RC Snubber Calculator — Damp Switch-Node Ringing from Two Measurements

When a MOSFET or diode switches off, the voltage across it rings. The parasitic inductance of the loop around the switch (traces, package leads, a transformer's leakage) trades energy with the capacitance across the switch (its output capacitance, the diode's junction, the layout), and with almost no resistance in that loop the ringing lasts many cycles, overshoots the supply and radiates. An RC snubber across the switch damps it: a resistor equal to the loop's characteristic impedance absorbs the energy, and a capacitor in series with it keeps the resistor from dissipating at DC.

Measuring the parasitic loop

The inductance and capacitance are rarely known well enough to calculate, but they can be measured. Note the ring frequency $f_0$ with no snubber, then put a known capacitor $C_{add}$ across the switch and note the new frequency $f_1$. The inductance has not changed, so

If the added capacitor halves the frequency, the circuit capacitance has quadrupled and $C_p$ is exactly one third of $C_{add}$, the procedure Todd gives in Unitrode's Snubber Circuits: Theory, Design and Application (1993). The inductance and the characteristic impedance follow:

Sizing the snubber

Setting the snubber resistor to $Z_0$ damps the loop close to $Q = 1$. The snubber capacitor must be larger than $C_p$ for the resistor to take control of the ringing; Todd recommends two to four times $C_p$. A larger capacitor damps harder but costs power, because the capacitor charges to the switched voltage $V$ and discharges once each per cycle, and the energy $\tfrac{1}{2}C_sV^2$ is lost in the resistor at each of those transitions:

The power does not depend on the resistor's value, which is why the capacitor, not the resistor, sets the loss. The resistor must be a low-inductance type: a wire-wound part has too much inductance at the ringing frequency.

Validity

The measurement assumes a single parasitic inductance and capacitance and an inductance that does not change when the test capacitor is added. The calculator flags a capacitor multiple outside two to four times $C_p$, Todd's range. It also flags a snubber whose time constant $\tau = R_sC_s$ is too long for the capacitor to discharge within the minimum on-time: Severns (Cornell Dubilier, Design of Snubbers for Power Circuits) allows two time constants, after which 0.14 of the voltage remains. Outside these bounds the values are still shown, with a warning. The power estimate also assumes the capacitor charges to the switched voltage and no further; overshoot left after damping adds to it.