Why two HMI panels fail on one PLC serial port

Which of the two layers is failing

Every symptom in a job like this traces to one of two layers, and the two need different fixes.

The physical layer. The RS-232 guideline behind that arithmetic is a capacitance rule set at 19,200 baud. It stretches on low-capacitance cable and shrinks as the baud rate rises. A run of that length sits past it at any speed, so the link turns marginal and errors climb with length. RS-422 exists for exactly this job. The interface standard allows about 4,000 ft (1,200 m) of cable at data rates up to 100 kbps, one driver per pair and up to 10 receivers. That trades speed for length, down to roughly 40 ft at 10 Mbps. At the line speed, the run is a small fraction of that length, so a converter pair at each end of the link clears the distance with margin.

The protocol layer. DF1 on a PLC serial channel is built for one initiating station on the link. The channel answers requests; it has no mechanism to arbitrate between two initiators. When both panels transmit, transactions interleave and each side discards corrupted frames. RS-422 conversion leaves that count alone: the standard gives the pair one driver, which is one transmitter slot, still taken by whichever panel speaks.

Two masters on one bus fail the same way in other protocols too; the Modbus case, with the same cure, is in two Modbus RTU masters on one RS-485 bus: why and the fix. DH-485 is the multi-master counterpart. Stations pass a token and only the holder transmits, so several master stations share one network in turn. Vendor network documentation puts the limit at 32 nodes at 19.2 kbps. Each node then needs its own station address; the addressing setup is in DH-485 network node addressing for several panels and PLCs.

The pure distance side of these choices is tabulated in HMI to PLC serial cable maximum distance by interface, and the DF1 RS-232 case in particular in how far can an HMI be from the PLC on DF1 RS-232.

Symptoms and what each points to

• Worked on the bench, fails on the long run with one panel connected. Physical layer. The run is past the RS-232 guideline; plan an interface change, not a settings hunt.
• One panel works on the long run after conversion, both together fail or garble. Protocol layer. Two initiating stations on a channel built for one.
• Panels work one at a time, fail together even on a short cable. Same protocol problem; distance was never part of this failure.
• Nothing works with the converters wired in, even one panel. Converter wiring. Check each link's pair assignment, the signal reference between the two ends, and any termination or biasing the converter manual calls for. A swapped pair or a missing common kills a link with no error message.
• Random dropouts that worsen when both panels poll hard. Interleaved transactions with no arbitration. Slower polling hides the symptom; the network change removes it.

Example: two panels, one PLC port, 150 ft at 19.2 kbps

Inputs, with where each comes from:

• Run length: 150 ft, from the job's plan.
• Port protocol and speed: DF1 at 19,200 baud, the speed the channel and both panels already share.
• RS-232 guideline: about 50 ft (15 m) at 19,200 baud, the accepted capacitance-based rule.
• RS-422: about 4,000 ft (1,200 m) at up to 100 kbps, one driver per pair, from the interface standard.
• DH-485: up to 32 nodes at 19.2 kbps, from the network documentation.

The rule, in words: an interface carries the run when the length sits inside its guideline at the line speed. A protocol carries the stations when the number of initiating stations sits inside what it arbitrates.

The arithmetic:

1. RS-232: 150 ft ÷ 50 ft = 3.0. The run is three times the guideline, so plain RS-232 at 19,200 baud is out.
2. RS-422: 150 ft ÷ 4,000 ft = 0.0375, so the run uses 3.75% of the standard's length, and 19.2 kbps ÷ 100 kbps = 19.2% of the rate that length is rated for. A converter pair clears the distance.
3. Masters on DF1: the channel serves 1 initiating station; 2 panels is 1 over. No converter changes that count.
4. DH-485 nodes: controller + 2 panels = 3 nodes, against a 32-node limit. That leaves 29 spare addresses for a future panel, a text display or a supervisory station.

Result: RS-422 at both ends makes the installed run electrically sound, and the link still fails with both panels attached, because step 3 never moved. Decision: keep DF1 and give each panel its own port, or move all three stations onto DH-485 at 19.2 kbps.

Checks before you rewire

1. Write down the PLC serial channel settings and each panel's driver settings: protocol, baud, station address, data format. Change one value at a time from there.
2. Take the second panel off the link. Bring one panel up on a short cable, then on the installed run through the converters. Testing two variables at once is how a wiring fault hides as a protocol error.
3. Read the splitter's documentation before trusting it. The deciding question is whether it presents two independent ports, two separate networks, or joins both drop cables onto one link; the second arrangement solves nothing. The architecture question behind that choice is walked through in connect two operator panels to one PLC with no plant network.
4. Wire each converter pair exactly as its manual shows: transmit and receive pair assignment, the signal reference between the two ends, and the termination or biasing it calls for. Never tie two converters' outputs together onto one pair.
5. Keep a saved copy of the working short-cable, single-panel configuration so you can return to it at any point.

The three working architectures

The controller maker's DH-485 documentation names the interface module for its port. The panel's documentation names the option the panel takes for the network; match both to each station's role before ordering. The master-PLC arrangement is drawn out for panels polling many controllers in two HMI panels polling many PLCs: use a master PLC; the same design carries data between several stations on one link here.

Recommendation: when both panels must live on one PLC serial channel and the run sits past the RS-232 guideline at the line speed, put the whole link on a token-passing network such as DH-485 at 19.2 kbps. That means one interface module on the controller port and one at each panel: three nodes with spare addresses to grow into. Take the splitter route only when its documentation states two independent ports. Take the master-PLC route when the panels come from different protocol families.

Restore service in stages

1. Put the known-good short-cable, one-panel link back as a temporary restore, but only if the machine can run safely while operators work from that single view.
2. Bring up the chosen architecture with one panel attached. Verify reads and writes of representative data, then watch the communication error counters on both ends for 30 minutes of polling and make sure they hold at zero.
3. Add the second panel. Exercise both at once: write values from each panel in turn while the other polls.
4. If PLC-triggered messages start dropping once both panels poll, the remaining work is in the message and poll arrangement, which fix dropped PLC-triggered messages on a serial HMI panel covers step by step.

Questions