How to Measure Power Supply Ripple and Noise: Avoid False Readings
A power supply ripple and noise measurement from a supplier may not match your in-house capture of the same adapter. That gap does not automatically mean the sample changed. The two teams may have used different probe connections, bandwidth limits, loads, cables, or measurement points.
A defensible power supply ripple and noise measurement defines the complete test path: where the signal is measured, how the probe is connected, which bandwidth and acquisition settings are active, and what input, load, cable, and temperature conditions apply. Without that record, a waveform is difficult to reproduce and a pass/fail decision can be misleading.
This guide focuses on the DC output of an external AC/DC adapter during device integration. It does not replace conducted or radiated EMC testing, nor does it treat load-transient response as the same test.
What are you measuring when a specification says ripple and noise?
Ripple and noise describe small AC variations riding on a DC output. Ripple usually has a recurring pattern related to energy storage, switching, or control behavior. Noise often refers to faster spikes and less regular high-frequency content. A scope trace may contain both, plus energy picked up by the measurement path itself.
The wording in a datasheet matters. Peak-to-peak voltage measures the distance between the highest and lowest observed points within the capture. RMS expresses effective AC magnitude. These values answer different questions and cannot be compared as if they were interchangeable.
Keep four effects separate
- Periodic ripple: A repeating component associated with conversion and regulation behavior.
- High-frequency spikes: Narrow events that may come from switching edges, diode recovery, parasitic paths, or the test connection.
- Load-transient deviation: A temporary output change after load current moves from one level to another. It needs a defined load step and recovery criterion.
- Common-mode pickup: Energy coupled into the probe loop or oscilloscope reference path rather than a true differential voltage across the output.
Start with the exact metric named by the adapter specification or device requirement. If one document states RMS ripple and another records peak-to-peak ripple plus noise, their numbers do not establish agreement or disagreement.
Why can the same adapter produce different readings?
Millivolt-level measurements are sensitive to the path between the adapter and oscilloscope. A change that seems minor on the bench can alter the displayed peaks more than a real difference between two samples.
Three groups of variables deserve separate control: probe and fixture behavior, oscilloscope configuration, and actual operating conditions.
Probe loop and measurement-path pickup
A passive probe with a long ground clip creates a relatively large loop. That loop can pick up magnetic and electric fields near a switching adapter, electronic load, or mains cable. Fast edges can then appear as exaggerated spikes on the trace.
Tektronix’s probing guidance recommends a short ground path for low-level power measurements. A ground spring, tip-and-barrel contact, or a suitable coaxial fixture can reduce loop area. The chosen method still needs to match the supplier’s stated test procedure.
Bandwidth and acquisition settings
Bandwidth defines which frequency content reaches the measurement. Applying a bandwidth limit may suppress high-frequency content and lower a peak-to-peak result. Leaving the full bandwidth active can reveal faster events, but it can also include more instrument and environmental noise.
There is no universal bandwidth setting for every adapter. Some power-supply test methods specify a 20 MHz limit, yet it should be used only when the applicable specification calls for it. Tektronix notes that bandwidth filters can also remove harmonics needed for other measurements, while Keysight’s ripple guidance stresses signal fidelity, low loading, and a controlled low-noise path.
Acquisition mode also matters. Averaging can hide non-repetitive peaks. High-resolution modes can reduce displayed noise by combining samples. Sample mode, peak detect, averaging, and high-resolution capture may all produce different views of the same output. Record the selected mode instead of relying on a screenshot alone.
Operating conditions and measurement location
An adapter can produce different results at minimum load, normal operating load, and rated load. AC input, temperature, cable resistance, connector contact, and a device’s pulsed current draw may also change the waveform.
Location changes the question being answered. Measuring at the adapter output connector characterizes the adapter and its stated cable configuration. Measuring at the device input includes the cable, connector, device-side capacitance, protection network, and load behavior. Both can be useful, but only the point named in the acceptance requirement should determine pass or fail.
What setup gives a repeatable power supply ripple and noise measurement?
A practical setup needs an AC source suitable for the test plan, an electronic load or representative device load, an oscilloscope, and a probe or fixture suited to low-level measurements. Record ambient conditions and allow the sample to reach the required operating state before capturing results.
Use instruments and probes rated for the nodes being measured. Never defeat protective earth on a grounded oscilloscope or improvise isolation to remove a ground-loop symptom. If the reference arrangement is uncertain, use an appropriate differential method or qualified test laboratory.
Control the connection
Connect at the defined measurement point with the shortest practical return path. Keep the measurement loop away from the electronic-load leads, adapter mains lead, and switching nodes. Avoid changing cable routing between samples if the goal is a direct comparison.
Some specifications call for external capacitors, a particular termination, or a dedicated test fixture. Record the component type and value, placement, cable, and fixture. Adding an unspecified capacitor may produce a cleaner trace, but that result no longer represents the same method.
Record the oscilloscope configuration
The setup record should include:
- Oscilloscope and probe or fixture model
- Probe ratio, input termination, and coupling
- Bandwidth limit or full-bandwidth setting
- Vertical scale, timebase, and sample rate
- Acquisition mode and number of acquisitions
- Measurement statistic, such as peak-to-peak or RMS
- Any filtering, math channel, persistence, or software processing
When the result approaches the noise floor of the oscilloscope and probe, perform a setup check. One useful comparison is a capture with the probe connection shorted at the measurement point using the same loop geometry, where the method permits it. The purpose is to estimate what the test path contributes, not to subtract an arbitrary number from the final result.
Which line, load, cable, and temperature conditions should you test?
Use a test matrix tied to the device and adapter requirement. Testing every imaginable combination wastes time, while one room-temperature capture at nominal input can miss the operating point that matters.
The matrix below is a planning template. Replace each entry with project-specific values and conditions.
| Variable | Conditions to define | Why it changes the decision |
|---|---|---|
| AC input | Required low, nominal, and high input points | Conversion and control behavior can shift across the input range |
| DC load | No-load or minimum load, typical device load, rated load | Ripple and burst-mode behavior can vary with load |
| Device state | Standby, normal operation, radio or motor event, peak processing state | Real loads may be pulsed rather than constant |
| Output cable | Approved length, conductor, connector, and routing | Cable and connector affect voltage at the device and can alter coupled noise |
| Measurement point | Adapter connector, cable end, or device input | Each point includes a different part of the power path |
| Temperature and time | Required ambient, warm-up time, and thermal state | Control mode and component behavior may change after warm-up |
If the device has a dynamic load, keep steady-state ripple/noise separate from transient-response testing. A load-step test needs start current, end current, slew rate, pulse duration, repetition rate, and recovery window. Folding those events into one unidentified peak-to-peak number makes supplier comparison harder.
Engineers working with instrumentation or sensitive acquisition hardware can also review the site’s test and measurement application overview. Application context should shape the test matrix, not replace the electrical acceptance method.
How should you read and report the waveform?
Read the trace in layers. Check for a repeating lower-frequency pattern, switching-related content, narrow spikes, and events that appear only when a cable or probe moves. Compare timebases before assigning a cause: a long record can show low-frequency modulation, while a shorter record can reveal switching detail.
Do not label every fast peak as adapter noise. If changing the probe loop, cable position, bandwidth, or nearby load wiring changes the peak dramatically, repeat the test with a controlled setup before judging the sample.
Build a reproducible result record
Every reported value should carry its conditions. A useful record includes:
- Adapter model, rating, hardware revision, and sample ID
- AC input voltage and frequency
- Load type, current, mode, and timing behavior
- Output cable and connector configuration
- Measurement point and any external components
- Ambient temperature, warm-up time, and enclosure state
- Oscilloscope, probe or fixture, bandwidth, coupling, and acquisition mode
- Metric, measured value, screenshot identifier, and pass/fail criterion
A screenshot without this record is supporting evidence, not a complete result. The same principle applies when comparing catalog samples against a customized cable or connector. The site’s guide to matching power adapter specifications provides the broader compatibility context; this measurement record covers one narrow performance check.
How do you turn the measurement into an OEM acceptance criterion?
“Ripple and noise below 100 mV” is incomplete. It does not identify the metric, bandwidth, input, load, measurement point, cable, probe method, temperature, or permitted external components. A supplier can meet that sentence under one setup and fail it under another without changing the adapter.
Write the limit and method as one requirement. That requirement can then travel from inquiry through sample approval and production control.
| Field | What to state | Risk if omitted |
|---|---|---|
| Metric and limit | Peak-to-peak or RMS, with units | Results using different metrics get compared |
| Measurement point | Exact connector, cable end, or device node | Adapter and system behavior become mixed |
| Probe or fixture | Connection type and required external components | Pickup or loading changes the waveform |
| Scope settings | Bandwidth, coupling, termination, acquisition mode | High-frequency content is included inconsistently |
| Input and load | AC condition and defined load state | A favorable operating point can mask another condition |
| Cable and temperature | Approved configuration and thermal state | Sample and production tests no longer match |
| Pass/fail rule | Samples, statistic, retest rule, and evidence | One unexplained peak can trigger an arbitrary decision |
Place this block in the technical requirement package, not only in an email thread. The existing guide to power adapter RFQ requirements shows how performance limits fit alongside AC input, load profile, connector, cable, evidence, and change control.
For a standard desktop adapter, confirm that the tested cable, connector, output rating, and model relationship match the configuration being quoted. If certification evidence also matters to the project, review it against the exact model rather than assuming that one family document covers every option; Merryking’s certification overview provides a starting point for that document discussion.
What should you check when a result exceeds the limit?
Treat an initial failed capture as a finding to reproduce. Confirm the adapter model and limit, then repeat the same operating point without changing multiple variables at once.
Use this order:
- Confirm model, revision, sample ID, cable, and applicable requirement.
- Recheck the measurement point, short return path, probe ratio, termination, coupling, bandwidth, and acquisition mode.
- Estimate whether the oscilloscope and probe setup is near its own noise floor.
- Repeat at the same AC input, load, temperature, and device state.
- Compare adapter output and device input only if both locations are part of the diagnostic plan.
- Change one controlled factor at a time, such as cable routing, representative load, or device-side filtering.
If a high reading remains repeatable, send the full record to the supplier. A clean record helps separate adapter behavior from cable drop, device load interaction, local decoupling, and measurement pickup. When the failure points toward a broader interference path rather than output ripple alone, the next reference is the site’s power adapter EMI and EMC troubleshooting guide.
What evidence should you send for adapter technical review?
A ripple value becomes useful when another engineer can reproduce it. Send the device input range, continuous and peak load behavior, target markets, cable and connector, operating environment, proposed limit, measurement method, and the complete result record. Include raw waveform files when available, not only a cropped screenshot.
Merryking can review those requirements against a relevant AC/DC power adapter family and the proposed production configuration. The review still depends on the exact model, cable, test method, and device conditions; no single ripple limit can represent every adapter and application. When the package is ready, contact Merryking for a technical discussion tied to your B2B project.
Is ripple the same as EMI?
No. Output ripple/noise and electromagnetic interference can interact, but they use different setups and acceptance criteria. A differential output measurement does not replace conducted-emissions, radiated-emissions, or immunity testing.
Why does oscilloscope bandwidth change the ripple reading?
A bandwidth filter includes or removes frequency content before the value is calculated. Two settings can therefore produce different peak-to-peak readings from the same adapter. Use the bandwidth named by the applicable test method and record it.
Should I measure at the adapter connector or the device input?
Measure at the point stated in the requirement. During integration, recording both points can help separate adapter-output behavior from cable, connector, and device-side effects, but only the defined point should control acceptance.
Can a multimeter measure power-supply ripple accurately?
Usually not for broadband noise, switching content, or narrow spikes. A multimeter may provide an AC reading within its own bandwidth, but it cannot replace a suitable oscilloscope setup when waveform shape and peak-to-peak deviation matter.
Does a lower ripple number always mean a better adapter?
No. The number matters only when it comes from a comparable method and supports the device’s actual needs. Safety, regulation, transient behavior, EMC, thermal performance, cable drop, reliability, and evidence for the exact configuration also affect approval.