Resolving Trend Curve Mismatches and High-Frequency Noise Discrepancies on Redundant WinCC OS Servers

In SIMATIC PCS 7 redundant OS server pairs, minor high-frequency trend curve mismatches between Server 1 and Server 2 are caused by independent cyclic data acquisition over separate S7 Named Connections, not archive corruption. Each OS server independently polls the AS controller at unsynchronized cycle boundaries; rapid analog signal variations between these poll intervals cause slight amplitude and timing variations in Tag Logging. Redundant synchronization (CCExpressSync) only recovers historical data during server outages and does not harmonize live acquired points. To eliminate trend discrepancies, configure deadband filtering (hysteresis) in WinCC Tag Logging, or switch to AS-side cyclic archiving using APL block MonAnL with AR_SEND.

Executive Summary

Operators and process engineers monitoring redundant SIMATIC PCS 7 OS servers occasionally observe that trend curves for rapid analog variables on OS Server 1 do not match OS Server 2 down to the exact fraction of a millimeter or decimal point. While long-term trends and steady-state values match, high-frequency signal noise creates minor graphic divergence. This discrepancy often triggers false alarms regarding archive database failure or Redundancy synchronization breakdown. This guide explains the independent data acquisition architecture of redundant WinCC servers, clarifies the function and scope of CCExpressSync, and provides engineering solutions to eliminate trend noise discrepancies.

Architecture and Fundamental Mechanics

The data archiving architecture of WinCC Redundancy explains why independent trend divergence occurs:
1. Independent Live Data Acquisition:
In a PCS 7 redundant pair, OS Server 1 and OS Server 2 are active, independent systems. Each server maintains its own S7 Named Connection via Industrial Ethernet (CP 1623 or Softnet-IE) to the S7-400 AS. Tag Logging on each server executes its own polling thread (e.g., 500 ms cyclic acquisition). Because network packet arrival and CPU scheduling are independent, Server 1 and Server 2 read analog peripheral values at slightly different milliseconds.
2. High-Frequency Process Noise Impact:
If an analog signal fluctuates rapidly (e.g., turbulent pressure or flow with 50 Hz electrical noise), Server 1 might sample the peak while Server 2 samples the trough 20 milliseconds later. Both values are completely valid samples, but their respective Tag Logging databases record slightly different numbers, producing visible trend divergence.
3. Scope of Redundancy Synchronization (CCExpressSync):
The WinCC Redundancy synchronization mechanism (CCExpressSync / WinCC Redundancy) is designed strictly for fault recovery. When Server 1 goes offline, Server 2 continues archiving. Upon Server 1's return, CCExpressSync copies the missing historical archive segments from Server 2 to Server 1. It does not compare or overwrite live, successfully acquired archive records between active servers.
4. Solution Pathways:
- Deadband / Hysteresis: Suppresses high-frequency noise so that archive entries are written only when signal variation exceeds physical significance.
- AS-side Archiving (AR_SEND via MonAnL): The controller samples the signal, generates timestamped data packets in S7 memory, and sends identical telegrams to both servers simultaneously.

Step-by-Step Implementation Procedure

Verify WinCC Redundancy Operational Health:
a. Open WinCC Explorer on OS Server 1 and OS Server 2.
b. Check the Redundancy status dialog (Redundancy -> Status).
c. Verify that both servers report 'Redundancy partner online' and internal system tags @RM_MASTER and @RM_SERVER_STATE reflect one Master and one Standby server.
d. Confirm that CCExpressSync reports 0 pending archive synchronizations.
Configure Deadband Filtering in WinCC Tag Logging (Method 1 - Client/Server Side):
a. Open WinCC Tag Logging on the OS project.
b. Select the fluctuating archive tag and open its Properties.
c. Under 'Archiving Type', ensure 'Cyclic-selective' or 'Acyclic' is selected.
d. Set 'Hysteresis' (Deadband) to a meaningful process threshold (e.g., 0.5% or 1.0% of span).
e. This ensures small noise ripples are filtered out, harmonizing trends across both servers.
Implement AS-Side Archiving with AR_SEND (Method 2 - Exact Bit-Level Parity):
a. In CFC, open the chart containing the process transmitter monitoring block (MonAnL, FB1845).
b. Set parameter ArchvEn = 1 to enable integrated S7-side archiving.
c. Configure the archive cycle parameter (ArchvCycle) to define the logging interval.
d. The AS packages identical timestamped values into AR_SEND telegrams (SFB37) and broadcasts them to both OS servers.
e. Because both servers receive identical historical data blocks directly from the CPU, trend curves on both servers are 100% identical.
Recompile and Download Changes:
a. Run AS-OS compilation to generate archive tags automatically.
b. Download OS changes to both servers.

Verification

In WinCC Runtime, open the Trend Control displaying the analog variable on OS Client 1 connected to Server 1.
Open the same Trend Control on OS Client 2 connected to Server 2.
Compare the curves: verify that both trends follow the exact same visual contour without noise divergence.
Review the Tag Logging table view: verify identical timestamps and values across both server databases.

Limitations and Operational Precautions

Enabling AS-side archiving (AR_SEND) increases S7-400 CPU communication workload and consumes controller memory for intermediate data buffers. For non-critical trends, deadband filtering in Tag Logging is the preferred solution without adding AS overhead.

Engineer FAQ

Does CCExpressSync synchronize live tag logging values?

No. CCExpressSync only copies missing archive blocks after a server outage or communication failure; it does not reconcile live data simultaneously logged by both running servers.

Why do steady-state signals match while noisy signals diverge?

Steady-state signals remain constant between polling intervals, so both servers read the same value. High-frequency noisy signals change rapidly, causing independent polling cycles to capture different points of the waveform.

Can network latency cause trend divergence?

Yes. Network jitter or varying Industrial Ethernet traffic between the AS and the two servers shifts polling phase angles, exaggerating noise differences.

How does deadband hysteresis resolve the issue?

Deadband ensures that the server records a new value only when the process variable changes by more than the configured percentage, eliminating sub-threshold noise differences.