Oscilloscope waveform of MOSFET drain-source voltage overshoot and ringing at turn-off
VDS overshoot and ring frequency measured at the MOSFET pins.
Posted in

How to Measure MOSFET Ringing and Overshoot Correctly

Every fast switching edge overshoots and rings. The question is whether the peak stays inside the MOSFET’s safe limits under every condition, and whether you are measuring the circuit or your probe. This guide shows engineers and technicians how to measure MOSFET ringing and overshoot correctly: full-bandwidth setup, probing at the pins, sample rate, what to read from the waveform and over which time frame, how to read the datasheet limits, and how to decide pass or fail.

The scope illustrations in this guide are live: click the buttons under each one to change the setup and watch the reading change.

1. Ringing and Overshoot: Definitions

  • Plateau: the voltage the node settles to after the edge, for example the 48 V bus across an off low-side MOSFET.
  • Overshoot: how far the first peak goes above the plateau: Vpk − Vplateau, often given in % of the plateau.
  • Ringing: the damped oscillation that follows, described by its frequency and how fast it decays.

The same method applies to VDS at turn-off, VGS (gate ringing), diode reverse-recovery ringing, the switch node of a buck converter and the leakage spike of a flyback.

2. Why Full Bandwidth?

Ripple is measured with a 20 MHz limit. Ringing is the opposite: it typically sits at 50 to 300 MHz, and a bandwidth limit filters out exactly the peak you need. The rules of thumb:

  • Scope and probe bandwidth ≥ 0.35 / rise time, and at least 3 × the ringing frequency.
  • Sample rate ≥ 5 × bandwidth, at the timebase you actually use.
Bandwidth limit:
A 48 V switch node with 110 MHz ringing. Change the bandwidth limit and watch the measured peak.

3. What Sets the Frequency and the Peak

The parasitic loop inductance (PCB traces, package leads, capacitor ESL) resonates with the MOSFET output capacitance COSS:

fring = 1 / (2π √(Lloop · COSS)) and overshoot ≈ Ioff × √(Lloop / COSS)

The overshoot grows with the current being switched off and with the loop inductance. That is why the worst case is usually at maximum load, during start-up or in a short circuit.

Loop inductance: RC snubber:
48 V bus, 10 A turn-off, COSS 400 pF, 100 V MOSFET.

4. How to Read the MOSFET Datasheet Limits

Datasheet itemWhat it means for ringing
VDS (absolute maximum)Never to be exceeded, including the shortest spike. “Absolute maximum” is not an operating point.
V(BR)DSS (drain-source breakdown)Specified at TJ = 25 °C, VGS = 0, small ID (for example 250 µA). It has a positive temperature coefficient: it is lower when cold. Read the “normalized V(BR)DSS vs TJ” curve; a typical drop is around 5 to 8 % at −40 °C.
EAS, IAR, EAR (avalanche)An avalanche-rated part survives occasional events (single pulse EAS, repetitive EAR). This is a safety net for faults, not a design margin: normal operation should never avalanche.
VGS (typically ±20 V)Gate ringing must stay inside it at both polarities. Some datasheets allow a higher transient value only under stated pulse conditions.
VGS(th) min and its temperature curveThreshold falls as the part heats up. Gate bumps while the MOSFET is off (Miller coupling) must stay well below the hot minimum threshold, or the switch can turn on falsely (shoot-through).
COSS (vs VDS)Use the value at your operating voltage to predict fring and the overshoot.
Diode / rectifier VRRM, controller SW pin abs. maxThe same peak applies to every part on that node: check them all.
Derating: most companies limit the worst-case measured peak to 75 to 90 % of V(BR)DSS (80 % is common), applied to the rating at the minimum operating temperature.

5. Probes and Tapping Point

  • Low-side MOSFET: a 10x passive probe rated for the voltage (for example 300 V), 200 to 500 MHz.
  • Tip on the drain pin, ground spring on the source pin, at the package, not on a trace or heatsink further away.
  • Never use the long ground clip. Its ~150 nH with the probe’s ~10 to 15 pF resonates near 100 to 130 MHz and adds ringing that does not exist.
  • High-side or floating nodes: high-voltage differential probe with enough bandwidth and good high-frequency CMRR; for very fast GaN/SiC edges, an optically isolated probe.
  • Gate (VGS): measure gate to source pin (Kelvin source if the package has one), not gate to ground.
  • Compensate the probe first, and deskew channels if you compare VDS with VGS or current.
Probe ground:
The circuit rings at 80 MHz. Watch what the long ground clip adds.

6. Scope Settings and Sample Rate

SettingUseWhy
Bandwidth limitOFFAny limit near the ring frequency lowers the peak.
Coupling / inputDC, 1 MΩ for passive probesYou need the absolute peak.
Vertical scaleWhole peak on screenAn overdriven input distorts the edge and recovers slowly.
TriggerEdge on VDS rising, ~50 %Stable view of the turn-off edge.
Sample rateHighest; ≥ 10 samples per ring cycleFewer samples miss or alias the peak.
AcquisitionSample (edge view), peak detect (long views)Averaging and high-resolution modes smooth the peak.
Sample rate:
The same 110 MHz ringing sampled at three rates. White dots are the actual samples when they are sparse.

7. What to Measure: Level and Time Frame

QuantityHowTimebase
Rise time10 to 90 % of the edge; sets the bandwidth you need.2 to 10 ns/div
Peak VpkMaximum of the waveform (Max measurement, with statistics over many edges). This is compared with the rating.10 to 50 ns/div
Plateau and overshootPlateau = settled level after ringing; overshoot = Vpk − plateau, and in %.10 to 50 ns/div
Ring frequencyCursors between two peaks: f = 1/T.10 to 50 ns/div
DampingRatio of two successive peaks (above the plateau): δ = ln(A1/A2), ζ = δ / √(4π² + δ²).10 to 50 ns/div
Settling timeFrom the edge until the waveform stays within ±5 % of the plateau. Must be well under the on/off time and dead time.20 to 100 ns/div
Worst edge in an eventPeak detect over start-up, short circuit or load step; then zoom in.0.1 to 10 ms/div
Show: Snubber:
Reading one turn-off edge: peak, overshoot, ring frequency, damping and settling time.
Report it fully: “VDS peak 96 V (plateau 60 V, overshoot 60 %), ringing 110 MHz, ζ 0.08, settles in 70 ns, worst case at max Vin and full load, full bandwidth, 5 GS/s, spring ground at the pins.”

8. Catching the Worst Case

The highest spike rarely happens in steady state. Measure at:

  • Maximum input voltage and maximum load (highest turn-off current).
  • Start-up (soft-start, pre-biased output), output short circuit and current limit, load steps, and shut-down.
  • Temperature extremes: V(BR)DSS is lowest when cold; switching speed changes with temperature.

Use a long capture with peak detect to find the worst edge, a level or window trigger set just below your limit to catch rare exceedances, and segmented memory or persistence to collect thousands of edges.

Acquisition at 1 ms/div:
A 10 ms start-up of a 48 V converter. The ringing peaks are nanoseconds long; the worst ones happen in the first millisecond.

9. Pass or Fail: Checking Against the Limits

CheckPass if
Peak VDS, every condition≤ derating × V(BR)DSS at the minimum operating temperature
Never into breakdownVpk ≤ V(BR)DSS at min temperature, even in faults (otherwise check avalanche ratings and redesign)
GateVGS peaks within ±VGS max with margin; off-state bumps well below hot VGS(th) min
Other parts on the nodeDiode VRRM, controller pin abs. max, capacitor voltage ratings
Ringing durationSettles well before the next edge and within the dead time
EMIRing frequency consistent with your EMI results (ringing often shows as a peak in radiated emissions)
MOSFET rating: Derating: Temperature:
Measured worst peaks of a 48 V (60 V max) converter under four conditions, against the derated limit.
If it fails: confirm the measurement (spring ground, full bandwidth, pickup test). Then reduce loop inductance (layout, decoupling capacitor right at the switch), slow the edge (gate resistor), add an RC snubber or clamp, or choose a higher-voltage MOSFET.

10. Verify: Is the Ringing Real?

  1. Source-to-source test: touch the tip and the ground spring both to the source pin with the circuit running. The trace should be almost flat.
  2. Move the probe a few centimetres. Real ringing does not change; pickup does.
  3. Check the frequency against 1/(2π√(L·COSS)) and against the probe ground resonance.
  4. Check compensation on the calibrator output.
Test point: Ground:
The source-to-source test shows how much of your waveform is pickup.

11. Step-by-Step Procedure and Report

  1. From the datasheet: V(BR)DSS, its value at minimum temperature, your derating, VGS max and VGS(th) min (hot). Calculate the limit.
  2. Estimate fring and the peak with the calculator; choose probe, bandwidth and sample rate.
  3. Compensate a voltage-rated 10x probe; fit a ground spring. Tip on drain, spring on source.
  4. Bandwidth limit OFF, DC coupling, whole peak on screen, highest sample rate, edge trigger on VDS.
  5. At 10 to 50 ns/div: measure Vpk (with statistics), plateau, overshoot, ring frequency, two successive peaks, settling time.
  6. Do the source-to-source pickup test.
  7. Repeat at max input, max load, start-up, short circuit, load steps and shut-down, using peak detect on long timebases to find the worst edge.
  8. Measure VGS gate-to-source at the pins under the same conditions.
  9. Compare every worst-case peak with its limit; record margins and PASS/FAIL.

Report template: part number and rating, derating rule, temperature, conditions, probe and ground, bandwidth and sample rate, Vpk, plateau, overshoot %, fring, ζ, settling time, VGS peaks, margins, PASS/FAIL, with screenshots.

12. Common Mistakes and Fixes

MistakeResultFix
20 MHz limit left onPeak reads too lowBandwidth limit OFF
Long ground clipFake ~100 MHz ringing, wrong peakGround spring at the pins
Low sample rate or long timebase in sample modePeak missed or aliasedHighest sample rate; peak detect for long views
Only steady state measuredStart-up / short-circuit peak missedTest every worst-case condition
Comparing with V(BR)DSS at 25 °C onlyNo margin when coldUse the rating at minimum temperature
Plateau reported instead of peakLooks safe, isn’tCompare the peak with the limit
VGS measured to groundSource bounce addedMeasure gate to source pin
Input clippedDistorted edgeScale so the whole peak fits

13. Ringing and Overshoot Calculator

Tab 1 estimates the ring frequency, required bandwidth and sample rate, expected peak and probe-ground resonance. Tab 2 checks your measured peak, damping and gate voltages against the datasheet limits.

Ringing and Overshoot Calculator: Setup and Pass/Fail
Tab 1 sizes your scope and probe setup and estimates the peak. Tab 2 checks your measured overshoot against the MOSFET datasheet limits.
Ringing frequency
1/(2π√(LC))
Scope + probe BW
Sample rate
at least 5 × bandwidth
Estimated peak VDS
Timebase
edge + about 8 ring cycles
Probe ground resonance

Related Guides

14. FAQ

Should I use the bandwidth limit when measuring MOSFET ringing?
No. Turn the bandwidth limit off. Switching ringing is typically 50 to 300 MHz, and a 20 MHz limit filters it out so the peak reads too low. Use scope and probe bandwidth of at least 0.35 divided by the rise time and three times the ringing frequency.
What is the difference between overshoot and ringing?
Overshoot is how far the first peak rises above the settled (plateau) voltage. Ringing is the damped oscillation that follows, described by its frequency and decay. The peak voltage, not the plateau, is compared with the MOSFET rating.
How much VDS margin should a MOSFET have?
Most design rules keep the worst-case measured peak at 75 to 90 % of V(BR)DSS, with 80 % common, using the breakdown voltage at the minimum operating temperature because it falls when the part is cold.
Why does the long ground clip create fake ringing?
The clip inductance of about 150 nH resonates with the probe input capacitance of about 10 to 15 pF near 100 to 130 MHz. A fast edge excites it and adds ringing that is not in the circuit. A ground spring has only a few nanohenries and moves that resonance above the probe bandwidth.
How do I measure the damping of the ringing?
Measure two successive peaks above the plateau, A1 and A2. The logarithmic decrement is ln(A1/A2) and the damping ratio is that value divided by the square root of 4 pi squared plus its square. Low damping means long ringing and stronger EMI.
What time frame should I use for ringing measurements?
Use 2 to 10 ns/div for the rise time, 10 to 50 ns/div for the peak, ring frequency, damping and settling time, and 0.1 to 10 ms/div with peak detect to find the worst edge during start-up, short circuit or load steps.
Is it OK if the MOSFET briefly avalanches?
An avalanche rating (EAS, EAR) lets the part survive occasional fault events, but normal operation should never reach breakdown. If the measured peak exceeds V(BR)DSS at minimum temperature, reduce the overshoot or use a higher-voltage MOSFET.
How do I check that the ringing I see is real?
Do a source-to-source test: touch both the probe tip and the ground spring to the MOSFET source pin with the circuit running. The trace should be almost flat. Also move the probe slightly and compare the ring frequency with 1/(2 pi times the square root of L times Coss).

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

electronics believer