The Symptom That Misleads Every Shift

Every SCADA engineer has seen this failure mode. The OPC server reports a healthy connection. The network team finds no faults. However, several tags sit frozen at their last value on the ABB System 800xA faceplates, driven by supervisory controllers such as the ABB PM865K01 processor unit. Meanwhile, other tags from the same GE PACSystems RX3i controller keep updating normally.

This partial freeze confuses operations. A recent instrumentationtools.com analysis listed eight distinct causes behind this exact symptom. The key insight is simple: a connected server does not guarantee flowing data. The fault usually hides in tag configuration, subscription settings, or controller load. Therefore, stop checking cables first. Start checking configuration.

An Eight-Step Diagnostic Sequence

• Step 1 — List every frozen tag: Group them by PLC, data block, and device group in the OPC configuration.
• Step 2 — Compare current PLC addresses against the OPC tag database: Program changes move variables during commissioning.
• Step 3 — Check tag quality codes: A BAD quality points to mapping. A GOOD quality with stale timestamps points elsewhere.
• Step 4 — Review the deadband setting on frozen analog tags: Small process moves may be filtered out deliberately.
• Step 5 — Audit scan rates and subscription update intervals: Sum the request load per controller.
• Step 6 — Check the PLC communication load and connection statistics: Monitor these metrics live while the problem occurs.
• Step 7 — Verify whether the PLC logic actually writes those variables: Confirm execution under current operating states.
• Step 8 — Inspect cache refresh counters and network fragmentation statistics: Evaluate these counters before restarting anything.

Address Mapping: The Number One Culprit

First, suspect the address map. PLC programs change constantly during optimization projects. Someone reorders a data block, shifts an array index, or converts an integer to a real value. The OPC configuration stays untouched. Consequently, the server keeps polling an address that no longer carries fresh data.

Sometimes the address still exists, but the logic no longer writes to it. The tag then freezes at its last received value while the connection stays green. This failure appears constantly in GE RX3i projects after SRTP memory reorganization. Therefore, version-control your PLC symbol exports and publish them to the SCADA team after every download. Cross-check one frozen tag manually: force the PLC variable in test mode, then watch the 800xA value. If nothing moves, the mapping is broken.

Deadband, Scan Rates, and Subscription Load

Second, examine filtering and timing. Deadband tells the server to ignore small value changes. A 0.5 percent deadband on a slow temperature loop from a GE RX3i analog input module looks like a frozen tag for minutes.

Moreover, aggressive scan rates create their own failure. Polling 5,000 tags every 100 ms saturates the RX3i CPU module communication processor. The controller then delays or drops selected read responses; some tags update, while others stall. Furthermore, subscription buffers overflow when clients demand fast refresh cycles.

A practical rule works well in the field: poll fast only for critical control-adjacent tags, and poll everything else at 1,000 ms or slower. Set deadbands per signal class, not globally. Finally, verify that the OPC UA publishing interval matches the negotiated keep-alive interval on the 800xA client side.

Cache, Fragmentation, and Conditional Logic

Third, look behind the server. Many OPC servers cache PLC values and serve clients from memory. If the cache stops refreshing, tags freeze while diagnostics stay silent. Restart the device group only after confirming this, because restarts interrupt every connected client.

Moreover, large read responses fragment across multiple Ethernet frames. On congested industrial networks, fragments arrive late or out of order, causing the server to miss partial data within the polling cycle. Segregate OPC traffic on a dedicated VLAN to control jitter.

However, do not overlook the program itself. Batch sequences write many variables only when specific steps are active; a frozen tag may simply mean the sequencer is waiting. Therefore, read the logic before you blame the network. Check the force-disable status too, because a leftover force table freezes outputs silently.