Troubleshooting
Why HMI Indicators Stay Off Over Modbus RTU Polling
Short answer
Indicators on internal tags animate from HMI memory with no serial traffic; an indicator bound to a PLC tag needs a completed Modbus RTU read of the right address in the right data area. Confirm the PLC bit changes, then read the request and response, then fix the address at one known point.
On this page
- Why do internal indicators switch while PLC-bound indicators stay off?
- Why do PLC indicators sit dark in simulation?
- How do I confirm the PLC bit is really changing?
- What should the Modbus RTU request and response show?
- Does a Mod Start/End range prove the tag is mapped?
- Which Modbus data area does the PLC expose the bit in?
- Which serial settings and wiring must match at both ends?
- How do I prove the fix end to end?
- Questions asked
An indicator wired to an internal HMI tag toggles with no PLC in the room; an indicator on a PLC tag lights only after a Modbus RTU read completes and returns the right bit. When internal objects animate and PLC-bound objects sit dark, the screen object itself is fine and the failure is in one of three places: the PLC bit, the serial exchange, or the address mapping between the two. Work them in that order. The checks below separate the three, the addressing arithmetic catches off-by-one and wrong-data-area errors, and the end-to-end test closes the fault.
Why do internal indicators switch while PLC-bound indicators stay off? #
An internal tag lives in the HMI's own memory and animates with no serial traffic, while a PLC-bound indicator also needs a completed Modbus RTU poll of the right address, so the fault sits in the request, the mapping, or the data area.
A tag whose storage is inside the panel toggles the moment a screen button or script writes to it, controller or no controller. A tag bound to PLC data receives its value only in the data field of a response frame: the Modbus application protocol specification defines a request and a reply for every transaction, with exactly one master initiating and servers only answering. Nothing is pushed, so a dark indicator means one of four states: no request sent, no reply, an exception reply, or a good reply carrying another location's data. Local rendering is also what keeps screen motion fluid regardless of the controller scan, the mechanism behind how moving analog indicators animate in HMIs.
Why do PLC indicators sit dark in simulation? #
Simulation animates internal objects from the HMI's local memory and drives PLC-bound tags only when the simulator is actually exchanging frames with the controller, so PLC indicators that sit dark in simulation mean the poll is not reaching the PLC.
Most development packages can run the project on the programming PC, and that run animates internal objects every time. Whether it also polls the controller depends on the simulator's connection setup: pointed at the physical serial port it exchanges real frames; left on a software controller or no connection, PLC-bound tags hold their initial values. Confirm what the simulator is connected to before drawing conclusions, and treat the result as open until the same project runs in the panel wired to the PLC. The setup for a real-controller run is covered in testing an HMI simulation against a live PLC.
How do I confirm the PLC bit is really changing? #
Watch the exact tag in the PLC's online monitor while you cycle its source condition, because if the bit itself never changes, the indicator is correctly reporting a dead tag and no HMI setting will fix it.
If the monitored value does not change, stop here: the fault is in the logic, the input, or the process condition, and the panel is reporting honestly. If it changes, record the tag name, its assigned Modbus address, and its data type; that recorded pair is what every later check compares against.
For the cycling test, use a diagnostic bit rather than an output bit: forcing an output can move the machine.
What should the Modbus RTU request and response show? #
The HMI, as master, must send function-code reads to the PLC's node address in the 1-247 range, and the PLC must answer with data or an exception, so the three failure patterns below each point to a different layer.
The Modbus specification reserves address 0 for broadcast and 248-255 for reserved use, leaving 1-247 for individually addressed servers; a node address outside that span can never be polled alone. Reads use function 01 for coils, 02 for discrete inputs, 03 for holding registers, and 04 for input registers, and every RTU frame ends in a 16-bit CRC. A frame that fails the CRC check is discarded silently, which shows on screen as no response rather than an error message. Read the exchange from the HMI's communication diagnostics and from the PLC's port counters where they exist:
| What you read | What it means | Next check |
|---|---|---|
| No request leaves the HMI | The runtime is not using that driver, port, or tag | Confirm the active connection and the indicator's tag binding |
| Requests out, nothing back | The PLC is not answering that port | Confirm the server role and node address, then the serial settings and cable |
| Exception or unexpected value in the reply | The exchange works; the area, address, or type is wrong | Compare the PLC's map entry against the HMI tag's area, number, and type |
Compare the values the loaded runtime reports, not the offline project: a port can be configured correctly in the file while the running panel uses another one. If the indicators do update but the screen stalls between refreshes, the fault is poll load and retries rather than mapping, covered in why an HMI freezes or updates slowly over serial.
Does a Mod Start/End range prove the tag is mapped? #
No: a Mod Start/End or First/Last Modbus Address range only sets the window of addresses the PLC exposes, and an indicator still fails when its tag has no assigned address inside that window or points at the wrong entry.
The range bounds what the port serves and where serving starts; it does not assign your tag a slot, and it does not tell the panel which slot inside the window is yours. Open the PLC's Modbus map, find the tag by name, and read its assigned address and data area. A tag with no entry in the map is simply not on the wire, and no HMI setting will find it.
Where the serial port itself is configured as the Modbus RTU server, the port answers requests from the mapped data on its own; ladder Modbus read and write instructions serve ports acting as a client, and adding them will not light a dark indicator.
Which Modbus data area does the PLC expose the bit in? #
Modbus serves booleans from coils or discrete inputs and 16-bit values from holding or input registers, so an HMI tag reading the wrong area gets either an exception or another variable's data at the same number.
| Data area | Access | Carries | Reference form |
|---|---|---|---|
| Coils | Read/write | Single bits | 0xxxx, first entry 00001 |
| Discrete inputs | Read-only | Single bits | 1xxxx, first entry 10001 |
| Input registers | Read-only | 16-bit values | 3xxxx, first entry 30001 |
| Holding registers | Read/write | 16-bit values | 4xxxx, first entry 40001 |
The five-digit forms are display conventions. On the wire the frame carries a zero-based offset, so 00001 is offset 0 in the coil area and 40001 is offset 0 in the holding-register area. A PLC boolean may be served as an individual coil or packed as one bit of a 16-bit holding register, depending on the map, and the HMI tag must match the area, the number, and the bit extraction. An imported tag name is not an address: the binding carries area, number, and type, or it carries nothing.
Worked example. The PLC map lists Pump1_Run at coil 00005. A read-coils request, function 01, must carry starting address 00005 - 1 = 4, written 0x0004 in the frame. The HMI project is set to zero-based entry and the tag was typed as 5, so the frame carries 0x0005, which is coil 00006, the neighboring bit, and the indicator waits on a bit that never turns on with the pump. Enter 4 in zero-based mode or 00005 in reference mode; both put 0x0004 on the wire. Registers follow the same arithmetic: holding register 40010 travels as 40010 - 40001 = 9, and the same digits under the input-register prefix, 30010 - 30001 = 9, address a different area, another variable, or nothing mapped, which returns exception 02.
Once one known point reads correctly, fix the convention once and apply it across the project; walking the whole database by trial offsets hides a systematic error behind dozens of edits.
Which serial settings and wiring must match at both ends? #
Both devices must run identical baud rate, parity, data bits and stop bits on the connected ports, with the HMI polling the PLC's configured node address, verified in the loaded runtime rather than the offline project.
The PLC must be the server at its configured node address, and the HMI's station list must target that same number. Read each value where the running devices report it, because offline copies drift from what is actually loaded.
On RS-232, each device's transmit lands on the other's receive with a common signal ground; confirm both pinouts in the port manuals instead of trusting connector shape. A pin-by-pin serial example on a related protocol is in DH-485 wiring for an HMI replacement: the RS-232 pin map. If requests leave and nothing returns after settings and node address check out, the remaining suspects are the cable and the signal reference; isolate them with the test the equipment vendor documents for that port.
How do I prove the fix end to end? #
Correct the single worst mapping error, reload both projects, then prove the same point through four checks in order, because an indicator is fixed only when it follows the PLC bit through both transitions on the running machine.
- In the PLC online monitor, watch the tag step when its condition activates and clear when it drops.
- In the PLC Modbus map, confirm the entry's address and data area are exactly what the HMI tag requests.
- On the loaded panel runtime, open the tag's live value and watch it follow the same two transitions.
- On screen, confirm the indicator follows both the on and the off transition, then remove temporary diagnostic displays and test bindings.
Several indicators failing together point at the shared layer, the connection, driver, node address, or the served window, so repair that before touching single objects. One indicator failing alone points at that binding's area, number, or type. An internal bit left standing in for PLC status shows a condition the controller never had, so it does not belong on a production screen.
Questions asked
What do Modbus exception codes 01, 02 and 03 mean on this link?
Code 01 means the server does not support the function code the HMI sent, 02 means no coil or register is mapped at the requested address, and 03 means a written value is outside the range the server accepts. Code 02 is the one that points at address mapping: correct the map entry or the requested address and retest one known point.
Can I attach a second Modbus master to the same link while testing?
No. A Modbus RTU serial link carries one master, and a second master's requests interleave with the HMI's polls so both devices see corrupted exchanges. Diagnose with the one real master plus the port diagnostics in the PLC and the HMI driver, and add a protocol analyzer only as a passive tap where the wiring supports one.
Is there a fast way to tell whether the simulator is polling the real PLC?
Unplug the serial cable and watch the screen. Internal objects keep animating because they never touch the link, while a working PLC-bound indicator freezes or flags a communication error the moment the cable comes out. If nothing on the PLC-bound side reacts to the cable, the simulator was never exchanging frames with the controller.
Does importing the PLC tag database into the HMI project guarantee the addresses?
No. An import only carries the addresses that were in the export, and a name list exported without the Modbus map leaves the HMI side pointing at defaults or empty slots. After any import, verify one known tag against the PLC's online Modbus map before trusting the rest of the database.