What a Standby Supply Changes About the System

Why this matters

Equipment that works all year fails during the monthly generator test, and the ticket says the generator broke it. Sometimes that is true and sometimes the generator is just the only condition under which an existing weakness becomes visible. Telling those apart is worth doing properly, because one of them is a design conversation with the set supplier and the other is a repair on the machine, and guessing wrong costs a mobilisation either way.

Before any of it: a circuit fed through an automatic transfer switch can re-energize without warning, from a source that starts itself when the utility drops. Isolation has to include the standby source and any means of starting it, not just the utility breaker. Panel and branch-circuit lockout runs under 29 CFR 1910.333(b)(2) in general industry or 29 CFR 1926.417 on construction work, since 29 CFR 1910.147 excludes exposure to electrical hazards from work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C), and every isolation gets proved dead live-dead-live per NFPA 70E-2021 120.5 as adopted by your employer's electrical safety program. A circuit proved dead with only the utility open is not proved dead.

One physical fact drives everything below: the source is soft

A utility service behaves as a very stiff source. Its impedance back to the system is small, so it holds voltage under an abruptly applied load and it can deliver enormous fault current.

An engine-driven alternator is a soft source. Its internal impedance is a large fraction of the load impedance rather than a small one, so applying load pulls its terminal voltage down, and the fault current it can deliver is bounded first by its own reactance and then by what its excitation system can sustain. Sustained short-circuit contribution from a typical set is a small multiple of its rated current rather than the tens of multiples a utility transformer delivers, and on sets whose excitation collapses under a fault the sustained value can fall much further. The authority for any specific set is the manufacturer's decrement curve, which is a document you can ask for by name and which most suppliers will provide.

Three consequences follow, and every one of them comes out of that single fact.

Consequence one: protective devices clear differently, or not at all

A fault that puts a branch device deep into its instantaneous region on utility power may sit below that pickup on the standby source. The instantaneous pickup of a standard molded-case breaker lands in the region of ten times rating with the product curve as the authority, and if the alternator cannot deliver ten times that device's rating, the fault does not clear in a fraction of a cycle. It clears on the thermal element in seconds, or the alternator's own protection acts first, or nothing clears at all and the fault burns.

That is why selective coordination requirements for emergency systems, legally required standby systems and critical operations power systems, in NEC Articles 700, 701 and 708 of the edition your authority having jurisdiction has adopted, are satisfied against the sources the system can actually run on, not against the utility alone. The sibling article on selective coordination covers how a pair is evaluated at a stated fault current; the point here is that a standby supply gives you a second, much lower current at which the same pair must be checked.

The reverse does not follow. Lower fault current on the generator does not relax the interrupting-rating question, because a device has to be adequate for the highest available fault current from any source that can feed it, and on a closed-transition transfer where two sources are briefly paralleled, the contributions add. The article on interrupting ratings owns that.

Consequence two: motor starting is where it bites first

Across-the-line starting on a standard induction motor typically draws 5 to 8 times full-load current for a few seconds, with the nameplate code letter as the authority for a specific machine. On a stiff source that surge produces a small voltage dip. On a soft source it produces a large one.

The reason it matters is the torque relationship: for an induction motor at a given slip and constant frequency, developed torque varies roughly with the square of applied voltage. Drop the terminal voltage to 80 percent and you have roughly 64 percent of the torque you had. If the driven load needs more than what is left, the motor does not accelerate; it sits at partial speed drawing near locked-rotor current until something opens.

Two other things fail in that same window. Control relays and contactors drop out somewhere below their rated voltage, in a band that is a property of the specific device and belongs on its datasheet rather than in a remembered number, and a dropout mid-start shuts the equipment down in a way that looks like the generator rejecting it. And engine speed sags on load application, so on a soft source voltage and frequency move together, which is not something that happens on utility power in any practical sense.

Consequence three: nonlinear load distorts the voltage, not just the current

Higher source impedance means the distorted current drawn by rectifier-front-end equipment - drives, uninterruptible supplies, switching supplies, LED drivers - produces a proportionally larger distortion in the voltage waveform at the terminals. On utility power the same load distorts the voltage far less because the source is stiff.

The practical result is that a load mix which is unremarkable on the utility can produce voltage distortion on the standby source that upsets other equipment sharing the bus. Sizing a set for a heavily nonlinear load is a calculation the manufacturer's sizing tools do, taking the load mix as an input; it is not a matter of matching kilowatts.

The gate: is this a property of the equipment or of the source it is running on?

Run it as one question, at one moment, on one piece of equipment. The measurement that answers it is the terminal voltage at the equipment during the event, on each source, because that is the quantity the source impedance actually changes.

Case one: the gate answers source

A packaged air handler that runs all year. On every monthly generator test its motor overload opens within a minute of start.

Live readings during a load-acceptance test fall under 29 CFR 1910.333(a)(1), which requires de-energizing before working on or near exposed energized parts unless the employer can demonstrate that de-energizing introduces additional or increased hazards or is infeasible; a voltage measurement during a start does not exist de-energized, and the demonstration belongs to the employer's program. Where it proceeds, boundary and arc-rated PPE come from the risk assessment at NFPA 70E-2021 130.5 and 130.7 in the adopted edition, and note that the arc-flash parameters on the standby source differ from the utility values on the label, so an unlabelled generator-source condition is a reason to get the study before opening the door rather than a reason to proceed carefully.

The engine running is its own hazard, on a different route. Exhaust carries carbon monoxide, which is an inhalation hazard that no glove or face shield addresses: work in the generator room only with the exhaust system confirmed intact and the room's mechanical ventilation running, wear a personal CO monitor set to alarm, and leave immediately if it alarms rather than finishing the reading. Keep clear of hot exhaust surfaces and the rotating assembly, and treat the set as capable of starting at any time.

The numbers. Motor nameplate full-load current 40.0 A. At 6 times full-load current, a start draws roughly 240 A, which against a set rated 100 A is about 2.4 times the alternator's entire rating in one block.

Measured line-to-line voltage at the equipment during the start, as a percentage of nominal:

• On utility: dips to 96 percent, recovers within the start.
• On the generator: dips to 78 percent, holds there for about 1.5 seconds.

Apply the squared torque relationship stated above, at constant frequency and a given slip:

• On utility, available starting torque is roughly 0.96 squared, or 92 percent of the full-voltage value.
• On the generator, roughly 0.78 squared, or 61 percent. Read that as an estimate rather than a figure: the arithmetic holds frequency constant and on a soft source it is not. An engine that sags in speed as it takes the block load moves volts and hertz together, so capture frequency alongside voltage during the start. Where frequency falls in step with voltage the machine holds more torque than 0.78 squared predicts; where volts fall and hertz do not, it holds less. The recorded pair is what the set supplier sizes against, and a voltage number alone forces them to assume.
• The generator start therefore gives the motor about 0.61 divided by 0.92, or roughly two-thirds of the starting torque it has on utility.

That is enough to explain everything on the ticket. The motor does not accelerate cleanly, it sits near locked-rotor current longer than the overload element tolerates, and the overload opens. The overload is correct and the motor is healthy. Nothing on this air handler will fix it.

The fixes are all on the source side or the starting side: a reduced-voltage or soft starting method so the block load is smaller, load sequencing so this motor starts before other loads are picked up, or a larger set. All three are design decisions for the generator supplier with the load list in hand, and the numbers above are exactly what they need from you.

Case two: the same gate answers equipment

Same complaint shape at a different site: a pump that only ever fails during the monthly test.

Measured terminal voltage during the start: 95 percent on utility and 94 percent on the generator. Effectively no difference, which means the source is stiff enough for this load and the squared torque relationship gives essentially the same available torque on both. The gate answers equipment, and it answers it in one measurement.

The finding, once the source is excluded: the pump fails on utility too. Nobody had noticed because the only time anyone stands there watching it is during the test. The generator was a scheduling coincidence, not a cause.

This is the outcome the gate exists to produce, and it is the one that gets skipped, because a generator in the story is a satisfying explanation that nobody feels the need to test. The cost of skipping it is a set supplier being asked to quote a larger machine to solve a failed bearing.

Where the standby transition itself belongs

Everything above concerns steady operation on the alternate source. The transfer event has its own distinct failure population - open-transition ride-through, control power dropout during the gap, equipment restart sequencing, motors reconnected out of phase - and the sibling article on the backup power transition as its own fault source covers it. Keep the two separate on a ticket, because they are separated by a single question: did the fault happen during the changeover, or during the run afterwards?

How to verify you got this right

• Measure on both sources, in the same conditions, or you have not run the gate. One reading on the generator proves nothing without its utility counterpart, because the whole method is a comparison.
• State the dip as a percentage of nominal, and convert it to a torque ratio before drawing a conclusion. A voltage number alone invites the wrong argument; the squared relationship is what makes a modest-sounding dip into a large torque loss.
• Ask for the decrement curve and the set's transient dip data before recommending anything. If you are proposing a change to the source or the starting method, those documents are what the supplier will size against, and the request itself tells you whether anyone has ever done the calculation.
• Check whether the arc-flash labelling covers the standby source. A label derived only from the utility condition is incomplete on a system with two sources, and that gap belongs in writing to the owner rather than in a technician's judgment on the day.

References

• NEC Articles 700, 701, 708 and 445, in the edition your authority having jurisdiction has adopted
• The generator manufacturer's decrement curve, transient voltage dip data and load-acceptance sizing documentation for the specific set
• 29 CFR 1910.333(a)(1) and (b)(2); 29 CFR 1910.147(a)(1)(ii)(C); 29 CFR 1926.417 for construction work
• NFPA 70E-2021, 120.5, 130.5 and 130.7, as adopted by your employer's electrical safety program
• See related: The Generator or Backup Power Transition as Its Own Fault Source; What an Interrupting Rating Is and Why It Is Not a Trip Setting; What Selective Coordination Means on a Service Call