Access Official Technical PDF

Share Summary:LinkedInEmail

In Class-D amplifier power stages, fast MOSFET switching can produce output-voltage overshoot, undershoot and high-frequency ringing. These transients can increase device stress and contribute to radiated emissions, particularly when amplifier outputs connect to a speaker through an unshielded cable. RC snubber design offers a way to damp the ringing at its source, using measured circuit behavior to guide component selection.

Why Class-D amplifier ringing occurs

A Class-D output stage combines switching devices with unavoidable parasitic inductance and capacitance. Relevant inductances include the supply-decoupling capacitor’s equivalent series inductance, PCB traces and IC package or internal routing. The MOSFET output capacitance, together with stray package and board capacitance, supplies the capacitive part of the resonant circuit.

When one MOSFET switches, current in the circuit’s inductance cannot change instantaneously. During switching transitions, that inductance interacts with the capacitance at the output node, creating an LC resonance. The resulting waveform may overshoot or undershoot and oscillate for several cycles before energy dissipates in the circuit.

In a full-bridge amplifier, the two switching transitions produce related effects. When the n-channel MOSFET turns off and the p-channel device turns on, the n-channel device’s parallel capacitance is charged through the parasitic inductances, producing rising-edge overshoot and ringing. On the opposite transition, speaker-inductance flyback turns on the n-channel body diode as the p-channel MOSFET turns off. The p-channel device’s output capacitance then participates in a resonant transient that can cause undershoot.

Why an output LC filter may not be enough

An LC filter at the amplifier output can attenuate ringing that would otherwise couple into nearby PCB structures or cables and contribute to radiated emissions. However, filtering the output does not remove the voltage stress experienced by the switching MOSFETs. An RC snubber instead damps the resonance near its source. The trade-off is additional switching loss, so the component values and resistor power rating need to be considered together.

RC snubber design begins with measurement

In practice, the small, nonlinear parasitic capacitances and application-dependent board effects make it difficult to determine all circuit parasitics from component data alone. A typical approach is to provide PCB locations for snubber components, measure the ringing on the bench, and use those observations to estimate the effective capacitance and inductance.

First, measure the original ringing frequency, f0. Then add a known external capacitor, CEXT, between the output and ground and measure the new frequency, f1. Define the frequency ratio as χ = f0/f1. The effective MOSFET output capacitance can then be estimated as:

COSS = CEXT / (χ² − 1)

With COSS estimated, calculate the equivalent parasitic inductance using the original resonance:

Lp = 1 / [(2πf0)² × COSS]

These calculations represent the effective parasitics of the measured circuit and layout; they are not a substitute for validating the snubber on the actual board.

Selecting the snubber resistor and capacitor

For a parallel RLC model, the damping factor is expressed as ζ = 1/(2RSN) × √(Lp/COSS). Choosing a damping factor of one, corresponding to critical damping, gives the starting resistor value:

RSN = ½ × √(Lp/COSS)

A practical starting point for the snubber capacitor is approximately three times the effective MOSFET output capacitance: CSN ≈ 3 × COSS. The capacitor helps damp the transient, while its value also affects the energy dissipated in the resistor during switching. A C0G/NP0 ceramic capacitor is identified as a suitable choice in the example design.

Estimate average resistor dissipation with PRSN = CSN × VSPK² × fSW, where VSPK is the relevant supply voltage and fSW is switching frequency. Select a resistor with an appropriate power rating, applying the organization’s component derating guidelines. The example calculation uses a 560 pF capacitor, an 11 V supply and a 430 kHz switching frequency, resulting in approximately 29 mW of average dissipation.

Example: turning measured ringing into component values

In the documented example, the unsnubbed ringing frequency is 111.11 MHz. Adding a 1 nF external capacitor changes it to 45.87 MHz. Applying the frequency-shift method gives an estimated COSS of 205.45 pF; the corresponding parasitic inductance is 9.99 nH.

These values produce a calculated snubber resistance of 3.45 Ω. The selected preferred value is 3.3 Ω, paired with a 560 pF capacitor, close to three times the estimated output capacitance. The snubber is connected directly across the drain and source of the n-channel MOSFET. In this example, the resulting waveform has substantially reduced overshoot and ringing that decays after only a few oscillations. The measured common-mode spectrum also shows approximately 12 dBm attenuation at the ringing frequency under the stated test conditions.

Practical implications for the design process

For engineers and integrators addressing switching transients, the key is to treat snubber selection as a measured design task rather than relying on nominal component values alone. Capture the ringing, use a known added capacitance to estimate the resonant parasitics, calculate initial RC values, and then verify the result on the target layout. Keep the snubber physically close to the MOSFET connection and check both waveform damping and resistor power dissipation.

An RC snubber can address two related concerns: excessive switching-device voltage excursions and high-frequency ringing that may contribute to emissions. Unlike an output LC filter, it damps the parasitic resonance at the switching stage itself, although it adds switching loss. Measurement-led design and board-level validation help balance these effects and establish whether the chosen values suit the application.

Access Official Technical PDF

About The Author