The Fault That Would Not Be Found [/ðə ˈfɔlt ðət ˈwʊd ˈnɑt ˈbi ˈfaʊnd/] n - The symptom. A ground-fault monitor upon a distribution panel raised a warning every few days, each clearing of itself within a minute. Nothing tripped, nothing was harmed, yet an intermittent Grounding and bonding alarm is not a thing one may ignore; the next might be true.

The dead end. [/ðə ˈdɛd ˈɛnd/] n - We meggered cables, inspected bonding jumpers, and proved terminations across the panel. All passed. An intermittent fault that will not reproduce on demand is the worst of its kind, and “replace it until it stops” is costly and unsatisfying alike when no bad part can be pointed to.

The actual cause. [/ðə ˈæktʃuəl ˈkɔːz/] n - I plotted every alarm timestamp, and they clustered: early mornings, a few times each week, never on weekends. Laid against the facility schedule they fell in line with a large The Standby Engine monthly and weekly load test. During the transfer, momentary leakage current through line filters and the brief change of the neutral-to-ground relationship at the Automatic transfer switch sufficed to nudge the sensitive monitor past its threshold for a second. It was a measurement artifact of a normal, wholesome test, and no fault of the wiring at all.

The fix: limited and clever. [/ðə ˈfɪks/] n - No rewiring. I contrived with the controls settings a short, fitting response delay upon that one monitor, so a sub-second transient during a known transfer would not latch an alarm, while a genuine sustained fault still would. Then I correlated the monitor’s log against the test calendar, that any future “phantom” might explain itself. The nuisance alarms ceased; true protection was untouched.

The lesson. [/ðə ˈlɛsən/] n - Before you conclude the system broken, ask whether the alarm merely witnesses something normal. Timestamps are evidence.

Part of Field Notes.