Troubleshooting
Why an HMI Freezes or Updates Slowly Over Serial
Short answer
A panel-wide freeze or slow updates over serial is a link fault — driver, PLC port, wiring, or startup timing — not a screen-object fault. Test the two field-proven fixes first: a 10-second startup delay before the first poll, and a port set to four-wire RS-422 at 19,200 baud. Fix an isolated stale tag separately, by address and data type.
On this page
- Is the whole panel frozen or is it one bad tag?
- Why does the HMI hang only when the machine powers up?
- Which PLC port should carry the serial link?
- How should four-wire RS-422 and RTS/CTS be wired?
- Can the HMI's diagnostic requests fault the PLC CPU?
- How slow is 19,200 baud for a full screen refresh?
- What should I check before calling the link fixed?
- Questions asked
An operator panel that freezes or crawls over a serial link is complaining about the link, not about your screen objects. Two field-reported fixes settle most of these cases: a 10-second delay between HMI startup and the first poll, and a PLC port set to four-wire RS-422 at 19,200 baud. Work those before you open the screen editor. The sections below separate whole-connection faults from single bad tags, then cover port selection, wiring, protocol behavior, and update-rate arithmetic in the order you meet them.
Is the whole panel frozen or is it one bad tag? #
If every value the connection serves is stale, the fault is in the shared link — driver, PLC port, wiring, or startup timing; if one value fails while its neighbors update, the fault is that tag's address or data type. Pick three or four read-only tags you know move: a timer accumulator, a step number, a motor-current value. All frozen together means the shared path is down. One frozen among live neighbors means the tag definition is wrong.
Distinguish the two slow modes as well. Widespread values that refresh occasionally point to a link that is retrying — wiring, noise, or a port mismatch — while a fully frozen panel with a communication alarm is a link that never opened. A driver that has opened the link cannot make an invalid address valid, and correcting a tag will not wake a port that is not answering.
| Symptom scope | Likely fault class | First test |
|---|---|---|
| Every tag on the connection stale, from power-up | Startup timing | Enable a 10-second startup delay, then cold-start both devices |
| Every tag stale at all times | Port, driver, or wiring | Verify port mode and wiring, then move to the RS-422-capable port |
| One tag stale, neighbors updating | Tag definition | Check that tag's address and data type in the project |
Why does the HMI hang only when the machine powers up? #
Set a 10-second delay between HMI startup and the first poll: at power-up the panel can fire serial requests before the PLC port is ready to service them, and a reported installation fixed repeated startup hangs exactly this way. The mechanism is a race. The panel boots faster than the PLC port initializes, sends its first reads into a port that is not servicing requests yet, and some drivers then wait on a response that never arrives instead of retrying cleanly.
The setting lives in the startup or communication-delay control for the PLC communication task in the HMI project software. Test it like this:
- Set the communication startup delay to 10 s.
- Download the project and cold-start the panel and PLC together.
- Repeat a cold start with the delay disabled, and keep the delay only if it fixes a repeatable startup failure.
The delay costs one 10-second wait per power-up and nothing at runtime. It will not cure a protocol, port, or wiring mismatch — those hang the panel during operation, not only at boot.
Which PLC port should carry the serial link? #
Use the port the CPU manual documents for four-wire RS-422 and set it to 19,200 baud: in the reported installation, one port hung the interface and crawled, while the other, set for RS-422 four-wire at that rate, ran clean. The two ports on a rack PLC CPU are often different devices — different electrical interfaces, different protocol support, different firmware — and a driver bound to the wrong one retries forever, which on the screen looks like a freeze with slow values.
So the decision is to move the cable before you edit the project: put the link on the RS-422-capable port, set the documented baud, and confirm the driver selection names your PLC family and protocol. One bad port on one CPU is a first thing to test, not a law about every port 1 in the family; it simply makes port and interface mode the first configuration items on the list.
How should four-wire RS-422 and RTS/CTS be wired? #
Wire two twisted pairs — transmit on one, receive on the other — plus signal common, and handle RTS/CTS exactly as the port's documentation shows: the reported working link also needed the RTS/CTS pair jumpered. On a four-wire link, the HMI's transmit pair lands on the PLC's receive pair and vice versa, each pair rides its own twisted pair, and signal common ties the two devices' references together. A polarity swap — TX+ to RX− — is the most common wiring fault and produces total silence.
RTS and CTS are hardware flow-control lines. An interface configured to wait on them stalls forever when they float, which is why the jumper exists — but pin functions differ between ports and cables, so install it only where that port's wiring diagram calls for it. Verify the pinout conductor by conductor; the habit shown in the RS-232 pin map for an HMI replacement is the discipline every serial pinout deserves. If this link will later carry more than one panel or run past a couple hundred feet, the wiring rules change — read up on connecting multiple HMIs to one PLC over 200 ft of serial before adding drops. Safe the machine before you rewire: a CPU that resets mid-cycle can drop outputs without warning.
Can the HMI's diagnostic requests fault the PLC CPU? #
Yes — one field case had the CPU fault on a port because the HMI kept requesting diagnostic information the serial protocol permits, so treat diagnostic polling as the prime suspect whenever ordinary reads fault the CPU. That case was traced with a serial analyzer to a diagnostic request the port did not tolerate; the vendor corrected the behavior for the affected CPU, and other CPUs in the family were still pending review at the time.
When the CPU faults while the panel is connected, read the PLC diagnostic status and the HMI communication statistics together, and note the port, the time, and whether the fault lands at startup or after a specific screen action. If the driver exposes a diagnostic-polling option, disabling it is the controlled test. A port that answers reads cleanly but faults on connect is a protocol-behavior problem, not a noise problem — confirm the CPU's firmware level against the vendor's release notes before spending days on wiring that is already proven.
How slow is 19,200 baud for a full screen refresh? #
A 10-bit character at this baud rate moves 1,920 characters per second, so 30 one-word polls cost about 0.31 s of wire time per refresh — before PLC scan time, turnaround, or a single retry. Run the numbers yourself with the arithmetic below.
Assume a common asynchronous frame — 1 start bit, 8 data bits, no parity, 1 stop bit — which is 10 bits per character; the frame and every character count here are illustrative inputs, while the baud rate comes from the reported working setup.
- 19,200 bits/s ÷ 10 bits per character = 1,920 characters per second.
- One single-word poll: 10-character request + 2 characters of data + 8 characters of response overhead = 20 characters → 20 ÷ 1,920 = 0.0104 s ≈ 10.4 ms.
- Thirty tags polled one word at a time: 30 × 10.4 ms = 312 ms ≈ 0.31 s per full refresh, wire time alone.
- The same 30 words contiguous in the data table, read as one block: 10 + 60 + 8 = 78 characters → 78 ÷ 1,920 = 0.0406 s ≈ 41 ms.
The decision: packing consecutive tags into block reads buys a 312 ÷ 40.6 ≈ 7.7 times cut in wire time. One retry on a 1-second timeout — an illustrative figure — costs more than the entire consolidated poll, which is why the port and wiring come first. If a display-side fault is also in play, troubleshooting serial freezes at 19,200 baud covers failure modes specific to that rate.
What should I check before calling the link fixed? #
Cold-start both devices together, confirm the driver reports an active connection with no CPU fault, and watch several live values update through at least one full machine cycle — a successful project download proves nothing. Start from a saved state: learn how to back up a PLC program the right way and archive the HMI project beside it, and if the download cable itself is the unknown, uploading an HMI program without the right cable pinout covers the pinout workarounds.
Work the list in order:
- Baud, character format, and station address identical at both ends; driver names the right PLC family and protocol.
- Port mode matches the wiring — four-wire RS-422 where documented — and the link sits on the RS-422-capable port.
- Transmit and receive pairs verified conductor by conductor, polarity correct, signal common bonded.
- RTS/CTS wired or jumpered exactly as the port's diagram shows.
- Startup delay enabled if any cold start ever hung.
- Diagnostic-polling setting recorded; CPU fault log cleared and re-read after the change.
- Three or more live values observed updating through one full machine cycle, including values that move only at cycle end.
Then decide by case: one stale tag — fix the tag's address and data type; hang only at power-up — keep the 10-second startup delay; persistent hang — move to the RS-422-capable port at 19,200 baud; CPU faults on connect — test diagnostic polling and confirm firmware; clean link but sluggish screens — consolidate tags into block reads.
Questions asked
Will raising the baud rate above 19,200 fix the slow updates?
Not until the link is stable, and only if both the PLC port and the HMI driver list the higher rate in their documentation. The reported working setup ran at 19,200 baud, and a link that is retrying looks frozen at any speed. Clean up the wiring and the polling structure first, then raise the baud only where both ends document support.
Where does the RTS-to-CTS jumper actually go?
At the connector of whichever device has hardware flow control enabled, tying the request-to-send pin to the clear-to-send pin so the transmitter always sees permission to send. Install it only where that port's wiring diagram calls for it, because pin assignments differ between ports and between cables, and a jumper on the wrong pair changes nothing.
Can a marginal cable cause slow updates instead of a hard comm fault?
Yes. Intermittent continuity and induced noise produce checksum errors, retries, and timeouts, which the screen shows as stale values rather than a red communication alarm. Ground the cable shield at one end only, route it away from drive and motor wiring, and prove continuity on every conductor, including signal common, before you blame the port.
How do I tell a protocol mismatch from a wiring fault without a serial analyzer?
Match the baud and character format at both ends and retest: garbage characters that persist point to a driver or station-address mismatch, while total silence that clears after corrected wiring was polarity or a broken conductor. If you can borrow an analyzer, characters arriving at the right baud that the PLC ignores almost always mean the driver, station address, or protocol selection is wrong.
What should I record before calling the vendor's tech support?
The CPU model and firmware revision, the port number and its electrical mode, baud and character format, the HMI driver name and software version, whether the failure lands at startup or in running operation, and any CPU fault codes with timestamps. A serial analyzer trace of one failing transaction, if you have one, usually shortens the call to minutes.