Using an Oscilloscope on Switching and Power Circuits
Switching power supplies, motor drives, class D amplifiers and LED drivers are everywhere, and when they misbehave the oscilloscope is the tool that finds out why. They are also among the hardest circuits to measure well. Their nodes swing tens or hundreds of volts in nanoseconds, many of them are not referenced to ground, currents of several amperes flow in short pulses, and the circuit's own switching noise gets into careless measurements so that you see problems that are not really there — or miss ones that are.
This lesson shows how to measure these circuits safely and accurately: what to look at, how to probe it, and how to interpret what you see.
Measuring a switching converter: the switch node, the output ripple with a short ground spring, and the inductor current with a clamp-on probe. A long ground lead turns the converter’s magnetic field into spikes that are not really there.
View LargerSafety First: Ground References and Isolation
A mains-powered bench oscilloscope's probe ground clips are all connected to mains earth through the scope's power cord. Every measurement you make with a standard probe is therefore a measurement relative to earth, and clipping a ground lead to any point that is not at earth potential connects that point to earth through the probe.
- Off-line (mains-powered) supplies: the primary side — the bridge rectifier, the bulk capacitor at about 325–400 V DC, the switching MOSFET — is referenced to the mains, not to earth. Clipping a ground lead to the primary-side “ground” shorts the mains through the probe. The result can be a bang, a destroyed probe and scope input, a tripped breaker, and a shock hazard.
- High-side nodes in low-voltage circuits: the switch node of a buck converter, the gate of a high-side MOSFET, or one leg of a bridge-tied audio amplifier output are not at ground either. Connecting a ground clip to them short-circuits the circuit, even at 12 V.
The safe ways to measure floating or mains-referenced nodes are:
- A high-voltage differential probe, which measures the voltage between its two inputs without connecting either to earth. This is the right tool for primary-side measurements and for any node that is not at ground.
- Powering the circuit under test (not the scope) from an isolation transformer, so that its mains-referenced side no longer has a path to earth. This makes a ground-clip measurement possible on the primary side, with care — but the circuit is now floating at whatever voltage you connect it to, and only one ground clip may be used.
- A battery-powered scope with isolated inputs, rated for the voltage and measurement category involved.
The Switch-Node Waveform
The single most informative point in a switching converter is the switch node — the junction of the switching transistor, the diode or synchronous switch, and the inductor. In a buck converter it is a rectangular wave that swings between nearly the input voltage (switch on) and slightly below ground (switch off, current flowing through the diode or lower switch). Its frequency is the switching frequency, and its duty cycle sets the output voltage.
A buck converter makes 5 V from 12 V at a switching frequency of 500 kHz.
Ideal duty cycle D = Vout / Vin = 5 / 12 = 0.417 (41.7%).
Period = 1 / 500 kHz = 2 µs, so the switch should be on for 0.417 × 2 = 0.83 µs and off for 1.17 µs.
If the scope shows a slightly higher duty cycle (say 44%), that is normal: the converter compensates for the voltage lost in the switch, inductor and diode. If it shows the duty cycle jumping between widely different values from cycle to cycle, or bursts of switching separated by gaps, the converter is either in a light-load skip mode (normal for many chips) or unstable (a fault).
Look closely at the edges. A clean switch node rises and falls in a few to a few tens of nanoseconds with a little ringing. Large ringing — tens of volts at tens to hundreds of megahertz — comes from the circuit's stray inductance and the transistor capacitance, and is a common source of interference. A snubber (a small resistor and capacitor across the switch or diode) or a better layout reduces it. A slow, rounded edge means the gate drive is weak and the transistor is wasting power as it switches.
Measuring Output Ripple Properly
Output ripple is the most commonly mis-measured quantity in power electronics. With a standard probe and its 15 cm ground lead, the loop formed by the lead picks up the converter's switching magnetic field, and the scope shows large spikes that are not actually on the output. Datasheets specify ripple measured a particular way, and you must measure the same way to compare:
- Remove the long ground lead and use the ground spring (or wrap a bare wire around the probe's ground barrel), touching the tip and spring directly across the output capacitor. This shrinks the pickup loop to a few millimeters.
- Turn on the scope's 20 MHz bandwidth limit, which most ripple specifications assume.
- Use AC coupling and a sensitive range (10–20 mV/div), because the ripple is a few tens of millivolts riding on several volts.
- Use a 1× probe setting or a dedicated low-attenuation probe if the ripple is very small — a 10× probe adds ten times the scope's own noise.
A 5 V rail feeding a microcontroller and an ADC should have ripple below about 1% (50 mV peak-to-peak), and below a few millivolts at the ADC reference.
Measured with the long ground lead: 180 mV peak-to-peak, mostly sharp spikes at each switching edge.
Measured with the ground spring and 20 MHz limit: 22 mV peak-to-peak, a smooth triangle at 500 kHz with small spikes.
The rail is fine; the first measurement was mostly the probe. If the spikes remain after correct probing, add a small LC post-filter or a ferrite bead and a ceramic capacitor, as described in the Bypass and Decoupling lesson.
Measuring Current
Current waveforms tell you what voltage waveforms cannot: whether the inductor is saturating, how large the current peaks are, and whether the converter is in continuous or discontinuous mode. There are three common methods:
- Clamp-on current probe: a Hall-effect or transformer probe clamped around a wire. It measures DC and AC (Hall type) without breaking the circuit and without any ground connection. To measure inductor current, lift one end of the inductor and add a short wire loop for the probe.
- Current-sense resistor: a small resistor (10–100 mΩ) in the ground return, measured with a probe. It is cheap and fast, but it must be on the ground side unless you use a differential probe.
- The converter's own sense resistor: many controllers already have one; probing across it (with care about its ground reference) shows the switch current.
In a buck converter in continuous mode, the inductor current is a triangle centered on the output current. Its peak-to-peak ripple is:
ΔIL = (12 − 5) × 0.417 / (500,000 × 10 × 10⁻⁶) = 2.92 / 5 = 0.58 A peak-to-peak
At a 2 A load the inductor current should be a triangle from about 1.71 A to 2.29 A. If the scope shows the triangle's slope suddenly steepening near its peak, the inductor is saturating — its inductance is collapsing — and it needs a higher current rating.
Gate Drive and Switching Losses
Probing the gate of a low-side MOSFET (gate to source, with the source at ground) shows how well it is being driven. As the gate voltage rises, you will see a flat step part-way up — the Miller plateau — during which the drain voltage is falling and the gate driver is busy charging the gate-drain capacitance. A long plateau means slow switching. Gate voltage that never reaches the full drive level (typically 10 V for standard MOSFETs, 4.5 V for logic-level types) means the transistor is not fully on and will run hot.
Each switching edge wastes power while the transistor is partly on with both voltage across it and current through it. A useful estimate is:
P = 0.5 × 12 × 3 × 40 × 10⁻⁹ × 500,000 = 0.36 W — on top of the conduction loss I²RDS(on). Doubling the switching frequency or halving the drive strength doubles this loss.
Math, FFT and Power Measurements
Modern DSOs make power measurements much easier:
- Multiply (V × I): with a voltage probe on one channel and a current probe on another, the math function shows instantaneous power. Averaging it gives the power dissipated in a switch or delivered to a load.
- Measurement statistics: duty cycle, frequency, rise time and peak-to-peak with minimum, maximum and standard deviation over thousands of cycles reveal jitter and instability that a single screen does not show.
- FFT: shows the spectrum of the switch node or the input current, revealing the switching harmonics that cause conducted interference. It is a useful first look before formal EMC testing (covered in the Test Equipment: Advanced module).
- Deskew: voltage and current probes delay the signal by different amounts. Before multiplying them, use the scope's deskew adjustment to line them up, or the power reading will be wrong.
Experiment: Inside a Buck Converter Module
⚖ Experiment: Measure a Low-Cost Buck Converter
Inexpensive LM2596 or MP1584 buck converter modules are ideal for practicing power measurements safely: they run from 12 V and every node is easy to reach.
- A buck converter module (LM2596 or MP1584 type) set to 5 V output
- A 12 V bench supply
- A load: a 10 Ω 5 W resistor (0.5 A at 5 V)
- Oscilloscope with two 10× probes, ground springs, and a 20 MHz bandwidth limit
- Connect the supply and the load. Probe the switch node (the junction of the chip's output pin, the diode and the inductor) with the ground clip on the module's ground. Measure frequency and duty cycle, and compare the duty cycle with Vout/Vin.
- Zoom in on a rising edge. Measure the rise time and note the ringing.
- Measure the output ripple with the long ground lead, DC coupling and full bandwidth. Note the peak-to-peak value.
- Repeat with the ground spring across the output capacitor, AC coupling and the 20 MHz limit. Compare.
- Remove the load. Watch the switch node change as the converter enters light-load (pulse-skipping) operation.
The switch node is a rectangular wave at the module's switching frequency (about 150 kHz for an LM2596, 500 kHz to 1.5 MHz for an MP1584), with a duty cycle a little above 5/12. The ripple measured with the long ground lead looks several times worse than with the ground spring; the difference is pickup in the probe's ground loop, not real ripple. With no load, the switching becomes irregular bursts — normal behavior that saves power at light load.
Frequently Asked Questions
Why does the output ripple look like a row of sharp spikes?
Usually because of the measurement, not the supply. Each fast switching edge creates a burst of magnetic field and a current spike in the ground; the loop formed by a long probe ground lead turns that into a voltage spike on the screen. Measure again with the ground spring touching the capacitor directly and the 20 MHz bandwidth limit on. Whatever spikes remain are real, and are reduced with better layout, a snubber, or a small LC filter on the output.
My scope shows noise even when the probe tip is touching its own ground clip. Why?
The probe and ground lead form a loop that acts as a small antenna. Near a switching converter it picks up the magnetic field directly, so even with the tip shorted to ground you see the switching noise. This is a useful test: whatever the scope shows with the tip shorted to its own ground clip at the measurement point is the noise floor of your measurement setup, and anything smaller than that on the real signal cannot be trusted.
What bandwidth do I need for power electronics?
The switching frequency is rarely the problem; the edges are. A converter switching at 500 kHz may have 5–20 ns edges, which need 50–200 MHz of scope and probe bandwidth to see accurately. Gallium-nitride (GaN) transistors switch in about 1 ns and need several hundred megahertz. For ripple measurements, on the other hand, you deliberately limit the bandwidth to 20 MHz so the measurement matches the specification.
Can I use two probes as a differential probe by subtracting the channels?
For small, slow signals that are both near ground, sometimes. Put one probe on each side, leave both ground clips on circuit ground, and use the math function A − B. But the two channels never match perfectly, so common-mode voltage leaks into the result, and the method is useless for fast edges or high common-mode voltages. It also does nothing for safety: both ground clips are still connected to earth. For floating or mains-referenced measurements, use a real differential probe.
Test Your Knowledge
Answer the questions below to check your understanding. Every answer can be found in the lesson above.