Generator Load-Step Response and Transient Recovery Testing


A standby generator that carries a steady 60 percent load for an hour can still stumble when a chiller block starts. The steady-state rating describes the power the set can deliver once conditions have settled; load-step response describes what happens in the first cycles, when the engine, the alternator and the excitation system have to find that power faster than the load asks for it. Transient recovery testing measures that gap, and it shows whether the whole installation, not only the generator, can absorb a real block of load.

The short answer: a load step is a test of stored energy

When a block of load closes onto a running generator, the current rises before the engine has changed anything. The engine cannot increase torque instantly: fuel has to reach the cylinders, the turbocharger has to respond and the governor has to see the speed error. For a short interval the energy comes from the rotating mass of the engine and the alternator and from the magnetic energy stored in the machine. The frequency dips while the governor calls for more fuel, and the voltage dips while the automatic voltage regulator calls for more excitation. Both recover as the prime mover and the excitation system catch up.

This is why the nameplate rating is a poor predictor of step response. Two sets with the same kVA can have different inertia, different turbocharger behaviour and different excitation ceilings. The manufacturer publishes a block load acceptance capability, usually as a set of steps at stated conditions, and that data together with the project specification is where the acceptance criterion comes from. It is not a universal property of a 500 kVA or a 1000 kVA machine, and it should not be treated as one.

The electrical side matters as much as the mechanical side. A step applied at the generator terminals meets the machine reactance directly; the same step through a transformer and a long cable meets additional impedance, so the load end sees a deeper dip than the terminal recorder. Reports should state where the measurement was taken.

The load is not passive either. A motor draws a starting current several times its running current while it accelerates, with the exact multiple set by the motor and starter data. A rectifier draws a current that depends on its control and on the DC link. A step made of motors therefore behaves differently from an equal kW step made of resistance, and the difference shows in both the depth of the dip and the time taken to recover.

Four mechanisms decide the first seconds after a load step. All four appear in the same recording, which is why a single voltage or frequency number rarely explains a failure.
Mechanism What carries the step What it looks like when it is short of margin Channel that shows it
Stored rotational energy Engine and alternator inertia Frequency falls quickly until the governor restores torque Frequency or engine speed
Fuel and air delivery Governor, fuel system and turbocharger Frequency recovery is slow even though the dip is not deep Speed and fuel demand
Excitation AVR, exciter and field circuit Voltage dip is deeper than the load step alone suggests Terminal voltage and field current
Load behaviour Motors, rectifiers, drives and transformers The step is larger or longer than its kW figure implies Current waveform and power factor
Protection and controls Breakers, ATS sequence and drive ride-through A sound generator still loses the load during the step Event log and breaker status

What transient recovery means in measurable terms

Transient recovery is easier to argue about when it is described by a small set of numbers read against time. The recording should start from a steady baseline before the step, show the switching instant, show the worst point of the excursion and then show the return to a band around nominal. Each of those parts has a name, and a report that uses the names consistently can be compared with the manufacturer’s data and with the next test on the same site.

The baseline is the pre-step condition: voltage, frequency, load and power factor held steady long enough for the controller to settle. The excursion is usually described by its depth as a percentage of nominal and by how quickly it occurs. The recovery time is measured from the switching instant to the moment the value re-enters the acceptance band and stays there; the band itself has a width, and a trace that crosses the band, leaves it again and then settles has not recovered at the first crossing. Overshoot and any steady-state offset after the event belong in the report as well, because they point at different causes.

Frequency deserves its own wording. A governor can recover voltage quickly and still leave a frequency dip that lasts longer, and the rate of change at the start of the dip is often more informative than the lowest value. Loads with motors and drives are sensitive to both. The acceptance numbers for depth, time and band are set by the manufacturer’s published transient data and by the project specification, and they must also suit the equipment on the bus: a contactor, a drive and a UPS rectifier each have their own tolerance, stated in their own data.

A step test is judged from a set of channels recorded on one time base. The list below is the minimum needed to explain a result rather than only to record it.
Channel Where to measure What it explains
Voltage, three phases Generator terminals and load bus Dip depth, phase balance and recovery time at both ends of the installation
Frequency or engine speed Controller or independent speed sensor Frequency nadir, rate of change and governor response
Current, three phases Generator terminals Step size, phase imbalance and sharing where sets run in parallel
Real and reactive power Generator terminals Whether the step is mainly kW or mainly kvar
Field current or excitation voltage AVR or exciter Excitation headroom at the moment of the step
Ambient and coolant temperature Room and engine Whether the test sits inside the rating basis
Event marker Test controller and recorder Alignment of the traces with the switching instant

Why the load matters more than the set

Two sites with the same generator can pass and fail the same step test because the load is different. A step of 300 kW made of resistance is a change in real power only. A step of 300 kW made of motors is a change in real power plus a much larger change in reactive current for the duration of the start. The generator has to supply both, and the excitation system does most of the work on the reactive part.

A UPS is a special case. A double-conversion UPS normally brings its rectifier back in a controlled walk-in after a transfer, so the generator does not see the full rectifier load at the instant of transfer. That behaviour is a setting, and it should be confirmed rather than assumed. The UPS covers the seconds between mains loss and generator load; the generator has to be ready to accept the rectifier load inside that window. Where the battery autonomy is short, the acceptance has to be proven by test, because a slow walk-in or a generator that is not ready will put the load back on a battery that is already discharging.

Data halls make the point clearly. In the 1000 kVA example published on this site, the set burns about 200 litres an hour and about 4800 litres over 24 hours, while the PUE 1.4 example shows that the power drawn by the whole facility is larger than the IT load alone. A step test that uses only the IT load, or only a resistive bank sized to it, does not reproduce what the generator sees when cooling plant, pumps and rectifiers start together.

Step size is normally expressed as a fraction of rating, and the fraction a set can take in one block comes from the manufacturer’s block load acceptance data at the site conditions. Motor-heavy loads usually reduce that block, because the starting current is larger than the running current. The 10 percent to 20 percent light-load band is a different subject: it describes continuous running, not a step.

Different loads present very different steps to the same generator. The size in kW is only the first part of the description.
Load at the step What the generator sees at the instant What usually limits acceptance What to record
Resistive load bank Current in phase with voltage, real power only Engine fuel response and frequency recovery kW step, frequency nadir, recovery time
Direct-on-line motor High starting current at low power factor Voltage dip and excitation headroom Terminal voltage, starting current, acceleration time
Soft starter or star-delta Reduced current that lasts longer Duration of the dip rather than its depth Current profile across the ramp
Variable speed drive Controlled current with harmonic content Rectifier walk-in and harmonic distortion Current waveform, drive status, dc-link voltage
UPS rectifier Controlled walk-in after transfer Rectifier settings and battery autonomy Rectifier current ramp and UPS mode
Transformer energisation Inrush current set by flux saturation Protection settings and voltage dip Current peak and breaker operation
Mixed site load Combination of the above, applied together The largest single step inside the mix Step size and order of application

Instrumentation and test set-up

The recorder is the instrument that decides whether the test can be judged. It should be independent of the generator controller, time-stamped, and connected so that the channels are labelled before the first step rather than reconstructed afterwards. Voltage and current at the generator terminals are the minimum; voltage at the load bus shows what the load actually experienced and separates a machine problem from an installation problem.

Pre-test conditions belong in the record as well. Fuel level, coolant level, battery condition, ambient temperature and the state of the ATS all change the result; the 40 °C reference used elsewhere on this site is a comparison point, so the report should say whether the test ran in the morning or in the afternoon heat.

Set-up items that decide whether the recording can be used as evidence. Each one should be closed out before the first step is applied.
Item Purpose What to record
Transient recorder Capture the excursion and the recovery Channel list and sampling settings in the procedure
Voltage at generator terminals Machine performance Phase-to-phase and phase-to-neutral as required
Voltage at load bus What the load actually sees Same time base as the terminal channels
Current, three phases Step size and balance CT ratio and location
Frequency or engine speed Governor response Source: controller or independent sensor
Excitation current or field voltage AVR headroom AVR or exciter terminals
Ambient and coolant temperature Rating basis Room and engine, before and after
Event marker Align traces with switching Test controller and recorder

Running the test: sequence and acceptance

A step test is run in a fixed order so that each result can be attributed. The set is started and warmed at a light load, the baseline is recorded, and the first block is applied with the recorder already running. Each block is held long enough for the recovery to be complete and for the controller to settle, then removed, and the set is allowed to stabilise before the next block. The largest step is normally the one that matters most, but the smaller ones show whether the governor and the AVR behave consistently across the range.

The annual load test at 50 percent to 100 percent of rating is the routine proof that the set can carry load; a step test is a different exercise, aimed at the transition rather than the plateau. The two are complementary. A set that passes the annual run at 75 percent can still fail a 40 percent block, and a set that accepts the block can still be unhappy at 10 percent to 20 percent load for long periods. Records should make clear which test was performed and which acceptance criteria applied.

Fuel planning is part of the test plan. A 500 kVA set at full load burns 101 litres an hour, 80 litres an hour at 75 percent and 60 litres an hour at 50 percent, and the specific figure lies between 0.24 and 0.29 litres per delivered kWh. A step test that holds those blocks for a short period is a modest fuel quantity; a 1000 kVA set running at about 200 litres an hour for a full day is about 4800 litres, which is a different scale of logistics. The procedure should state the fuel available and the level at which the test will be stopped.

Acceptance is a comparison, not an opinion. The recorded depth, recovery time and settling behaviour are compared with the manufacturer’s published transient data and with the project specification, and the test also has to show that no protection operated, no drive tripped, no contactor dropped out and no UPS transferred to battery because of the step. A trace that meets the voltage limit while a drive trips has not passed the site test, even if it passes the machine test.

A step test sequence that keeps each result attributable. The blocks and hold times are taken from the project specification, not from habit.
Stage Action What it proves
Preparation Fuel, coolant, battery, protection and ATS checks; recorder connected and calibrated The test can be run and stopped safely
Start and warm-up Run at light load until temperatures stabilise Baseline conditions are representative
Baseline Hold a steady load and record a settled period The pre-step reference exists
First block Apply the agreed step with the recorder running Behaviour at the smallest step
Larger blocks Apply, hold, record and remove each block in turn Consistency across the load range
Critical step Repeat the largest or most onerous step The result is repeatable
Shutdown and review Cool down, stop, download traces and compare with the criteria The judgement is documented

What a failed recovery looks like

A failed step usually has a recognisable shape. A deep voltage dip with a fast recovery points at excitation headroom or at impedance between the machine and the load. A shallow dip with a slow recovery points at the AVR response or at a sensing and wiring problem. A frequency dip that recovers slowly points at fuel delivery, turbocharger response or governor settings, and a frequency that oscillates after the step points at control stability rather than at capacity.

The first response should be to read the traces, not to change settings. Governor and AVR gains interact with protection, with load sharing and with the other machines on the bus, and an adjustment made to pass one step can degrade another. Where a setting change is justified, it should be made by the manufacturer’s service engineer against the published procedure and followed by a repeat test with the same instrumentation and the same load.

Common shapes in a step-test trace and the direction the investigation should take. The trace narrows the question; it does not by itself authorise a setting change.
Symptom in the trace Likely cause First check Who decides the fix
Deep voltage dip, fast recovery Excitation ceiling or impedance to the load AVR limit, cable and transformer data Manufacturer and design engineer
Shallow dip, slow recovery AVR response or voltage sensing Sensing wiring and AVR settings Manufacturer’s service engineer
Frequency dip, slow recovery Fuel or air delivery lag Fuel pressure, filter condition, turbocharger Manufacturer’s service engineer
Frequency oscillation after the step Governor stability Governor settings and load sharing Manufacturer’s service engineer
Breaker or drive trips during the step Protection set below the starting current Protection curves and drive ride-through Protection and drive specialists
UPS transfers to battery Generator not ready inside the transfer window Transfer sequence and rectifier walk-in UPS and generator specialists

Records, retest triggers and the UPS link

The record of a step test is worth more than the pass mark. It gives the next engineer a baseline for the same set at the same site, which is what makes a later change visible. A useful record includes the load list and the step sizes, the single-line diagram, the measurement points, the recorder files, the ambient and fuel conditions, the acceptance criteria and the name of the person who judged the result.

Retesting is triggered by change rather than by the calendar alone. A new AVR, a governor adjustment, an alternator rewind, an engine overhaul, a change in protection settings, a new largest load or a change in the ATS sequence all justify a repeat step test. Between those events, the annual load test at 50 percent to 100 percent of rating and the trend from remote monitoring show whether the set still behaves as it did. A drift in the recorded dip at the same load and ambient is a reason to look before the next outage provides the proof.

The link to the UPS is the practical reason most sites care. The UPS covers the seconds between mains loss and generator load, and the step test is what proves that the generator is ready inside that window with the real rectifier load connected. When the two are tested separately, each can pass while the sequence fails; when the step test includes the UPS walk-in and the transfer, the site has evidence for the case that matters.

Send the load list, the single-line diagram and the recorder traces

Tell us the largest single step in the load, the motor and drive data behind it, the generator rating basis and the ambient at site. The reply sets out the step sequence to test, the channels to record and the acceptance items to agree before the next load test.

Request a quote WhatsApp the sales desk

What is the difference between load acceptance and load-step response?

Load acceptance is whether the set can take a defined block in one step without stalling or tripping. Load-step response is the recorded voltage and frequency behaviour during and after that change. Acceptance is a pass or fail against published data; the response is the trace that explains the result.

How large a load step can a generator accept?

The figure comes from the manufacturer’s block load acceptance data at the stated conditions, adjusted for ambient and for the project specification. It is not a fixed percentage of rating, and motor-heavy loads usually reduce the acceptable block.

Why does voltage dip when a motor starts but frequency stays stable?

Motor starting current is largely reactive, so it depresses voltage through the machine and cable reactance before the engine sees much real power. The excitation ceiling and the AVR response decide how deep the dip is and how quickly it recovers.

Can a resistive load bank prove motor-starting performance?

Not on its own. A resistive bank draws current at unity power factor, so it exercises the engine and governor but not the excitation system in the same way as a motor. It is useful evidence for block load acceptance of the prime mover; motor starting needs a test that reproduces the reactive current.

How often should transient recovery be tested?

At commissioning, after any change to the AVR, governor, alternator, protection settings or the largest load, and whenever the annual load test at 50 percent to 100 percent of rating shows a change from the baseline. Routine annual testing does not replace a step test when the load mix has changed.

What is the UPS doing while the generator recovers from a step?

A double-conversion UPS covers the seconds between mains loss and generator load, so the generator has to accept the rectifier load inside that window. If the UPS is already on battery when the step arrives, the acceptable generator recovery time has to sit inside the remaining battery autonomy.

KVA Grid sales engineering desk. The figures used here are published elsewhere on this site: the annual load test at 50 % to 100 % of rating and the 10 % to 20 % light-load band from the maintenance schedule; 0.24 to 0.29 litres per delivered kWh, and 101, 80 and 60 litres an hour for a 500 kVA set at full, 75 % and 50 % load, from fuel consumption; the 1000 kVA example at about 200 litres an hour and about 4800 litres for 24 hours, the UPS role between mains loss and generator load, and the PUE 1.4 example from data center backup power; the 40 °C reference ambient from hot and humid climates; load-bank practice from load-bank testing; UPS interaction from generator and UPS interaction; load sharing from paralleling and load sharing; and commissioning records from the commissioning checklist. Block load acceptance, voltage and frequency limits, recovery times, protection settings and acceptance criteria are set by the manufacturer’s data and the project specification. Reviewed 7 November 2026.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top
WhatsApp