Automatic transfer switch commissioning field guide

The automatic transfer switch, and what it does

An automatic transfer switch is the device that moves the load from the normal source to an emergency source and back, on its own, when the normal source fails. It senses loss of normal power, signals the engine to start, waits out its time delays, transfers the load to the generator once that source is good, and then on restoration waits, retransfers the load to normal, and runs the engine through a cooldown before it stops. Nobody has to be standing there for any of it.

The ATS is a separate listed device from the generator, with its own controller, its own timers, and its own failure modes. The generator can pass its own load test clean and the system still misses the transfer because a time delay was left at the factory default or a retransfer was never forced. The switch is where the sequence the building depends on actually lives.

Commissioning the ATS is proving that sequence under a real loss of power, not on the bench and not on paper. You pull the normal source, you time how long the load is dead, you force the retransfer, you operate the bypass, and you write down every time-delay setting as you leave it. The acceptance is the day you find out whether the switch transfers in time or just transfers.

Open transition or closed transition?

Open transition is break-before-make: the switch opens the load from the first source before it closes to the second, so the load sees a brief dead interval during the transfer. Closed transition is make-before-break: the switch closes to the second source while still connected to the first, overlaps the two for a fraction of a second, then opens the first, so the load sees no interruption at all. Most transfer switches in service are open transition.

Open transition is the default because it is simpler and it cannot accidentally tie the utility and the generator together. The dead interval is short, commonly well under a second, and most loads ride through it. The catch is that interval still exists, and on motor and UPS loads the momentary outage and the out-of-phase reconnection can cause trouble, which is what the in-phase monitor and the transition type are chosen to manage.

Closed transition exists to remove the outage entirely, which matters on critical and process loads that cannot take even a blink. The price is that for the brief overlap, commonly held under about 100 milliseconds, both sources are paralleled, so the generator has to be synchronized to the utility in voltage, frequency, and phase, and the closed-transition gear and the utility interconnection have to be set up to allow that parallel. Pick the transition type from the application and the spec, then commission what was actually installed, because the test for each is different.

Soft loading and the closed-transition ramp

Soft loading is closed transition that does not just overlap and switch, but ramps the load between sources while the two are paralleled, so the load shifts gradually instead of all at once. The generator picks up load on a controlled ramp and the utility sheds it on the same ramp, and the transfer back works the same way. The point is to avoid a sudden step of load on either source during the parallel.

This shows up where the utility or the design will not tolerate a hard step. A large block of load slammed onto the generator, or dumped back onto the utility, is a transient on both. Ramping it over a few seconds keeps the step inside what the machine and the interconnection allow, and on a peak-shaving or utility-parallel scheme it is what lets the plant run paralleled at all.

Soft loading needs more than a transfer switch. It needs the generator controls, the paralleling and protection, and usually a utility interconnection agreement that permits the parallel and sets how long it can last. Commissioning it is closer to commissioning a small paralleling scheme than a plain ATS, so confirm the parallel-time limit and the protection settings against the interconnection terms and the spec, not against a default.

What is an in-phase monitor?

An in-phase monitor is a control feature on an open-transition switch that holds the transfer until the two sources are close to synchronized, then transfers at the moment the phase difference is near zero. It is break-before-make like any open transition, but instead of transferring on a fixed timer, it waits for the windows where the normal and emergency sources line up in phase, so the load reconnects to a source that is electrically close to the one it just left.

The reason is motors. A running motor is still spinning and still generating a back-voltage when the switch opens. Slam it onto a second source that is far out of phase and the motor draws a huge inrush to pull back into step, which can trip the breaker and produce a torque shock that damages the shaft, the coupling, or the driven load. The in-phase monitor transfers at the moment that phase difference is small, so the inrush and the torque are kept down.

It has limits. In-phase transfer suits motor loads where a brief synchronized transfer is acceptable, and it has acceptance windows on phase angle and frequency difference that vary by manufacturer. Above that, or where no synchronized window is acceptable, the design moves to a timed open transition with a deliberate delay to let the motors slow down, or to closed transition. If the installed switch has an in-phase monitor, commission it by confirming it actually waits for synchronism instead of transferring on a fixed delay. That is the function, and it is the one that gets assumed rather than verified.

The sequence of operation

The sequence of operation is the script the ATS runs automatically on a loss of normal power, and commissioning is proving the switch follows it step by step under real timing. The steps are always the same in shape, and each one has a time delay that exists for a reason.

Normal source fails or drops below the dropout setting. The switch starts an engine-start time delay to ride through momentary dips, so a one-cycle blink does not start the plant. If the outage outlasts that delay, the switch sends the start signal to the generator. The engine cranks, starts, and comes up to voltage and frequency. Once the emergency source is within the pickup settings, the switch waits a transfer time delay to confirm the source is stable, then transfers the load to the generator.

On the way back the sequence reverses with its own delays. When the normal source returns and holds within pickup, the switch waits a retransfer time delay, commonly several minutes, before moving the load back, so it does not dump the load onto a utility that is still hunting. After the retransfer it runs an unloaded engine cooldown for a set period, then stops the engine. Walk every one of these in the test, both directions, and time them. A switch accepted on a one-way transfer with the retransfer never forced is a switch with half its sequence unproven.

The time delays and why each one exists

The time delays are the heart of the ATS settings, and the most common commissioning miss is leaving them at the factory default instead of setting them to the project sequence of operation. Each delay solves a specific problem, and setting one wrong either causes nuisance operation or blows the transfer-time requirement.

The engine-start delay rides through momentary sags so a brief blink does not crank the generator for nothing. The transfer delay lets the generator stabilize before it takes load. The retransfer delay keeps the load on the generator until the utility has proven it is back to stay, commonly a minimum in the range of several minutes, so the plant does not bounce back to an unstable source. The cooldown delay runs the engine unloaded after retransfer to let the turbo and the engine shed heat before shutdown. The exerciser clock starts the plant on a schedule so the set that has to run in an outage actually gets run.

Set them to the sequence, not the box. On a life-safety system the engine-start and transfer delays have to be short enough that the whole transfer still lands inside the required transfer time, while on a less critical load a longer start delay cuts nuisance starts. Record every delay as you leave it, because the next technician needs the baseline to know whether a setting drifted, and the owner needs it to defend the transfer time.

The controller and its sensing settings

The ATS controller is the brain that watches both sources and runs the sequence. It senses voltage and frequency on the normal and emergency sources, decides when each one is acceptable using pickup and dropout settings, runs the time delays, drives the transfer mechanism, and on a modern switch reports status and alarms to a building management system. Commissioning the controller is confirming those settings match the design, not the factory default.

The sensing settings are pickup and dropout. Dropout is the voltage or frequency at which the controller declares a source failed and starts the sequence. Pickup is the higher level at which it declares a source acceptable again, set above dropout on purpose so the switch does not chatter between the two on a marginal source. Typical controllers let you set undervoltage dropout as a percentage of nominal and pickup a few percent above it, with overvoltage, underfrequency, and sometimes phase-imbalance sensing layered on. The exact values come from the spec and the equipment, so set them to the design and confirm them, do not accept the default.

On a networked switch the comms are part of acceptance too. Force the conditions and confirm the controller reports the correct status and alarms to the BMS or the annunciator, because a point that reads right on the local controller but never reaches the monitoring system leaves the operator blind. That broken mapping is the classic miss, and it only surfaces when someone forces the alarm and watches both ends.

What is a bypass-isolation ATS?

A bypass-isolation transfer switch is an ATS built with a manual bypass section that keeps the load energized while the automatic switch is isolated for test, service, or replacement. You bypass the load onto one source through the bypass switch, then isolate the automatic switch from the line so it can be worked on or pulled entirely, all without dropping the load. It is the standard choice on data centers, hospitals, and other places where the load cannot be interrupted to maintain the switch.

The reason it exists is that a plain ATS forces a choice nobody wants. To maintain or replace the automatic switch you have to de-energize the load it feeds, which on a critical system means a planned outage of the thing that is not supposed to go out. The bypass section removes that choice by routing the load around the automatic switch first, so the switch can be isolated live and serviced while the load stays up.

Commission the bypass, do not assume it. The function only counts if it actually transfers the load to bypass and isolates the automatic switch without a blink, and the only way to know is to operate it during acceptance, both into bypass and back, while watching the load. A bypass that was never exercised at commissioning is a feature on the nameplate, and the first time it gets used in anger is the worst time to find out the sequence was wired wrong.

The codes and the application: NEC 700, 701, 702, 708

Which NEC article governs the transfer equipment is set by how the load is classified, and it drives the transfer-time requirement the commissioning test has to meet. Article 700 covers emergency systems, the life-safety loads where loss of power could cost a life, and emergency systems are commonly required to restore power within 10 seconds. Article 701 covers legally required standby, where the alternate source commonly has to come online within 60 seconds. Article 702 covers optional standby, where the time is set by the owner and the load, not a code limit. Article 708 covers critical operations power systems, COPS, for facilities whose loss would hit public welfare or national security.

The transfer equipment itself has to be automatic, and for emergency and legally required standby it has to be listed and marked for that use. The listing for the switch is UL 1008, the standard for transfer switch equipment, which sets the safety and performance requirements the switch is built and tested to, including its short-circuit ratings. A switch used on an emergency system that is not listed and marked for emergency use is a finding regardless of how well it transfers.

Name the article that actually applies to the load in front of you, because the transfer-time number rides on it, and confirm the exact section and the time against the adopted NEC edition and the AHJ. The 10-second emergency figure and the 60-second legally required figure are widely held, but the section numbers and the precise wording shift between code cycles and the jurisdiction can amend them. Cite the article by its scope, verify the number on the adopted edition, and let the project spec override when it is stricter.

3-pole or 4-pole? The switched neutral and the grounding

A 3-pole transfer switch switches the three phase conductors and leaves the neutral solidly connected through, common to both sources. A 4-pole switch switches the neutral along with the phases, so the neutral is broken and re-established with the rest of the load. The choice is not a preference. It decides how the generator is grounded, and getting it wrong leaves either a missing bond or multiple neutral-ground bonds that put current on the grounding system.

When the switch leaves the neutral solid, a 3-pole, the generator is a non-separately derived system: the neutral stays bonded to ground back at the service, and there is no neutral-ground bond at the generator. When the switch breaks the neutral, a 4-pole, the generator becomes a separately derived system, and it needs its own neutral-ground bond, a system bonding jumper, established at the generator under the grounding rules for separately derived systems. The relevant requirement, commonly cited at NEC 250.30(A), governs how that bond and the grounding electrode connection are made.

This is where ground-fault protection and grounding part ways from the wiring. Put a system bonding jumper at the generator on a 3-pole solid-neutral system and you create a parallel neutral path, with neutral current flowing on the grounding conductors and a ground-fault sensor that misreads it. Leave the bond off a 4-pole switched-neutral system and the generator source has no established neutral-ground reference. Confirm the pole configuration against the bonding scheme and the ground-fault protection during commissioning, and see the grounding and bonding guidance for how the separately-derived bond is made. This is one decision, not two, and the inspector checks it.

Commissioning the ATS: the loss-of-normal test

The core acceptance test is the loss-of-normal test: take away the normal source and watch the switch run the whole sequence on its own. Open the normal source ahead of the switch, start a clock, and time from the loss until the load terminals have acceptable power from the generator. That elapsed time is the transfer time, and it is the headline number the test exists to prove against the required limit for the system class.

Run it more than once, and run it the way the building will see it. Simulating the loss with a test switch on the controller proves the logic, but opening the actual normal source proves the sequence the way a real outage delivers it, including the engine start. Confirm the engine-start delay fires, the start signal reaches the generator, the engine comes up to voltage and frequency, the transfer delay times out, and the switch transfers. If the transfer lands outside the limit, time the pieces separately, the crank-to-running, the running-to-stable, and the ATS transfer delay, so you know whether the engine or the timers ate the seconds.

Test under load where the spec calls for it, not just dead. A switch that transfers fine with nothing connected can behave differently with the real load on it, and the block of load the generator has to accept in one step is part of what is being proven. Capture the transfer time, the source readings, and the conditions on every run, because a transfer time with no record of how it was measured is a story, not an acceptance.

Retransfer, cooldown, and the fail alarms

Half the sequence is the way back, and it is the half that gets skipped. After the transfer to generator, restore the normal source and prove the retransfer: the switch waits its retransfer delay, confirms the utility is holding, transfers the load back to normal, then runs the engine through its cooldown and stops it. Time the retransfer delay against the setting and confirm the cooldown runs. A switch accepted with the retransfer never forced has a timer nobody ever watched against a real source return.

Force the failure alarms, do not wait for them. The system has to annunciate when the engine fails to start and when the switch fails to transfer, because those are the two failures that leave the load dead in a real outage with no warning. Simulate a failure to start, within what you can safely simulate, and confirm the overcrank or fail-to-start alarm raises. Confirm a fail-to-transfer condition annunciates. On a switch with an in-phase monitor, confirm it transfers at synchronism rather than timing out and transferring out of phase.

Prove the not-in-auto signal hardest of all. A switch or generator left in off or manual after a service visit will not run on a real outage, and the not-in-auto alarm is the one thing that tells the operator the system is disarmed. If that point does not annunciate, the most common cause of a no-start in the field has no warning, and the operator finds out when the lights go out.

Testing the bypass without dropping the load

On a bypass-isolation switch the bypass operation is its own acceptance test, and it has to be proven with the load energized, because keeping the load up is the entire reason the bypass exists. Transfer the load to bypass, confirm the load never blinked, then isolate the automatic switch and confirm it is dead and safe to work on while the load stays fed. Then reverse it: bring the automatic switch back in, transfer off bypass, and confirm again that the load did not drop.

Walk the manufacturer's bypass sequence exactly. Bypass-isolation switches have a defined order of operations, and doing it out of order can interrupt the load or fail to isolate the switch, which defeats the point. Confirm the interlocks that enforce the sequence actually work, because the interlock is what stops a tired technician at 3 a.m. from dropping the critical load by throwing the handles in the wrong order.

This is the test schedule pressure quietly drops, because it takes a live load and careful steps and it feels redundant once the transfer test passed. It is not redundant. The transfer test proves the automatic switch. The bypass test proves you can maintain that switch for the next twenty years without an outage, which on a data center or a hospital is the whole reason a bypass switch was bought.

The integrated test: pulling the utility on the whole plant

The standalone ATS test proves the switch. The integrated systems test proves the switch as one link in the whole power chain, by pulling the utility on the complete plant at load and watching everything ride through together. The ATS is where the integrated test lives or dies, because it is the device that actually moves the load when the utility drops, and its timers set whether the transfer lands inside the window the sequence allows.

Bring the switch to the integrated test already proven on its own. Accept the ATS transfer and retransfer standalone, with its time delays recorded, then run it in the integrated test where it has to coordinate with the generator start, the UPS or BESS carrying the gap, the paralleling scheme, and the load-shed sequence. A failure there is then a coordination or timing fault between systems, not an unproven switch nobody timed, which is far faster to isolate.

The integrated test belongs to the broader power-QA scope of commissioning, and the generator acceptance and load bank work feed it from the source side. The handoff is clean when each piece arrives proven: the generator load-tested, the switch transfer-tested, the delays set to the sequence. Then the pull-the-utility test is checking the seams, which is exactly where the failures that take a building down actually live.

Is the switch rated for the load and the fault current?

A transfer switch has to be rated for two things: the continuous load current it carries, and the available fault current at its location. The fault-current rating is the withstand and closing rating, the WCR, which is the short-circuit current the switch can survive and, importantly, close into without coming apart. Confirm the WCR against the available fault current at the switch and against the upstream overcurrent device, because a switch under-rated for the fault current is a hazard the day a fault happens, not a paperwork item.

The closing part is what people miss. A transfer switch does not just have to withstand a fault, it has to be able to close on one, because the switch can transfer onto a source while a fault is still present on the load. If the switch closes into that fault, it has to hold together long enough for the upstream overcurrent device to clear it. UL 1008 sets these withstand and closing ratings, and they appear on the switch label, often tied to a specific upstream breaker or fuse and a time rating.

Read the rating the way it is qualified. A WCR is commonly stated with the specific protective device it was tested with, or as a short-time or time-based rating, and the rating only holds if the installed upstream protection matches the qualification. Confirm the available fault current from the study, confirm the switch WCR and how it was qualified, and confirm the upstream device matches. A switch whose WCR is below the available fault current, or qualified with a breaker that is not the one installed ahead of it, is not properly applied, and that is a finding.

Multiple transfer switches and load shed

On a plant with one generator and several transfer switches, the switches cannot all slam their load on at once, because the sum of the block loads can exceed what the engine can accept in a single step. The design handles this with sequencing and priority: the switches transfer in a set order, with the highest-priority load going first, so the generator picks up load in steps it can carry instead of one block that stalls it.

Load shed is the other half. When the generator is near capacity, the controls shed lower-priority loads to protect the higher-priority ones, and add them back as headroom returns. On a paralleling plant this lives in the master control rather than the individual switch, but each ATS still has to honor its place in the priority and shed order. Commissioning it means proving the order, not assuming it.

Test the sequence as a system, because the failure mode is a coordination fault that no single switch reveals. Each ATS can transfer fine on its own and the plant can still trip when all of them transfer onto one engine in the wrong order, or when the load-shed step sheds the wrong bus. That only shows up when you run the whole set against one generator under load, which is why multiple-ATS sequencing folds into the integrated test.

Maintaining it once it is in service

Commissioning proves the switch once. The reliability the building lives on comes from the testing program the owner inherits at turnover, and for emergency and standby systems NFPA 110 governs that ongoing testing. Name the line so it does not get confused. Acceptance proves the new switch before occupancy. The monthly routine keeps proving it for the life of the system.

NFPA 110 calls for the transfer switch to be operated monthly, commonly cited around 8.4.6, by electrically transferring from the normal position to the emergency position and back. The associated monthly engine exercise is run under load, commonly for at least 30 minutes, loaded enough to reach the manufacturer's recommended exhaust gas temperature rather than to a fixed percentage, because light loading lets a diesel wet stack. Where opening the live utility is unsafe, the standard allows a load bank to stand in, with the requirement that building load automatically replaces the bank if the normal source actually fails during the test.

Hand the owner the schedule, not just the switch. Set up the monthly transfer test, confirm the exerciser clock is set, and record the as-left time delays so the maintenance crew has the baseline. Confirm the durations and the load requirement against the adopted NFPA 110 edition and the AHJ, because the numbers have shifted between editions. A switch accepted clean and then never transferred under load again is a switch waiting to fail on the one outage it was bought for.

What to document

The commissioning record is what a future operator trusts when the question is whether this switch was ever actually proven. Capture enough that a reviewer who was not there can reconstruct the test and check each result against the spec, the NEC article, and the adopted NFPA 110 edition.

Record the switch identification and rating, the pole configuration, the transition type, the measured transfer and retransfer times against the required limit, every time-delay setting as left, the pickup and dropout settings, whether the bypass was operated, the WCR and the upstream device it is coordinated with, the alarm and annunciator point checks, and who witnessed it. If anything was adjusted and re-tested, record the before and after, because the next person needs to see what changed and that it was proven after the change.

Common mistakes

• Leaving the time delays at the factory default instead of setting them to the project sequence of operation.
• Using an open-transition switch on motor loads with no in-phase monitor, so motors take an out-of-phase reconnection.
• Specifying the wrong neutral switching, so a 4-pole switch has no system bonding jumper or a 3-pole switch has one.
• Accepting a transfer time that lands outside the required limit for the system class instead of trimming the delays.
• Testing the transfer one way and never forcing the retransfer or watching the cooldown.
• Skipping the bypass operation test on a bypass-isolation switch because it takes a live load and careful steps.
• Applying a switch whose withstand-and-close rating is below the available fault current, or qualified with the wrong upstream device.
• Confirming an alarm on the local controller and never proving it reached the annunciator or the BMS.
• Confusing the NFPA 110 monthly transfer test with the one-time commissioning acceptance test.

Field checklist

Standards and references

Several bodies govern different parts of an ATS commissioning, and naming the right one for the point is the difference between a credible record and a guess. The NEC, NFPA 70, sets which rules apply by the system class: Article 700 for emergency systems, Article 701 for legally required standby, Article 702 for optional standby, and Article 708 for critical operations power systems. The transfer-time expectations, commonly around 10 seconds for emergency and 60 seconds for legally required standby, ride on those articles, and the exact section numbers and times shift between code cycles, so confirm them against the adopted edition and the AHJ.

The switch itself is listed to UL 1008, the standard for transfer switch equipment, which sets the safety and performance requirements including the withstand and closing ratings that appear on the label. The grounding and neutral switching follow the NEC grounding rules, with the separately-derived-system bond on a switched-neutral generator commonly cited at NEC 250.30(A). NFPA 110, the standard for emergency and standby power systems, classifies the system by Type, Class, and Level and governs both the installation acceptance test and the ongoing monthly transfer test the owner inherits, with the monthly switch operation commonly cited around 8.4.6.

For field acceptance testing of the electrical gear, ANSI/NETA ATS gives the inspection and test requirements before energization. Above all of these sit the manufacturer's instructions and the project specification, which set the actual numbers: the pickup and dropout settings, the time delays, the transition type, and the WCR coordination. When a standard and the spec disagree, the stricter controlling document wins, and the authority having jurisdiction has the final say on what is enforceable.

Units and terms

The transfer-switch vocabulary reads differently across a spec, a submittal, and a controller manual, and the same idea carries a few names. Transfer time is in seconds, the count from loss of normal to acceptable power at the load. The transition type is the switching action, break-before-make or make-before-break. The WCR is the short-circuit rating. The terms below are the ones that travel across the whole acceptance.

Pickup and dropout are the source-sensing thresholds, set as percentages of nominal voltage and as frequency limits. Poles count the switched conductors, three for a solid neutral and four for a switched neutral. The articles, 700, 701, 702, and 708, classify the system and set the transfer-time expectation. Read the controller manual for how that switch names its delays, because the labels differ by manufacturer even when the function is the same.

ATS

Automatic transfer switch, the device that moves the load between the normal and emergency sources

Open transition

Break-before-make transfer, with a brief dead interval as the load moves between sources

Closed transition

Make-before-break transfer, with a brief source overlap and no interruption to the load

In-phase monitor

Open-transition control that transfers at the moment the two sources are near synchronized, to protect motors

3-pole / 4-pole

Solid neutral switched through, or neutral switched with the phases, which sets the generator grounding

Bypass-isolation

An ATS with a manual bypass that keeps the load up while the automatic switch is isolated for service

WCR

Withstand and closing rating, the short-circuit current the switch can survive and close into, per UL 1008

Pickup / dropout

The voltage and frequency thresholds at which the controller accepts or rejects a source

NEC 700 / 701 / 702

Emergency, legally required standby, and optional standby system articles that set the transfer-time class

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is an NFPA 110 acceptance test?
• How long does data center electrical commissioning take?
• What size standby generator do I need?
• How long must emergency lights stay on?
• What is the difference between emergency and standby power?