OPC Server Connected but Tags Frozen: A Yokogawa and ABB Field Method

Why a Frozen Tag Tricks Operators

OPC servers bridge PLCs to SCADA, HMIs, and historians. Operators trust the green connected icon. Yet tags can freeze at the last good value — the channel looks healthy while the data is stale. This gap is dangerous. A Yokogawa CENTUM VP or ABB 800xA screen can show steady pressure while the real value drifts silently. Treat a stale tag as a quality fault, not a link fault, and work it with a layered method.

Q: How Do I Confirm It Is Not a Real Link Failure?

Separate link health from data quality first. Quality bits and time stamps reveal the truth — a fixed time stamp means no fresh sample is arriving.

The ABB AC 800M Controller exposes both link status and tag quality in its diagnostic views, making it easier to split the two faults apart.

• Step 1 — Ping the PLC and open its transport port (e.g. Modbus TCP 502 or the vendor Ethernet port).
• Step 2 — In the OPC server, read one frozen tag and one healthy tag in the same group.
• Step 3 — Compare quality codes and time stamps. A good tag advances every scan cycle.
• Step 4 — Rule out the client. The server monitor may update while the HMI view is stuck.

Q: Why Does an Address Map Audit Matter?

PLC edits during maintenance silently break tags. A variable moves to a new data block or register; array indexes shift; data types change. The OPC item descriptor still points to the old address and returns a cached value or zeros.

On an ABB AC 800M, check the control structure path carefully. The ABB CI853A Communication Interface Module passes structured data between the controller and external OPC clients — a type mismatch here produces bad quality codes without an obvious error message.

• Step 1 — Browse the live server namespace instead of the saved project file; compare names.
• Step 2 — Re-map changed tags to the current DB number, offset, or register.
• Step 3 — Match data types exactly. Reading a DINT as REAL yields bad quality, not a clean value.
• Step 4 — Save the map, then restart only the affected group — not the whole server.

Q: How Do Scan Rates and Subscriptions Cause Frozen Tags?

Large tag counts stress the driver. Aggressive polling floods a serial gateway and the server drops late samples — some tags update while others lag. Group by criticality and set realistic rates.

For Modbus TCP polling to an ABB system, the ABB CI867AK01 Modbus TCP Interface Module and ABB CI867 Modbus TCP Module each have per-channel scan limits that must match your OPC server scan class settings.

• Step 1 — Set scan classes: 250 ms for interlock points, 500 ms for control, 1 s for trends.
• Step 2 — Enable a deadband of 0.2–0.5% of span to cut low-value traffic.
• Step 3 — Never request a client update faster than the driver can deliver.
• Step 4 — Move bulk non-critical points to Modbus multi-register reads to lighten load.

Q: How Does DCOM Hide Partial Access Faults?

Classic OPC DA rides on Windows DCOM, which blocks access silently. Wrong launch rights freeze some items while the connection still reads "connected". Fix the transport layer, then align clocks so time stamps match downstream historians.

• Step 1 — Grant Launch and Access rights to both client and server service accounts.
• Step 2 — Open the RPC dynamic port range (1024–5000) on the host firewall.
• Step 3 — Set DCOM authentication to Connect across domains, not Packet Privacy.
• Step 4 — Sync server and PLC to one NTP source so stale rows are immediately obvious.

Q: When Should I Migrate to OPC UA?

OPC UA removes DCOM pain entirely. One TCP port (4840) passes the firewall, with a browsable address space and real security built in. Wrap legacy DA with a DA-UA tunnel, or adopt a native server on the next planned shutdown.

For cross-vendor integration between Yokogawa and ABB systems, the Bently Nevada 3500/90 Communication Gateway can serve as a bridge while a full UA migration is planned.

• Step 1 — Open TCP 4840 and issue a certificate to each client.
• Step 2 — Select Basic256Sha256 with SignAndEncrypt for wide-area links.
• Step 3 — Browse the UA address space and re-map tags to stable NodeIds.
• Step 4 — Validate quality and time stamps before retiring the DA wrapper.

Conclusion & Action Advice

A frozen tag is a quality fault wearing a healthy mask. Work the layers in order: confirm the link, then the address map, then subscriptions, then the DCOM transport. Keep a baseline of scan rates and quality codes for each server, and log one bad point as a lesson for the whole fleet. Do this on your next Yokogawa or ABB data gap, and a stale screen becomes a solved problem before an operator ever acts on it.

Author: Wang Zhiqiang is an industrial automation engineer with over 10 years of experience in PLC, DCS, and control systems.