A generator is often sized for the load it carries, then asked to start a motor that briefly behaves like a short circuit. The running load decides the kVA on the nameplate; the start decides whether the set holds voltage through the first two seconds. The two questions are different, and a design that answers only the first one is usually discovered during the first emergency.
A motor start is a short circuit with a useful outcome
At standstill a squirrel-cage induction motor has no back EMF to oppose the supply, so its impedance is close to the locked-rotor impedance and the current is limited mainly by that impedance. The result is a starting current several times the full-load current: commonly in the region of five to seven times for a standard induction motor, with the exact multiple taken from the motor datasheet rather than assumed. The same effect produces the starting torque that gets the load moving, so the current cannot simply be designed away without also reducing the torque.
For a generator the important number is not the multiple but the product of that multiple and the motor rating, expressed as starting kVA. A motor started direct-on-line can present several hundred kVA to the set for the duration of the start. The set sees that as a load step with a poor power factor, and the alternator voltage falls because the current flows through the alternator’s internal impedance. The fall is fast, the recovery depends on the excitation system, and the engine has its own slower response as the governor adds fuel and the turbocharger catches up.
Two consequences follow. First, the dip reduces the torque available at the motor terminals, because torque falls roughly with the square of voltage; a start that looked comfortable on paper can become a long, hot acceleration, or a stall. Second, the dip is seen by every other load on the same bus, not only by the motor being started. Both effects are transient, but they are the reason a set is sized for the start and not only for the running load.
Acceleration time links the two. During acceleration the motor draws a current that falls as speed rises, and the engine and alternator must supply it while the voltage recovers. If the load torque rises steeply with speed — a loaded conveyor, a positive-displacement pump, a crusher — the motor may reach a point where motor torque and load torque cross below full speed. The set then holds a high current at low speed until protection operates. That is a mechanical problem that appears as an electrical one, and it is settled with the load torque curve, not with a larger breaker.
| Data item | Why it matters | Where it comes from |
|---|---|---|
| Motor rating and full-load current | Defines the steady load the set carries after the start | Motor nameplate and datasheet |
| Locked-rotor current and starting kVA | The transient the set must supply | Motor datasheet at the stated voltage |
| Starting torque and load torque curve | Whether the motor can accelerate the driven equipment | Motor data and driven-equipment data |
| Starting method and current limit | The shape and size of the current step | Starter or drive datasheet |
| Acceleration time | How long the transient lasts and how much heat it adds | Calculation, checked at commissioning |
| Base load on the bus | The load already on the set when the start begins | Site load schedule, including UPS recharge |
The dip is set by the source, not by the motor alone
On a utility supply the source impedance is small compared with the motor, and a direct-on-line start produces a dip that is usually measured in a few per cent at the point of common coupling. On a generator the source impedance is the alternator, and it is not small. The alternator’s subtransient reactance, taken from its datasheet, dominates the first cycles of the start; the transient and synchronous reactances and the excitation system shape the recovery afterwards. A set with a low subtransient reactance and a fast, force-fed excitation system holds voltage through a start that would collapse a set of the same kVA built differently.
That is why two sets with the same nameplate rating can behave very differently on the same motor. The kVA rating describes continuous capability; it says little about the first two hundred milliseconds. When comparing options, ask for the subtransient reactance, the excitation type, the short-circuit ratio and any published motor-starting capability, and ask for the conditions under which those figures apply.
The dip should be checked in three places. At the motor terminals it decides whether the motor accelerates. At the main busbar it decides what the rest of the site sees. At the terminals of sensitive equipment — a UPS input, a drive, a control relay — it decides whether that equipment rides through or trips. A single figure for the whole installation hides these differences.
What counts as acceptable is a project decision, not a universal number. A planning figure in the region of 15 per cent at the motor terminals is sometimes used in starting studies, but the governing value is set by the project specification and by the tolerance of the connected equipment, and it has to be confirmed against the manufacturer’s data for each item. Contactors may drop out, relays may chatter, drives may trip on undervoltage, and a UPS may transfer to battery, each at its own threshold. The study is only useful if those thresholds are written down first.
| Equipment | What a dip can do | Design response |
|---|---|---|
| Contactors and control relays | May drop out on a deep or repeated dip | Confirm hold-in voltage; use delayed-dropout relays where the sequence allows |
| Variable-speed drives | May trip on DC-bus undervoltage | Check ride-through setting and restart logic |
| UPS input | Transfers to battery when the input leaves its window | Confirm the UPS covers the seconds before the set carries load |
| Lighting and IT load | Flicker, reboot or a transfer to battery | Confirm the dip is inside equipment tolerance |
| Motors already running | May slow, stall or trip | Sequence starts so only one large motor accelerates at a time |
| Generator protection | May read the start as an overload | Set definite-time and instantaneous elements to ride through the start |
Sizing the set for the start, not only for the load
Sizing a set for motor starting is a set of checks, not a single margin. The first check is steady state: the running load, plus the largest motor running, plus the site base load, must sit inside the set’s continuous rating at the site ambient. The second is the transient: the starting kVA of the largest motor, added to the load already on the bus, must produce a dip that the motor and the rest of the site can accept, and the engine must accept the kW step without a frequency excursion that trips equipment.
The load already on the bus matters more than it is often given credit for. In a data hall the IT load multiplied by a PUE of 1.4 is the base load the set is carrying before a chiller pump or a fire pump starts; the starting motor is added on top of that, not on an empty set. UPS recharge, battery chargers, lighting and other running motors belong to the same sum. A start that is comfortable on a lightly loaded set can be marginal at night, when the base load is higher and the same motor still has to accelerate.
The engine side has its own limit. A diesel engine accepts a load step according to its turbocharger, governor and fuel system, and the acceptable step is published by the manufacturer for the set, not for the engine alone. The transient speed dip and its recovery are described in load step response and transient recovery testing. A start that is acceptable electrically can still produce a frequency dip that trips a drive or a UPS, so the two checks belong together, on the same load schedule and in the same test.
Ambient derating applies to the transient as well. At the 40 °C reference ambient used on this site, the continuous rating already includes a margin for the site condition; the start adds a short overload on top of a rating that has been adjusted for heat. The cooling system has to reject that heat afterwards, which links the start to radiator siting and ambient derating. Repeated starts in a hot room heat the alternator and the engine, so the number of starts per hour and the interval between them are part of the duty, not an afterthought.
Fuel and test planning belong in the sizing note as well. Specific consumption of 0.24 to 0.29 L/kWh is a steady-load figure; a start burns fuel that does no useful work and adds heat that must be removed. A 500 kVA set at 50 per cent load uses about 60 L/h, and a 1000 kVA set at full load uses about 200 L/h, roughly 4800 L over 24 hours, so a test programme with repeated starts is a fuel and logistics exercise as well as an electrical one. The figures come from diesel generator fuel consumption.
| Check | Input | What it proves |
|---|---|---|
| Continuous load | Running load plus base load, at site ambient | The set carries the steady state with margin |
| Starting kVA | Largest motor starting kVA plus base load | The dip is inside the accepted limit |
| Alternator transient data | Subtransient reactance, excitation type, published starting capability | The set can supply the first cycles |
| Engine load step | Published block load acceptance for the set | Frequency dip and recovery are acceptable |
| Thermal duty | Starts per hour, acceleration time, ambient | Windings and coolant stay inside limits |
| Fuel and test plan | Consumption at test load, tank capacity | Tests can be run without disrupting the site |
Starting methods and what each one trades away
The starting method is a design choice that trades current against torque, cost and complexity. Direct-on-line is the simplest and gives full starting torque, at the price of the largest current step. Star-delta reduces the current to roughly one third and the torque to roughly one third during the star phase, which suits loads that are easy to accelerate, such as a fan or a centrifugal pump with a closed valve. The transition to delta produces a second transient, and a poorly timed transition can be worse than the start itself.
A soft starter reduces voltage during acceleration by phase-angle control, so the current falls and the torque falls with the square of the applied voltage. It fits where the load needs a gentle start but not a high breakaway torque. The starter is bypassed once the motor is at speed, so it carries no current in normal running, but it produces harmonics during the start, which links it to the harmonic mitigation discussion for the rest of the installation.
A variable-speed drive gives the most control: current can be limited to a set value, torque can be shaped to the load, and the motor can be accelerated slowly. The drive draws a rectified current with its own harmonic signature, and it may need to be oversized for the motor or fitted with an input filter. Not every motor is suitable for drive operation, and the motor and drive datasheets have to be checked together. An autotransformer starter uses taps to reduce voltage and, like the soft starter, reduces torque with the square of voltage; the taps are fixed at design, so it is less flexible than a drive.
Each method changes the question the set has to answer. A soft starter or a drive limits the current the set must supply, which can allow a smaller set or a larger motor on the same set, but it adds equipment that has to ride through the dip it creates elsewhere and that must itself be protected. The cheapest method on the starter is rarely the cheapest on the whole installation once set size, protection settings and commissioning time are counted.
| Method | Starting current | Starting torque | Best fit | Watch-outs |
|---|---|---|---|---|
| Direct-on-line | Full starting current, commonly several times rated | Full starting torque | Small motors, or sets with ample transient capability | Largest dip; check the breaker magnetic setting |
| Star-delta | About one third during the star phase | About one third during the star phase | Fans, centrifugal pumps, light starts | Transition transient; not for high breakaway torque |
| Soft starter | Reduced by voltage control | Falls with the square of voltage | Pumps and compressors started unloaded | Harmonics during the start; bypass contactor |
| Variable-speed drive | Limited to a set value | Controlled and shapeable | Where acceleration must be controlled | Input harmonics; motor suitability; cost |
| Autotransformer | Reduced by tap | Falls with the square of voltage | Larger motors with moderate torque demand | Fixed taps; transition step |
Sequence, protection and the rest of the site
Sequence is part of the design. Starting the largest motor first, on a bus that is otherwise lightly loaded, produces a smaller dip than starting it after the rest of the site has come on. The control system should allow only one large motor to accelerate at a time, and it should hold the next start until voltage and frequency have recovered. After a transfer, the sequence has to work with the ATS transfer sequence rather than against it: the set accepts the site load first, then the motors, with a deliberate pause between the steps.
Protection has to distinguish a start from a fault. The breaker’s instantaneous element must sit above the starting current, including asymmetry and the offset of the first cycles, or the breaker will trip on a healthy start. The thermal element must still protect the motor and the cable during a long acceleration or a stall. Where the set is paralleled, the load sharing and droop settings decide how the starting current is divided between sets, and a start on two sets is not simply half the dip on each unless the sharing has been set up and tested.
The UPS has a specific role that is often misunderstood. It covers the seconds between a mains interruption and the set carrying load; it is not a motor starter and it is not a cure for a starting dip. If the dip pulls the UPS input outside its window, the UPS transfers to battery during the start and the battery becomes part of the transient. The interaction is covered in generator and UPS interaction. The design should state what the UPS is expected to do during a start, and the commissioning test should confirm it rather than leave it to the first real event.
Settings have to be agreed before the test, not adjusted during it. The list includes the breaker curve, the motor protection relay, the under-voltage relay, the ATS timing, the drive ride-through and restart settings, and the generator’s own protection. Each has an owner and a recorded value, and each change after commissioning should be traceable. A start that works only because a protection is set outside its recommended range is not a design; it is a fault waiting for the next maintenance visit.
| Item | Setting question | Evidence |
|---|---|---|
| Start sequence | Largest motor first, one large start at a time | Sequence drawing and functional test |
| Breaker instantaneous element | Above starting current with margin | Coordination study and trip test |
| Motor protection | Thermal capacity and locked-rotor time respected | Relay settings and stall test |
| Under-voltage relay | Rides through the start, trips on a real fault | Setting record and injection test |
| ATS and drive restart | Start order and delay after transfer | Transfer drill record |
| Paralleled sets | Load sharing during the start | Load sharing test |
Commissioning and the annual check
Commissioning should measure the start rather than assume it. A power quality recorder at the motor terminals and at the main busbar captures the dip, the starting current, the acceleration time and the recovery. The measured values are compared with the study, and any difference is explained before handover. The test should be repeated with the site base load in place, because a start on an empty bus proves less than a start on the load the site will actually carry.
The annual load test is the natural home for a repeat of the start check. The practice on this site is a load test once a year at 50 to 100 per cent of rating, with a check of the light-load band at 10 to 20 per cent to confirm the set still runs cleanly when lightly loaded; the basis is in the maintenance schedule. A start test fits into the same visit: run the set at load, start the largest motor, record the dip, and confirm that protection did not operate. Records from the commissioning checklist give the baseline for comparison.
Between tests, monitoring earns its place. A controller that records each start event, with pre-start voltage, minimum voltage, recovery time and starting current, turns a complaint about flickering lights into a trend. The same data shows a motor whose acceleration is getting slower, a battery that is weaker, or a set that is losing transient capability as its excitation ages. The approach is described in remote monitoring for standby generators. A load bank can reproduce the base load for a start test when the site cannot be loaded, as described in load bank testing.
The annual check should also look at the things that change slowly: terminal torque, cable condition, the breaker’s trip record, the drive’s fault log and the condition of the starting equipment. A dip that was acceptable at handover can grow as contacts age, as a motor is replaced with a larger one, or as the base load grows. The review is cheap; the alternative is discovered during an outage, when the site needs the set to do the one thing it was bought for.
| Check | When | What to confirm |
|---|---|---|
| Voltage dip at motor terminals | At commissioning and annually | Inside the accepted limit |
| Voltage dip at main busbar | At commissioning and annually | Other loads ride through |
| Starting current and duration | At commissioning and annually | Matches the study, with no upward trend |
| Frequency dip and recovery | At commissioning and annually | Inside the load step limit |
| Protection operation | Every start test | No trip on a healthy start |
| Start event log | Continuous, reviewed annually | Trends visible and explained |
How much voltage dip is acceptable when a motor starts on a generator?
There is no single figure that suits every site. A planning value in the region of 15 per cent at the motor terminals is sometimes used in starting studies, but the governing limit is set by the project specification and by the tolerance of the connected equipment. Check the motor torque, the contactors, the drives and the UPS input separately, and confirm each threshold against the manufacturer’s data before the set is ordered.
Why does a generator dip more than the grid when the same motor starts?
Because the source impedance is different. The utility supply is stiff compared with the motor, while a generator presents the alternator’s subtransient reactance, which is not small. The current flowing through that reactance produces a voltage drop, and the excitation system then takes time to restore the voltage. A set with a low subtransient reactance and a fast excitation system holds the dip better than another set of the same kVA.
Can a soft starter or a drive let me use a smaller generator?
It can reduce the starting kVA the set must supply, which may allow a smaller set or a larger motor on the same set. It does not change the steady-state load, and it adds harmonics, ride-through settings and protection that have to be engineered. The saving has to be checked against the cost of the starter or drive, the input filter and the extra commissioning, using the manufacturer’s data for both the motor and the starting equipment.
Should the largest motor always start first?
On a bus that is otherwise lightly loaded, starting the largest motor first usually produces the smallest dip, because the base load is not yet on the set. The control sequence should allow only one large motor to accelerate at a time and should hold the next start until voltage and frequency have recovered. Where the process requires a different order, the study has to be run for that order rather than for the convenient one.
Why did the breaker trip on a healthy motor start?
The instantaneous element is probably set below the starting current, including the asymmetry and the offset of the first cycles. Check the coordination study and the breaker curve against the measured starting current before changing anything, because raising the setting to stop a nuisance trip can leave the cable and the motor under-protected. The measured start, not the calculated one, should be the basis for the final setting.
How often should a generator-fed motor start be tested?
Test it at commissioning and again with the annual load test at 50 to 100 per cent of rating, alongside the 10 to 20 per cent light-load check described in the maintenance schedule. Test sooner after a motor is replaced, after the base load grows, or after an unexplained trip. Record the dip, the starting current, the acceleration time and the recovery, and compare each test with the one before it.
Check the start before the set is sized
Send the motor datasheet, the driven load curve, the site base load and the proposed starting method. The review will set out the starting kVA, the dip to check at each busbar, the protection settings to agree and the records to take at handover.
KVA Grid sales engineering desk. Reused site figures: the annual load test at 50 % to 100 % of rating and the 10 % to 20 % light-load band from the maintenance schedule; 500 kVA at 101/80/60 L/h, 1000 kVA about 200 L/h and about 4800 L over 24 hours, and 0.24 to 0.29 L/kWh from fuel consumption; 40 °C ambient from hot and humid climates; the PUE 1.4 base-load example and the UPS seconds between mains loss and set load from generator and UPS interaction; load step acceptance from load step response testing; transfer sequencing from ATS transfer testing; load sharing from paralleling and droop protection; harmonics from harmonic mitigation; cooling and derating from radiator siting; start event records from remote monitoring; load bank work from load bank testing; handover records from the commissioning checklist. Starting current multiples, dip limits, alternator reactance values, load step limits, protection settings, thermal duty and test intervals are determined by the manufacturer’s data and the project specification. Reviewed 9 November 2026.
