Troubleshooting
Why HMI Alarm Beeps Sound Irregular Over Serial PLC Links
Short answer
The panel samples the toggle bit once per serial request round, and the PLC services one request per scan, so the round takes about request groups × scan time. When that exceeds the 100 ms toggle interval, transitions are missed and the beep stutters. Move dependable timing to the panel's own buzzer or a PLC-driven physical buzzer.
On this page
The beep is irregular because the panel never hears the flasher bit live. Over a serial link the PLC services one read request per scan, so the panel refreshes the alarm bit once per request round — and when that round runs longer than the toggle interval, the panel skips transitions. Estimate the round as request groups × PLC scan time. In the documented case the PLC toggled the alarm bit every 100 ms, so any request round past that mark makes the beeps bunch and gap while the link itself stays healthy. The fix is a decision, not a repair: keep the uneven beep if operators accept it, and when timing must be dependable, drive the panel's own buzzer from a steady alarm bit or drive a physical buzzer from a PLC output.
Why the panel misses toggle transitions #
The PLC side is regular: a counter-done condition qualifies the alarm, and a fixed-cycle system relay toggles a spare bit — in the documented case, system relay SC5 toggling bit DS13 every 100 ms, giving the bit a 200 ms on-off period.
The beep fires when the panel's poll reads the bit as high or as changed, and a poll is a sample, not a wire. Between two consecutive reads the bit may have toggled twice, and the panel cannot know; it compares snapshots. Once the request round reaches the toggle interval, an entire on-phase can fall between two samples, and whether consecutive reads differ depends on where each sample lands in the square wave. Beeps then merge, drop, and drift with every change in poll timing. That is sampling behavior on a healthy link — not a failing buzzer, a bad cable, or a slow PLC.
Estimate the panel's request cycle #
Treat this figure as a lowball estimate; wire time, retries, and panel processing add on top:
Request-cycle time ≈ request groups × PLC scan time
The basis: the panel groups its reads by contiguous register ranges, every gap in the address map starts a new group, and background work — the alarm table, trends, the event log — adds groups even when the visible screen is simple. Read the scan time from the programming software's scan-time display; scans of a few milliseconds are typical for compact PLCs.
Worked example with illustrative numbers:
- PLC scan time: 5 ms, taken from the scan-time display.
- Request groups: 30, from a screen with scattered addresses plus alarm and event logging.
- Request-cycle time ≈ 30 × 5 ms = 150 ms.
- Toggle interval: 100 ms.
- 150 ms > 100 ms, so an on-phase of 100 ms can fall entirely between two samples.
Decision from those numbers: no 100 ms flasher will ever sound regular on this link. Either stretch the flasher so each on-time is at least three times the request-cycle estimate — 3 × 150 ms = 450 ms, in practice a 0.5 s on / 0.5 s off pattern — or take the beep off the link.
For contrast, 20 groups on a 2 ms scan gives 20 × 2 ms = 40 ms, comfortably inside the interval: every on-phase is sampled at least twice and the beep holds a steady one-beat-per-200 ms rhythm. The product moves the answer, not the baud rate — a faster wire shortens each frame but does not reduce the number of frames or the scan-by-scan servicing. When the same overload appears as crawling numbers instead of stuttering sound, read why an HMI freezes or updates slowly over serial; on Ethernet the wire is fast yet request count still governs, which is why HMI point updates run slow on an Ethernet PLC link.
Why screen navigation changes the beep #
Each screen carries its own tag set, so the poll round changes with the display:
- Paging to the next or previous screen swaps the foreground request groups while the background groups keep running.
- In the documented case, navigating with the F4 and F5 keys made the alarm beep faster and less even — the signature of a changing request round, not of a defect.
- A screen change can also activate or idle background features, adding or removing groups from the round.
Use the behavior as a diagnostic. Watch the toggle bit in the PLC while someone pages through the screens: if the bit holds its interval and only the sound changes, the PLC is fine — stop debugging it and go back to the request-cycle estimate. Never diagnose PLC timing by ear; trend the bit.
Fixed-cycle system relays are not a tuning knob #
System control relays run at fixed cycle durations you cannot adjust — SC4, SC5, and SC6 in the documented case. Their recorded results:
- SC4, the fastest: the sampled pattern degenerated into a Morse-code-like jitter.
- SC5, toggling DS13 every 100 ms: the alarm worked, the cadence stayed uneven.
- SC6: too slow to read as an alarm.
Swapping among fixed relays is not calibration. When you need a chosen pattern, generate it in logic — the documented approach was a drum instruction stepping a dedicated flasher bit. That controls the PLC-side sequence, but it does not synchronize the panel's sampling; over a long request round the panel still samples it unevenly. Treat the bit as a sound command traveling a variable path.
Where dependable beeping lives #
Three placements, two of them immune to the poll round:
- The panel's own alarm buzzer. Trigger the panel's alarm system from a steady alarm-active bit — the count-reached state latched until reset — never from the flasher. The panel then generates its audio cadence locally, on its own clock. If the panel also lets PLC logic write to its buzzer directly, check the documentation for how that interacts with the alarm object before combining the two.
- A PLC output driving a physical buzzer. The cadence comes from PLC timers and exists whether or not the panel is listening. This is the documented fallback and the right call when the sound must be loud or exact; it costs an output point.
- A slowed flasher. With no buzzer available, the 0.5 s on / 0.5 s off pattern from the worked example stays legible over a 150 ms request round.
If operators rely on this beep, confirm the replacement is audible at the machine before removing the old indication. And if the panel's speaker clicks but never sounds, that is a different fault — start with why HMI alarm speakers click but audible alarms stay silent.
The decision, case by case #
| What you observe | Decision |
|---|---|
| The bit toggles on schedule in the PLC, the beep is uneven, reset clears both | The link is healthy. Keep the alarm if the sound is acceptable; otherwise move the beep to the panel's buzzer or a PLC output. |
| The beep cadence changes when screens change | Request-load variation. Pack tags into contiguous blocks and trim per-screen reads; do not chase relay swaps. |
| The request-cycle estimate exceeds the toggle interval | No flasher at that interval can sound regular. Stretch the on-time to at least three times the estimate, or take the beep off the link. |
| The toggle bit itself chatters | Fix the trigger logic first; a panel-side change only masks it. |
| The cadence must be exact, or the alarm must be loud | A PLC output drives a physical buzzer; timing comes from PLC logic, not from the poll round. |
Questions asked
Does raising the serial baud rate make the beep regular?
Usually not. Baud rate sets the wire time per frame, but the round-trip is dominated by request groups multiplied by PLC scan time, because the PLC services one request per scan. Shorter wire time helps only when each frame is long; with many small, scattered reads the scan-by-scan servicing dominates. If you raise the baud and the cadence does not change, the request-cycle estimate told you why.
How do I find how many request groups my panel uses?
Open the HMI project and count what it polls: every tag on the active screen, plus background data such as the alarm table, trends, and the event log. Registers in one contiguous block form a single group; every jump in the address map starts another. If your software shows a request or transaction list, read the count from there instead of estimating.
Why did the fastest fixed relay sound like Morse code?
Aliasing. When the toggle interval is at or below the request cycle, consecutive samples land on shifting phases of the square wave, so the panel reports a pseudo-random sequence of edges instead of a steady rhythm. The shorter the flasher interval relative to the poll round, the more random it sounds, which is exactly what the fastest fixed relay produced in the documented case.
Is an irregular beep a sign the serial link is failing?
No. A failing serial link reports timeouts, garbage values, or frozen screens. An uneven beep with correct values and a clean reset is normal behavior for a fast flasher riding a loaded poll round. Treat the sound alone as a fault only when the panel logs communication errors at the same time.
Can I trigger the panel buzzer from the flasher bit instead?
You can, but you defeat the purpose: the panel sounds its buzzer each time the poll reads the bit, so the cadence still follows the request round. Trigger the buzzer from a steady alarm-active bit that stays set until reset, and let the panel time its own sound.