Troubleshooting
Why HMI operation records miss PLC start stop transitions
Short answer
An HMI operation record writes an entry only when a touch changes a value, so background transitions of PLC output Y5 never reach it. Build the start/stop history from a state-driven event log on Y5 that stores both the activation and the return transition, verify the tag reads back live, and latch in the PLC any state shorter than the panel's polling period.
On this page
Operation Record on a DOP100 panel writes an entry when a touch changes a value — it is an operator-action log, not a recorder for PLC variables that change in the background. That is why the start and stop times of output Y5, which the controller drives from M167 through its own logic, never appear, and why a second press produced no new entry. Keep the requirement, change the mechanism: drive the history from the state of Y5, store both transitions, and confirm the tag reads back live. The alarm page this was moved to temporarily is the right mechanism once the return event, the timestamp source and retention are configured deliberately.
What Operation Record actually records #
The Operation Record function (chapter 24 in the DOP100 documentation) reports changes made by touching HMI objects: an entry is written when the touch changes the recorded value. A PLC bit toggling in the background stays invisible to it, even when that bit is displayed on the same screen. A simulation run tells you little here, because the recorder only reacts to touches and no controller is driving Y5; the check that means anything is testing HMI simulation against a live PLC.
The installed panel is a DOP-B, and Operation Record is a DOP100-series function; its absence on the DOP-B family is expected rather than a fault. Confirm what history functions the exact DOP-B model provides in its own documentation before copying a DOP100 configuration across — a function carried over by name does not carry over by behavior.
Four events inside one start #
Decide which event the record must mean before touching the configuration:
| Event | Where to inspect it | What it proves |
|---|---|---|
| Operator touches a control | HMI object and its recording settings | an operator interaction |
| M167 changes | PLC online monitor | the internal start condition moved |
| Y5 transitions | PLC output status and HMI readback | the command this history must record |
| Machine actually runs | running feedback, if fitted | physical operation |
In normal running these coincide; an interlock, an automatic sequence or an output fault separates them. For a command history, record Y5; for operating-time accounting, record running feedback when the machine has it.
Then verify the path before blaming the recorder:
- Watch M167 and Y5 online in the PLC through one authorized start and stop; note which changes first and whether Y5 returns to inactive on the stop.
- Display the readback of Y5 on the HMI beside the controller monitor; the two must agree on both edges.
- If the PLC changes and the HMI does not, check the connection, station address, address syntax and data type of the history tag. A working Start button proves nothing here, because the button can write an internal command while the recorder reads a different variable — the same fault class as why HMI indicators stay off over Modbus RTU polling.
- If both agree and entries are still missing, the address path is sound and the fault is in event generation or storage.
Record both edges of Y5 #
A start/stop history is two events per cycle: the 0-to-1 transition as the start and the 1-to-0 transition as the stop. A history that shows starts without stops is storing the activation edge only, and a trigger that stays active cannot produce a new activation until it clears. Building the edge pair in controller logic follows the same discipline as shifting both edges of a discrete signal delay: the rising and the falling transition are separate events, and each needs its own capture.
Two arrangements do the job:
- One state-driven event on Y5 with both an activation entry and a return entry — preferred, because start and stop stay tied to one trigger.
- Two separate events, each generated by one edge — use this when the history needs distinct labels or cannot present a return as a stop record. Generate each from the transition, never from the level: a continuously inactive output is not a new stop.
Read the live trigger state next to the history display. If it stays active after the stop command, go back to PLC logic and address mapping. If it clears and no stop entry appears, the history is not storing returns. Acknowledgment is an operator action, not the machine stopping, so keep acknowledgment, activation and return as distinguishable entry types.
Compare the shortest state with the polling period #
The panel only sees Y5 at each read of the tag. If the output can change and return between two consecutive reads — a shortest state shorter than the tag's acquisition period — both edges are lost with no error raised. A momentary command that is true for one PLC scan is the classic case. Write down the machine's minimum on-time and off-time from the sequence, read the acquisition period from the tag settings, and apply one rule: any state that can be shorter than the acquisition period gets latched in the PLC and transferred as an event, not polled as a level. Capturing a short pulse on both transitions is the same problem as configuring both transitions when debouncing a PLC input — arm the capture on each edge.
Checking Y5 coverage with a three-cycle test #
Inputs:
- Y5 produces 2 transitions per start/stop cycle — the 0-to-1 start and the 1-to-0 stop — by this machine's logic.
- The test runs 3 supervised start/stop cycles with the machine cleared, because it commands real motion.
- The rule in words: expected entries equal cycles times two, one activation and one return per cycle.
Arithmetic: 3 cycles × 2 transitions per cycle = 6 entries — 3 starts and 3 stops.
Decision by the count:
- 6 entries — both edges recorded; the state-event history is sound, keep it.
- 3 entries — only the activation edge is stored; enable the return entry or define the separate stop event.
- 1 or 2 entries, matching touches only — the history is still touch-driven; move it to the state-driven event configuration.
- 0 entries — the history tag is not reading Y5; fix connection, address syntax and data type first.
Keep personnel clear of the motion envelope and product out of the machine during the test, since it performs real starts and stops.
Turn the alarm page into the retained record #
Moving the indication to the alarm page was the right mechanism; what made it rough is configuration, not the choice. State-driven alarm and event functions on industrial panels trigger on bit or word tags, PLC or local, and some platforms expose the same mechanism as value-change logging on a tag — that is what to use for Y5.
- Fix the meaning first: command history from Y5, operating history from feedback; keep M167 as a diagnostic signal unless the required event is exactly that internal condition.
- Configure Y5 as a state event: the activation entry is the start, the return entry is the stop.
- Set and verify the clock the recorder timestamps from; if the PLC supplies timestamps, check the controller clock and their transfer.
- Size storage and overwrite for the longest retention you need, and use the retained log — a live alarm list that clears when runtime stops is not a history.
- Keep activation, return and acknowledgment distinguishable in the display and in any export.
- Where states can be shorter than the acquisition period, add the PLC latch with a queued transfer, and clear queued events only after the history has accepted them.
The decision for this machine #
- Stop looking for Operation Record on the DOP-B: it is a DOP100 function, and even on a DOP100 it records touches, not Y5's background transitions.
- Keep the alarm-page history, reconfigured as a state event on Y5 with both edges, a verified clock and retained storage.
- Record running feedback instead of Y5 if the log is meant to be operating time rather than command time.
- Latch any transition shorter than the tag's acquisition period in the PLC and transfer it as an event.
Questions asked
Why did the panel log the first press and nothing after that?
Operation Record writes an entry when a touch changes the recorded value. The first press changed it, so one entry appeared; later presses either left the value unchanged, or the value was being changed by the controller's logic through M167 and Y5, which a touch log never sees. In simulation the result is the same silence, because there is no controller toggling the bit.
Is an acknowledged alarm the same as a recorded stop?
No. Acknowledgment is an operator action, timestamped when someone presses ack, which can be minutes after the machine stopped; the stop record is Y5 returning to inactive. Configure acknowledgment as its own entry type so nobody reads an ack time as a stop time, and so a late ack cannot mask a missing return entry.
What if Y5 switches faster than the panel polls it?
The miss is silent: the tag reads steady and no error appears, because the HMI never sampled the bit while it was on. Prove it by forcing Y5 on for less than the tag's acquisition period; if nothing logs, latch each edge in the PLC with a one-shot and a retained pending flag, transfer the events, and clear them only after the history accepts them.
Should the history record Y5 or the machine's running feedback?
Y5 is the command; feedback proves motion. Record feedback when the log is operating-time accounting, and keep Y5 as the command history. The two differ exactly when it matters: an interlock can drop Y5 while the machine coasts, and a jam can stop motion with Y5 still on. Only the feedback record shows true run time.