Troubleshooting
Trace an HMI command that never reaches the PLC output
Short answer
When the screen shows a command but the machine does not act, walk five boundaries in order: the screen object's tag binding, the communication connection and controller identity, the input interface, the ladder logic with its permissives, and the output interface. Because a PLC reads inputs once per scan and a scan is typically milliseconds, a brief command can be read and overwritten within one scan. The first boundary where indications disagree marks the fault.
On this page
When the HMI accepts a command and the machine does nothing, walk the command chain instead of guessing at the program. A screen indication and a physical output are different checkpoints, and the fault sits at the first boundary where the indications disagree. Check five boundaries in order: the screen object's tag binding, the communication connection and controller identity, the input interface, the ladder logic and its permissives, and the output interface. One timing fact explains the strange half of these cases: a conventional PLC reads its inputs and writes its outputs once per scan, and a scan is typically milliseconds, so a command shorter than the scan can be read and overwritten — or never read at all — inside one cycle.
What the screen indication actually represents #
First decide what the lit object claims to be. An operator command requests an action, an output command shows the controller's decision, and device feedback reports a measured state. A highlighted button proves only that the object drew itself lit.
Two binding faults live here. An object can animate from a local screen variable or an expression with no controller involved. And a correctly configured connection can carry the wrong controller variable. Neither will ever show up in the ladder.
| Setting | Where it lives | What it proves |
|---|---|---|
| Write binding | The object's command configuration | Which controller tag receives the press |
| Read or animation binding | The object's display configuration | What the indication actually reports |
| Connection | The driver configuration | Which controller supplies the data |
| Data type and representation | Tag configuration and controller declaration | Whether the received value decodes correctly |
| Data quality | Runtime communication diagnostics | Live value versus stale or unavailable data |
Watch the intended tag online in the programming tool while an authorized operator presses the button. This boundary passes when the press changes the intended tag, on the intended controller, with good quality. If the tag never moves, fix the binding or the connection before touching logic.
Confirm the connection and the controller identity #
Separate connection-wide failures from single-tag failures first. If every value on one connection shows bad quality or stops updating together, the fault is the connection: the configured destination, the communication path, the driver selection, or the controller's own communication diagnostics. The screen symptom of that state is why HMI values show question marks during comms timeouts.
If other tags on the same connection update normally and only one fails, the fault is local to that object: its binding, the variable's availability, its access permissions, or its data type.
Then check identity, not just connectivity. A connection to the wrong controller returns plausible, live values and can command the wrong machine — common where two identical lines share an address range. Read the identity from the installed controller, not from the configuration file.
The boundary passes when the connection carries good-quality data from the controller you can physically point at, and an authorized write reaches the intended tag. If the value changes and immediately snaps back, the connection is fine; keep walking.
Prove the input image matches the field #
When the command depends on a field condition — a guard closed, a part present — the screen shows the controller's input image, not the terminal. IEC 61131-2 defines the voltage ranges a digital input must recognize as ON and OFF, so a signal floating between the guaranteed-ON band and the guaranteed-OFF band is undefined for any conforming module: the wiring decides what the logic sees, not the program.
| What you observe | Inspect next | Fault boundary |
|---|---|---|
| Field device changes, module channel indication does not | Field supply, wiring, terminals, module input specification | Field circuit or input interface |
| Channel indication changes, controller input value does not | Module diagnostics and input mapping | I/O configuration or data assignment |
| Controller input changes, screen indication does not | Screen binding and data quality | Display path |
| Input appears briefly but logic misses it | Input update behavior, filtering, task interval | Signal duration versus scan |
Operate the field device through a controlled change and watch the chain in one pass: device, channel indication, controller input, screen. The first disagreement is the fault. For analog feedback, the same boundary method is how you detect a failed analog input channel and report it to SCADA.
Walk the ladder: permissives, rung order and the scan #
The logic can refuse the command for reasons that look alive on screen. Monitor the command tag and every permissive feeding the output rung: operating mode, interlocks, sequence state, fault bits, and the feedback from the previous section. A set command with one permissive off is a correct refusal, not a fault.
Execution order is the other trap. IEC 61131-3 standardizes the PLC programming languages, ladder diagram among them, and the task model — tasks with configured intervals and priorities — so on a modern controller the scan is a configured task, and the project's task settings decide when code runs. An instruction uses the values present when it executes, and a later rung can overwrite what an earlier rung computed; the same effect is why a CODESYS TON works in structured text but not ladder. Cross-reference every write to the output tag and its intermediates: the last write before the output image update wins.
Worked example with illustrative numbers. The HMI writes a start command as a momentary 1 held for 250 ms, and the task diagnostics report a 30 ms scan.
- 250 ms / 30 ms = 8.3, so the command is visible to at least 8 scans. Timing is not swallowing this command — read the permissives.
- The same machine has a sensor pulsing 25 ms per part. 25 ms < 30 ms, so that pulse can fall entirely between two input updates and never be read. That one is a timing fault.
- Decision rule: hold any pulsed command at least two scan times (2 x 30 ms = 60 ms here) or latch it in logic with a set/reset pair so a single scan catches it.
This boundary passes when the output tag in the controller goes to 1 on command and stays there as long as the conditions allow.
Finish at the output interface #
The last chain is output tag, module channel indication, terminal voltage, device. Watch the tag while the command is issued — do not force it.
- Tag at 1, channel indication off: the module or its configuration. Check channel protection or the electronic fuse, the output common, and whether the point is configured for the sink or source wiring actually installed.
- Channel indication on, terminal dead: wiring. Output circuits are often line voltage, so measure at the terminal with a meter rated for the circuit, machine in a safe state, and never force an output to test motion with anyone near the machine.
- Terminal live, device dead: the device side — coil, overload, its own fuse, mechanical jam.
This boundary passes when the terminal voltage changes with the command and the device itself operates.
The ordered checklist #
- Decide what the screen object claims to be: operator command, output command, or feedback.
- Confirm the write binding points at the intended tag on the intended controller.
- Watch that tag online while an authorized operator presses the button.
- Judge quality across the connection: every tag bad means connection, one tag bad means binding.
- Verify the online controller identity against the machine you are standing at.
- For each field permissive, follow device, channel indication, controller input, screen, and stop at the first disagreement.
- Cross-reference every write to the output tag; the last write before the output update wins.
- Compare command hold time with the measured scan time; hold at least two scans or latch in logic.
- Watch the output tag and channel indication together on command — never by forcing.
- Measure the terminal last, machine in a safe state.
Questions asked
How long must an HMI pulse command stay set for the PLC to see it?
At least one full input update period, and in practice two scan times for margin. With a 30 ms scan, hold the command 60 ms or longer (2 x 30 ms), or better, write to a set/reset pair and let the ladder clear it when the action completes. A command shorter than the scan can fall between two input updates and never be read at all.
The tag value changes and immediately returns to zero. Is the driver faulty?
Probably not. Something else is writing the same tag after the HMI does: a ladder rung later in the scan, a reset instruction, or a second client on the network. Cross-reference every write to that tag in the project and watch it online; the last write before the output image update is the one that wins.
The button lights up on screen. Does that prove the PLC received the command?
No. Many screen objects animate from a local state, a macro or an expression, and a write can be bound to a tag on the wrong controller. The proof is watching the intended tag change in the programming tool while online with that controller. If the tag never moves, the fault is the binding or the connection, not the logic.
How do I tell a connection-wide fault from one bad tag?
Look at quality across the whole connection. If every tag on that connection shows bad quality or stops updating together, troubleshoot the path, driver and controller diagnostics first. If other tags on the same connection update normally and only one fails, the fault is in that object's binding, data type or access permissions.
Can a slow scan hide a genuine field input pulse?
Yes. When the input image updates once per scan, a pulse shorter than the scan time can sit entirely between two updates and never be seen by the logic. Fix it at the source: latch the signal with a seal-in circuit, route it to a high-speed or event-driven input, or shorten the task interval.