine-places the SMD parts, you hand-solder THT + 4 unconfirmed SMD parts — read this before submitting
jlcpcb_bom.csv / jlcpcb_cpl.csvJLCPCB SMT-assembly BOM + pick-and-place, 42 of 46 SMD parts (real positions from board/, real LCSC part numbers)
UPSTREAM_BUGS.mdcircuit-synth bugs found, drafted for upstream filing
kicad-symbols/, kicad-footprints/libraries — custom/community committed, stock via ./fetch_libs.sh
.github/workflows/CI: firmware build + tests, spell check, KiCad ERC/DRC, firmware releases (see below)

Build

sh
./fetch_libs.sh                        # one-time: pull stock KiCad 9.0 libs
python3 gen_custom_symbols.py          # if the custom symbols changed
python3 gen_project.py digifant2_pruefstand_cs_v1

Placement + routing needs KiCad's pcbnew Python API (KiCad ≥ 10) and a non-headless JRE for Freerouting — see the header of place_and_route.py.

Status

Schematic ERC-clean; VW harness pinout verified against the cabby-info.com Digifant II reference and the user's own hands-on HiL notes (see DESIGN.md). Board placed + autorouted (144 × 150 mm, 2-layer, 4 × M3 holes, isolated OLED + USB clearance zones), real Power/Gnd netclasses, GND re-routed as one connected net (was 4 islands), dedicated Kelvin sense traces for both shunts. The tight U1/U3/RS1/RS2 pocket (2 shorting_items, 2 clearance, 5 tracks_crossing left by the scripted Kelvin routing) was cleaned up by hand in KiCad's interactive router, and a B.Cu GND pour was added. Added the two AD9833 decoupling caps (C7, C8), fully routed. Fixed RS2 (idle-valve shunt) to be truly in series with the ECU's VW-23 driver instead of a parallel path to GND that was corrupting the current reading — this reopened one small pocket at U3 pin 9 (2 shorting_items, confined to that one spot after extensive hand-routing/Freerouting attempts, same class of residual as the historical U1/U3/RS1/RS2 items). Gave the AD9833 knock output a defined path out (TP1/TP2 test points for VW-4/VW-5) instead of a dead end. Added 12 more THT scope-probe test points (TP3–TP14, real drilled holes — an earlier pass used bare SMD pads and put 10 of them physically underneath the Pico module, both corrected) on the main GPIO signals, both buses, and both 12 V rails. Found and fixed a real bug along the way: footprints cloned via pcbnew.FOOTPRINT(template) inherit the template's UUID unless you explicitly reset it, which had corrupted DRC's item-identity tracking for every cloned part this session (C7/C8 and all of TP1–TP14) — several "violations" turned out to be real defects reported against the wrong footprint. See DESIGN.md's "Known open points" for details if you script further additions. Remaining warnings are cosmetic (silkscreen overlap, lib_footprint_mismatch, and one starved_thermal note on a single Pico GND pad with only 1 pour spoke instead of 2 — connected fine, just a slightly weaker thermal/mechanical joint). Thermal copper for D1/Q2 still optional to add by hand. Added a 12×16 perfboard-style proto area (PROTO1, isolated THT pads on 0.1″ pitch) in the empty space bottom-right of J0. Checked the whole board against JLCPCB's published fab minimums (trace/spacing, via/hole sizes, hole-to-hole, copper-to-edge, board thickness) — everything copper- and hole-related had 1.5×–14× margin already. The one real miss was silkscreen: KiCad's default 0.1mm line width and 0.8mm text height are both below JLCPCB's stated minimums (0.15mm / 1.0mm), risking thin/dropped silkscreen. Bumped the project defaults and widened/enlarged every existing silkscreen line and text item on the board to meet those minimums.

Checked +3V3 for adequate trace width (it's on the thinnest Default netclass, 0.2mm, despite feeding ~15 chips) — worked out the real numbers (thermal capacity ~0.77A at 10°C rise, worst-case drop a few mV) and it's fine as-is; not worth the rerouting risk in this densely-packed board. Checked decoupling instead: per each part's own datasheet, U1 and U3 (the two INA226 current-sense amps) had no bypass cap at all, unlike U2/U4 which already have theirs (C6, C2/C7/C8). Added C9 (U1) and C10 (U3), 100nF 0805, tapped directly off the existing +3V3 B.Cu trunk and the GND pour a few mm south of the packed U1/U3/RS1/RS2 pocket (no room to place them right at the pins) — verified DRC-clean (the only residuals are the same new-copper-vs-stale-zone-fill noise that clears on a zone refill, same as after the TP resort). U6/U7/U8 (the sensor-disconnect analog switches) are still without a bypass cap; lower priority since they're not noise-sensitive, but could use the same treatment.

Added the thermal copper for D1/Q2 that was flagged as a real (not cosmetic) risk in USE_CASES.md — D1 dissipates ~1.25W at sustained ~2.5A with no thermal relief otherwise. Both pours verified electrically connected to their pads (not just overlapping) and DRC-clean. Q2's is small since there's almost no open board space around its DPAK tab; D1's is generous since F1/RS1 (same nets, either side) leave a wide open corridor.

All 14 test points (TP1–TP14) are now fully routed — the user hand-routed 7 of the 10 remaining GPIO test points after the resort (TP3, TP10, TP11, TP4, TP12, TP5, TP6); the last 3 (TP9, TP8, TP7 — all on U5's east side, the tightest part of the board) were routed with a grid-based pathfinder (A* over a 0.1mm grid, spatial-bucket-indexed for speed) purpose-built for this corridor after several rounds of manual/heuristic routing kept snagging on real obstacles: the Pico's own pin column has no gap wide enough for a second trace anywhere along its length (confirmed by direct probing — routed escape stubs from other GPIOs fill in what would otherwise be a 0.49mm theoretical gap between pin pads), so both TP8 and TP9 had to detour around the south end of the Pico footprint entirely, and — since that detour corridor is itself only wide enough for one trace at a time — needed genuinely different physical crossing points, not just parallel lanes. Verified via real kicad-cli pcb drc, not just the pathfinder's own clearance model (which undercounted real collisions during development in two ways worth noting for next time: it modeled non-circular pads by their half-width instead of their half-diagonal, and it didn't check hole-to-hole spacing against a net's own via near its own pad — both are now fixed in the approach, see DESIGN.md). GND zone refilled after all of the above (TP9/TP8 routing, C9/C10, D1/Q2 thermal pours) — DRC dropped from 136 to 101 violations, all pre-existing cosmetic categories plus the one known starved_thermal note, zero clearance/short/crossing issues. (The "whole-board refill introduces ~115 spurious violations" finding earlier this session was a testing artifact — running kicad-cli against a bare copy of the .kicad_pcb without its paired .kicad_pro silently falls back to stricter default netclass rules; with the project file present, the refill is clean.)

Trimmed the board from 144×169mm to 144×150mm — the bottom 19mm was completely empty (bare GND pour, nothing below PROTO1/J3 until the old mounting holes). H3/H4 moved up to keep the same 4.5mm hole-to-edge margin as H1/H2; DRC stayed clean (99 violations, same pre-existing cosmetic set).

Added functional silkscreen labels next to J0 (the VW harness connector) and every test point, instead of leaving bare unlabeled pads. J0 gets each pin's real VW harness number (VW-2, VW-9, ... or GND / +12V SW) in the clear strip just outside its footprint. The TPs already had descriptive text (e.g. GP2_CRANK) sitting on them from earlier in the session — turned out it was on F.Fab, a documentation-only layer that doesn't actually get printed on the physical board, so none of it was ever going to be visible with a probe in hand. Moved it to F.SilkS and shortened the long ones (e.g. GP6_THROTTLE_IDLE_SW → IDLE SW) since the west-margin TP column only has about 10mm of clearance between the board edge and the Pico's pin column — the original full-length strings would have run off the board edge once actually rendered as real silkscreen (checked via real bounding-box widths, not by eye). This did add some new silk_overlap hits — mostly TP labels sitting close to the Pico module's own tiny per-pin silkscreen text in that same tight corridor — all cosmetic, same accepted category as the rest of this board's silkscreen warnings, verified via DRC to introduce zero clearance/short/crossing issues.

Added fab-identification text on the back silkscreen (B.SilkS, previously completely empty — no back components, nothing on that layer at all): project name, revision, date, and designer, so a bare board can still be identified later. Set via the board's title block (${TITLE}, ${REVISION}, ${ISSUE_DATE}, ${COMPANY}) rather than hardcoded text, so updating the revision/date for a future respin is a one-place edit in KiCad's Page Settings, not a silkscreen re-edit. Verified the variables resolve correctly (checked the exported SVG's embedded text, since this sandbox can't render the actual glyphs) and confirmed zero new DRC violations. Not yet fabbed.

Continuous integration

Every push and pull request runs on GitHub Actions:

WorkflowWhat it checks
CI (ci.yml)Builds the firmware with Pico SDK 2.1.1, warnings as errors, and uploads pruefstand.uf2 as a build artifact. Runs firmware/test/run_all.sh: all PC unit tests, the command parser, host/bench.py over a virtual serial port, and the crank PIO program in an emulator.
Spell check with codespell over code, comments and docs (.codespellrc).
Board (board.yml, only when board/ changes)KiCad 10 ERC, and DRC with schematic parity. Fails on errors (unconnected nets, clearance, shorts); the known cosmetic warnings are only reported. Full reports are attached to the run.

Before opening a pull request, the same checks run locally with firmware/test/run_all.sh (after a firmware build) and codespell.

Firmware releases: bump FW_VERSION in firmware/src/cmd.h, commit, then push a matching tag, e.g. git tag fw-v0.9.0 && git push origin fw-v0.9.0. CI checks the tag against FW_VERSION, runs everything above, and publishes a GitHub Release with pruefstand-<version>.uf2 to download and flash.

License

MIT (see LICENSE) for everything authored here — scripts, custom symbols, the board design. Bundled third-party files (the Raspberry Pi Pico symbol/footprint from ncarandini/KiCad-RP-Pico, CC-BY-SA-4.0) keep their own licenses — see NOTICE.

Parts

78
QtyValueFootprintReferences
8100nFCapacitor_SMD:C_0805_2012MetricC2, C4, C5, C6, C7, C8, C9, C10
5100kResistor_SMD:R_0805_2012MetricR16, R19, R22, R23, R24
4220RResistor_SMD:R_0805_2012MetricR2, R3, R4, R10
4MountingHole_3.2mm_M3MountingHole_3.2mm_M3H1, H2, H3, H4
310kResistor_SMD:R_0805_2012MetricR8, R13, R21
31kResistor_SMD:R_0805_2012MetricR5, R15, RPU1
32N7002Package_TO_SOT_SMD:SOT-23Q1, Q3, Q5
3330RResistor_SMD:R_0805_2012MetricR9, R11, R17
34.7kResistor_SMD:R_0805_2012MetricR6, R7, R14
3PUSHButton_Switch_THT:SW_PUSH_6mmSW1, SW2, SW3
3TS5A3159APackage_TO_SOT_SMD:SOT-23-6U6, U7, U8
1+12V ECUTestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP13
1+12V SWTestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP14
10.02R 1% shuntResistor_SMD:R_2512_6332MetricRS1
10.033R 1% shuntResistor_SMD:R_2512_6332MetricRS2
1100RResistor_SMD:R_0805_2012MetricRG1
1100uF/25VCapacitor_THT:CP_Radial_D8.0mm_P3.50mmC1
110nFCapacitor_SMD:C_0805_2012MetricC3
112V BENCH INTerminalBlock:TerminalBlock_bornier-2_P5.08mmJ1
1200RResistor_SMD:R_0805_2012MetricR12
120kResistor_SMD:R_0805_2012MetricR20
125MHzOscillator:Oscillator_SMD_EuroQuartz_XO91-4Pin_7.0x5.0mmY1
13AFuse:Fuse_1206_3216MetricF1
1AD9833Package_SO:MSOP-10_3x3mm_P0.5mmU4
1CRANKTestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP3
1I2C SCLTestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP11
1I2C SDATestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP10
1IDLE SWTestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP4
1IDLE VALVE N71Connector_JST:JST_XH_B2B-XH-A_1x02_P2.50mm_VerticalJ2
1IGNITIONTestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP5
1INA226 (0x40)Package_SO:MSOP-10_3x3mm_P0.5mmU1
1INA226 (0x41)Package_SO:MSOP-10_3x3mm_P0.5mmU3
1INJECTORTestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP6
1KNOCK GND (to VW-5)TestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP2
1KNOCK OUT (to VW-4)TestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP1
1LED DATATestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP12
1MCP4728 (0x60)Package_SO:MSOP-10_3x3mm_P0.5mmU2
1OLED / UI I2CConnector_PinHeader_2.54mm:PinHeader_1x04_P2.54mm_VerticalJ3
1P-MOS -30V >=5APackage_TO_SOT_SMD:TO-252-2Q2
1PROTO AREAPROTO1
1Raspberry Pi PicoRPi_Pico:RPi_Pico_SMD_THU5
1SK6812LED_SMD:LED_SK6812_PLCC4_5.0x5.0mm_P3.2mmD2
1SPI CSTestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP7
1SPI MOSITestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP9
1SPI SCKTestPoint:TestPoint_THTPad_D1.5mm_Drill0.7mmTP8
1SS54Diode_SMD:D_SMCD1
1VW ECU HARNESSTerminalBlock:TerminalBlock_MaiXu_MX126-5.0-15P_1x15_P5.00mmJ0

Similar boards

  • 2 layers99 × 90 mmHomeyPCB, the printed circuit board in 3D

    HomeyPCB

    jadam3k/Homey

    Home Asistance/Industrial PLC

    99 parts
  • 2 layers80 × 65 mmRLCMeterH750 v2, the printed circuit board in 3D

    RLCMeterH750 v2

    balmerdx/RLCMeterH750

    Good RLC meter

    161 parts26
  • 2 layers50.8 × 22.9 mmWaveform Generator, the printed circuit board in 3D

    Waveform Generator

    CedarGroveStudios/AD9833_FeatherWing

    Precision Waveform Generator FeatherWing

    25 parts15
  • 2 layers50.8 × 22.9 mmWaveform Gen ADSR, the printed circuit board in 3D

    Waveform Gen ADSR

    CedarGroveStudios/PCB_AD9833_ADSR_Featherwing

    An amplitude-controlled version of the original AD9833 Waveform Generator FeatherWing.

    28 parts2
  • 2 layers65 × 85 mmFunction Generator, the printed circuit board in 3D

    Function Generator

    MUDAL/Function-Generator

    Custom Function Generator designed and developed by Olaoluwa Raji. This project is a complete embedded system solution with both hardware and software implemented by me. Relevant reference materials are included.

    30 parts1GPL-3.0
  • 2 layers127.1 × 61.8 mmImpedance analyzer, the printed circuit board in 3D

    Impedance analyzer

    Wang-ZiWen2006/impedance-analyzer-pcb

    Custom in-home impedance analyzer pcb

    16 parts1
  • 2 layers28 × 25.5 mmSlowdds, the printed circuit board in 3D

    Slowdds

    brightcloudy/pcb

    Some PCBs, hopefully

    34 partsMIT
  • 2 layers295 × 150 mmRippleCharger, the printed circuit board in 3D

    RippleCharger

    HabonRoof/SRC-Charging

    Sinusoidal Charging for Li-ion Battery

    129 parts

Questions

How big is the Digifant2-pruefstand PCB?

It measures 144 × 150 mm and has 2 copper layers, 1.6 mm thick.

What parts does the Digifant2-pruefstand use?

78 parts, built around the TS5A3159ADBV, AD9833BRM, SK6812 and SS54. 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.

Where can I download the Digifant2-pruefstand KiCad files?

From the designer2k2/digifant2-pruefstand repository on GitHub, at the commit this page was made from: the board is board/digifant2-pruefstand.kicad_pcb and its schematic board/Digifant2_Pruefstand.kicad_sch. PCBKit opens the board in the browser, with nothing to install.

Can I use the Digifant2-pruefstand design in my own project?

Its repository publishes it under the MIT licence: you may reuse it on that licence's terms (keep the author's notice, and for a copyleft licence share your changes alike).