Drogheda/03-Architecture-Hardware/watchdog-proto-bench-bringup-v3.md

title	tags	project	author	date	status
Watchdog v3 — Proto Bench Bring-up & Build-Readiness (DEV board)	project-drogheda hardware watchdog prototype breadboard bring-up readiness	Project Drogheda	Helgen CTO	2026-06-24	Active — tomorrow's bench guide (supersedes the procedure half of watchdog-protoboard-bringup.md for v3)

Watchdog v3 — Proto Build Readiness & Bench Bring-up

Scope: breadboard the DEV watchdog on the existing Blackpill and prove the state machine + sensing + load switch + fan/temp loop. PCB layout is explicitly out of scope for tomorrow (do it after the proto works). Authority: watchdog-electrical-design-state (canonical) · firmware 04-Source-Code/firmware/blackpillrtkcode_v2/ (v2.3) · schematic kicad/hbox-watchdog-v3.kicad_sch.

---

0. Verdict — are you ready to build? Yes, with three caveats

You can build and bring up the core watchdog tomorrow. Nothing about the schematic or its ERC state blocks a breadboard build. The caveats are about firmware coverage and scope discipline, not readiness:

1. ERC "issues" are cosmetic — not a blocker. File-based ERC on hbox-watchdog-v3.kicad_sch = 0 errors, 1 warning (a Q_NPN_BCE symbol-cache mismatch). The ground net is unified (4 GND symbols, 0 placed GNDREF instances — the GNDREF you may see is just an orphaned library-cache entry, not a real second ground). None of this affects a breadboard. (Clean it only before you go to PCB layout — see §2.)
2. Two v3 sub-circuits aren't in firmware yet. The flashed v2.3 build does not drive the 4-LED battery gauge (PB12–15) or CHRG_EN (PA8). You can wire them, but they'll sit dark/idle until a firmware pass. Decide per §4 whether to build them now or defer.
3. v2.3 has never been hardware-validated. Repo invariant: bench-test before flash. Tomorrow is its first real bench test — treat step 1 (flash + confirm telemetry) as the gate before trusting any later step.

The one thing I couldn't verify: your exact purchased parts. The two proto BOM .xlsx (watchdog-protoboard-BOM-3boards.xlsx, watchdog-LISTA-DE-COMPRAS-ES.xlsx) are locked open in LibreOffice, so I rebuilt the checklist (§3) from the design docs. Close LibreOffice and I'll diff your actual list against it.

---

1. What you can prove tomorrow (firmware v2.3 supports it) vs. not

Sub-circuit	Pins	v2.3 firmware	Testable tomorrow?
Boot + $PHLG/STREAM telemetry (USB)	USB-C	✅ ENABLE_PHLG=1	Yes — this is the gate
Status LED (blink codes)	PB0	✅	Yes
RTK-stack load switch (the recovery heart)	PA1 → 2N3904 → IRF9540	✅ OFF-at-reset, MCU-asserts-ON	Yes
Sense: battery / 5 V / ignition	PA0 / PA4 / PA5	✅ hysteresis + debounce	Yes
Smart fan + tach	PA6 / PB10	✅ 25 kHz PWM, RPM, stall fault	Yes (3-wire fan)
I²C temp/RH (closed-loop fan)	PB6/PB7 → SHT40 0x44 + TMP102 0x48	✅ + MCU-temp fallback	Yes (fallback works even with no sensor)
4-LED battery gauge	PB12–15	❌ not coded	wire OK, stays dark
Temp-qualified charge cut (CHRG_EN)	PA8	❌ not coded	No (also LTC4040/PROD-only)
Reset (hardware)	NRST	n/a (HW only)	Yes — button just works

Fan note: v2.3 reads a 3-wire tach on PB10, so use a 3-wire 5 V fan on the breadboard. (The "2-wire production fan" decision in the canonical doc is a PCB change; it doesn't apply to tomorrow's bench.)

---

2. The ERC situation (so you can stop worrying about it tonight)

• What ERC actually reports on the DEV file: 0 errors, 1 warning (Q_NPN_BCE not found in lib Device). That warning means the symbol in your schematic is from the project cache and the installed Device lib has a slightly different name/pins — purely a library-link cosmetic.
• The GNDREF you may have seen: there are no placed GNDREF symbols — only the leftover symbol definition in the embedded cache. Your ground is one net.
• Why it doesn't block the breadboard: ERC validates a netlist for fabrication. On a breadboard you tie all grounds together physically; off-grid placement and a cache-name mismatch are invisible to copper you wire by hand.
• Before PCB layout (later), clean these in KiCAD: Tools → Update Symbols from Library (clears the Q_NPN_BCE cache warning), then Annotate and re-run ERC; nudge any off-grid parts onto the 1.27 mm grid. Don't let me edit the schematic while you have KiCAD open (file-lock conflict) — close it first and I'll do the cleanup pass.

---

3. Pre-build parts checklist (reconstructed — confirm against your box)

Already have (don't buy): Blackpill STM32F411, USB-C cable, 18650 + PowerBoost (or bench PSU), Mean Well SD-15B-05 (or bench 5 V), multimeter, a heat source (hair-dryer/cup of hot water) for the fan test. F9P + 4G master are NOT needed for this bring-up.

Core (the "brains" — must-have):

• Breadboard ×2 (830-pt) + jumper kits (M-M and M-F)
• Resistor assortment incl. 330 Ω, 1 k, 4.7 k, 10 k, 100 k, 220 Ω
• 2N3904 ×3 (load-switch driver, TO-92)
• IRF9540 / IRF9540N ×2 (P-FET high-side 5 V switch, TO-220)
• LED ×5 (status) · 100 nF ceramic ×10 (ADC filters/decoupling) · 100 µF ≥16 V ×4 (bulk)
• 10 kΩ trim pot ×2 + tactile button/slide switch ×2 (sweep VBAT / simulate ignition)
• Male header strip 1×40 (to pin any bare module)

Thermal (v3 new):

• 3-wire 5 V fan, 40–60 mm (3-wire so PB10 reads RPM — matches v2.3)
• IRLZ44N ×2 (logic-level N-FET, fan low-side) — plain IRF540/IRFZ44 won't turn on from 3.3 V; if out of IRLZ44N use 2N2222A/TIP120 + 1 kΩ base
• 1N5819 ×3 (fan flyback; 1N4007 acceptable)
• I²C SHT40/SHT31 breakout (0x44) — or LM35/DS18B20 fallback (tell me which; I'll patch the read)

v3 additions (build only if you want them wired now — see §4):

• Reset: tactile button → NRST↔GND (+ optional 2-pin header for external reset)
• Battery gauge: 4× LED + 4× 330 Ω on PB12–15 (will stay dark until firmware)
• Battery-backup stand-in: TP4056 module + MT3608 boost + 2× 1N5819 (OR-ing) + 10 kΩ NTC (reproduces charge / power-loss→backup / BAT_SENSE; the temp-qualified cutoff is LTC4040-only → firmware-validated, not on the proto)

---

4. Decision before you sit down: how much of v3 to wire

• Recommended for tomorrow — build the testable core + thermal (everything in §1 marked Yes). It proves the whole watchdog logic and the new fan/temp loop, which is the point of the proto. Add the reset button (free — works with no firmware).
• Gauge (PB12–15): wire it if you want the layout/feel, but it won't light until I write the gauge driver. Cheaper to add later. — I can write that firmware pass tonight/tomorrow if you want the LEDs live.
• Battery-backup stand-in (TP4056+MT3608): worth wiring if you specifically want to watch power-loss → backup hand-off and BAT_SENSE track the cell. Skip if tomorrow is about the watchdog logic only.

---

5. Bring-up sequence (run in order — each step gates the next)

1. USB only — flash & gate. Compile v2.3 (Arduino IDE / STM32duino, board = Blackpill F411CE), flash via DFU (hold BOOT0). Open serial @115200. Confirm BOOT / COLS / TLM lines and a $PHLG,...*CS every 30 s. If this fails, stop and fix firmware before wiring anything.
2. Status LED — PB0 → 330 Ω → LED → GND. Watch blink codes change with state.
3. Load switch (the recovery heart) — PA1 → 1 kΩ → 2N3904 base; collector → IRF9540 gate; 10 kΩ gate→5 V; drain → switched 5 V; put an LED+330 Ω on the switched output. Confirm: OFF at reset, ON when PA1 asserts. (Power-off is the safe state.)
4. Sense dividers — 10 k/10 k ÷2 into PA0 (battery via 18650/pot) and PA4 (5 V rail); 10 kΩ pot into PA5 (ignition). Sweep and watch veh/ign/vbat flip in TLM with hysteresis (V5V_ON 4.20 / V5V_OFF 3.80; BAT_CRIT 3.40 / BAT_REC 3.70).
5. Fan + tach — PA6 → 220 Ω → IRLZ44N gate; fan(−) at the drain; 1N5819 ∥ fan; 100 kΩ gate→GND. 3-wire tach → 10 kΩ pull-up to 3V3 → PB10. Watch it spin/ramp; unplug fan → fan_fault in telemetry.
6. I²C temp — SHT40/TMP102 on PB6/PB7 (+4.7 kΩ pull-ups each). Heat it → fan ramps on the 45→70 °C curve. No sensor? The MCU-temp fallback still ramps the fan — warm the whole board.
7. (Optional) reset button — NRST↔GND momentary; press → clean reboot, reset-cause line in log.
8. (Optional) battery-backup stand-in — vehicle 5 V → TP4056 IN; cell on TP4056 BAT; TP4056/cell → MT3608 → 5 V via 1N5819 OR-ing into the rail; NTC by the cell. Pull vehicle 5 V → watch backup hold the rail and BAT_SENSE track.

---

6. Wiring cheat-sheet (Blackpill side, v3)

PA0  ← BAT  ÷2 (10k/10k) from 18650+ / pot      PA6  → 220Ω → IRLZ44N gate; fan(−); 1N5819 ∥ fan
PA4  ← 5V   ÷2 (10k/10k) from 5V rail            PB10 ← fan tach, 10k pull-up → 3V3
PA5  ← IGN  ÷2 (10k/10k) from pot/switch         PB6/PB7 → I²C SCL/SDA (+4.7k each → 3V3)
PA1  → 1k → 2N3904 base; collector → IRF9540     PB0  → 330Ω → LED → GND
        gate; 10k gate→5V; drain → switched 5V    NRST ← reset button → GND   (optional)
        → stack (LED+330Ω as a test load)         PB12–15 → 330Ω → LED → GND  (gauge; dark till FW)
GND / 5V / 3V3 rails common.  Fan 5V from a SEPARATE 5V (2nd PSU / LM2596) — not the switched stack rail.

Proto↔PCB part map (don't get confused): the schematic uses SMD equivalents (Q1 AO3401A P-FET, Q5 MMBT3904, Q2 AO3400A fan FET). On the breadboard you use the through-hole stand-ins above (IRF9540N / 2N3904 / IRLZ44N) — same topology, same firmware.

---

7. After the proto works — next steps (not tomorrow)

• Firmware pass: gauge-LED driver (PB12–15 from BAT_SENSE) + CHRG_EN cut-charge-when-hot logic.
• KiCAD: cosmetic ERC cleanup → footprints → then PCB layout (DEV first, then build PROD per watchdog-production-schematic-delta).
• Pack Ah confirmed ~12 Ah (2026-06-30; "40 Ah" label was wrong) — sizes the LTC4040 charge current (≈C/5, ≤2.5 A).

Related

watchdog-electrical-design-state (CANONICAL) · watchdog-protoboard-bringup (Jun-22 shop list) · watchdog-production-schematic-delta (PROD build spec) · kicad/hbox-watchdog-v3.kicad_sch