Rules of Thumb That Hide the Physics
Every magnetics engineer has heard “4 A per mm² for natural cooling, 6 for forced”. The numbers are useful starting points and dangerous when treated as limits, because the permitted current density is an outcome of the thermal design, not a property of the wire. Two windings at the same 5 A/mm² can run 25 K apart if one is potted to a cold plate and the other sits in a sealed box.
What Actually Limits a Winding
- Temperature rise. Loss per unit of surface area, not current density, determines the rise. A small part has more surface per unit of copper and can run at a higher density than a large one.
- Insulation class. The permitted rise is what is left between ambient and the hot-spot limit, with a margin for thermal ageing.
- AC loss. At high frequency the effective resistance rises with the proximity effect, so the loss is set by the layer arrangement and the strand or foil choice, not by the DC cross-section.
- Fill factor. The window can only hold so much copper; above roughly 40–50 % fill with hand winding, or 60–70 % with machine winding and thin insulation, the assembly becomes difficult and the risk of damage rises.
- Handling. Wire below about 0.1 mm is fragile in production; the choice may move to a finer strand count or multifilar construction for manufacturability rather than electrical reasons.
A Workable Selection Method
| Cooling | Starting density (A/mm²) | What must then be checked |
|---|---|---|
| Free air, small part (<50 cm² surface) | 3.5–4.5 | Rise by resistance change at worst-case ambient |
| Free air, large part | 2.0–3.0 | Hot spot inside the winding; thermal resistance estimate |
| Forced air 1–2 m/s | 5–7 | Airflow actually reaching the part, not the enclosure rating |
| Potted to a metal surface | 6–9 | Interface material, flatness, and the potting layer thickness |
| Liquid cooled cold plate | 8–12 | Coolant temperature, PD and the CTE stress on the winding |
The method is: pick a starting density, compute the loss, compute the rise with the actual thermal resistance, then check whether the result respects the insulation class margin. If it does not, reduce the density, add copper, or improve the thermal path — in that order of cost.
Frequency Changes the Answer
At 50/60 Hz the winding is essentially DC and the density rules above apply directly. Above roughly 20 kHz the AC resistance factor becomes the dominant term: a winding at 4 A/mm² with Fr = 2.0 dissipates like one at 5.7 A/mm² at DC. The fix is construction, not wire size: fewer layers, interleaving to reduce the magnetomotive force between layers, Litz or foil where the current is high.
What to Put on the Drawing
- Wire diameter and insulation grade, with the strand construction if Litz.
- Permitted temperature rise and the measurement method.
- Maximum fill factor or the maximum winding build height, since this constrains the supplier’s choices.
- Whether the part will be potted or varnished, because it changes the thermal path.
- Any handling constraint, such as a minimum wire diameter for production reliability.
Related reading
- Window Utilisation and Fill Factor
- Estimating Temperature Rise
- GaN and SiC High Frequency Inductor Design
- Winding Tension and Layer Alignment
Send us the ambient, the cooling condition and the insulation class and we will return a winding proposal with the rise calculated, not estimated from a rule of thumb. sales@sirui-tech.com.
