Portable power station test protocol

Scope and evidence boundary

Version 1 covers single-phase 120 V / 60 Hz portable power stations with integral rechargeable storage and at least 100 W of claimed continuous AC output. Lower-rated products are outside v1 so its 100–200 W reference-load rule never exceeds a unit's continuous rating. It measures delivered AC energy, runtime, paired idle depletion, an observed rated-output interval, an advertised EPS/UPS mode, wall recharge, and noise. A result describes one physical sample under stated conditions.

• It is not a safety certification or conformity assessment.
• It is not proof that every unit of the same model performs alike.
• AC outlet Wh is usable AC output energy, not internal battery capacity.
• Low-load retention is not inverter efficiency; the internal DC bus is not measured.
• An observed switch interval is not UPS certification or a compatibility guarantee.

Specimen identity and equipment record

Assign one immutable public unitId. Record manufacturer, model, hardware and firmware revision, acquisition condition, claimed Wh, continuous W, surge W, and the manufacturer specification URL. Never publish the full serial number. A retest receives a new test ID; it is not a correction of the original run.

Every quantitative instrument record includes its public instrument ID, make and model, range, accuracy statement, calibration date, and uncertainty where available. Use “measurement result supported by documented metrological traceability” only when the calibration chain and uncertainty are actually documented. “NIST-certified” and an unsupported “NIST-traceable” label are prohibited.

Conditions, conditioning, and repeat policy

1. Check the exact unit for recalls, damage, swelling, leakage, odor, unusual noise, or abnormal heat. A failed check ends the procedure before energizing it.
2. Use the supplied charger unless the manufacturer explicitly identifies an approved replacement. Follow the manual's charging, ventilation, clearance, and temperature rules.
3. Target 23 ±2 °C, record temperature and relative humidity, and remain inside the manufacturer's allowed range. The 23 °C target is this protocol's repeatability choice.
4. Reach the manufacturer's full indication, wait until input is stable near maintenance level for 15 minutes unless the manual says to disconnect sooner, unplug, then rest the sample for 60 minutes.
5. Disable unrelated lighting and wireless outputs. Record AC mode, eco/auto-off, charge mode, and every changed setting; never silently alter a setting.
6. Use a resistive AC load with measured power factor above 0.99. Reference load isclamp(20% × rated continuous W, 100 W, 200 W); low load is a measured 15 W.
7. Run two valid capacity/runtime trials. Add a third when the first two differ by more than 3%. Publish every run and the median—never only the favorable result.

Eight measurement modules

Usable energy, runtime, and low-load retention

From the conditioned full state, apply the measured reference load and log AC output until automatic cutoff. Repeat at 15 W. If factory auto-off ends the low-load run, publish that behavior. Calculate low-load retention only when auto-off can be disabled through an ordinary user setting and that setting is disclosed.

Paired AC-on idle depletion

Perform two paired A/B trials and alternate the order. In A, leave the inverter off for eight hours, then discharge at the reference load. In B, leave AC on with no external load for eight hours, then use the same discharge. If the inverter shuts itself off, publish the shutdown time and do not extrapolate eight-hour draw. A negative paired delta is reported as below detection—not silently clamped to zero.

Observed rated-output hold

From full, hold a resistive load at 50% of the claimed continuous rating for 10 minutes, then at 100% for 15 minutes. Publish actual average W, voltage, duration, and stop reason. Version 1 performs no surge, overload, short-circuit, protection-bypass, or destructive test.

Advertised EPS/UPS switchover

Run this module only when the manufacturer advertises an EPS or UPS-style mode. Capture five input-loss events under a documented resistive load with a properly isolated, mains-rated two-channel recorder at sufficient sample rate. A connected device “staying on” is not a timing measurement. Never connect a grounded USB oscilloscope directly to mains. Without a qualified operator and correctly rated equipment, publishnot_tested.

AC recharge

After automatic cutoff and cooling, recharge with all outputs off. Record wall-input Wh, peak input W, elapsed time to displayed 80%, displayed 100%, and the following 60 minutes of taper separately. This is wall-to-station input—not cell charge efficiency.

Noise

Use an IEC 61672-1 Class 1 or Class 2 sound-level meter. A phone, laptop, or unqualified microphone does not become compliant by calibration software. At one metre with the microphone level with the unit, record three 60-second LAeq and LAFmax observations for AC on/no load, reference load, and normal wall recharge; record background the same way. One metre and 60 seconds are v1 protocol choices, while NIOSH supports keeping distance and duration consistent. If source minus background is under 10 dB, mark the resultbackground_limited; under 3 dB, publish no product noise number.

Safety stop rules

Stop immediately for a recall, damage, swelling, leakage, odor, odd noise, excessive heat, unstable output, an instrument overload, a loose connection, or any manufacturer warning. Never open the enclosure, bypass the battery-management system, defeat thermal protection, use a non-approved charger, or leave charging and high-load modules unattended. Rated-output and switchover work require a qualified operator and properly rated equipment. Publishingnot_tested is the correct result when those conditions are not met.

Certification claims stay separate from performance

ANSI/CAN/UL 2743 covers construction, safety, AC-pass-through/UPS, and specified rating-test requirements for products within its scope. PowerPreflight's protocol is not UL conformity testing and cannot grant, reproduce, or imply certification. A claim check records the exact mark, standard, certifier, listing or file number, lookup URL, and checked date.

• verified_current: the exact model is found in the certifier's current listing.
• mark_seen_only: a mark is visible, but the exact listing is not yet verified.
• not_found: listing not verified as of the stated date—never translated to “uncertified.”
• not_claimed: no certification claim was identified for checking.

OSHA explains that proper NRTL certification involves a registered mark, a product and standard inside the laboratory's recognized scope, and certification from a recognized site. OSHA recognition is not a product endorsement, and not every consumer power station is automatically subject to workplace NRTL requirements.

Reporting formulas and uncertainty

usable_ac_delivery_pct = usable_ac_output_wh / manufacturer_rated_capacity_wh × 100
runtime_hours          = elapsed_seconds / 3,600
low_load_retention_pct = median_low_load_output_wh / median_reference_output_wh × 100
idle_depletion_w       = (control_remaining_wh - ac_on_remaining_wh) / dwell_hours
round_trip_ac_ratio_pct = matched_usable_ac_output_wh / wall_recharge_input_wh × 100

“Usable AC delivery percentage” compares outlet energy with the manufacturer's label Wh; it is not battery efficiency. “Round-trip AC ratio” compares a matched outlet discharge with wall recharge input; it is also not battery efficiency. Missing or zero denominators produce no derived number. Every result carries instrument accuracy and uncertainty notes, and raw 1 Hz logs stay outside the Astro build; only compact summaries are published.

Blank result templates

The JSON template is canonical and contains no sample result values. The CSV contains one header row and zero data rows. Neither download is a dataset of tested products, and this site does not accept community uploads.

Schema 1.0.0 · Protocol 1.0.0 · Template 1.0.0

Versioning and corrections

• Major: a measurement definition, load, conditioning rule, threshold, or formula changes; old and new results are not directly comparable.
• Minor: an optional module or field is added without redefining an existing result.
• Patch: copy or source correction with no effect on a result.

Schema and template versions move independently. A published result artifact is never overwritten. A correction creates a new result version withsupersedesResultId, correctionReason, and a correction date. Invalidated results remain accessible as tombstones but stay out of summaries. Every result records protocol and schema version, source commit, instrument IDs, test date, and immutable result ID.

Primary references and what each supports

• US Department of Energy — current UPS test-procedure hubRegulatory context for products that meet the federal UPS definition; it does not make every portable station a UPS.
• ENERGY STAR UPS Version 2.0 specificationReference-load, stabilization, simultaneous-measurement, and instrument-accuracy concepts; not a portable-station certification path.
• IEC 62040-3:2021Performance and switching reference only for a marketed UPS/EPS continuity mode within scope.
• IEC 61672-1:2013Performance specifications for Class 1 and Class 2 sound-level meters.
• NIOSH — comparing equipment noiseSupports holding microphone distance and measurement duration constant across comparisons.
• NIST policy on metrological traceabilityTraceability requires a documented unbroken calibration chain and uncertainty; NIST does not certify third-party traceability claims.
• ANSI/CAN/UL 2743 Edition 3Current standard for covered portable power packs; this protocol is not a conformity assessment or certification.
• OSHA NRTL program FAQExplains the certifier, mark, applicable-standard, recognized-site, and scope requirements behind an NRTL certification claim.
• US Fire Administration — battery fire safetySupports following product instructions and stopping use for odor, excessive heat, deformation, leakage, or odd sounds.

Calculator boundary

The site's calculators still use adjustable planning assumptions published onHow We Estimate. Until physical results exist, they do not silently substitute test data. Even after results exist, one sample's measurement cannot certify a user's exact unit, load, wiring, medical device, or outage plan.