The Physics of Class-D Reconstruction Filters
Class-D audio amplifiers do not amplify linear analog waveforms directly. Instead, they convert the incoming audio signal into high-frequency Pulse-Width Modulation (PWM) square waves switching at 350 kHz to 1.2 MHz. A passive second-order LC low-pass filter strips away the violent high-frequency PWM switching carrier, reconstructing the smooth analog audio waveform before it reaches the loudspeaker terminals.
The LC Filter Design Formulas
The low-pass filter consists of an Inductor ((L)) and Capacitor ((C)), with the speaker impedance ((R_{ ext{load}})) providing the necessary damping:
Quality Factor: Q = R_load / (2 × π × f_c × L) = 2 × π × f_c × C × R_load
Inductance: L = R_load / (2 × π × f_c × Q)
Capacitance: C = Q / (2 × π × f_c × R_load)
The Load-Dependence Dilemma (4Ω vs. 8Ω)
Because the speaker impedance forms an integral part of the passive LC filter circuit, the filter's Q factor and frequency response change whenever a different speaker is connected:
- An LC filter calculated for a 4(Omega) load will have its Q factor double to (Q approx 1.414) when an 8(Omega) speaker is plugged in, causing a prominent +3 dB to +5 dB resonant treble spike near 30-40 kHz.
- If an amplifier runs with no speaker connected (open circuit), the Q factor approaches infinity, causing high-voltage ultrasonic ringing that can arc across PCB traces or destroy the amplifier.