HMI DH+ comm error after days: check power and ground first

7 min read · Published November 14, 2026

When nine DH+ panels run for days and then drop with a comm error that only a power cycle clears, suspect power and ground before the cable. Measure AC volts between like comm terminals with the trunk open, check the 120 V neutral bond, then verify the 150 ohm end termination.

Nine operator panels hang on one DH+ trunk off a PLC. They run fine for days, then a communication error prompt shows up, and the only thing that brings the link back is a power-down and restart. That detail sets the direction. An open cable recovers the moment you reconnect it, so a panel that stays down until it is restarted has been disturbed, not disconnected. Work the fault in this order: read the failure pattern, measure the supply references between panels, then check the trunk termination. The measurements, not the message on the screen, pick the layer.

Why does the panel only come back after a power cycle?

Because the fault is in the panel's power or ground reference, not in the cable: an unplugged DH+ connector gives the other error you have already seen, and plugging it back restores that link without a restart.

You already own the baseline test. Unplug the DH+ connector at a panel, note the exact error, plug it back in, and the link returns. The failure you are chasing behaves differently: days of clean running, a different prompt, and a panel that holds the fault state until its power is removed. When that repeats across a group of panels, the upset is shared, and shared means supply, ground reference, or trunk termination. This is the same evidence that sends you to check supply and ground when an HMI panel needs a periodic power cycle. Four checks settle it, in this order: shared supply, inter-panel ground voltage, the 120 V control neutral bond, end-of-trunk termination.

What does each symptom point to?

The pattern on the floor tells you which layer to open first: how many panels drop at once, whether the message matches the unplug error, and what it takes to recover, and each of those three answers points at power and ground before it points at the cable.

  • Several panels drop together. A shared cause: the common supply, the ground reference, or the trunk itself. Check the feed and the ground path before anything local.
  • One panel drops, the rest keep updating. A local cause: that panel's supply, connector, or wiring. Compare its feed and connections against a working panel.
  • The message differs from the unplug error. This is not an open cable. Something disturbed the panel; photograph both error texts for the log.
  • Only a power-down restores the link. The panel holds a fault state from an electrical upset. Supply quality and ground reference come first.
  • It returns every few days. A drifting or intermittent reference. Check the neutral bond and the inter-panel AC voltage.

Your case sits in the third and fourth lines, and nine panels failing over days points at the shared rows.

Do all nine panels share one supply and one ground?

If every panel hangs off one supply, that supply and its return path are the first suspects; if the panels sit on separate feeds, the question is whether their grounds meet at one point, and the answer decides what you measure.

Walk the feeds before you touch the trunk. Panels on different branches of one control transformer ride on different ground potentials, and the comm wiring ties those references together whether you planned it or not. Pull the electrical drawings and record what you find: which branch feeds each panel, where each neutral bonds, where the comm shield lands. Two neighboring patterns confirm the layer: a sagging supply shows up as a dim screen, the case in HMI screen dim on PLC power: check the supply drop, and ground noise bad enough to restart a controller when inductive loads switch, the case in PLC restarts when solenoid valves switch: suppress and ground first. Where the drawings and the as-found wiring disagree, that disagreement is your finding, and the bond correction goes to the electrician who owns those drawings.

How do you measure the ground-loop voltage between panels?

Set the meter to AC volts, open the trunk at one end, leave both panels powered, and read between the same terminal on each panel, blue to blue then clear to clear: a steady voltage there is a ground loop, and volts mean you fix references before anything else.

  1. Open the trunk at one end so the panels share no signal path.
  2. Leave both panels powered, because the offset exists only with power on.
  3. Measure between the two panels farthest apart, blue terminal to blue terminal, then clear to clear.
  4. Record the steady reading and ignore the meter's residual flicker.
  5. At each panel's feed, measure the 120 V neutral to the ground bus and note which panel reads what.

DH+ is a differential multi-drop link, in the same family as RS-485, and a differential link wants both ends sitting near one potential. Any steady reading you can quote in volts between like terminals on two powered panels is a ground-loop finding, and no termination resistor compensates for it. At the feed, neutral to ground reads near zero where the bond exists and tens of volts where it does not. This is a live measurement on powered control equipment: a qualified person takes it, with a meter rated for that circuit and the leads on the comm terminals only.

How does the whole check run in a worked example?

Nine panels daisy-chained on a 57.6 kbaud trunk, one span every 250 ft, fail with the delayed comm error every 3 days; this is the whole check with its numbers.

Clear the length first: 9 panels with 250 ft between neighbors gives 8 spans x 250 ft = 2,000 ft of trunk, inside the 10,000 ft (3,048 m) limit published for 57.6 kbaud, so length is out.

Open the trunk at panel 9, leave everything powered, and read blue to blue between panel 1 and panel 9: the meter holds 6 V AC. That is the ground loop. Now the feeds: panel 1 reads 0 V AC from neutral to ground bus, and panel 9 reads 47 V AC on the same test. Panel 9 hangs on a second branch whose control transformer neutral was never bonded. The electrician bonds it to the drawings, and the blue-to-blue reading falls to 0.3 V AC.

Close out the termination: 150 ohm at 1/4 W between blue and clear at panel 1 and at panel 9, the two physical ends, and nothing on the seven panels between them. The rate is 57.6 kbaud, so the 82 ohm value that belongs to 230.4 kbaud links stays in the drawer.

Then the calendar. The old failure interval is 3 days, so the repair must clear 6 days to meet the two-interval rule, and the line runs 21 days with no prompt: 21 / 3 = 7 old failure intervals. Solved.

Is the DH+ trunk terminated correctly?

A 57.6 kbaud trunk wants one 150 ohm, 1/4 W resistor at each physical end, wired between the blue and clear terminals, and no resistors anywhere in between.

The Data Highway installation manual, publication 1770-6.2.2, is the source for that value and for placement at both ends of the network. The ends are physical devices: the first and the last on the trunk. After any panel swap or trunk extension the last device changes, so the resistor moves with it, and the far panel, the one nobody visits, is the one to check. The same end-terminated, middle-bare practice governs every differential serial link, the reasoning behind RS-232 to RS-422 cable distance for an HMI panel link.

What do you fix first, and when is it solved?

Fix references first, termination second, and judge the repair against the calendar: the link must run past at least two of the old failure intervals, so a 3-day failure needs 6 clean days before you call it fixed.

References first: bond corrections and single-point ground repairs, done to the drawings. Termination second: the documented value at the two physical ends. During diagnosis, restart only the affected panel and make one change at a time, so each recovery tells you which change worked. Keep a log with four columns: panel number, exact screen text, uptime at failure, work performed. When the prompt comes back, that log separates a repair that failed from a new fault wearing the old fault's face.

Recommendation: before anyone orders a cable or a panel, open the trunk at one end and measure AC volts between like comm terminals on the two farthest powered panels; a steady reading in volts means fix the ground references and the 120 V neutral bond first, re-terminate the two trunk ends, and judge the repair against two clean failure intervals on the calendar.

Questions

What resistor value terminates each end of a DH+ trunk?

150 ohm at 1/4 watt, wired between the blue and clear terminals at the first and last physical devices on the trunk, and nowhere else. The 82 ohm value belongs to 230.4 kbaud high-speed links; on a standard 57.6 kbaud trunk it loads the bus the wrong way.

How far can a 57.6 kbaud DH+ trunk run?

The published trunk limit at 57.6 kbaud is 10,000 ft (3,048 m). Nine panels daisy-chained at 250 ft per span total 2,000 ft, well inside it, so a link that fails at that length points at grounding or termination, with length cleared by the arithmetic alone.

Why measure the inter-panel voltage with the trunk disconnected?

With the trunk open and both panels powered, the only thing left connecting the two comm references is the building's grounding, which is exactly what you are testing. Power stays on because the offset exists only while the panels run, and a qualified person takes the reading with a meter rated for that circuit.

How long should the link stay clean before you trust the repair?

At least two of the old failure intervals, so a 3-day history needs 6 clean days. Log the panel number, the exact screen text and the uptime at each failure; if the prompt returns, that log shows whether the old fault survived or a new one arrived.

Can an ungrounded 120 V control neutral really cause comm errors?

Yes. The comm interface rides on the panel's supply reference, and an unbonded neutral lets that reference wander as other loads on the transformer switch. The check is direct: neutral to the ground bus reads near zero where the bond exists and tens of volts where it does not, as the worked example shows.

Planning information, not legal or engineering advice. Rules differ by place: check codes and permits with your local authority and your licensed trades.

Keep reading

Build your signAsk for a quote