The waveform changes when you move the probe
Fictional engineering case. All case measurements and outcomes are constructed for this series; no company or reader is being quoted.
Last week’s answer: the pin-compatible op-amp
The replacement fitted the footprint and worked from a cold start. As the board warmed, the sensor reading moved upwards and the available ADC headroom narrowed.
Those observations did not identify a unique cause. Offset drift, input current, PCB leakage and operating-range limits all remained credible. Here is how the fictional investigation separated them.
The team replaced the sensor with a stable voltage source set to the same input voltage. It repeated the temperature sweep with a low source resistance, then with a resistance representative of the sensor. The low-resistance test produced only a small residual drift. Restoring the higher resistance brought back most of the error.
For this worked example, the amplifier is a unity-gain buffer. At a fixed warm temperature and input voltage, increasing the source resistance by 100 kΩ increases the settled output error by +2.0 mV. Source drift, resistor temperature coefficient and instrument loading have been checked. Assuming the net input current is approximately constant over this small voltage change, the result corresponds to:
2.0 mV ÷ 100 kΩ = 20 nA
The positive sign means that the net current raises the input voltage in this particular circuit. It is not a universal direction for op-amp bias current. Nor does the calculation identify whether that current originates inside the amplifier or elsewhere on the board.
The next comparison therefore moved the replacement device to a clean, guarded test fixture. The temperature-dependent error remained. Several replacement samples showed the same trend; the original device produced a much smaller error under the same conditions. Separate leakage checks on the production board and repeat measurements after cleaning did not account for the observed shift.
Low-impedance measurements bounded the offset contribution. The measured input and output levels were compared with the selected device’s specified input common-mode and output-swing limits at the actual supply, temperature and load. They remained within those limits. The buffer was stable under the ADC’s actual sampling conditions.
In this fictional case, temperature-dependent input current was the dominant cause. It was not necessary to assume a defective batch or a firmware fault. The replacement’s temperature behaviour had been omitted from the comparison used to approve it.
The team selected an amplifier against its permitted error over temperature, including source resistance and board leakage. Room-temperature calibration could remove the initial error but could not prevent it changing as the assembly warmed.
This week’s challenge: which waveform belongs to the circuit?
A low-voltage, isolated bench prototype contains a 12 V synchronous buck converter. Two engineers investigate positive overshoot at its switching node, which is the drain of the low-side MOSFET. Its source is connected to the local circuit ground used as the measurement reference. One capture suggests a peak of 18 V; the other shows approximately 14 V.
Both use the same compensated 10:1 passive probe and oscilloscope channel, with unchanged bandwidth, sampling and acquisition settings. The input supply, load and probe-tip location are also unchanged.
The first setup uses a 150 mm ground lead. The second uses the probe’s ground spring. Both return to the same local ground pad adjacent to the low-side MOSFET, so the intended voltage reference is unchanged. Moving the long lead changes the apparent ringing, even while the probe tip remains in place. There is no observed change in the converter’s output voltage.
The team has verified that the circuit ground can safely be connected to the oscilloscope’s earth-referenced ground. This is not a mains-referenced measurement.
What would you test next?
How would you separate measurement-loop pickup and probe resonance from genuine switching-node ringing? What independent check would make you confident enough to compare the peak with the low-side MOSFET’s drain-source voltage limit?
The lower reading is not automatically the correct one. Explain the measurement you would trust and the conditions that would make it trustworthy.
The answer follows on 15 October, together with a connector that runs too hot despite apparently adequate current ratings. Submit your reasoning or a problem of your own through the Troubleshooter page.
