Loadbank testing sits at the end of a generator commissioning sequence as the point where performance claims become measurable. The structure is straightforward — stepped load progression, dwell at each increment, sustained full-load run. What varies considerably across test packs is whether the records actually demonstrate stable performance at each stage, or whether they document that load was applied and something was recorded.
Dwell time and the stability question
The standard progression — 25%, 50%, 75%, 100% of rated kVA — is less significant than what happens between steps. Dwell periods exist because coolant temperatures, governor response, and alternator excitation all require time to reach equilibrium after a load change. The steady-state readings that appear in test records are only meaningful if taken after that equilibrium is reached — typically five to ten minutes minimum per step, depending on machine size and ambient conditions.
A pattern that appears in test packs is compressed step intervals: time column entries showing transitions within ninety seconds or less of the prior step. Voltage and frequency are recorded at each increment, the columns populate, and the record reads as a completed stepped load test. But where the time between steps is insufficient for thermal stabilisation, the recorded values may represent transient rather than steady-state conditions. The test pack documents load progression — it doesn't necessarily demonstrate stable performance at each step.
Step timestamps. Elapsed time between each load increment compared against the test procedure's specified dwell. Where no minimum dwell is stated in the procedure, the absence of that requirement is itself worth noting — the test plan should define it.
Reading timestamps relative to load application. A steady-state voltage and frequency reading taken at T+90 seconds after a step change may or may not reflect equilibrium. Engine parameters — oil pressure, coolant temperature — are slower to stabilise than electrical values and can indicate whether the machine had settled.
Consistency across steps. Compressed dwell at one step in an otherwise compliant sequence might reflect a specific operational constraint. Compressed dwell uniformly across all steps suggests the test campaign was structured around a time allocation rather than a technical requirement.
Power factor and what the loadbank was actually testing
Generator nameplate ratings typically reference 0.8 power factor lagging — the condition the machine is designed and sized for. A resistive-only loadbank operates at unity power factor, which is a materially different loading condition. The two sit at different points on the generator's capability curve; reactive current demand at 0.8 PF lagging places additional burden on the alternator excitation system that unity loading doesn't exercise.
Reactive loadbank elements are available precisely to replicate rated power factor conditions during testing. Where test records don't specify loadbank configuration — resistive only, or combined resistive-inductive — it's not possible to determine whether the test was conducted at nameplate PF or at unity. That distinction may not affect whether the machine holds voltage under resistive load, but it does affect whether the test was representative of the operational loading profile the generator will actually see.
On an anonymised data centre commissioning review, test records presented load progression across four steps, each within 90 seconds of the prior, with voltage and frequency noted at each increment. Loadbank type was not recorded. All values were within limits. The test pack was presented as demonstrating full load acceptance to nameplate specification. Whether the machine had reached thermal equilibrium at any step, and whether reactive load was applied at any point, couldn't be determined from the documentation provided. The records weren't fabricated — they just documented less than the test specification required.
Transient behaviour under load steps
ISO 8528 defines limits on voltage and frequency transients during load application — the dip and recovery envelope that occurs at the moment load is stepped. Those limits exist because downstream equipment, particularly UPS systems and sensitive loads, may be affected by transient depth and recovery time even where steady-state values are within tolerance.
Test records routinely capture steady-state values at each step but not transient behaviour at the moment of application. That gap means the record doesn't address the ISO 8528 performance criteria — it shows the machine settled within limits, but not whether it stayed within limits during the step itself.
Worst-case voltage dip and recovery time at each step. Manual recording during testing can capture these if the test engineer notes deviation and time to recovery. Where a data logger was connected, the trace should be included — steady-state tabular readings alongside a logged trace are complementary, not redundant.
Zero-to-full-load step, where performed. Some test procedures include a single large step as a specific transient performance test. Where this is referenced in the test plan, the corresponding transient record warrants checking against the applicable limit class in ISO 8528 Part 5.
Absence of transient data where a data logger was present. If a logger was connected during the test — and most loadbank rigs are instrumented — a test pack containing only tabular steady-state values may suggest the trace data exists but wasn't submitted. That's worth querying before the pack closes.
Full-load duration and thermal evidence
The distinction between a generator that held full load for a brief interval and one that sustained rated output for the full duration specified in the test plan isn't always apparent from test records. Typical requirements for critical standby applications run to two to four hours at full load — not because an extended run changes the steady-state electrical performance, but because the cooling system, oil circuit, and fuel system are all still approaching equilibrium at thirty minutes.
Coolant temperature continues to rise for the first thirty to sixty minutes at sustained full load, and stabilises only once heat rejection balances heat input. A test that concludes before that equilibrium is reached hasn't exercised the cooling system at the condition it will operate in during a sustained outage. Whether the temperature was still climbing or had plateaued at test completion is typically visible in the engine parameters log — where one was kept.
A complete generator test record would be expected to show: load applied at each step with timestamps, power factor configuration, dwell duration at each step, steady-state voltage and frequency at each step, transient voltage and frequency behaviour at each step, and engine parameters — oil pressure, coolant temperature, exhaust temperature — recorded at intervals during the full-load run. Where any of those elements is absent, the record may demonstrate that a generator ran under load. It doesn't necessarily demonstrate that it meets rated specification under the conditions it will actually be called upon to perform.