Formulas
| Skin depth δ | = √(2ρ / (ωμ)) = 1 / √(πfμσ) |
|---|---|
| Eddy loss (thin plate) | P ∝ (B · f · t)² / ρ (t = thickness) |
| Hysteresis loss | P_h ∝ f · B^α (α ≈ 1.6–2.5) |
| Core loss (Steinmetz) | P ≈ K · f^a · B^b |
Skin depth in copper (20 °C)
| Frequency | Skin depth |
|---|---|
| 60 Hz | 8.5 mm |
| 1 kHz | 2.1 mm |
| 10 kHz | 0.66 mm |
| 100 kHz | 0.21 mm |
| 1 MHz | 66 µm |
| 10 MHz | 21 µm |
| 1 GHz | 2.1 µm |
Mitigation
- Laminations: thin insulated sheets stacked to reduce eddy-path area (transformers, motors).
- Powdered cores: ferrite or iron powder — particles isolated by binder.
- Litz wire: many thin insulated strands twisted — each smaller than skin depth at f.
- Ferrite instead of steel at high frequencies — high resistivity ≈ no eddy currents.
- Air-core: no core at all — used in high-frequency RF when losses dominate.
Proximity effect
- Nearby conductors carrying AC induce currents in each other — raises effective resistance.
- Matters at HF and in transformer / inductor windings.
- Mitigation: separate wires, use Litz, or bundle small strands.
Worked examples
Skin depth in copper at 100 kHz: δ = √(ρ / (π f μ0)) with ρ = 1.68×10-8 ohm-m. The denominator is π × 100,000 × 4π×10-7 = 0.3948, so δ = √(4.255×10-8) = 2.06×10-4 m = 0.21 mm, which matches the table. Current is concentrated in the outer 0.2 mm, so a 2 mm solid wire uses only a thin shell and its AC resistance is several times its DC resistance.
Picking Litz strands: a strand diameter at or below about the skin depth uses its whole cross-section. At 100 kHz (δ = 0.21 mm) strands of AWG 38 (0.10 mm diameter) are comfortably small. At 1 MHz (δ = 66 µm) go finer, around AWG 44 (0.05 mm), or switch to foil or a single wide trace.
Lamination scaling: because eddy loss varies as the square of thickness, halving the sheet from 0.50 mm to 0.25 mm cuts it to one quarter. Doubling the frequency at the same flux density quadruples it. This is why 50/60 Hz transformers use laminations of about 0.3 to 0.5 mm, while switch-mode transformers above tens of kilohertz use ferrite.
Common mistakes
- Using DC resistance to estimate winding loss at switching frequencies. Always correct with the AC resistance ratio, which includes skin and proximity effects.
- Using Litz wire above its useful range. Litz strands finer than needed add cost and lower the copper fill factor; above a few MHz it stops helping.
- Assuming a thick winding on a high-frequency inductor behaves like its low-frequency self. Each additional layer increases proximity loss, so a single layer or interleaved winding is often better than many layers.