Generator Runs but Transfer Switch Does Not Transfer, Response

Purpose

This procedure guarantees that the callback where the generator starts and runs but the house stays dark ends with a verified, complete transfer cycle in both directions, confirmed by the switch's own control logic, not by a swapped part that seemed to work on one manual test. A switch that half-recovers, transfers to generator but will not retransfer, sends the crew back out for a second emergency visit mid-outage. Worse, a switch mis-diagnosed or left in a state where both sources can close onto the same bus is a backfeed event that can injure a utility lineman told the line was de-energized.

Scope

Covers on-site diagnosis, repair, and verification for a confirmed "generator runs, transfer switch does not transfer" callback, on residential and light commercial automatic transfer switches, contactor-based or relay-controlled. Does not cover the component-level fault-finding logic; the ATS troubleshooting tree and the utility-loss-but-no-transfer decision tree own identifying which part failed, and this procedure assumes that diagnosis has run before its repair steps begin. Does not cover new-install commissioning or the manual interlock backup method, owned by their own files.

Roles and responsibilities

Stop before anything else moves

Before opening the ATS enclosure, meter which source or sources are actually live right now. Do not rely on the customer's description or the controller's display alone; a stuck or partially welded contactor can show one state on the display while physically passing power from both sides. Never force, jog, or manually move the switch mechanism while any possibility exists that both sources are live, a stuck ATS attempting to close a second contactor onto an already energized bus is a bolted-fault event with arc-flash energy behind it. If metering shows voltage present on the utility side and the generator side simultaneously at the transfer mechanism, stop this procedure entirely, open the generator's output breaker at the genset immediately to physically separate the two sources, and treat it as a suspected welded contactor requiring a full teardown inspection, not a normal diagnostic sequence.

The procedure

1. Meter the utility-side input, the generator-side input, and the load-side output of the ATS before touching anything, to establish the true present state independent of the controller display. Acceptance: three recorded voltage readings, one per point. Wrong looks like reading only the controller's status indicator. Stop rule: any reading showing voltage on both source sides at once triggers the entry stop rule above, generator breaker off, full teardown, escalate, not a continued normal diagnosis. Hazard: an ATS enclosure holds two independent live sources; treat every conductor inside it as energized until this reading proves otherwise.
2. With both sources confirmed safely separated, open the enclosure and visually and mechanically inspect the contactor: position, contact condition, and whether the manual operator handle moves it freely. Acceptance: a specific documented state, contacts clean and mechanism free, contacts visibly pitted or welded, or mechanism binds, not a general "checked the switch." Wrong looks like assuming the contactor is the fault without confirming its actual mechanical condition. Stop rule: if manual operation with both sources confirmed off does not move the contactor at all, or it moves but will not seat, condemn the contactor assembly outright and do not attempt to force it further. Hazard: even de-energized, the transfer linkage can pinch or crush fingers during manual operation; keep hands clear of pivot points and follow the manufacturer's manual-operation sequence exactly.
3. Use the ATS troubleshooting tree or the utility-loss-but-no-transfer decision tree to identify the specific failed component, then confirm that diagnosis with a direct test before ordering or swapping anything. A direct test might mean simulating a controlled utility loss and confirming whether the controller issues a transfer command, or briefly energizing a transfer coil with sources isolated to confirm the mechanism is sound. Acceptance: a component-level cause confirmed by an actual test result, not inferred from symptoms alone. Wrong looks like a swap-and-pray parts order based on the phone description. Stop rule: no part gets replaced without a confirmed test pointing to it specifically. Hazard: simulating a utility-loss event on a live, occupied panel commands a real transfer; warn occupants and keep hands and tools clear of the load panel first.
4. De-energize and lock out both sources at their respective disconnects, and prove zero energy at the actual terminals being worked, live-dead-live on a known live source before and after, per NFPA 70E-2021, 120.5, with work practices under 29 CFR 1910.333(b)(2). Acceptance: a meter reading of zero at the specific work point, proven by the live-dead-live method, not by a breaker label. Wrong looks like trusting a labeled breaker without metering the actual work point. Stop rule: any reading above zero stops the repair; re-trace and re-isolate, and if the proving sequence fails on its own known-live check, the meter is suspect, so swap meters before proceeding rather than trusting a second unverified attempt. Hazard: unverified stored energy at the work point is a direct shock and arc-flash hazard; the only response to a failed proving step is stop and re-isolate.
5. Execute the repair: replace the failed component, reconnect per the manufacturer's wiring diagram, and torque every disturbed lug to spec, recording the value used. Acceptance: a torque record per connection and every wire matched to the diagram, none left unlabeled. Wrong looks like a lug tightened by feel, or a wire left disconnected because its function was not understood. Stop rule: any wire whose function is not confirmed against the diagram gets traced before the enclosure is closed. Hazard: a loose lug on a transfer switch's high-current terminals is a heat and arc source that fails later under real load, exactly when the switch is needed; the repair disturbed these connections, so it owns verifying them before closing up.
6. Re-energize one source at a time, utility first, and confirm normal operation before the generator ever enters the same test, watching for the dual-source-closed condition named in the opening stop rule. Acceptance: load side reads correct utility voltage with the generator side confirmed still open. Wrong looks like re-energizing both sides together to save a step. Stop rule: any voltage at the load side traceable to the generator input before the generator has started means the mechanism is not isolating correctly; same response as step 1, breaker off immediately, full re-isolation, no live diagnosis attempted. Hazard: identical to step 1, now live at the load side rather than at rest.
7. Run a full, real transfer cycle in both directions using an actual simulated utility loss, not the controller's manual test button, since a manual test can bypass the sensing logic that originally failed. Time and record the sequence: start, transfer with load carried through a brief engineered open-transition gap, utility restoration, retransfer, cooldown and shutdown. Acceptance: the full cycle completing on the controller's own logic in both directions, timed and logged. Wrong looks like confirming only the transfer-to-generator half, exactly the half most likely to still be broken if the original fault sat in sensing or timing rather than the contactor. Stop rule: any half that does not complete on its own control logic, meaning the tech has to force it manually, fails verification outright; the switch is not returned to service on a partial cycle. Hazard: a real cycle re-arms the dual-source hazard from steps 1 and 6 under live commands; keep a second person at the generator breaker able to open it if a dual-close condition appears.
8. Document the confirmed fault, the repair, the torque log, and the full cycle timing, then confirm the auto-exercise schedule is active on the controller's live display, not the settings menu. Acceptance: a written record filed and the schedule observed active. Wrong looks like leaving without confirming it resumed, so the next real outage becomes the first test since the repair. Stop rule: the job does not close until the schedule is confirmed active by direct observation.

The record this produces

The three-point pre-repair reading, the mechanical inspection finding, the confirmed component-level cause and how it was confirmed, the torque log, the full-cycle timing in both directions, and the confirmed exercise schedule state. Filed against the unit's service history so a second no-transfer call at the same address is checked against what was actually replaced, not re-diagnosed from zero.

Worked pass: a real outage, a control-board fault behind a sound contactor

A homeowner calls; the generator is audibly running in the yard, the house is dark. The tech arrives and meters the ATS before touching it: utility side reads 0 V, a genuine outage, generator side reads 240 V, load side reads 0 V. No dual-source condition, and the fault sits downstream of a genset that is producing power correctly.

The tech stops the generator from the controller display so the enclosure can be opened with nothing live inside it, then opens it. Step 2's mechanical inspection finds the contactor free-moving on the manual operator with clean, unpitted contacts, ruling out a mechanical or welded-contact failure. Step 3 turns to the ATS troubleshooting tree: the generator's own start and run signals succeeded, but the transfer relay never issued a close command. A direct test, briefly and safely energizing the transfer coil with both sources isolated, closes the contactor cleanly, proving the mechanism itself is sound and pointing the fault at the control board's transfer logic.

Step 4 locks out and proves zero at the board terminals. Step 5 replaces the control board, reconnects it against the manufacturer's diagram, and torques every disturbed lug to spec with the values logged. Step 6 re-energizes utility only; the load side shows correct utility voltage with the generator side still open, no dual-source condition.

Step 7 is where this pass fails. The transfer-to-generator half runs clean, load carried through the transition. But on retransfer, the switch does not return to utility within its own set time delay; it sits on generator for several extra minutes past the value the controller itself displays. Under the stop rule, verification fails outright, not treated as good enough because the first half worked. The tech traces the retransfer delay and finds it left at a shortened service-test value from a previous card swap rather than the manufacturer's field default. Correcting it and rerunning the full cycle produces a clean retransfer on schedule. Step 8 documents both fixes, then confirms the exercise schedule active on the live display before leaving.

References

• See related: ATS troubleshooting tree, for the component-level fault-finding logic this procedure assumes has already run.
• See related: Utility loss but no transfer decision tree, for the diagnostic branch on this exact symptom.
• See related: Interlock versus ATS reliability reference, for the manual backup method this procedure does not cover.
• NFPA 70E-2021, 120.5, for the live-dead-live proving sequence used before any repair.
• 29 CFR 1910.333(b)(2), for electrical work practices on de-energized conductors and equipment.
• NEC Article 700, 701, or 702, current locally adopted edition, for transfer equipment requirements matching the install's classification.
• Manufacturer service manual for the specific ATS and contactor make, for torque specifications and wiring diagrams.