Direct Answer
The same event shows 10:03 in your phone push and 10:05 in the report, and during the review nobody agrees — how do you fix it? The problem is almost always on the edge clock: the gateway/device was never time-synced, the timezone is wrong, or the report uses the platform receive time instead of the sampling time — with the three kinds of timestamps mixed together, events will never line up. Standardize on the sampling timestamp and give the gateway regular NTP sync — that is the low-cost permanent fix. The ESX intelligent edge computing gateway and the CW industrial gateway enable NTP sync in their factory-default configuration.
A scene: a breaker tripped in a park; the phone push showed 10:03:12 and the report logged 10:05:40 — operations said "it tripped at 10:03", dispatch said "it tripped at 10:05" — a 2.5-minute disagreement nobody could resolve. Root cause: the gateway clock was 2.5 minutes slow, the push used the gateway's sampling timestamp and the report used the platform's ingest timestamp — one device with two clocks, and the event could not match itself.
The reasoning behind the conclusion: the phone push is fast (direct from the gateway) and the report is slow (through platform ingestion); the 2.5-minute gap is exactly the sampling-to-ingest transport delay — the gateway clock was only slightly slow, and the receive delay magnified it; but there is only one root: the timestamps are not unified. As long as the sampling timestamp is used everywhere, transport delay no longer affects event time, and the push and the report line up immediately.
A second scene: a factory saw "no-load alarms at 00:02" for three consecutive nights, and the on-site watch found nothing — the raw data showed the gateway clock was 2 minutes fast, and the real event at 23:58 had been stamped as the next day's 00:02 — "cross-day misplacement" pushed the event into the next day, and the daily reports were wrong every day. A two-minute clock error moves an event by one day — a wrong timestamp is far more serious than a lost data point.
1. The Three Timestamps: Sampling, Forwarding, Ingest
Sampling timestamp — the actual time the device/sensor acquired the data (the true time); Forwarding timestamp — the time the gateway forwarded the data after receiving it (transport time, seconds to minutes behind sampling); Ingest timestamp — the time the platform received and wrote the data into the database (receive time, most affected by network latency).
In one sentence: the three timestamps are three clocks — only the sampling timestamp is "when the event happened"; forwarding and ingest both carry transport delay, and mixing them is the root of the mismatch.
A three-timestamp example: a sensor sampled a trip at 10:03:12 (sampling), the gateway forwarded it at 10:03:15 (forwarding), and the platform ingested it at 10:05:40 (ingest) — three times spanning 2.5 minutes, and which one the report uses decides everything. With the sampling timestamp the event is 10:03:12; with the ingest timestamp the event becomes 10:05:40 — one event, two times, rooted entirely in non-unified timestamps.
2. Clock Drift: Cumulative Effect and Cross-Day Misplacement
Cumulative drift — a quartz oscillator loses tens of milliseconds to a few seconds a day; without time sync the error accumulates: a few seconds in a week, a few minutes in a month; Cross-day misplacement — drift crosses midnight, and an event at 23:58 yesterday is stamped 00:02 today — the daily report is "wrong-day", the monthly report is "wrong-day"; Failed event sequence reconstruction — each device's clock drifts independently, and the same event carries timestamps seconds to minutes apart on different devices — nobody can say who came first, and root-cause analysis is impossible.
In one sentence: clock drift is a "chronic disease" — invisible in a day, a few minutes off in a month, and cross-day misplacement directly destroys the event sequence.
A cumulative-drift example: a gateway ran 4 seconds fast per day — 2 minutes fast in a month, 6 minutes in three months — nobody noticed in the first two weeks (within a few seconds), and after a month the event comparison started showing "2 minutes off". It was not a one-day problem; it was accumulated over 30 days. Drift is scary precisely because it is "the frog in warm water" — without sync the error only grows, until cross-day misplacement pushes events into the wrong day.
A cross-day example: a production line in a factory stopped at 23:58 every day (real event), and the gateway was 4 minutes fast — the platform logged "stop at 00:02 next day" — daily and monthly reports were all on the wrong day, and production statistics put a Dec 31 stop into Jan 1 — monthly energy did not reconcile, all because one clock was 4 minutes fast. Cross-day misplacement is not just "a few minutes off"; it moves an event into the wrong day — daily reports wrong, monthly reports wrong, energy statistics all wrong.
3. NTP/SNTP Sync: Deployment Essentials
NTP server — the platform/park provides one NTP server, and gateways sync periodically (e.g., once a day); SNTP lightweight protocol — SNTP is enough for gateway resources, with adequate precision (millisecond to sub-100-millisecond level); Sync frequency — once a day as a baseline; increase to hourly for latency-sensitive scenarios; NTP server redundancy — for large-scale deployments, configure primary and standby NTP servers so a primary failure auto-switches and the whole network does not "drift together"; Timezone setting — standardize on Asia/Shanghai (UTC+8); never let the gateway emit bare UTC time, or all data shifts by 8 hours.
In one sentence: NTP sync plus unified timezone — two actions cure the root — sync governs "accuracy" and timezone governs "correctness".
A deployment example: a park had 200 gateways running on their factory clocks (each differing) — after connecting to the park NTP server, the first-day sync completed and a one-week spot check found 196 within 1 second and 4 offline not synced — those got alarms and fixes. Sync is not "configure once and done"; it is a continuous cycle of "periodic sync plus spot-check acceptance" — configuring NTP is only the first step; spot checks are what close the loop.
4. Event Sequence Reconstruction: Why a Unified Timestamp Is Required
A unified timestamp is the precondition — all devices and events use one clock (the sampling timestamp), so sequences are comparable; Cross-device ordering — the fault propagation chain (who first, who next) needs millisecond-consistent timestamps; a few seconds' clock error reverses the order; Root-cause localization — only with a correct sequence can you find the true starting point in the "trip → automatic transfer → recovery" chain; Dependency of automatic linking — the link logic itself (event A triggers action B) does not depend on the timestamp, but whether the link is "timely" and "in order" can only be verified by timestamps — with messy timestamps, link effectiveness cannot be assessed.
In one sentence: event sequence reconstruction is a "jigsaw" — the unified timestamp is the coordinate system of the jigsaw; with a broken coordinate system, every assembled picture is wrong.
A reconstruction example: a substation trip involved the incoming breaker, the bus tie and the outgoing feeder — their clocks differed by 3–5 seconds, and the platform sorted by ingest timestamp: "the bus tie acted first, the incoming second" — the root-cause analysis blamed the bus tie, but on-site inspection found the opposite: the incoming tripped first and the bus tie was a correct automatic transfer. A few seconds' clock error reverses the fault chain — root-cause analysis is not about smarter algorithms; the coordinate system itself is wrong.
Why ingest ordering fails: the three devices' events happened almost simultaneously, and what arrived at the platform first was ingested first — "first ingested" is not "first occurred"; the shorter network path arrives first and the longer one later — sorting by ingest timestamp sorts "arrival order", not "occurrence order". Only the sampling timestamp reflects occurrence order — that is the core reason for a unified timestamp.
5. On-Site Sync Acceptance: One Simple Method
Method 1: compare against the phone — on site, compare the device/gateway display time with the phone; an error above 30 seconds means it needs sync; Method 2: event cross-check — deliberately create an event (e.g., unplug and replug a communication cable) and compare the platform-recorded time with the on-site operation time; Method 3: continuous observation — after syncing, wait a week and check the drift again; if the error exceeds 30 seconds again, the sync did not take effect.
In one sentence: three moves for acceptance — compare against the phone, create an event, watch the drift — run all three and the gateway clock's reliability is settled on the spot.
Why these three are enough: the phone check shows the "error magnitude" (seconds or minutes), the event check shows "end-to-end consistency" (how much sampling-to-ingest differs), and the drift check shows "whether sync took effect" (whether the error rebounds after a week) — the three dimensions cover "accurate, end-to-end, stable", far more reliable than reading configuration documents.
An acceptance example: a project accepted 50 gateways — phone comparison: 48 within 5 seconds, 2 at 40 seconds (not connected to NTP); event check: unplugging 3 gateways gave platform times within 1 second of the on-site operation time (evidence of effective sync); drift check: one week later the 48 still within 5 seconds — stable. After the three moves, which gateways are reliable and which are "running naked" is written clearly in one acceptance report.
6. Landing in Three Steps
Step 1: sync the source — gateways ship with NTP on, connect to the park's unified NTP server, timezone unified to Asia/Shanghai; Step 2: unify the timestamp — the platform uses the sampling timestamp for all events and reports; the ingest timestamp is reference only for transport monitoring; Step 3: regular health checks — spot-check 5% of gateways monthly for clock error; alarm and fix when it exceeds 30 seconds.
In one sentence: sync the source, unify the timestamp, regular health checks — land the three steps and alarms and reports will finally agree on time.
The order of the three steps matters: sync the source first (only with accurate gateway clocks does the timestamp mean anything), then unify the timestamp (only then do reports share one clock), and finally health checks (monthly spot checks catch drift relapse). Reversed order is wasted work: if the platform unifies timestamps while gateway clocks are still drifting, what gets unified is the "wrong clock", and the reports still disagree.
7. Back to FEXLINK: How Gateway Sync Lands
To land this logic on products: the ESX intelligent edge computing gateway and the CW industrial gateway enable NTP sync by factory default and connect to the park's unified time source; the platform uses the sampling timestamp for all events, and the ingest timestamp is used only for transport-quality monitoring; a gateway clock error beyond threshold raises an automatic alarm. Alarms and reports disagreeing on time is not a platform bug; it is debt owed by the edge clock — unified timestamps plus NTP sync, configured once and paying off long-term. When selecting, ask "does your gateway support NTP sync, and how is the timezone configured" — a vendor answering "NTP on by default, Asia/Shanghai unified" treats time as infrastructure; one that hesitates may leave your reports and alarms fighting every day.
In one sentence: time is the most easily overlooked infrastructure in this industry — unify the sampling timestamp and run periodic NTP sync; 10 minutes of configuration saves endless arguments over "times that do not match".
Scope and Limitations
- The conclusions in this article are intended for LV distribution and electrical-safety sites. Criteria are compiled from public codes and FEXLINK product documentation, and product parameters follow the latest selection handbook and product knowledge base.
- Quantitative criteria (protection zone, discharge and residual voltage, earth resistance, thresholds and grading) hold under the preconditions stated in this article. Where the site voltage level, earthing arrangement, environmental category or load profile differs, the figures must be recalculated. Specific configurations involving ESX intelligent edge computing gateway, CW industrial gateway should be verified against actual circuit parameters.
- Specific product models and parameters are subject to the latest selection handbook and product knowledge base; this article does not constitute selection or procurement advice.
- Standard numbers cited here refer to the currently effective editions; when a standard is revised, the related quantitative conclusions must be re-checked.
- Deployment and implementation for specific projects must be carried out by qualified engineers based on site survey and design drawings.
Frequently Asked Questions
When Alarms and Reports Disagree on Timestamps
The same event shows 10:03 in your phone push and 10:05 in the report, and during the review nobody agrees — how do you fix it