EMI Tuning for Inductorless Class D Audio Outputs
A medium-power Class D amplifier can replace a conventional output inductor with a ferrite bead, reducing cost and board area. This is not a drop-in substitution. Bead impedance, current rating, output-edge speed, resonance damping, and PCB return paths must be designed together to preserve both EMI margin and audio performance.
What the ferrite bead does
A ferrite bead is mainly inductive at lower frequencies, becomes strongly lossy near self-resonance, and is increasingly affected by parasitic capacitance at higher frequencies. At a Class D output it slows high-frequency current edges and dissipates part of the interference energy. Unlike an ideal inductor, its impedance changes with frequency and DC bias, so the complete impedance curve, DC resistance, and current derating all matter.
In the cited 2012-package examples, a 120 Ω at 100 MHz bead corresponds to roughly 0.39 μH, while a 600 Ω part corresponds to about 1.59 μH. Higher impedance can slow the output edge more effectively, but it often comes with a lower current rating. The example 600 Ω part is rated at 2 A, making load-current verification the first selection constraint.
Selection factor Circuit effect Verification
Impedance at 100 MHz Shapes high-frequency attenuation and edge rate Compare the full impedance curve
Current rating Bias reduces impedance and raises temperature Derate for peak current and ambient temperature
DC resistance Adds conduction loss Calculate I2R loss and temperature rise
Parasitic capacitance Controls behavior above self-resonance Review curves and measure on the board
Use the load condition to bound the choice
For the source example with PVDD = 12 V, an 8 Ω speaker, BD modulation, and negligible trace and bead resistance, the simplified current is 12 V ÷ 8 Ω = 1.5 A. A 2 A, 600 Ω at 100 MHz bead has nominal current margin in that case. A production design must still account for peak audio current, simultaneous channel operation, thermal conditions, and component tolerance.
Edge rate and conducted emissions
With an amplifier switching near 300 kHz and a bead filter acting in the low-megahertz region, reducing edge rate lowers harmonic energy below about 5 MHz. A higher bead impedance generally produces a slower edge, but may raise loss or limit output power. Spread-spectrum operation and phase shifting between stereo channels can distribute spectral peaks, yet neither technique replaces control of the power loop and filter resonance.
Damp resonance with a Zobel network
The bead and output capacitor form a resonant network, while package and PCB parasitics create additional ringing near the amplifier. The source uses separate RC branches for device-side and filter-side energy. A typical starting point for the device-side branch is 10 Ω and 330 pF. Those values are not universal; the final network should follow the measured ringing frequency, decay rate, and component temperature.
A representative waveform has a ringing period of about 350 ns, or approximately 2.85 MHz. That resonance falls inside the conducted-emissions band. Poor damping can amplify noise from 2 MHz to 4 MHz, so tuning should compare ringing amplitude, edge time, and spectral margin rather than optimizing a single oscilloscope metric.
PCB layout requirements
?Place each ferrite bead close to the amplifier output to minimize the unfiltered switching node.
?Return the output capacitor directly to power ground without forcing its current through a narrow signal-ground path.
?Maintain a continuous reference plane below the amplifier and filter so high-frequency return current does not detour.
?Place the filter-side Zobel branch near the bead and capacitor and the device-side branch near the output pins.
?Run conducted and radiated tests with representative speaker loads and cable lengths.
Verification sequence
Begin at low output power to confirm waveform quality and temperature, then cover maximum supply voltage, load variation, and simultaneous channel operation. When comparing beads, RC values, or layout revisions, record efficiency, audio distortion, component temperature, and emissions in the 2 MHz to 5 MHz region and above. That evidence determines whether the lower-cost inductorless design retains adequate margin throughout its operating range.