Class-D audio amplifiers put a magnetic component directly in the signal path, in a way that switching power supplies do not. The output inductor’s linearity, saturation behaviour and self-resonance all appear as distortion, and a part that measures well on a bench load can sound wrong on a loudspeaker at 2 kHz. This guide covers the selection rules for 10 W–1 kW Class-D output filters.
1. What the output inductor does
The LC filter after the switching bridge removes the PWM carrier (typically 384 kHz for a self-oscillating design, or 1–2 MHz for a fixed-frequency one) while passing the audio band. The inductor carries the full load current with a large AC component at the carrier frequency and a small one at the audio frequency, plus a DC offset in single-supply designs.
2. Why core material matters more here than in a power supply
The inductor’s value must not change with current, or the filter cut-off moves and the frequency response becomes load-dependent. Powder cores with soft saturation hold inductance better than gapped ferrite up to their rated current, which is why iron powder and sendust dominate this application. Beyond the rated current, inductance falls smoothly — audible as compression at high volume rather than as a hard clipping artefact.
| Material | Suitability for Class-D output | Note |
|---|---|---|
| Iron powder | Good, low cost | Higher loss at the carrier frequency; fine below 500 kHz |
| Sendust | Very good | Lower loss than iron powder, stable permeability |
| High flux | Good for high bias | Higher saturation, useful at high output current |
| Gapped ferrite | Workable | Low loss but sharp saturation; needs generous current headroom |
| Nanocrystalline | Not typical | High permeability is the wrong property here |
3. The numbers that set performance
- Inductance: typically 10–33 µH for 4–8 Ω loads; higher values reduce carrier residue but increase size and DCR.
- Saturation current: at least 1.5× the peak output current, and checked at the highest ambient the amplifier will see.
- DCR: below 20 mΩ for a 10 A design if efficiency and damping matter; DCR also sets the output impedance at low frequencies.
- Self-resonance: well above the carrier frequency, or the filter behaves unpredictably at high frequency.
4. Distortion mechanisms in the inductor
- Inductance modulation with current. The dominant cause of third-harmonic distortion at high output. A material with a flat bias curve fixes it more cheaply than a larger core.
- Hysteresis nonlinearity at low signal. At low volume the flux swing is tiny and the B-H loop dominates; some cores are noticeably nonlinear there.
- Thermal drift. As the inductor heats, inductance shifts and the response changes — audible as a tonal shift after twenty minutes of playback.
- Winding movement. Loose turns vibrate at the carrier frequency and can produce a mechanical buzz that couples into the enclosure.
5. Practical build notes
Use a single-layer toroidal winding where possible: it minimises capacitance and gives predictable inductance. Varnish or pot the part to stop winding movement. Keep the inductor away from the input stage, since the carrier leakage field is strongest at the core’s edge. For high-power amplifiers, a shielded toroid reduces radiated carrier noise into the input cabling.
6. Measuring what matters
Measure inductance at the actual DC bias, not at zero current. Measure THD at 1 kHz at several power levels into a resistive load, and again at 20 kHz where the filter attenuation is lower. Then listen with the real loudspeaker: a load that is reactive at the crossover frequency will expose behaviour the resistor test cannot.
Shaanxi SiRui Technology supplies Magnetic Core series iron powder, sendust and high flux cores for Class-D output filters, with bias curves published at the currents audio designers actually use.
Send your output power and load impedance to sales@sirui-tech.com.