Time sync keeps automation trustworthy. Devices rarely share one exact clock. A small deviation is enough. It misorders logged events. Engineers then chase phantom sequences. On a machine trip, a lagging clock reorders protection signals. The root cause looks wrong. So fix the clock first. Then read the events. I align SOE across turbine and vibration platforms often. The lesson holds everywhere: pick the right protocol for the accuracy you need.
First: Choose NTP, SNTP, or PTP by Accuracy
NTP syncs a master to slaves. The master holds the reference. Clients poll it at intervals. NTP even estimates network delay. SNTP is simpler. It ignores delay compensation, so it is looser. A typical budget: SNTP reaches roughly 50 ms. NTP reaches about 10 ms. PTP, or IEEE 1588, is the fastest. It targets about 1 ms or better. Here the master pushes timestamps continuously. The slave never waits to be asked. Therefore it tracks offset in real time.
- Step 1 — Use SNTP or NTP for operator logs and historical reporting.
- Step 2 — Reserve PTP for trip logic, SOE, and phase measurement.
- Step 3 — Remember the hidden cost. PTP needs PTP-capable switches or NICs. Ordinary switches add random delay and destroy the millisecond target.
Second: Build a Clean Clock Hierarchy
Use a stratum model. Stratum 0 is the source. A GPS receiver sits there. It gives accurate UTC. The plant server becomes the master below it. PLCs and HMIs act as clients. They ask the master every few seconds. A single source prevents drift conflicts. Never let two masters fight. Define one GPS grandmaster. Then feed the DCS, the historian, the safety system, and the monitors. Confirm clients point to the server, not the internet.
- Step 1 — Set the GPS unit or PTP grandmaster as the sole stratum 0 to 1 anchor.
- Step 2 — Point the SCADA and DCS servers at the anchor via UDP port 123 for NTP.
- Step 3 — Configure each controller and monitor as a client of that server, not of a public pool.
- Step 4 — Where a hardwired code is required, feed IRIG-B to legacy racks. Check AC or DC coupling and the encoding format.
Moreover: Align Woodward and Bently Nevada Time Bases
A gas turbine trips on many signals. The Woodward controller logs one timestamp. The Bently Nevada 3500 monitor logs another. The historian stores a third. If these clocks differ, the sequence is fiction. Phase and keyphasor data are worse. A few degrees of phase error can hide real motion. Therefore give both platforms the same time source. Then cross-trigger a known event. The trip channel and the probe waveform should agree within the SOE resolution.
However: Verify with a Real Cross-Check
Do not trust the settings page alone. Prove it with an event. Inject a test pulse into a shared digital input. Then read the time-stamp from each system. The difference is your true offset. Watch for silent NTP failures too. A stranded client keeps drifting once its server disappears. Poll the offset field, not just the status. A growing offset means a broken path, even while the link stays green.
- Step 1 — Trigger one known event across the controller, monitor, and historian.
- Step 2 — Compute the maximum timestamp spread. Aim for under one SOE scan.
- Step 3 — Log the client offset alarm so a drifting clock cannot hide.
Conclusion & Action Advice
Match the protocol to the fault you must catch. Use SNTP or NTP for reporting. Use PTP for trip and phase data. Anchor everything to one GPS grandmaster. Give Woodward and Bently Nevada platforms that same source. Then prove it with a shared test pulse and measure the real offset. Log the offset alarm permanently. A synchronized clock turns scattered events into one honest timeline. Do it before your next SOE investigation.