Ringing from a 15cm oscilloscope ground lead

The ground lead and probe tip form a series resonant circuit

A clip-style ground lead is, electrically, just an inductor. A straight wire has roughly 1nH per mm in practical estimates (thickness and nearby metal change it a little, but not the order of magnitude). A 15cm lead is about 150nH of wire alone, and 100-200nH once the clip routing and loop shape are included.

The probe tip also has input capacitance. A 10:1 passive probe has a compensation network at the tip that reduces the scope input (1MΩ, a dozen or so pF) to 10MΩ and a few pF, and the capacitance you see is roughly 8-16pF on datasheets. We use 10pF as a typical value.

The lead's L and the tip's C are in series in a single loop that runs from the test point through the ground lead back to the probe input. Every edge excites this LC, and it rings at its resonant frequency. With f = 1 / (2π√(LC)), L = 150nH and C = 10pF give f ≈ 130MHz. Across 100-200nH it is 113-159MHz.

Q and damping: why it rings for many cycles

The characteristic impedance of this series circuit is Z0 = √(L/C), which is 122Ω for L = 150nH and C = 10pF. Q = Z0 / R, where R is the source impedance at the test point plus the lead's own resistance (usually under 1Ω for thin copper wire).

Near a decoupling capacitor or on a power plane the source impedance falls below 1Ω. With R = 1Ω, Q = 122/1 ≈ 122, so Q exceeds 100 and the excited energy hardly decays. The 130MHz ringing lasts tens of cycles, a few hundred ns.

A larger R lowers Q. Critical damping (the point beyond which it no longer oscillates) is R = 2Z0 ≈ 245Ω. A series resistor of 100-200Ω at the probe tip (a damping resistor) brings Q down to about 1 and the ringing settles in 1-2 cycles. The resistor also divides the signal, so the amplitude must be rescaled.

A 1ns edge rings on the screen even when the board does not

The number 130MHz suggests that MHz-range signals are safe, but what triggers the ringing is the speed of the edge, not the repetition rate. A digital signal with a 1ns rise time has frequency content up to several hundred MHz and excites the 130MHz resonance.

The board sees a plain rising edge, yet the screen shows several cycles of overshoot and ringing. Concluding that "the board is ringing" is the classic misdiagnosis. What rings is the measurement system formed by the ground lead and probe tip.

There are two ways to tell. (1) Shorten the lead. Since f ∝ 1/√L, halving the lead roughly halves L and raises f by about 1.4 times. If the ringing frequency changes, the lead is the cause. If it does not, suspect the board side, such as poor supply regulation or reflections from a real transmission line (see also when a trace becomes a transmission line). (2) Touch the tip and ground to the same point (the short test in the next section).

The short test: see the measurement system's own noise with no source

Take the probe off the test point and touch the tip directly to the ground clip (or ground spring). There is no signal and no return current, so ideally the screen shows a straight line at zero volts.

If the trace wobbles, the loop formed by the probe and ground lead is picking up ambient fields. With a 15cm clip lead, the dangling lead loop becomes a small antenna and picks up noise from a fluorescent-lamp switching supply, nearby digital circuits and the power line. Switch to a ground spring and ground the tip right beside it, and the loop nearly disappears, so the trace usually gets close to flat.

A ground spring raises the resonant frequency

The fix is to lower L and push the resonance above the band you are measuring. A "ground spring" slipped over the probe tip (an accessory that ties the coax shield to a ground pad or via a few mm from the tip) shrinks the loop from 15cm to a few mm and brings L down to 1 to a few nH.

L = 3nH and C = 10pF give f ≈ 919MHz; L = 5nH gives f ≈ 712MHz. That is 5-7 times the 130MHz of the clip lead and is outside the band of most digital signals and switching noise.

Keep the probe's catalog bandwidth (typically several hundred MHz for a 10:1 passive probe) separate from the effective bandwidth of the setup. With a 15cm ground lead, the effective bandwidth tops out at that resonant frequency. The catalog figure assumes proper grounding, usually a spring or an equally short connection.

Active probes (FET input) have a tip capacitance below 1pF and many models make a short ground reference near the tip, so for the same reason their resonance is already in the GHz range.

Lead length, resonant frequency and measurable rise time

To see a 1ns edge at face value, even a 3cm lead is not enough, as the table shows. A 3mm spring still gives 3.3ns, so seeing a 1ns edge accurately takes an even shorter connection or an active probe. A signal slower than the table's 23ns (most analog circuits and slow logic) is rarely a problem even with a 15cm clip lead.

"Measurable rise time" is a rule of thumb, not an RLC simulation result. On the bench, the ringing amplitude and how fast it settles depend on the source impedance (R), so confirm with the short test and by comparing different lead lengths.

Switching-supply ripple: the loop picks up a magnetic field

So far this was the series resonance formed by the test-point voltage and the probe's input capacitance. Measuring the output ripple of a switching supply brings in another mechanism: the ground lead's loop acts as an antenna and picks up the magnetic field of the inductor and SW node.

Around the SW node and inductor of a synchronous buck, current changes sharply each time the switches toggle (as in the loss breakdown of a synchronous buck, the switching loss comes from the speed of this transition). If a long ground lead's large loop sits in the field of this large dI/dt, a voltage is induced through mutual inductance and shows up on the screen as a sharp spike at every switching event.

It is the same wiring inductance as in via inductance, but a different phenomenon. The earlier resonance is the probe's own response to the test-point voltage; this spike is coupled noise from the loop picking up the surrounding field. It appears even with no source at the test point, as long as the loop is in the field.

A typical mistake is to place the probe with a long lead right over the inductor or near the SW trace and conclude "the ripple is large." The fix is to shrink the loop with a ground spring and also to place it right next to the point you want to measure (the output capacitor's terminal), away from the inductor and SW node.

What the 20MHz bandwidth limit does

The "BW LIMIT" or "20MHz" button on many scopes inserts a low-pass filter ahead of the input and cuts high-frequency content. The 130MHz and 900MHz ringing, and the sharp spikes picked up from the magnetic field, carry their energy far above 20MHz, so the bandwidth limit makes them disappear from the screen.

If the ripple suddenly looks clean after enabling the limit, the measurement did not get better; everything above 20MHz was thrown away. If harmonics of the switching frequency really extend above 20MHz (which can matter for EMI), they vanish in the same way.

The limit is useful for separating noise from ringing (if it disappears, high-frequency content is the main cause). The final ripple compliance check, however, should be measured without the limit and with proper grounding. Reporting a number taken with the limit on gives a value smaller than the real one.

Common traps on the bench

• The ringing survives a board redesign. Before revisiting damping resistors, swap to a ground spring and measure the same point. If the frequency does not change, the cause is on the board; if it rises and the ringing disappears, the cause was the probe.
• Looking at a single edge of a differential or high-speed serial signal. With edges faster than 1ns, even a few-mm ground spring may not be enough. Consider switching to an active probe.
• The trace still wobbles in the short test with a ground spring. Suspect the probe's own noise floor or strong nearby fields (a switching supply, a motor, a radio). You may have to shield the measurement point or move away from the source.
• The diagnosis comes down to three checks: change only the lead length and see whether the frequency moves, touch the tip to ground and look at the self-noise, and see whether the bandwidth limit removes it.