Automatic Transfer Switch Commissioning and Test
Purpose
A transfer switch that has been wired and switched by hand has not been commissioned. Its settings are the product: the sensing thresholds decide what the switch calls an outage, the time delays decide whether the set starts on every momentary blip or rides it out, and the retransfer delay decides whether a bouncing utility drops the house four times in ten minutes. All of that ships on factory defaults chosen for a catalog, not for this service.
Two things go wrong when the settings are never touched or never written down. The customer gets a generator that fires up for a two-second flicker until they stop trusting it, and the next technician arrives with no idea what the numbers were meant to be, so a controller that quietly lost a setting looks exactly like one that never had it.
Scope
Covers commissioning and functional testing of an automatic transfer switch, 600 V or less, serving an optional standby system under NEC Article 702 in the edition your authority having jurisdiction has adopted, on a dwelling or small commercial building.
Does not cover the generator set's own siting, fuel, bonding and first start, owned by the standby generator installation SOP; the choice of which circuits sit downstream, owned by the critical load panel SOP; manual transfer switches and breaker interlocks, owned by the portable generator inlet and interlock SOP; or emergency systems under NEC Article 700, where the transfer time itself is a code requirement rather than a preference.
Roles and responsibilities
| Role | Owns | Hands off |
|---|---|---|
| Office | The outage window and who in the building must be told | Names the on-site authority to the lead in writing |
| Lead electrician | The as-found reading, every setting entered, the simulated outage and the fail-to-start test | Hands the settings sheet and any deficiency to the office the same day |
| Customer | Confirms nothing critical rides through the window | Signs the settings sheet, which states what the delays mean in plain language |
Procedure
Confirm this is the right switch for this service before any setting is touched. Read the nameplate against the installation: ampere rating at or above the overcurrent device ahead of it, poles matching the bonding decision made for the generator, service-entrance rating present if the switch is the service disconnect. Acceptance: rating, pole count, whether the neutral is switched and the enclosure type recorded and matching the plan. Wrong looks like an indoor-rated enclosure on an exterior wall; stop rule, a mismatch is a scope change settled before energizing. Hazard: reading a nameplate is not panel entry, but if a cover must come off to reach it, step 2 applies first.
Verify the installation cold, before the first energization. Check both source terminations and the load terminations for torque, check control conductors against the wiring diagram terminal for terminal, and confirm the engine start contact lands on the set's start terminals and not a run terminal. Acceptance: every lug torqued to its marked value, control conductors verified point to point, NEC 110.26 working space clear at 3 ft by 30 in by 6.5 ft. Wrong looks like a start pair on the wrong terminal, which produces a switch that transfers to a source that never started; stop rule, any conductor you cannot trace on the diagram stops the energization. Hazard: open both sources, lock and tag them, and prove every conductor dead on a known live source immediately before and after, per NFPA 70E-2021, 120.5, with work practices at 29 CFR 1910.333(b)(2), which governs because 1910.147 excludes electrical utilization work at (a)(1)(ii)(C). A transfer switch has two live sides and killing one proves nothing about the other.
Read and record every as-found setting before you change one. Walk the controller's menu or dip switches and write down what is there now, including the exercise schedule and whether it runs with or without load. Acceptance: a complete as-found column, with any setting the controller does not display marked not readable rather than left blank. Wrong looks like a technician who changed four settings and can name two; stop rule, a controller that will not display its settings gets model and firmware recorded and the manufacturer called before anything is entered blind. Hazard: none if the deadfront stays on, and it should, because these settings are reachable from the front.
Set the voltage and frequency sensing so the switch calls an outage the way this service actually fails. Dropout is where the switch stops believing the normal source, pickup is where it believes it again, and pickup sits above dropout so a sagging utility cannot chatter the switch. Acceptance: dropout and pickup entered as a percentage of nominal and also written out in volts for this service, separated by at least the manufacturer's stated minimum. Wrong looks like a dropout so close to nominal that a motor-start sag on the street fires the generator; stop rule, a service with a history of low voltage gets the utility called before the sensing is widened to hide it, because a switch tuned to tolerate a real supply problem tolerates it during the emergency too. Hazard: none at the keypad, but these numbers are what will move a live load later, so they are re-read after entry rather than assumed.
Set the four time delays and the exercise schedule, and write what each one buys. Engine start delay rides through momentary blips, transfer delay lets the set stabilize, retransfer delay keeps the load on the generator until the utility has proven steady, cool-down runs the engine unloaded before shutdown. Acceptance: all four entered, each written on the sheet in seconds or minutes with one plain sentence saying what it protects, plus exercise day, time, duration and the with-load or without-load choice. Wrong looks like a zero cool-down, which shuts a hot engine down straight off load; stop rule, a delay the controller will not accept is a model limit to record, not a number to fudge. Hazard: none at the keypad; the consequence lands at step 6, which is why the numbers get read back first.
Run the simulated outage with a stopwatch and time the sequence against the settings you just entered. Open the normal source at the switch's own means or at the service disconnect and record each event: start signal, load pickup, then on restoration the retransfer and the engine stop. Acceptance: each measured interval within the manufacturer's tolerance of the step 5 setting, load confirmed energized on the alternate source with a meter rather than a lamp, and the mechanical interlock confirmed to prevent both sources closing. Wrong looks like a switch that transfers correctly and retransfers in seconds, meaning a delay is not doing what the sheet says; stop rule, any interval outside tolerance sends you back to step 3 to re-read the setting, and the switch is not handed over until the test re-runs clean. Hazard: this step puts a live load on a running engine with the customer in the building, so nobody is inside the switch or a panel while it cycles, deadfronts are secured first, everyone stands clear of the set, and the interlock is checked through its window or with the switch dead, never by forcing it.
Prove the failure path, then prove it is put back. Disable the start signal at the manufacturer's provided test means so the set cannot start, and run one more simulated outage. Acceptance: the controller annunciates fail-to-start after its stated crank cycles, the load stays on the normal source, the alarm is visible where the customer will see it; then the disable comes out and a final successful transfer is run and recorded. Wrong looks like a switch sitting silent on a set that never started; stop rule, no fail-to-start annunciation means the alarm circuit is a deficiency with an owner on the record. Hazard: never simulate a failure by defeating an overcurrent device, jumpering an interlock or disabling a protective function, and never leave the site with the disable in place. The final successful transfer is the verification that the starting function you interfered with still works.
The record this produces
One settings sheet per switch, filed with the job and left in a sleeve at the switch.
Header: date, technician, switch make, model, serial, ampere rating, poles, whether the neutral is switched, service-entrance rated yes or no, controller firmware if displayed, adopted NEC edition. Body: a three-column table of setting, as-found, as-left, covering dropout, pickup, engine start delay, transfer delay, retransfer delay, cool-down, exercise day, time, duration and load choice. Footer: the step 6 measured intervals beside the as-left values, the step 7 fail-to-start result, deficiencies with owners, the customer's signature.
The next technician reads the as-left column to see whether a setting has drifted or been changed by someone else, which is the only way to tell a controller fault from a service call somebody already made.
Worked pass: 200 A service-entrance rated switch, residential
A 200 A service-entrance rated automatic transfer switch, 240/120 V, feeding the whole house, with a 22 kW set ahead of it.
Step 1: nameplate reads 200 A, service-entrance rated, three-pole with the neutral switched, outdoor-rated enclosure, matching the plan and matching the generator bonding decision.
Step 3, as-found: dropout 80 percent, pickup 90 percent, engine start delay 3 seconds, transfer delay 5 seconds, retransfer delay displayed as TEST, cool-down 5 minutes, exercise Saturday 07:00 for 20 minutes without load.
Step 4: nominal is 240 V, so dropout at 85 percent is 0.85 times 240, or 204 V, and pickup at 90 percent is 0.90 times 240, or 216 V, leaving 216 minus 204, or 12 V of separation, above this controller's stated minimum. Dropout is raised from 80 to 85 percent because the house sits at the end of a long rural lateral and the owner reports the lights browning out before the power goes; 80 percent would have been 192 V, which this service reaches on a bad afternoon without ever failing.
Step 5: engine start delay 3 seconds, transfer delay 5 seconds, retransfer delay 10 minutes, cool-down 5 minutes, exercise moved to Wednesday 10:00 for 20 minutes without load at the customer's request.
Step 6 fails. Utility opened at 0 seconds: start signal at 3 seconds, load picked up at 19 seconds, both inside tolerance. Utility restored: the switch retransferred at 9 seconds against the 10 minutes entered at step 5. Re-reading the controller shows the retransfer delay reverted to the TEST value in the as-found column, because it was entered while the controller sat in test mode and was never committed. Stop rule fires. Settings are re-entered from the normal menu and the test re-runs: retransfer at 10 minutes 4 seconds, engine stopped 5 minutes 2 seconds later at 15 minutes 6 seconds.
Step 7: the start signal is lifted at the controller's test terminal. The set cranks its stated three cycles, does not start, and the controller annunciates fail-to-start with the load still on utility. The signal is landed again and a final simulated outage transfers and retransfers clean, which puts the starting function back before anyone leaves.
The customer signs the sheet with the sentence that matters to them on it: the generator waits three seconds before it decides an outage is real, and ten minutes after the power returns before it hands the house back.
References
- NEC (NFPA 70) Article 702 for optional standby systems, including 702.5 transfer equipment, and 110.26 for working space; confirm the adopted edition
- NEC (NFPA 70) Article 700 where the system is an emergency system, which carries transfer-time requirements this procedure does not address
- 29 CFR 1910.333(b)(2) for de-energizing and verifying both sources at step 2, with NFPA 70E-2021, 120.5 for live-dead-live
- Manufacturer's controller manual for the switch: setting ranges, minimum pickup-to-dropout separation, crank cycle count, and tolerance on each timer
- See related: the standby generator installation and commissioning SOP and the critical load panel design SOP