Thermal enclosure¶
The bench validation confirmed both bridges thermally matched under sustained load with the boards on an open bench — clip-on TO-220 heatsinks per MOSFET, free convection. The bridge case-temperature delta was ≈ 3 °C in the demo session, well within safe operating range for the IRFB4110.
This roadmap item is the path to enclosed deployment — a chassis that the inverter can ship inside, with the thermal performance maintained.
What changes when you put the boards in a box¶
Free convection on an open bench dissipates 5–10 W per device pretty easily. Inside an enclosed chassis with no airflow, the same 5–10 W can take the junction temperature past spec within minutes. Three things change:
- Convection path is broken. Hot air collects above the boards; ambient temperature inside the box drifts up; thermal margin shrinks every minute.
- The transformer / LC filter (if added per LC filter) adds its own losses inside the same enclosure.
- EMI that radiated freely on the open bench now has to be contained — the chassis itself becomes part of the EMI strategy.
Enclosure dimensions¶
A starting envelope based on the as-built component footprint:
| Component | Approx dimensions |
|---|---|
| 2× single-bridge v4 PCB (≈ 150 × 100 mm each, end-to-end) | 320 × 100 × 30 mm |
| STM32 Nucleo-F303RE + standoffs | 90 × 70 × 20 mm |
| 2× isolated bench supplies (in production, integrated power-supply board) | 200 × 100 × 40 mm |
| LC filter board (planned) | 150 × 80 × 60 mm (the 15 mH toroid is the bulky part) |
| Cable strain-relief + terminal blocks | 50 mm depth allowance |
Rough enclosure: 300 × 250 × 150 mm. Wall-mount or DIN-rail mounting bracket.
Forced air vs. larger heatsinks¶
| Approach | Pros | Cons |
|---|---|---|
| Forced air (axial fan, 60 mm or 80 mm, 12 V) | Cheap (50–100 TL per fan), proven, scales with the load | Audible noise; fan is a failure point — needs a FAN_FAULT bit in firmware; ingress protection (dust filter) needed |
| Larger heatsinks (vertical extruded with fins) | Silent, no moving parts, no extra failure mode | Significantly more BOM cost (200–400 TL per heatsink); takes up enclosure height; only works up to ≈ 5–10 W per device of dissipation |
| Combined (smaller heatsinks + slow fan) | Best of both — fan runs only when temperature climbs | Adds a temperature sensor + thermostat circuit + control logic to firmware |
For the project's ≈ 400 W power-stage load with ≈ 2.4 W of MOSFET dissipation total, larger heatsinks alone should be sufficient. Add a fan if production load could scale higher.
Forced-air supervision (if a fan is added)¶
The firmware should add a FAN_FAULT bit to the protection chain:
- Tach input from the fan → GPIO → simple interrupt-driven pulse counter → expected RPM range.
- If RPM falls outside range for > 2 s, latch
FAULT_FANand proceed through the normal FAULT path (MOE off, FAULT_OUT pulled LOW, telemetry frame emits$F,0x40,...). - The existing FSM + protection infrastructure already supports adding fault bits —
protection.hdefines the bitmask.
Mechanical isolation¶
The four-layer ground-separation design (grounding fix) must survive the chassis integration:
- PCB standoffs must be insulating (nylon, not metal) so the bridge-side ground islands don't bond to the chassis through the mounting screws.
- Chassis ground bonds to the controller's 5V_GND only — never to any bridge-side ground.
- Cable shields that enter the enclosure must terminate to chassis ground at the entry point, not run inside to a board ground.
Failing any of these defeats the isolation architecture that the entire design depends on.
EMI containment¶
Enclosed switching power stages need filtering on the DC input and AC output, plus chassis-bonded shielding:
| Filter | Where | Purpose |
|---|---|---|
| Common-mode choke + Y caps on DC input | Per bridge | Reject conducted EMI back to the supply |
| Common-mode choke on AC output | After the LC filter | Reject conducted EMI to the load |
| Chassis bonding via low-impedance straps | At each PCB | Bleed off radiated EMI before it escapes |
Standard EMI filter modules from Direnc.net or Motorobit cover the conducted-EMI requirements at this power level.
Effort estimate¶
| Sub-item | Engineer-time |
|---|---|
| Enclosure CAD (Inventor / Fusion 360) + sheet-metal fab | 2 weeks |
| Thermal characterisation in the enclosure (with + without fan) | 1 week |
| EMI filtering + chassis bonding | 1 week |
| Fan + supervision firmware | 3 days |
Total: ≈ 2 engineer-months including the enclosure fab turn.
Where this fits in the sequence¶
After LC filter (because the filter has to be inside the same enclosure) and after closed-loop control (because the control loop needs to be stable across the enclosure's thermal regime). Before grid tie (because a grid-tied unit must be enclosed for safety).