How to measure switching power supply ripple accurately

Overview

Ripple measurement looks simple until two engineers measure the same 24 V supply and get completely different numbers. One report says 35 mVpp. Another says 180 mVpp. The power supply did not change. The test method did.

That is the common trap with switching power supplies. Output ripple is a small AC signal riding on a much larger DC output. It also sits close to switching devices, transformers, rectifiers, long output leads, and control electronics that can radiate high-frequency noise. A poor probe ground or a careless test point can add more noise than the power supply itself is producing.

For B2B buyers, equipment manufacturers, and automation integrators, this matters because ripple numbers are often used to compare suppliers, check incoming quality, or diagnose field failures. A false high reading can reject a good unit. A false low reading can hide a real stability or EMI problem. The goal is not to make the waveform look pretty. The goal is to measure the rail the way the load actually sees it.

What ripple is, and what it is not

In a switching power supply, output ripple usually contains two visible parts. The first is the periodic ripple related to the switching action and output filter. This is the part designers expect to see around the converter switching frequency and its lower-frequency behavior.

The second part is higher-frequency spike noise. It often comes from diode recovery, MOSFET transitions, transformer leakage inductance, PCB layout, parasitic capacitance, and measurement loop pickup. In many real waveforms the two are mixed together, so the industry typically reports ripple and noise as peak-to-peak voltage, or Vpp, within a defined bandwidth.

Do not use RMS voltage as the main ripple value unless a customer standard specifically asks for it. A 20 mV RMS number and a 100 mVpp number can describe very different waveforms, and procurement teams can easily compare the wrong metric if the report is unclear.

Start with the oscilloscope settings

The oscilloscope does not need to be exotic for ordinary output ripple checks. The setup does need to be deliberate.

Use AC coupling when the task is to view small ripple on a DC rail. DC coupling keeps the full DC output on screen, so a 12 V or 24 V rail forces the vertical scale to stay large unless the scope has offset capability. AC coupling removes the DC component and lets you use a sensitive vertical range.

Turn on the 20 MHz bandwidth limit for normal output ripple measurement. Most power supply datasheets specify ripple and noise with a bandwidth limit, commonly 20 MHz. Full bandwidth can be useful when debugging switching node ringing, MOSFET transitions, or fast power integrity events, but it also lets more radiated and conducted noise enter the measurement. For a production-style output ripple check, the bandwidth limit makes the result more repeatable.

Set the vertical scale as low as the signal allows. Typical starting points are 1 mV/div, 2 mV/div, or 5 mV/div. If the scope is sitting at 100 mV/div, the display may still show a waveform, but the measurement resolution is poor for a tens-of-millivolts ripple specification.

Average acquisition can help when random ambient noise is present. Eight to sixteen averages is a practical range for a stable load condition. Avoid using infinite persistence as the main reading method; it can make occasional pickup look like continuous power supply noise.

Trigger from the measured channel and place the trigger level around the center of the ripple waveform. If the trigger is wandering, the displayed ripple may smear and the automated Vpp reading may become less stable.

The probe ground is where most bad readings begin

The standard long alligator ground lead is convenient. It is also one of the worst accessories for low-level ripple measurement on a switching supply.

A long ground lead forms a loop. That loop acts like a small antenna near fast-changing magnetic fields. In a switch-mode supply, those fields are everywhere: around the transformer, MOSFET, rectifier, snubber, and power traces. The loop can pick up switching noise that never existed across the output capacitor. The result is a ripple reading that may be several times higher than the real rail noise.

The best everyday method is to remove the long ground clip and use a ground spring. The probe tip touches the output positive point, while the spring touches the nearby output ground point. The loop area becomes tiny, so induced noise drops sharply.

When a spring ground is not available, use the shortest possible ground connection. A trimmed ground lead of 5 mm to 10 mm is much better than a 10 cm or 15 cm clip lead. If your team performs ripple tests often, it is worth building a small fixture or adding proper probe test points to the PCB.

1X or 10X probe: choose for the signal, not habit

Engineers often leave passive probes in 10X mode because it is the safe default for general measurements. For high-voltage nodes and fast switching waveforms, 10X is usually the right habit. Output ripple is different.

For low-voltage, millivolt-level ripple measurement, a 1X probe can give a better signal-to-noise ratio because it does not attenuate the signal before the oscilloscope front end. With a 10X probe, the ripple reaching the scope is ten times smaller, and the scope's own noise floor becomes more visible in the final displayed reading.

The tradeoff is bandwidth and voltage rating. Many 1X probes have much lower bandwidth than 10X probes, sometimes in the 6 MHz to 15 MHz range. For ordinary output ripple with a 20 MHz bandwidth limit, that can still be acceptable, but it should be checked against the test requirement. 1X mode is also only suitable for low-voltage outputs within the probe and oscilloscope ratings.

If the rail is 5 V, 12 V, or 24 V and the goal is ordinary output ripple, 1X is often a practical choice. If the task is to look at switch-node ringing, transformer voltage, MOSFET drain voltage, or a higher-voltage output, use the proper 10X probe, differential probe, or power rail probe.

Measure at the output capacitor, under real load

Ripple measurement only means something when the test point represents the rail delivered to the load. For a PCB power supply, the first choice is usually across the output electrolytic capacitor or output MLCC bank. The probe tip goes to the output positive side; the ground spring contacts the capacitor ground side.

Avoid measuring at the end of a long output wire if the purpose is to judge the power supply itself. Wire inductance and loop area can exaggerate high-frequency spikes. That measurement may still be useful if the question is, "What does this remote load see with this exact harness?" It is not the same as the converter's output ripple at the board.

The power supply should also run with the specified load. No-load ripple can be misleading because many switching supplies enter burst mode, skip-cycle mode, or other light-load behavior. Those modes are normal, but they do not represent rated-load operation. For production checks, use the load condition stated in the datasheet or customer test plan.

Keep the test area tidy. Do not lay the power supply board beside the oscilloscope power adapter, another switching converter, or a bundle of live motor cables. If the waveform changes when you move the probe cable or rotate the board, your setup is probably measuring environmental pickup as well as rail ripple.

How to read the waveform

Use Vpp as the reported value. The waveform may show a smoother periodic ripple plus narrow spikes. Unless the test standard separates ripple and spike noise, both normally contribute to the peak-to-peak reading inside the stated bandwidth.

As a practical example, many general-purpose 12 V switching supplies specify ripple and noise around 120 mVpp or lower, while 5 V outputs may be around 50 mVpp. Those are not universal pass/fail values. Medical, instrumentation, LED, PLC, and communication equipment can have tighter or different requirements. Always compare against the product datasheet and the customer's application standard.

When using automated measurement, verify that the scope is measuring the intended channel and that the acquisition window contains a representative waveform. A single switching spike at the edge of the screen can distort a Vpp reading. It is better to capture a stable section of waveform and repeat the reading after moving the timebase.

A quick confidence test: short the probe tip to ground

Before trusting a suspicious ripple number, short the probe tip to the ground spring while keeping the same oscilloscope settings and roughly the same physical location.

If the baseline noise drops to only a few millivolts, the measurement setup is probably usable. If the shorted probe still shows tens of millivolts of noise, the setup is picking up energy from the environment. Shorten the ground path, move away from noisy equipment, reduce loop area, and check that the bandwidth limit is enabled.

This quick test is especially useful in supplier audits and incoming inspection. It separates "the power supply is noisy" from "our probe is acting like an antenna." That distinction can save a lot of unnecessary arguments.

Floating outputs and grounded oscilloscopes

One safety issue deserves special attention. A typical bench oscilloscope has its BNC shell connected to protective earth. If the circuit point you call "ground" is not actually safe to connect to earth, attaching the probe ground clip can short part of the circuit.

For many isolated AC-DC power supplies, the secondary output may be floating, but that does not mean every possible measurement is safe with a grounded scope. For measurements across non-ground-referenced components such as MOSFET drain-source voltage, transformer windings, ungrounded shunts, or primary-side circuits, use an isolated oscilloscope or a properly rated differential probe. Do not float a bench oscilloscope by lifting its earth ground. That creates a shock hazard and can damage equipment.

Common mistakes that create false ripple

Most wrong ripple measurements come from the setup, not from the oscilloscope itself. The table below is a useful checklist for engineers, quality teams, and buyers reviewing supplier test reports.

What a good ripple test report should include

A useful report does more than state "ripple: 65 mV." It should tell the reader how the number was produced.

At minimum, include the output voltage, load current, input voltage, coupling mode, bandwidth limit, probe setting, test point, and Vpp reading. If the waveform includes large spikes, note whether the value includes them. For supplier comparison, keep the same load, same bandwidth, same probe method, and same test point across all samples.

For higher-volume B2B purchasing, it is worth defining a standard test fixture. Even a simple fixture with short coaxial connections, defined load terminals, and a fixed measurement point can make results far more repeatable between factories, incoming inspection, and customer labs.

Conclusion

Accurate ripple measurement is mostly a discipline problem. Use AC coupling, enable the 20 MHz bandwidth limit for standard output ripple checks, choose the right probe attenuation, keep the ground loop tiny, measure at the correct output point, and test under load. Then report Vpp with the test conditions.

When those details are controlled, ripple data becomes useful for engineering decisions and supplier comparison. SIPURUI supports switching power supply buyers with practical selection guidance, product documentation, and application-focused power supply solutions for industrial equipment and automation systems.