Troubleshooting
Why One PLC to Two HMIs Fails Over RS-422 Multilink
Short answer
A 1:1 serial test proves one cable and one framing set; it cannot prove a shared n:1 multilink. Fix two HMIs on one RS-422 trunk by running the multilink driver entry on both panels, giving each a unique link identity, matching every serial parameter, and terminating only the trunk ends. The documented limit is 16 displays, so two panels are a configuration problem, never a capacity one.
On this page
A 1:1 test proves one cable and one framing set — it says nothing about station identity, shared-medium behavior or trunk termination. That is why two GP2501 panels that each poll a Q02H alone stop communicating the moment both sit on the same RS-422 trunk after an A1SJ71UC24-S8 is replaced by a QJ71C24N. On an n:1 multilink the fix is almost always in the HMI projects, not in the wire: both panels must run the multilink driver entry for the Q-series C24 module, carry different link identities, and match the PLC channel field for field, with termination at the two trunk ends only. The Pro-face documentation for this driver allows up to 16 displays on one n:1 link, so two panels are a configuration problem, never a capacity problem. If the driver list does not name your exact module suffix, stop and give each HMI its own channel.
What the 1:1 test proves #
A point-to-point test exercises one request path, one leg of cable and one set of serial parameters. It confirms that the panel frames a request the C24 channel accepts and reads the reply. It never tests the three things a shared link adds:
- arbitration between two masters, because each HMI initiates requests and the PLC channel only answers;
- station identity, because on a private pair there is nobody else to distinguish you from;
- the multidrop cable itself, which is electrically not two point-to-point cables joined together.
TIA/EIA-422 defines a balanced interface with one driver and up to ten receivers. Two panels that both transmit sit outside that model unless something sequences them, and sequencing the panels is exactly what the n:1 multilink driver mode exists to do. When both panels run the 1:1 driver entry, or share one link identity, their frames overlap and neither holds a stable poll.
The module suffix checkpoint #
The GP2000-series connection material for this driver family names QJ71C24 for multilink operation, and the Pro-face notes state plainly that the PLCs supporting n:1 connection are limited. The upgrade module is the QJ71C24N, a later revision with its own user's manual. Read the driver list literally: select the entry that names the exact module in the rack, or get written confirmation from the HMI vendor that the entries are interchangeable, before touching the cable. An hour on that question beats a week of termination experiments against a driver that was never polling your module.
Five checks, in order #
- Driver mode. Both HMI projects use the n:1 multilink entry of the Q/QnA serial driver — not the 1:1 entry that passed the point-to-point test. Re-download both projects afterwards; a mode that does not survive the download looks exactly like a wiring fault.
- Link identity. Each panel gets a unique station number in the multilink fields the driver exposes. Leave the PLC station number identical on both panels; only the HMI side differs.
- Serial parameters. Speed, data bits, parity and stop bits identical across the PLC channel and both panels. One mismatched field rejects every frame from one panel while the other still polls.
- Trunk topology. One trunk with controlled drops, transmit pair into receive pair, polarity consistent at every drop, signal reference continuous end to end.
- Termination. At the two electrical ends of the trunk only, by the method the connection diagram prescribes. Termination mid-trunk loads the drivers; no termination at the ends invites reflections.
Worked example: the line budget for two panels #
These are example inputs, not documented limits: 19,200 bit/s, 8 data bits, one stop bit, no parity, which is 10 bits per character.
- Character rate: 19,200 ÷ 10 = 1,920 characters per second.
- One poll: an 8-character request plus a 16-character response = 24 characters = 240 bits; 240 ÷ 19,200 = 12.5 ms.
- Add a 5 ms turnaround: 12.5 + 5 = 17.5 ms per transaction.
- Both panels polling every 100 ms: 2 × 17.5 = 35 ms of line time per 100 ms, or 35 % utilization.
The decision those numbers force: leave the baud rate alone. Doubling the response to 32 characters moves utilization to about 52 %, and doubling the speed to 38,400 bit/s drops it to about 23 %, so capacity is not what is breaking this link — identity, driver mode or wiring is.
Diagnostic sequence #
- Write down the CPU, the module suffix, the HMI model, the driver entry, the channel and the protocol mode; settle the QJ71C24 versus QJ71C24N question first.
- Keep the working 1:1 settings as the baseline and record every field, not just the ones you changed.
- Put HMI 1 alone on the complete trunk with its intended connector and termination, then repeat with HMI 2. If one fails alone, swap the panels at the same drop to separate a project fault from a cable fault.
- Connect both and classify the failure: both dead points at mode, addressing or termination; only one dead points at that drop or that project.
- Watch the module's status indications while each panel polls, and separate no-receive, framing and no-response conditions before changing anything. Resolving the 7F68 framing error on this module's RS-232 channels applies the same read-the-indications-first discipline to a sibling fault.
- Change one configuration group at a time and rerun the two-panel test; bulk changes hide intermittent collisions.
Trunk wiring and termination #
Trace every conductor end to end — wire color proves nothing about signal identity. The discipline is the same as building an RS-232 Modbus cable between an HMI and a controller: verify what each conductor carries before trusting the jacket. Build the network as one trunk with controlled drops; do not join two point-to-point cables into a star unless the approved diagram shows exactly that.
With power isolated as the equipment requires, check continuity, crossed pair members, shorts to shield and any unintended termination resistance. After power-up, disconnect one panel at a time: if removing either one restores the other, you are looking at duplicate identity, reversed polarity at a drop, or termination in the wrong place. The trunk rules involved are the ones covered from the operator side in connecting multiple HMIs to one PLC over a long serial run.
The fallback and the decision #
If the driver list names your module, the fixes above carry it: multilink mode on both panels, unique identities, matched framing, ends-only termination, with two panels well inside the documented maximum and a trunk running at roughly a third of capacity. If the vendor cannot confirm the QJ71C24N under that driver, stop tuning and split the link: give each GP2501 its own channel of the C24 module — or one module per panel — and let each leg keep the 1:1 configuration that already passed. The panels stay, the migration proceeds, and nothing on the floor depends on an unconfirmed driver entry.
The long-term route is Ethernet-capable panels, where each HMI holds its own connection and shared-medium arbitration stops existing as a failure mode; selecting a Modbus TCP HMI panel for an existing controller walks that path for the day the GP2501 units are retired.
Questions asked
Both HMIs worked at 1:1 with identical settings. Why not just connect both?
The PLC channel never initiates traffic; it only answers the panel that addressed it. With two panels configured identically there is nothing in the exchange that tells the replies apart, and both panels transmit on the same pair whenever their poll timers fire. Unique link identities plus the multilink driver entry give each panel a distinguishable address and a shared turn-taking rule.
How many HMIs can share one n:1 multilink link?
The Pro-face documentation for the Q/QnA serial driver sets the maximum at 16 displays per n:1 multilink connection. Two panels sit far inside that limit, and the worked example shows about 35 percent line utilization at 19,200 bit/s, so capacity is never the reason a two-panel link fails.
Is the QJ71C24N supported where the connection manual says QJ71C24?
Treat them as different until proven the same. The connection material for this driver family names QJ71C24, and the N revision carries its own user's manual. Also confirm the PLC side: the channel must be set to the protocol the HMI driver expects, because an RS-422 electrical selection alone does not choose the application protocol.
Where do termination resistors go on the multidrop trunk?
Fit them only at the trunk's two electrical ends, following the connection diagram's method, and leave every mid-trunk drop unterminated. The failure signatures differ: termination in the middle loads the drivers, while none at the ends shows up as communication that runs for minutes and then drops, which is reflection noise rather than a settings fault.
Would Ethernet-capable HMIs avoid this whole problem?
Yes. Each panel then holds its own connection to the controller, so station identity, shared-medium arbitration and trunk termination simply stop existing as failure modes, and adding a third panel needs no re-architecting of the link. The cost is replacing the GP2501 units, which have no Ethernet port, and rebuilding or converting their projects.