Bad Clocks Create False History
The instrumentationtools.com article "Synchronizing Time in Industrial Automation Applications" begins with the problem every trip investigator knows. Devices with different clocks write event records with contradictory timestamps. Engineers then argue about which log tells the truth.
The article's argument deserves emphasis. Time errors do not merely annoy; they rewrite causality. A pump seal failure and a motor trip appear in the wrong order. The investigation follows the false order and blames the wrong system. Therefore, time synchronization is a safety-related function, even though no standard labels it that way.
NTP, SNTP, and PTP: The Article's Numbers Hold Up
The article walks through the three dominant protocols, and its accuracy claims match my measurements. First, NTP. A master server distributes time, and client devices poll it at regular intervals. The article quotes roughly 10 ms accuracy for NTP on a decent network. My plant NTP domains land between 1 ms and 10 ms on unloaded switches, so the number is realistic.
Second, SNTP. The article calls it a simplified NTP that skips network delay calculations, with accuracy up to about 50 ms. That matches the SNTP specification's intent. Use it for HMIs and displays, never for event capture.
Moreover, the article explains PTP, IEEE 1588, as the precision option at around 1 ms or better. In practice, hardware-timestamped PTP reaches sub-microsecond accuracy on PTP-capable switches. However, the article correctly notes the catch: you need PTP-capable switches or NICs to get there.
The article also explains stratum levels, from stratum 0 reference clocks through stratum 3 downstream clients. Finally, it recommends GPS time servers as the best plant reference. I agree without reservation.
Building the Architecture: The Article's Example, Extended
The article sketches a plant with one SCADA server, three PLCs, one HMI, a managed switch, and a GPS source. The SCADA server synchronizes to GPS and serves time to everything else. That architecture works, and I extend it with two rules from practice.
Step 1
Install a dedicated GPS time server as the stratum 1 source. Do not depend on internet NTP for a plant that must operate islanded.
Step 2
Configure at least two internal NTP servers for redundancy. Point every controller at both, so one dead server degrades nothing.
Step 3
Keep the time VLAN flat and low-latency. Chained router hops add jitter that NTP then absorbs as error.
Step 4
Set polling intervals sensibly. Controllers polling every 64 seconds stay tight; HMIs can poll every 15 minutes.
Step 5
Log clock offsets centrally. An alarm on offsets beyond 500 ms catches failing batteries and isolated switches early.
Emerson Ovation: Time Discipline in a Power Plant DCS
Emerson Ovation controllers carry their own time infrastructure, and it must agree with the rest of the plant. Ovation uses a time synchronization architecture built around GPS receivers feeding the controller network. Drop an Ovation controller off time sync and its sequence-of-events resolution suffers immediately.
Step 1
Verify the GPS receiver status and antenna path during every outage. Water in antenna cable kills sync slowly and silently.
Step 2
Confirm each Ovation controller reports synchronization healthy in the controller diagnostics, not just the operator station.
Step 3
After any controller replacement or software load, force a time re-synchronization and verify the SOE timestamps align with the plant reference within one scan.
Step 4
Align the Ovation historian clock to the same GPS source. A historian one second off makes cross-system event correlation guesswork.
Bently Nevada 3500: When Milliseconds Are the Whole Story
Machinery protection adds a harder requirement. A Bently Nevada 3500 rack with System 1 captures shaft vibration, bearing temperature, and speed changes at high sample rates. However, the diagnostic value collapses if the rack's clock drifts from the DCS.
My rule: sync the 3500 rack through its supported time interface, either NTP or an IRIG-B signal from the plant GPS system. IRIG-B delivers microsecond-level alignment over a dedicated wire, immune to Ethernet load. Moreover, it decouples machinery timing from network health entirely.
Step 1
Choose IRIG-B from the GPS master where cable runs allow it. Keep total run length within the driver's specification.
Step 2
If NTP is your only option, verify the 3500 rack's offset weekly against the plant reference.
Step 3
After a machine trip, compare the 3500 event list against the Ovation SOE list within the first hour. Record the observed offset.
Step 4
Investigate any offset beyond 10 ms before the next trip, not after it.
Conclusion & Action Advice
The instrumentationtools.com article gives a clear, correct map of NTP, SNTP, PTP, and GPS-based synchronization for automation systems. Its accuracy numbers and stratum explanation match field reality. Therefore, act on it in order: secure a GPS stratum 1 source, distribute NTP redundantly, and reserve PTP or IRIG-B for systems that justify the hardware. Finally, run a time audit this week. Compare your Emerson Ovation SOE log against your Bently Nevada 3500 event list at a known shared event. If they disagree by more than 10 milliseconds, you have already found your next trip investigation's first suspect.