The full power of TDengine, now free forever for up to 5,000 tags.

Explore More

Time Synchronization for Industrial Data Collection

Juno Qiu

November 4, 2026 /

Two plant systems can record valid measurements and still disagree about the order of events. If their clocks differ by three seconds, a historian trend may place a valve movement before the command that caused it. The resulting investigation can be wrong even when both values were transferred intact.

Time synchronization is the discipline of keeping relevant clocks within an agreed error of a reference and detecting when they are not. A timestamp’s number of decimal places is its representation, not proof that the underlying clock or measurement is equally accurate.

Start with an error budget

Ask what decision the timestamps support. A daily energy total may tolerate a larger timing difference than a fast sequence-of-events investigation. With the process and application owners, specify the maximum uncertainty or relative misalignment that the task can tolerate. Do not adopt a single plant-wide target without considering the use case.

The total error may include clock offset, the point at which a value is timestamped, device processing delay, sampling interval and uncertainty introduced by gateways. Synchronizing clocks does not eliminate delay between the physical event and the timestamp assigned by a collector.

Example: reconstruct a pump trip

Suppose an investigation requires a PLC trip indication and a gateway-observed pressure change to be aligned within one second. The team uses illustrative conservative bounds, not measured performance: each relevant clock may be offset by up to 100 ms from the same reference, giving up to 200 ms relative clock error. The gateway polls pressure every 500 ms, so a change occurring just after a poll may be observed up to 500 ms later under the assumed regular schedule. The team also allows 100 ms for timestamp assignment and processing variation, plus 100 ms for the uncertainty of the verification measurement. Adding these worst-case bounds gives 900 ms, leaving 100 ms of margin against the one-second requirement. The PLC trip timestamp is assumed to be assigned at the event; that assumption must be verified.

This conservative sum is an example, not a statistical uncertainty calculation or a guarantee. The clock offset describes disagreement between clocks; timestamp assignment describes which point in the path receives the time; sampling delay is the time until the change is observed; measurement uncertainty limits confidence in the verification itself. Schedule jitter, source update delay and other omitted components require their own bounds. If those bounds cannot be established or their sum exceeds the target, the team cannot claim the requirement is met.

To verify, measure both clocks against the approved reference with tools whose uncertainty is documented, then trace a known, safely generated or independently recorded event through both paths. Compare observed timing with the source update and polling schedule across representative conditions, including loss and recovery of synchronization. Keep the measurements, assumptions and untested conditions with the acceptance record.

Map the clocks and their dependencies

For each important path, record the device that timestamps the measurement, its time source, synchronization method, network path, time zone and owner. A sensor, PLC, gateway and historian may have separate clocks. If a gateway timestamps a polled value, its timestamp may describe the read or receipt, not the physical event.

NTP and IEEE 1588 PTP are possible time distribution approaches, but accuracy depends on implementation, network design, configuration and monitoring. NIST’s IEEE 1588 overview explains how synchronized clocks can support correlation; its examples do not guarantee a particular deployment’s accuracy. Select a method against the measured requirement and the devices that actually support it.

Verify the path in operation

  1. Confirm each device is configured to use an approved time source, including authentication or network protections appropriate to the site.
  2. Measure offset and synchronization status using tools appropriate to the devices; document measurement uncertainty and alarm thresholds.
  3. Check what happens when the source is unreachable, a clock restarts, or its time is corrected. A clock can continue running while drifting away from the reference.
  4. Trace a known event across systems, accounting for sampling and processing delays. Record whether the observed alignment satisfies the use-case budget.
  5. Assign responsibility for monitoring and responding to loss of synchronization. Mark affected records or time ranges as uncertain where the application supports it.

Use a consistent time basis for storage and retain enough information to interpret local time and daylight-saving transitions. Do not silently rewrite historical timestamps when a clock problem is discovered: preserve evidence, document the correction and assess downstream calculations. NIST’s discussion of time traceability explains why a reference and its uncertainty matter.

Related Articles