IMDTouchKnowledge BaseShop main store

Troubleshooting

Why Two PanelViews Slow Down an SLC 5/04 DH+ Network

HMI and touchscreens Published 6 min read

Short answer

The slowdown is display and alarm polling from the two terminals, not a fault in the SLC 5/04 — the sparse-screen test already proves it. Group the polled data into contiguous blocks, move alarming to the alarm-word method, and if the trunk is still saturated, move one PanelView to the controller DF1 port with pass-through routing, after confirming the firmware supports it.

On this page
  1. What the screen swap proves
  2. How much traffic two terminals put on the wire
  3. Cut the traffic at the source
  4. Move one PanelView to the SLC 5/04 DF1 port
  5. Verify each change
  6. The decision for each case
  7. Questions asked

Two terminals added to a DH+ trunk that also carries the SLC 5/04's own traffic will bog the whole link down when their screens poll dozens of embedded variables. Your sparse-screen result already convicts display polling: the network speeds up the moment the visible tag count drops. Work the problem in this order — trim and group the display variables, convert alarming to the alarm-word method, and, if the trunk is still saturated, move one PanelView to the 5/04 DF1 port with pass-through routing. The update-frequency setting is not your lever; in the reported case it changed nothing measurable.

What the screen swap proves #

The test you already ran is the diagnosis. Put the variable-heavy screen on both terminals and note how the link behaves; then switch both terminals to screens with few or no embedded variables and note it again. If response recovers on the sparse screens, the load is the per-tag polling those displays generate — not the node count, not the cable, not the 5/04 itself. Nothing on the physical network needs replacing.

The update-frequency result fits that reading: changing it produced no observable relief, so do not count that setting as a fix or as evidence that traffic was reduced.

Observation Meaning Next action
Response improves on sparse screens Displayed-variable polling drives the load Audit and group tags on the heaviest screens
Update-frequency change shows no effect That setting is not what limits the traffic Keep measuring with the screen swap
Alarms are configured tag by tag Alarm polling adds its own transactions on its own schedule Move to the alarm word and offset method
Trouble began when the second terminal joined Load adds per node on the shared trunk Apply the fixes to both terminals

How much traffic two terminals put on the wire #

DH+ is a token-passing, half-duplex network: a node transmits only while it holds the token, then passes it on. A terminal that constantly re-reads its screen tags stretches every token rotation, and every other node — including PLC-to-PLC traffic — waits. The wire is not slow; the request volume is. The same mechanism is why HMI point updates run slow on an Ethernet PLC link: the terminal's request pattern, not the medium's rated speed, sets the pace.

A 5/04 DH+ channel runs at 57.6, 115.2 or 230.4 kbit/s, and every node must be set to the same rate. At 57.6 kbit/s the raw ceiling is 57,600 ÷ 8 = 7,200 bytes per second, before token rotation, frame overhead and turnaround — which is why a loaded trunk feels full well before the arithmetic reaches 100%.

Worked example, assumptions stated (estimates, not protocol constants):

  • Each terminal's busy screen shows 60 embedded variables, each read as its own transaction, refreshed once per second.
  • One single-word read round trip costs about 24 bytes on the wire.
  • One terminal: 60 transactions/s × 24 bytes = 1,440 bytes/s.
  • Two terminals: 2 × 1,440 = 2,880 bytes/s, which is 2,880 ÷ 7,200 = 40% of the raw capacity of a 57.6 kbit/s trunk. With overhead on top, the link is effectively saturated and every node queues.
  • Grouped instead: the PLC copies those 60 values into contiguous integer files and each terminal reads four blocks of 16 words. A block round trip costs about 40 bytes (32 bytes of payload plus overhead). Two terminals: 2 × 4 = 8 transactions/s × 40 bytes = 320 bytes/s, about 4.4% of raw capacity.
  • Same data, 2,880 ÷ 320 = one-ninth the traffic.

That arithmetic is the whole argument for the code rewrite you were dreading.

Cut the traffic at the source #

Group the data. Copy scattered values into one contiguous file per screen group on each PLC scan, and point the screen objects at that file. Fewer distinct addresses per visible screen means fewer transactions, and the ratio above is what that buys.

Convert the alarming. PanelView alarms configured tag by tag generate their own polling on top of the displays. Use the alarm word and offset scheme described in the PanelView manual: one word carries 16 alarm bits and offsets extend the range, so the terminal polls a block instead of individual triggers. On a loaded trunk this is part of the fix, not a refinement.

Delete what nobody reads. Indicators kept just in case on busy screens are transactions on every refresh, and several separate indicators that always travel together can become one multi-state object with a single reference.

Move one PanelView to the SLC 5/04 DF1 port #

If screens and alarms are lean and the trunk is still saturated, take one terminal off DH+ entirely. Newer SLC 5/04 firmware can act as a bridge: connect the PanelView to channel 0, the DF1 RS-232 port, enable pass-through, and the 5/04 uses the DF1 destination (DST) byte to route the terminal's requests to other DH+ nodes on its behalf. Adjust the DF1 baud rate up or down to balance PanelView update time against DH+ impact. The trade is real — the local 5/04 does the routing work — but that costs far less than the system-wide impact of one more terminal pounding the shared trunk.

Channel 0 tops out at 19.2 kbit/s, a raw 19,200 ÷ 8 = 2,400 bytes/s for that one terminal. After the grouping work above, one terminal at 320 bytes/s uses 320 ÷ 2,400 ≈ 13% of the port's raw rate, which is comfortable, while the trunk drops to a single terminal's traffic.

One gate before you rewire: pass-through support is firmware-dependent, and at least one shipped SLC 5/04 lacked it; the revision that added it is not pinned down in the available material, so read the installed firmware revision and confirm the capability in the manufacturer documentation first. The alternative margin is rate — setting the whole trunk to 230.4 kbit/s multiplies raw capacity by 230.4 ÷ 57.6 = 4, with every node reconfigured to match and the cable verified. Treat extra bandwidth as margin after the traffic cuts, never instead of them.

The underlying discipline is older than this problem: keep operator terminals on the connection local to the controller they animate, and let the plant network carry machine-to-machine data — the same reasoning as connecting multiple HMIs to one PLC over 200 ft of serial.

Verify each change #

Run these tests when the line can tolerate slow screen updates; the HMI is an operator aid, not a safety function, so coordinate with production before you start.

  1. Baseline with heavy screens on both terminals and alarms exercising: time how long a freshly opened screen takes to settle on live values, and how long a command takes to reflect in the PLC.
  2. Change exactly one thing — group the tags on the single heaviest screen — and repeat the identical sequence.
  3. Convert the alarms to the alarm-word method and retest with alarms active.
  4. Work through the remaining screens the same way, one retest per change.
  5. Still slow after all of that: confirm the firmware supports pass-through, move one terminal to DF1, re-baseline, and tune the baud.
  6. Finish with both terminals on their normal screens, alarms active, measured the same way as the baseline — that is the pass condition.

The decision for each case #

  • The sparse-screen test improves response: display polling is confirmed, so do the grouping rewrite. The traffic ratio above is the payoff, and it justifies the code change.
  • Alarms are polled tag by tag: convert to the alarm word and offset method as part of the fix, not as a later cleanup.
  • Screens and alarms are lean and the trunk is still saturated: confirm the firmware, then move one terminal to the 5/04 DF1 port with pass-through routing.
  • The installed firmware lacks pass-through: trim traffic further, then set the entire trunk to its top rate and re-test.
  • Only one terminal is heavy: move that one to DF1 and leave the lighter terminal on the trunk.

Questions asked

Why didn't changing the PanelView update frequency setting help?

The update-frequency setting paces how often the terminal refreshes displayed values; alarm triggers are polled on their own schedule, so alarm traffic can keep the link loaded no matter what the screen setting says. In the reported case the change made no visible difference, so measure with the sparse-screen swap instead of trusting the setting.

Is the PLC rewrite to group data worth doing?

Yes, when the sparse-screen test shows clear improvement. Reading 60 individually addressed values as four contiguous 16-word blocks cuts a two-terminal load from roughly 2,880 bytes per second to about 320 on a 57.6 kbit/s trunk, the same data at one-ninth the traffic. The PLC side is a per-scan copy of scattered values into one contiguous file, cheap next to a network that queues every node.

Which SLC 5/04 firmware supports DF1-to-DH+ pass-through?

Pass-through routing is firmware-dependent, and at least one shipped SLC 5/04 lacked it; the exact revision that added it is not pinned down in the available material. Read the firmware revision of the installed controller and confirm the capability in the manufacturer documentation before moving a terminal to the DF1 port. If the revision lacks it, trim traffic further and consider running the whole trunk at its top rate.

Can we just raise the DH+ baud rate instead of restructuring screens?

It buys headroom, not a fix. Going from 57.6 to 230.4 kbit/s multiplies raw capacity by four, and every node on the trunk must be reconfigured to the same rate over a sound cable. A terminal polling hundreds of individual tags will still consume the headroom, so group the data first and use the higher rate as margin.

Do both terminals need the work, or only the newer one?

Both. Traffic adds per node: two identical terminals each contribute about half the load, which is why the trouble appeared when the second one joined. Fix the heaviest screens on both terminals, but change and test one thing at a time so each improvement stays measurable.

Keep reading

  • Troubleshooting

    Why a PanelView 550 Shows Nothing but the PLC Communicates

    A PLC link that works proves nothing about the PanelView: if the 550 shows no display, menu, or reset response, treat it as a terminal power or hardware fault. Verify a stable 24 VDC at its terminals under load before touching DF1, cable wiring, or pass-through settings.

    HMI and touchscreens

  • Troubleshooting

    How to Connect an HMI to an SLC 500 Over DH-485 Serial

    Most HMI-to-SLC 500 DH-485 failures live in the cable, the serial settings or the panel driver, not in the tags. Run every station at the same baud rate (19200 is the DH-485 default), give each a unique address in the 0–31 range, use the approved cable or the documented interface-module wiring, and prove the link by reading one known PLC address before trusting a single screen value.

    HMI and touchscreens

  • How-to

    Replacing a Fleet of PanelView 550 Terminals on DH-485

    There is no direct swap: the PanelView 550 is Silver Series, so each terminal is a migration. The PanelView Plus 6 600 keeps the 550 cutout but needs a third-party DH-485 adapter; a Plus 6 700 takes the 2711P-RN6 but modifies the cutout. Audit all 25 catalog numbers, convert one application in FactoryTalk View Studio ME, pilot one machine, then buy the fleet.

    HMI and touchscreens