Troubleshooting
Fixing HMI Modbus Device No Response Over RS-485
Short answer
Device no Response means the slave never returned a valid frame to that request. When a PC polling tool reads the same scanner, the bus is probably carrying two masters: run one master at a time, then copy the known-good read's serial settings, station address, and register offset into the HMI.
On this page
- Why does the HMI say Device no Response when the PC can read the sensor?
- How do I test with only one master on the bus?
- How do I check the RS-485 wiring and polarity at the HMI port?
- Which serial settings have to match between the HMI and the scanner?
- What number do I enter for register 40001 in the HMI driver?
- When should I increase the response timeout?
- Why does the temperature value look wrong once the read works?
- Questions asked
Device no Response on a numeric object means the panel transmitted a Modbus RTU request and nothing valid came back. When a PC polling tool reads the same scanner, the slave, its wiring, and its register map all work, and the difference is what is doing the asking. A Modbus RTU bus runs exactly one master at a time, so a PC tool and an HMI sharing one pair collide and each can silence the other. Walk the fault in the order you meet it: prove the read with one master, verify polarity, match the frame, enter the register as an offset, and touch the timeout last.
Why does the HMI say Device no Response when the PC can read the sensor? #
Because two Modbus RTU masters are on one RS-485 pair: the Modbus over Serial Line specification allows only one master on the bus at a time, and the PC tool's requests collide with the panel's until you disconnect one of them.
In the installation behind this article, one PC polling tool read the scanner's register while a second scanning tool and the HMI both reported no response. With only one master connected, either side could fetch the data. That signature — the device answers somebody, just not everybody — is a bus-contention fault until proven otherwise.
RS-485 lets many devices share the pair electrically: the specification guarantees 32 connected devices, and a multidrop bus can carry up to 127 slaves. Electrical sharing is not protocol sharing. The master schedules every transaction as request, silent gap, response, and the slave answers only a frame that arrives intact. Two independent pollers interleave their bytes on the same pair, both CRC checks fail, and each master sees silence or garbage. Whichever polls last or most often appears to work.
A message that appears while building or downloading the project exercises the same device link, so treat it as a communication fault, not a project error. Record the software's own build diagnostics separately from the runtime behavior.
| What you observe | Where the fault lives | Next move |
|---|---|---|
| PC reads the scanner alone; HMI alone gets no response | The panel's port, its wiring, or its settings | Copy the logged PC transaction into the panel |
| Both connected; only one side works | Two masters on one bus | Disconnect one and retest |
| PC and HMI both fail when tested alone | Scanner power, common wiring, wrong target | Check scanner status and the physical link |
| A response arrives but the value is wrong | Register choice, word format, scaling | Read the raw register before changing comms |
Your first move is fixed: take every other master off the pair and prove one clean transaction.
How do I test with only one master on the bus? #
Disconnect every other polling device, repeat the known-good PC read, and save one full logged transaction — that frame is the reference the panel has to reproduce before anything else is worth changing.
- With the HMI off the pair, run the PC tool and read the temperature register. Write down baud rate, data bits, parity, stop bits, station address, register category, the address exactly as entered, quantity, and the raw value returned.
- Save the transaction log if the tool has one. A logged request such as
01 03 00 00 00 01tells you station 1, function 03, offset 0, one register, with no guessing about conventions. - Disconnect the PC converter from the two terminals. Unplug the conductors, not just close the window, so nothing on the pair can transmit.
- Land the panel on the same two terminals with polarity observed, configure one numeric object to the identical address, download, and watch that object only.
- If the panel reads, add display logic back one element at a time. If it does not while the PC read repeated cleanly, the fault is in the panel's port path — port assignment, adapter, or connector — and no register number will fix it.
Copying the polled value into a panel-internal register for screen logic is fine, but it comes after the poll works. If you later add slaves, the wiring discipline is the one behind connecting multiple HMIs to one PLC over 200 ft of serial.
How do I check the RS-485 wiring and polarity at the HMI port? #
Trace the pair conductor by conductor and confirm polarity at both ends from each device's documented signal names, because terminal letters such as + and − are not standardized between a scanner and a panel connector.
In the installation that produced this fault, the pair landed on COM3 with pin 9 as the positive line and pin 6 as the negative line. Treat that as the description of one working job, and verify your own panel's connector drawing before moving any conductor — pin assignments do not transfer between models.
Put the machine in a state where a disturbed input is safe before you touch the terminals; a scanner you unplug reads as a failed sensor to whatever is watching it. Then work the checklist:
- Confirm the port you configured in software is the port the conductors physically reach.
- Check connector seating, screw-terminal torque, and end-to-end continuity of each conductor.
- With the bus powered and idle, measure DC volts between the two lines: the conductor sitting at least +200 mV above the other is the non-inverting line, so match it to the port's documented positive pin when labels disagree.
- Terminate the two physical ends of the pair with a 120 Ω resistor each — one pair of resistors on the balanced pair, not one at every drop — and set biasing per the equipment instructions.
A reversed pair or a wrong pin prevents a valid response from ever reaching the panel. Changing 40001 to a neighboring number cannot repair that.
Which serial settings have to match between the HMI and the scanner? #
All five must match: baud rate, data bits, parity, stop bits, and station address — start from 9600 baud, 8 data bits, even parity, 1 stop bit, the combination this scanner was found on, then confirm every value against the scanner's own configuration menu.
RTU's standard frame is 8 data bits with even parity, and 9600 baud with one stop bit is the most common factory setting, but the scanner's configuration owns the numbers, not the sensor datasheet. A two-wire Pt100 describes copper between the sensor and the scanner; it says nothing about the serial port. Your polling target is the scanner, and its manual owns the register map too.
The station address must match exactly, and never use 0 — address 0 is broadcast, and a slave does not answer a broadcast.
Change one variable per test. If you flip polarity, address, and offset together and it starts working, you have learned nothing reusable on the next machine. Before you widen a block read, check the scanner's permitted request size: compact I/O devices often answer only a few registers per request, so start with one register and widen.
What number do I enter for register 40001 in the HMI driver? #
Enter 0, not 40001, when the driver asks for the register offset: in the 4xxxx reference convention, 40001 is the first holding register and its protocol offset is 0, so the offset is always the reference minus 40001.
Two numbering conventions describe one memory location. The six-digit reference names the register for humans; the request frame carries the 0-based offset. Drivers differ — some take the full reference, others take the register category separately plus a plain offset — and scanner manuals often print both as columns.
The arithmetic for this job: the temperature value sits at 40001, so the offset is 40001 − 40001 = 0. If the map later puts a second value at 40113, the offset is 40113 − 40001 = 112. The category (holding registers, function 03) is selected as a type in the driver, not typed into the number.
Enter 40001 into an offset field and you ask for the register 40001 places past the first one — reference 80002 — far outside a small scanner's map, and the panel reports the same message whether the slave answers with exception 02 or nothing at all. The logged PC request settles the question in one look: the two address bytes in that frame are the offset, in hex.
When should I increase the response timeout? #
Almost never at 9600 baud on a bench-length bus: one register read takes about 25 ms of wire time, so a 1000 ms timeout carries roughly 40 times margin — lengthen it only when a captured transaction shows a valid response arriving late.
Work the timing for one read at 9600 baud, 8 data bits, even parity, 1 stop bit:
- Each character is 11 bits: 1 start + 8 data + 1 parity + 1 stop. One character takes 11 ÷ 9600 = 1.15 ms.
- RTU separates frames with a silent interval of 3.5 character times, about 4.0 ms.
- A function-03 request for one register is 8 characters: 8 × 1.15 ms ≈ 9.2 ms.
- The response with one register is 5 + 2 × 1 = 7 characters ≈ 8.0 ms.
- Total: 9.2 + 8.0 + 2 × 4.0 ≈ 25 ms, before the scanner's internal turnaround.
Against that, 1000 ms is enormous margin. Cable distance does not move the number either — propagation costs about 1 µs per 200 m. What actually eats time is repeaters, slew-limited transceivers on very long runs (the reason long-run installation guides advise a slightly longer master timeout), or a slave whose firmware answers slowly. Slaves answer from memory with the last converted value, so the sensor's own update rate is not your delay either.
Raise the timeout only on evidence from a captured exchange. A longer timeout cannot repair a wrong address, reversed polarity, or a colliding master; it only makes every failed attempt slower. Slow-but-working updates are a different fault, covered in why an HMI freezes or updates slowly over serial.
Why does the temperature value look wrong once the read works? #
Because the communication layer is now fine and the fault has moved to interpretation — read the raw register first and choose between word length, word order, and scaling before touching anything else in the project.
With a 16-bit signed integer and one decimal place in the display object, a raw 250 shows as 25.0 °C. See 250.0 instead and the format needs one decimal; see nonsense and the format is not the problem.
Word order only matters for 32-bit values. A floating-point temperature held in two registers can arrive high-word-first or low-word-first, and the symptom is a value that is absurd rather than merely mis-scaled — swap the word order in the driver and the number snaps into range.
Use the sensor as the test signal: breathe on the Pt100. A value that tracks but sits off by tens or a factor of ten is scaling; a frozen value is a stale register; a value jumping between plausible and absurd is word order. Once the link is healthy, an object that stops updating points at the polling logic rather than the wire — the pattern behind why HMI indicators stay off over Modbus RTU polling.
Keep reading
Questions asked
Does a successful project download over Ethernet prove the RS-485 port is good?
No. The Ethernet link between the panel and your PC is a separate interface; downloading the project confirms the application reached the panel, not that its serial port can poll the scanner. The download can succeed every time while COM3 sits miswired or misconfigured, so test the serial path on its own.
Can I leave a second PC tool connected as a listener while the HMI polls?
Only if it truly never transmits. A tool that polls, logs by requesting, or refreshes periodically is a master, and its requests will collide with the panel's on the same pair. A passive capture that only watches the pair is safe. Close or disconnect anything that sends, then retest the panel alone.
Why does the scanner manual list both 40001 and 0 for the same register?
They are two notations for one location. 40001 is the six-digit reference convention, where the leading 4 means holding register and 001 means the first one; 0 is the protocol offset carried in the request frame. Some manuals print them as separate columns, and a driver that asks for an offset wants the second column.
How many slaves can share the bus before I need a repeater?
The Modbus specification guarantees 32 connected devices, master included, and a multidrop bus can be built out to 127 slave addresses. Past that, or past the cable length your wiring rules allow, add a repeater or an isolated segment rather than pushing one pair further. One master per segment still applies.
Do the HMI and the scanner need a signal ground connection?
Yes, in almost every retrofit. RS-485 receivers hold a common-mode limit of roughly −7 V to +12 V between the two device grounds. Without a signal-common conductor the pair can work on the bench and fail on the machine, so run the documented common unless both ports are specified as isolated.