Common-Mode & Differential-Mode Choke Sizing, Explained

Two kinds of noise, two kinds of choke

A switching inverter generates conducted EMI that comes in two flavours, and they need different filtering:

• Differential-mode (DM) noise flows down one line and back the other — the same path as the useful current. A DM choke puts inductance in series with that loop.
• Common-mode (CM) noise flows the same direction on all lines together, returning through ground/chassis. A CM choke is wound so the load current's flux cancels (the balanced currents cancel in the core) while the common-mode current sees the full inductance.

That cancellation is the whole trick of a CM choke: because the wanted current produces no net flux, you can use a small high-permeability core and get large impedance to CM noise without the core saturating on the load current.

How the core sets the inductance

Inductance comes from the magnetic circuit: L = N² / R_m, where N is the turns and R_m is the core's reluctance (its "resistance" to flux, set by the magnetic path length, the effective cross-section area Ae, and the material permeability). A toroid has an exact closed-form geometry; U- and E-cores are approximated as a rectangular loop, but real datasheets publish part-specific Ae/le/window area because those families have no single universal formula — so a final design cross-checks the manufacturer's numbers.

The two limits that actually size the choke

• Saturation — push too much flux and the core saturates, its permeability collapses, and the inductance vanishes exactly when you need it. For a DM choke that's driven by the full load current; for a CM choke it's driven by the imbalance (a worst-case DC offset or common-mode current), since the balanced load flux cancels.
• Core loss — the AC flux swing heats the core every cycle. This is estimated with a Steinmetz-style law (loss rising with frequency and flux-swing amplitude to material-specific powers). Too much loss and the core overheats regardless of saturation.

The window also has to physically fit the turns, and the target impedance (a CISPR 25 class gives a starting rule-of-thumb) sets how much inductance you need in the first place.