RC Snubber

The problem: current that refuses to stop

Open a switch in a purely resistive circuit and nothing dramatic happens — the current simply drops to zero. Open a switch in series with any inductance and you have a fight on your hands. An inductor stores energy in its magnetic field, and the defining law V = L·di/dt says that the faster you try to change its current, the larger the voltage it generates to oppose you. A mechanical contact opening in a microsecond, or a MOSFET turning off in tens of nanoseconds, asks the inductor to change its current almost instantaneously — so it answers with an enormous voltage spike. That spike is the inductive kick, and it is the single most common destroyer of switches in power electronics.

The numbers are sobering. A modest 50 mH relay coil carrying 0.2 A, interrupted in 1 µs, sees di/dt = −200,000 A/s, so the ideal kickback is 50 × 10⁻³ × 200,000 = 10,000 V. In practice the wiring's parasitic capacitance clamps it far below that — but only by forming a resonant LC tank that rings: the energy sloshes back and forth between the inductor and the stray capacitance at f = 1 / (2π√(LC)), with the first peak reaching V_peak = I·√(L/C). That ringing both stresses the switch with repeated overvoltage and radiates broadband electromagnetic interference. The RC snubber is the two-component circuit that absorbs the energy and kills the ring.

The topology and the mechanism

An RC snubber is exactly what the name says: a resistor R_s in series with a capacitor C_s, the pair wired directly across the thing you want to protect. Across a MOSFET it goes drain-to-source. Across a rectifier it goes anode-to-cathode. Across a relay it goes straight over the contacts. Across a transformer it goes over the winding. Wherever a switching node rings, the snubber sits across it.

Two things happen the instant the switch opens. First, the capacitor presents a low impedance to the sudden voltage step, so the inductor current diverts into C_s instead of being forced through the opening switch. This slows the rate of voltage rise — the dv/dt across the switch — which directly reduces the peak the switch ever sees and softens the EMI-generating edge. Second, the resistor stands in the path of that diverted current and dissipates the trapped energy as heat. Without the resistor, the capacitor and the inductance would simply form a new, undamped LC tank and ring just as badly at a lower frequency. The resistor is what turns an oscillation into a single decaying pulse.

The design target is critical damping: enough resistance to stop the ring in one cycle, but not so much that the resistor itself becomes a high-impedance path that defeats the capacitor. The sweet spot is the characteristic impedance of the parasitic tank, R_s = √(L_par / C_par). Set R below it and the circuit stays underdamped and rings; set R far above it and the capacitor can no longer divert the current quickly enough.

Sizing the snubber — the numbers that matter

There are two unknowns to pin down before you can pick components: the parasitic inductance L_par doing the kicking, and the parasitic capacitance C_par it rings against (the MOSFET's C_oss, diode junction capacitance, transformer winding capacitance, plus board stray). You rarely know either precisely, so the standard procedure — popularised by Rudy Severns — is empirical:

Step 1.  Measure the unsnubbed ring frequency:   f_0
         f_0 = 1 / (2π·√(L_par·C_par))

Step 2.  Add a known capacitor C_add across the node until the
         ring frequency drops to f_0 / 2.
         Halving the frequency means total C is now 4×, so:
             C_par = C_add / 3

Step 3.  Knowing C_par and f_0, back out the parasitic inductance:
             L_par = 1 / ((2π·f_0)²·C_par)

Step 4.  Pick the snubber capacitor large enough to dominate:
             C_s ≈ 3 to 10 × C_par     (often C_s = C_add itself)

Step 5.  Set the resistor to the tank's characteristic impedance:
             R_s = √(L_par / C_par) = 1 / (2π·f_0·C_par)

Step 6.  Rate the resistor for the dissipation:
             P_R = C_s · V² · f_sw
         (energy C·V² is lost every full charge/discharge cycle)

A worked example makes the scale concrete. Suppose a 100 kHz buck converter switching node rings at f_0 = 60 MHz with a 30 V swing. You add 470 pF and the ring drops to 30 MHz, so C_par = 470/3 ≈ 157 pF. The parasitic inductance is L_par = 1/((2π·60e6)²·157e-12) ≈ 45 nH. The characteristic impedance is R_s = √(45e-9 / 157e-12) ≈ 17 Ω. Choosing C_s = 470 pF, the resistor dissipates 470e-12 × 30² × 100,000 ≈ 42 mW — a 0.25 W film resistor is plenty. Now push the same snubber onto a 400 V flyback node and the dissipation jumps to 470e-12 × 400² × 100,000 = 7.5 W, and suddenly you need a wirewound or several parallel resistors. That V² term is why snubber design at high voltage is dominated by thermal management, not by the ringing itself.

Layout — why the loop length is everything

A correctly calculated snubber will do nothing if it is placed badly. The snubber works by offering the ringing current a low-inductance shortcut, so any inductance you insert between the snubber and the switch becomes part of a new tank that rings with C_s. Ten nanohenries of trace — about a centimetre of PCB track — is enough to create a fresh resonance and undo the fix. The rules are blunt: place R and C directly across the device terminals, make the loop area as small as physically possible, and use a capacitor with low equivalent series inductance (a small ceramic or film part, never an electrolytic). The tight loop also minimises the magnetic area that radiates EMI, so good snubber placement helps you pass conducted and radiated emissions tests at the same time.

Variants — RC, RCD, and the lossless snubbers

The plain RC snubber is the workhorse, but it is one member of a family, each tuned to a different problem.

• RC snubber. Symmetric, non-polarised, damps ringing on both edges. The default for relay contacts, triac dimmers, rectifier diodes, and the high-frequency ring on a MOSFET drain. It does not hard-clamp the peak — it spreads the energy over time.
• RCD clamp (resistor + capacitor + diode). A polarised snubber that catches only the positive-going spike. The diode lets the capacitor charge to the peak and blocks it from discharging back, so the resistor bleeds the captured energy off slowly between cycles. This is the canonical fix for the leakage-inductance spike on a flyback converter primary, holding the MOSFET drain below its breakdown rating. It clamps the voltage to a defined level rather than merely damping.
• Freewheeling (flyback) diode. A single diode across a DC inductive load such as a relay coil. When the switch opens, the diode gives the coil current a recirculation path so it decays gently. Cheapest possible fix, but it works only on DC, and it slows relay drop-out — sometimes dangerously, in safety contactors. Adding a series resistor or Zener speeds the decay back up at the cost of a higher (but still bounded) clamp voltage.
• Lossless / energy-recovery snubbers. At high voltage and high frequency the C·V²·f loss of an RC snubber becomes unacceptable. Lossless snubbers use extra inductors, diodes, and a capacitor to return the captured energy to the supply instead of burning it. Common in multi-kilowatt converters where every watt of snubber loss costs efficiency points.
• TVS / MOV clamp. A transient-voltage-suppressor diode or metal-oxide varistor in parallel with (or instead of) the snubber, providing a hard voltage ceiling for the rare large transient while the RC handles the routine ringing. Belt-and-braces protection in industrial drives.

How the snubber options compare

Property	RC snubber	RCD clamp	Flyback diode	TVS / MOV
Limits dv/dt (ringing)	Yes — primary job	Partial	No	No
Hard voltage clamp	No (soft)	Yes	Yes (≈0.7 V over rail)	Yes (sharp)
Works on AC	Yes	No (polarised)	No (polarised)	Yes (bidirectional MOV)
Steady-state loss	C·V²·f (can be watts)	Bleed-resistor loss	≈ none	≈ none until it clamps
Slows load turn-off	No	No	Yes — significant	No
Best for	HF ringing, relay/triac arcing, rectifier ring	Flyback transformer leakage spike	Low-speed DC coils	Catastrophic transient ceiling

Where RC snubbers actually show up

• Relay and contactor contacts. An RC snubber across mains relay contacts suppresses the arc that erupts as the contacts part on an inductive load (a motor, solenoid, or transformer). Without it, the arc erodes the contacts and welds them over thousands of cycles. A typical mains snubber is 100 Ω in series with 100 nF, sometimes sold as a packaged "RC suppressor".
• Rectifier diodes. When a diode recovers (the reverse-recovery snap), it rings against the transformer leakage inductance and generates a buzz of EMI right at the rectifier. A small RC across each diode — say 10–100 Ω and 1–10 nF — damps it.
• MOSFET and IGBT switching nodes. In buck, boost, and bridge converters the drain or collector node rings every switching edge. An RC snubber damps it to keep the device within its voltage rating and to pass EMC.
• Triac and SCR dimmers. The classic snubber across a triac (often 39–100 Ω with 10–100 nF) limits the dv/dt that would otherwise falsely re-trigger the device when driving an inductive load like a motor or magnetic transformer.
• Flyback converter primaries. Here the RCD-clamp variant dominates, catching the leakage-inductance spike that would otherwise punch through the primary MOSFET. The plain RC may be added on top to clean up the residual ring.
• Motor drives and inverters. Snubbers across the inverter switches limit the reflected-wave overvoltage on long motor cables and protect the IGBTs from the bus-stray-inductance kick at turn-off.

Failure modes and trade-offs

• Undersized resistor power rating. The most common field failure. Engineers compute R and C correctly, then fit a 0.25 W resistor where C·V²·f demands 5 W. The resistor overheats, drifts, cracks, and eventually opens — at which point the snubber is just a capacitor and the node rings freely again. Always size the resistor for the full per-cycle dissipation plus margin.
• Capacitor with too much ESL. An electrolytic or large multilayer part has enough internal inductance to ring with itself. Use low-ESL film or C0G/NP0 ceramic capacitors, voltage-rated well above the peak swing.
• Long snubber loop. As covered above — added trace inductance creates a new resonance and the snubber stops working. This is the number-one reason a "correctly designed" snubber fails on the bench.
• Oversized capacitor. Bigger C_s damps better but slows the switching edge, increases switching loss in the device (it must charge and discharge the larger cap), and raises the C·V²·f dissipation. There is a real efficiency cost to over-snubbing.
• Snubber resonance with the load. On lightly damped systems the snubber capacitor can interact with the load inductance to create a sub-resonance audible as a whine or visible as ripple. Critically damping with the correct √(L/C) resistor avoids it.
• Capacitor failure to short. If the snubber capacitor fails shorted, the resistor sees the full node current continuously and can overheat catastrophically. In mains-connected snubbers, use safety-rated (X/Y class) capacitors that fail open.

The one-line intuition

Strip away the formulas and an RC snubber is a shock absorber for electricity. The inductive kick is a hammer blow of voltage; the capacitor is the cushion that catches it so the rate of change is gentle, and the resistor is the damper that turns the bounce into a single soft thud instead of a ringing clang. Size the cushion to dominate the stray capacitance, size the damper to the characteristic impedance √(L/C), mount it on the shortest possible loop, and rate it for C·V²·f of heat. Get those four things right and a 12 V coil never kicks 300 V into your transistor.