Product sourcing guide

USB-C Charger Ripple Reports: Match the Setup Before Comparing Millivolts

Two suppliers send output-waveform screenshots. One looks quieter, but its capture comes from a different load state and measurement point. Which charger is better? The images alone do not answer. Before millivolts influence a sourcing decision, establish whether both reports describe the same electrical quantity under a comparable method.

Classify the waveform before comparing its number

For this worksheet, treat ripple/noise as output-voltage variation measured during a declared settled operating interval. Identify whether the reported quantity covers repetitive ripple, broader noise or both; the label alone does not define a measurement. State the metric, bandwidth and observation window. A peak-to-peak result and an RMS result are not interchangeable purchasing figures.

Keep transitions in a separate dataset. A load step, output-voltage change, attachment event or restart may be important, but its excursion is not automatically a steady-state ripple result. If a capture straddles a transition, preserve it and label the event instead of silently selecting a quieter subsection. The agreed window-selection rule must be the same for both reports.

The USB-IF Source Power Requirements Test Specification, revision 0.79, dated May 24, 2021, explicitly places source-output voltage measurement at the connector receptacle in its assertion tables. It also distinguishes static from transient conditions and requires time for the output to settle between load steps. This older document illustrates why measurement location and event classification matter; it is not presented as the complete current certification program or a universal ripple acceptance limit.

Fix the electrical measurement plane

A report taken at the charger receptacle and one taken after a cable describe different boundaries. Declare which boundary the purchase requirement addresses. Retain a fixture diagram showing the voltage-sensing point, cable, connector path and load connection. Record any additional fixture capacitance or other loading introduced by the measurement arrangement rather than treating the fixture as invisible.

Do not relabel a load-end result as charger-receptacle ripple. If both boundaries matter to the application, request two explicitly named datasets using a controlled arrangement. They can answer complementary questions without being interchangeable. A small number cannot compensate for an unidentified sensing point or a missing explanation of the test fixture.

A closed charger, cable and load with separate source-receptacle and load-end sensing points, explicitly not a probe wiring diagram
Conceptual electrical boundaries only, not a wiring instruction. A qualified laboratory selects a safe output fixture; the charger remains closed.

Read component plots within their stated conditions

Texas Instruments’ TPS2585x-Q1 datasheet labels separate 6-A, 100-mA and no-load output-ripple figures in section 11.2.3. The associated plots and application-curve preamble specify operating conditions; nearby load-transient figures are separately titled. These are automotive synchronous-buck converter examples, not measurements of a mains USB-C adapter. Their amplitudes are not purchase limits, and they do not establish a ChargeKeku design or performance claim.

TI’s UCC28740 datasheet explains that valley-skipping operation can add output ripple whose frequency and amplitude depend on the shunt-regulator loop response. That is a controller-specific mechanism, not a diagnosis of an unknown charger from its screenshot. Use it as a reason to preserve operating conditions, not to assume all chargers share one ripple pattern or control method.

Use a two-report comparison worksheet

Fill both columns from the underlying records, not from assumptions about what a laboratory “normally” uses. A missing field is an evidence gap. A difference is not necessarily a defect, but it can prevent a fair numerical comparison.

Comparison fieldReport A must identifyReport B must match or explain
Sample and operating pointRevision, input source, fixed output profile and loadEquivalent declared state; sample identity remains separate
Thermal and port stateWarm-up/stability criterion and every port’s stateSame criterion and port configuration
Electrical planeSensing location, cable and fixture loadingSame boundary and equivalent fixture arrangement
Acquisition chainProbe, attenuation, coupling and effective bandwidthSame agreed method or documented comparability
Capture and processingSampling, timebase, record length, window and filteringEquivalent capture rule; disclose averaging or other processing
Reported metricPeak-to-peak or RMS definition, units and uncertaintySame calculation, uncertainty basis and acceptance rule

An oscilloscope screenshot is useful context, not a substitute for the acquisition record. Retain instrument and probe identification, configuration and raw waveform files. Ask the laboratory to document clipping, overload indications or other validity concerns. A cropped image that hides settings cannot establish that the two acquisition chains were equivalent.

Do not prescribe 20 MHz as a universal USB-IF ripple standard. If the buyer and laboratory choose that bandwidth for a particular method, state it as their agreed setting and document its applicability. The same applies to probing conventions: an appropriate technique depends on the fixture, equipment ratings and measurement objective, not a generic instruction copied between reports.

Commission matched captures safely

Use a qualified laboratory, a closed charger and suitable safe output fixtures. The laboratory must assess reference connections, instrument ratings and isolation requirements. This is not an invitation to open a mains adapter, defeat protective earthing or attach an unsuitable probe. Test only approved operating points and stop under the laboratory’s abnormal-condition rules.

  1. Agree the acceptance question. Specify the electrical plane, metric, operating cases and reporting window. Obtain the customer-supplied millivolt limit, if a pass/fail decision is required, and an explicit uncertainty decision rule. Without them, report observations or comparability rather than inventing a pass threshold.
  2. Match the operating state. Freeze input source conditions, output profile, load, thermal preparation and other-port states. Record actual conditions rather than only nominal labels. Use the same valid no-load, light-load and working-load definitions where those cases are included in the approved plan.
  3. Document the acquisition chain. Capture instrument/probe identities and all worksheet settings, including the sensing point and fixture loading. Let the laboratory justify suitable coupling, bandwidth and sampling. Save settings with the run so a later screenshot cannot detach the number from its method.
  4. Capture the declared interval. Start the comparison window only when the agreed settling criterion is satisfied. Keep attachment, load-step and voltage-transition records separately labeled. Preserve the complete relevant acquisition and explain any excluded interval instead of selecting only the most favorable trace.
  5. Calculate consistently. Apply the same agreed metric and processing to both datasets, retaining raw files and derived results. Document uncertainty and any limitations that affect interpretation. If a material setting differs, request matched acquisition or a defensible method comparison rather than arithmetic correction by guesswork.
  6. Issue a scoped decision. Join each result to sample, run, configuration and requirement identifiers. Retain unfavorable and repeated captures. Resolve mismatches before ranking suppliers; do not describe a method-comparability check as product certification or proof of long-term reliability.

Compare, retest or hold

FindingProposed decisionWhy
State, plane and method are matchedCompare the agreed metricApply the customer’s limit and uncertainty rule separately
Load, thermal state, plane or bandwidth differs materiallyRequest matched capturesThe numerical ranking is not yet justified
Trace is clipped or essential settings are missingHold the numeric claimMeasurement validity cannot be established
Transition is included in a settled-state resultSeparate and reclassify the datasetsRipple/noise and transient response answer different questions

Mean-output accuracy belongs in a PPS setpoint acceptance review; temperature-dependent output behavior has its own thermal-derating test scope. Neither replaces this waveform comparison. For an OEM sample-evidence enquiry, send the completed worksheet, intended operating cases and customer acceptance rule. Request paired raw captures and unresolved method differences before accepting a “lower ripple” headline.

Sources