Particle spike at 14:03, door alarm at 14:05, pressure drop at 13:58, on three clocks. Nobody can say which came first.
That is the normal state of a plant where the BMS, the EMS and the particle counting system came from different vendors, each keeping its own time. Internal clocks drift, and after a year the gap can be minutes. In an excursion investigation those minutes decide whether the door caused the spike.
ALCOA+ asks for contemporaneous records. A record stamped five minutes late sits next to the wrong cause.
What we write into the specification:
- Every server, controller, PLC and counter syncs to one NTP source with a secondary, and loss of sync raises an alarm.
- Data are stored in UTC, local time is a display setting, and the audit trail records the offset used.
- OQ reads every system's clock against the reference at the same moment and accepts within a stated tolerance. We would write ±2 seconds for NTP-synced systems, which a working NTP client holds easily. Then pull the sync source and confirm the alarm.
- Periodic review repeats the comparison, because a replaced controller can come back running on factory time.
UTC storage earns its place when one system is hosted abroad. Viet Nam is UTC+7 all year, while most US time zones move an hour twice a year, so mixed local timestamps can put events an hour out of order for part of the year.




