There's a calculation almost every sealed-enclosure design skips, and it's rarely skipped out of ignorance. The closed-box thermal budget gets computed after the first prototype runs hot — not because engineers don't understand dissipation, but because assembling it means pulling a dozen numbers out of four different datasheets, and that's exactly the kind of tedium that loses to a deadline every time.
Here's what one of those budgets looks like when you actually run it, from a recent design session: a 50 W BLDC motor controller — 12 V battery, isolated CAN, encoder feedback — living in a small sealed aluminium case.
The inventory
Six BUK7230-55A MOSFETs switch about 5.9 A RMS of phase current at 20 kHz. Worked through step by step — conduction from R_DS(on) at temperature, switching from gate charge and transition times:
| Source | Dissipation |
|---|---|
| 6× BUK7230-55A FETs (conduction + switching) | 4.0 W |
| AMS1117-3.3 LDO — (12 V − 3.3 V) × 180 mA | 1.57 W |
| AMS1117-5.0, encoder rail | 0.35 W |
| 2× 10 mΩ current shunts at 5.9 A RMS | 0.70 W |
| MCU, CAN isolator, RS-422 receivers, rest | 0.60 W |
| Total, inside the sealed case | ≈ 7.2 W |
None of these lines is a revelation on its own. An AMS1117 dropping 12 V to 3.3 V is a resistor with a reputation — every engineer knows it burns the difference. R_DS(on) losses are a first-week calculation. That's precisely why the budget feels skippable.
What the assembled budget shows that the individual facts don't
The information isn't in any single line — it's in the ranking and the total.
The ranking: in this box, the housekeeping regulator is the number-two heat source. Not the shunts carrying six amps, not the MCU, not everything else combined — the fifty-cent LDO, at 40% of what the entire power stage produces. You know it dissipates; you probably wouldn't have ranked it there. Prioritization is what the budget buys you, and prioritization is exactly what per-part intuition doesn't provide.
The total: 7.2 W against a small sealed enclosure's ~11 °C/W is an ~80 °C internal rise. On a 25 °C day the electronics sit above 100 °C. This design doesn't degrade gracefully in the field — it fails on paper, which is the cheapest place a design can fail.
The consequence was a one-line BOM change: TPS54302 synchronous buck in place of the LDO, converter losses in the hundreds of milliwatts, budget closed, case stays sealed.
The tedium is the actual obstacle — so remove it
Every number above came out of a datasheet: R_DS(on) and gate charge from the MOSFET doc, the buck's efficiency from its own, the case thermal resistance from the enclosure's. The reason this budget usually goes unrun isn't difficulty — it's that collecting the inputs is an afternoon of PDF spelunking.
In this session it was one sentence — "compute device temperature at full load inside a closed case" — typed into whatic, an AI design companion that reads component datasheets and answers with the page cited. It assembled the inventory, ran each step through its calculator, and returned the table above in about a minute, every input traceable to its source document. The physics was never the barrier; the legwork was.
Before your next board order: total the dissipation of everything in the box, divide by the case's °C/W, and look at the number. If you want the legwork done for you, with citations: whatic.io