The Green Light That Lies

Every control engineer eventually meets this failure. The OPC server reports a healthy connection. However, a block of tags refuses to update on the SCADA faceplate. Values sit frozen at the last reading, sometimes for hours. Operators lose trust first. Then the historian loses data, and compliance questions follow. I hit this on a GE Proficy iFIX node polling an ABB System 800xA connectivity server. The plant ran a batch process, and twelve analog tags went silent mid-shift. Therefore, learn this rule early: connection status only proves the channel is alive. It says nothing about data flow.

First Suspect: Address Mapping Drift

PLC programs change constantly during commissioning and turnarounds. Engineers move variables, reorganize data blocks, and shift array indexes. However, the OPC configuration usually stays untouched. The server keeps reading the old address, which may no longer carry live data. First, compare the PLC symbol table against the OPC server tag database. In iFIX, export the OPC item IDs and sort them by address. Second, open the 800xA Control Builder and verify the application references. A tag that still exists but no longer receives writes will read as a frozen value, not a bad quality. That distinction fools many engineers. Therefore, always diff the address mapping after any controller download, before you blame the network.

• Step 1 / Export the full OPC tag list with item IDs, addresses, and data types.
• Step 2 / Export the current PLC variable table after the last program change.
• Step 3 / Run a diff and flag every moved, retyped, or deleted address.
• Step 4 / Repoint the affected tags and force a quality refresh from the server console.

Subscription Rate and Communication Load

OPC UA subscriptions look self-tuning, but they are only as good as their configuration. A publishing interval set to 100 ms with thousands of monitored items will saturate the server. Moreover, the controller has its own limits. An AC 800M controller serves SCADA, HMIs, historians, and engineering stations from communication modules such as an ABB Coupler Module or ABB Analog Input Module. When request volume exceeds its capacity, it delays or drops the excess reads. The server stays connected, yet selected tags starve. This is the signature: some tags update, others freeze, and the pattern shifts under load. Therefore, audit the subscription parameters. Set the publishing interval to 1000 ms for slow analogs. Reserve 200 ms or faster only for critical control-adjacent tags. Moreover, check the monitored item queue size, because a queue of one drops intermediate updates by design.

• Step 1 / Open the OPC UA diagnostics page, default TCP port 4840, and read the subscription statistics.
• Step 2 / List all clients attached to the same controller connection.
• Step 3 / Sort monitored items by publishing interval and find the aggressive ones.
• Step 4 / Group tags by required speed, then set one publishing interval per group.
• Step 5 / Recheck the controller communication load value after the change.

Deadband, Cache, and Fragmentation Traps

Deadband filtering saves bandwidth, but it can fake a failure. If the deadband exceeds the natural signal movement, the server suppresses every update. A temperature drifting 0.2 degrees per minute will look frozen behind a 1% deadband. Second, watch the server cache. The OPC server reads the PLC on a cycle and serves clients from cache. If driver delays stall that cycle, cached values go stale while the connection stays green. Finally, consider network fragmentation. Large multi-tag read responses can fragment across frames. On congested networks, fragments arrive late or out of order. The server discards incomplete responses within the polling window. However, it never changes the connection status. Therefore, correlate the frozen tags with response-size boundaries. If every frozen tag sits in one large read block, fragmentation or a device-group timeout is your answer.

• Step 1 / Temporarily set deadband to zero for one problem tag and watch for updates.
• Step 2 / Compare the server cache timestamp against the PLC value timestamp.
• Step 3 / Split oversized read groups into blocks under 480 bytes per request.
• Step 4 / Verify device group timeouts, IP addressing, and rack or slot settings after any upgrade.