What a Transfer Switch Decides
Why this matters
During an outage, "the power is off" stops being a fact about the building and becomes a fact about a specific circuit. A transfer switch draws that line. Equipment on one side of it is dead until the utility comes back; equipment on the other side goes live when an engine starts, with nobody in the room and no warning. Techs get hurt on that boundary, and shops lose jobs on it too, because the customer who says "my freezer has backup power" is usually describing the building, not the panel the freezer is fed from.
The switch is a boundary, not a box
A transfer switch is the point where a load stops being fed by one source and starts being fed by another. Everything upstream of it belongs to the normal supply. Everything downstream of it belongs to whichever source the switch has selected at that instant. That is the whole idea, and it is the part that gets lost when the conversation turns into a catalog of switch types (a sibling reference covers the types, ratings and the mechanics of manual, automatic and service-entrance-rated units; this card is about what the switch decides for you).
Here is one common arrangement, the split-load case. It is not the only one - a service-entrance-rated switch can sit ahead of everything, so the whole building is downstream - but this one is where the boundary confuses people most:
utility service
|
meter
|
service disconnect
|
+--------------+
| |
normal panel transfer switch <--- generator
(dead through |
the outage) emergency panel
(live in an outage,
re-energizes with
nobody present)
The picture is worth carrying because the two panels can sit side by side on the same wall, fed from the same disconnect, and look identical with the covers on.
Four decisions the switch makes
One: coverage. Which loads are downstream of it at all. This is decided at install time, usually by the size of the generator and the rating of the switch, and it is almost never revisited. A customer who paid for backup power in one decade and added equipment in the next has no reason to know that the new equipment landed on the normal side.
Two: which conductors it opens, and specifically whether the neutral is switched. A switch that opens only the ungrounded conductors leaves the grounded (neutral) conductor solidly connected through to the normal supply. A switch that also opens the neutral separates the two systems completely. That choice decides whether the generator is a separately derived system, which in turn decides where the system bonding jumper and the grounding electrode connection belong. NEC Article 250 governs that, in the edition your authority having jurisdiction has adopted, and getting it wrong produces a system that works perfectly on utility power and carries current on grounding conductors on generator power. Grounding and bonding are covered in depth elsewhere in this library; the point here is only that the transfer switch is where the decision is made.
Three: how it transitions. An open-transition switch breaks the first source before it makes the second, so the load sees a real interruption, typically long enough to drop a contactor and short enough that a control board may or may not reset. A closed-transition switch briefly parallels the two sources, which the utility has to permit and which is a different animal entirely. A delayed-transition switch parks in the middle for a set interval, deliberately, so motor loads spin down before they are re-energized out of phase with what they were doing.
Four: what triggers it and what returns it. Sensing thresholds, transfer delays, retransfer delays and the scheduled exercise cycle. This is the decision with your name on it, because it says when a circuit you are standing in front of becomes live.
The decision that reaches the technician
Decision four is the one that gets a person hurt, and decision one is the one that gets a shop blamed.
A load downstream of an automatic transfer switch has two independent sources, and opening the utility disconnect removes one of them. That is not an electrically safe work condition, and it is exactly the sequence that reads as safe: the lights go out, the panel meters dead, and forty seconds later the engine starts because the switch saw the same loss of utility you just created. If the unit is set to exercise weekly, it will also do this on a schedule that nobody at the site remembers.
So the gate comes first, and it is not optional. Under 29 CFR 1910.333(a)(1) the conductors and parts you work on are de-energized before you work on or near them unless your employer can demonstrate that de-energizing introduces additional or increased hazards or is infeasible - and "there is a generator" is not that demonstration, it is the reason to isolate both sources. For electrical work, the lockout and tagging requirement is 29 CFR 1910.333(b)(2); 29 CFR 1910.147 does not govern here, because it excludes work on electric utilization installations at 1910.147(a)(1)(ii)(C), and the construction counterpart for electrical work is 29 CFR 1926.417. Prove dead with the live-dead-live sequence at NFPA 70E-2021, 120.5 - test the meter on a known live source, test the conductors, test the known source again - using NFPA 70E-2021, 130.5 and 130.7 as the basis for boundaries and PPE, in the edition your employer's electrical safety program adopts, because NFPA 70E binds you through that program and through your contract, not on its own.
Isolating both sources means the generator too: put its starting circuit out of service by the manufacturer's stated method, lock the transfer switch in a position, and confirm at the emergency panel with the meter rather than at the switch nameplate. A transfer switch that has failed mechanically can be showing you one position and holding another.
Worked case: the freezer that had backup power and did not run
A restaurant loses utility power for most of a day. The lights, registers, walk-in cooler and hood controls all ran. The walk-in freezer's condensing unit did not, and the owner calls to have the freezer looked at because he believes it failed during the outage.
The building has 40 branch circuits across two panels. Twelve of them are on the emergency panel, and the freezer condensing unit is not one of them. The transfer switch is rated 100 A against a 200 A service, which is why coverage was drawn where it was: the emergency panel was sized to the switch and the generator, not to the building. The condensing unit ran through the outage the same way it would have run through any other loss of supply, which is to say it did not run at all, and it came back when the utility did.
Nothing is wrong with the freezer. The diagnosis takes about the same effort as the walk to the panel, and the whole job is establishing which side of the boundary the load sits on - which is why it is worth doing before pulling a single cover.
The follow-on question is the real one: can the condensing unit be moved to the emergency side? That is not answered by the building's load or by the generator nameplate alone. It is answered on the emergency side of the boundary, against the switch rating and against what the generator can accept as a starting step. Twelve circuits carrying lighting, registers and control loads present a very different starting demand than twelve circuits plus a compressor, because a compressor's starting current is a multiple of its running current, and the multiple depends on the motor and its starting method rather than on any number I can give you here. Route that to the generator manufacturer's step-load data and the compressor's nameplate. What you can tell the owner today, with confidence: the freezer is fine, the coverage decision was made at install, and changing it is a design change to the emergency side, not a repair.
What changes the answer
Three conditions genuinely invert the reasoning above.
If the installation is a legally required or emergency system rather than an optional standby one, the coverage decision is not the owner's preference at all. NEC Articles 700 and 701 govern those, in the adopted edition, and what is on the emergency side is set by the occupancy and the authority having jurisdiction. Optional standby under NEC Article 702 is where owner preference actually lives.
If an inverter or battery system is doing the transfer instead of an engine, decision four changes character: there is no start delay, so the boundary can re-energize in a fraction of a second rather than after a crank cycle, and the absence of engine noise removes the one cue that used to tell people the building was on backup.
If the site has an interconnected source that exports or runs in parallel, NEC Article 705 is in play and the switch is only part of the story. Treat any parallel arrangement as outside the scope of a service call until the interconnection documents are in your hand.
How to verify you got this right
Do this while the site is on normal power and the work is de-energized, so the check itself does not become the hazard.
- Open the transfer switch's own documentation or its nameplate and record the pole configuration. Three poles on a single-phase 120/240 V system means the neutral is switched; two poles means it is not. That one field predicts most of the grounding questions that follow.
- Photograph the directory of the emergency panel and count the circuits against the building's total. Coverage is a number, not an adjective, and the customer should have it in writing.
- Find the exercise schedule in the controller and write down the day and time. Then say it out loud to whoever is responsible for the site, because that is the moment the boundary becomes live without an outage.
- Confirm the switch position sensing agrees with reality: with the site on normal power, the controller should report normal, and a meter at the emergency panel should agree with the controller rather than merely be plausible.
If any of those four cannot be answered from what is at the site, that is the finding, and it goes on the ticket. A backup system whose coverage nobody can state is a system nobody can be relied on to work around.
References
- 29 CFR 1910.333(a)(1) and 1910.333(b)(2), general industry electrical work practices and lockout for electrical work; 29 CFR 1926.417 for the construction counterpart
- 29 CFR 1910.147(a)(1)(ii)(C), the carve-out that sends electric utilization work to Subpart S
- NFPA 70E-2021, 120.5 (verification of an electrically safe work condition), 130.5 and 130.7 (risk assessment and PPE), as adopted by your employer's electrical safety program
- NFPA 70 (National Electrical Code), Articles 250, 700, 701, 702 and 705, in the edition adopted by your authority having jurisdiction
- See related: Generator Transfer Switch Types Reference; The Generator or Backup Power Transition as Its Own Fault Source