Building a Sensor Dashboard with Node-RED and MQTT
Read more...
A smooth historian trend is not an acceptance test for sensor data quality. Before adjusting a filter, identify which part of the path you are changing: the measurement, the events sent to the historian, or the events retained in its archive. AVEVA’s PI System presentation distinguishes interface exception reporting from archive compression (AVEVA, 2023).
Consider an illustrative commissioning problem: a pressure tag looks steady in the archive, yet an engineer wants to know whether brief changes are being retained. Resist the urge to make the plotted line look better. Establish what the tag must reveal, then compare that requirement with its acquisition, transmission and storage settings.
TL;DR: Map the complete tag path before tuning noise reduction. Check sampling and instrument filtering, preserve value quality and time context, then evaluate collector deadbands and archive compression separately. Commission with known inputs and permitted fault states. Accept the settings against the measurement task, with evidence from acquisition through archived retrieval.
Start with one tag and write down its actual path. Use the following as a worksheet, adapting it to the equipment installed:
Process → sensor/transmitter → input module or fieldbus → controller → collector/interface → historian → archived retrieval.
For each boundary, record the value available there, its engineering units, its timestamp origin and its quality representation. Leave an unknown explicitly marked for investigation. Do not fill a gap with an assumption about what a gateway “normally” does.
Keep a settings record alongside that map. Include the input range, sampling or refresh configuration, transmitter damping, low-cut functions, collector deadbands and archive compression. Bachmann’s GIO212 documentation, for example, lists input filtering and sample/refresh behavior as distinct specifications. That is a useful reminder to inspect both on the actual module, rather than recording only the PLC polling interval (Bachmann).
Define the acceptance requirement in process terms: which changes must this tag reveal, and what timing or amplitude differences are acceptable for its intended use? Agree those criteria with the people using the history. A diagnostic trend and a slow operating summary need their own stated requirements; this guide supplies no universal numeric tolerance.
If the source value already looks wrong, use the separate guide to troubleshooting 4–20 mA loop noise. Verify PLC raw-count scaling as part of the input record before treating the archive as the source of the discrepancy.
Review the acquisition settings before changing a smoothing filter. For the selected tag, record the configured sample or refresh interval, input filter and transmitter settings. Use the installed equipment’s documentation to identify what each setting controls. Complete missing configuration records before choosing settings for the tag.
Set an acceptance test before making changes. Identify a permitted input change that represents what the tag must reveal, agree the observation points, and save a baseline with the current configuration. Compare acquisition-side values with the collector output and retrieved history. Note any response difference for investigation; do not assign its cause from the appearance of one trend.
When testing a proposed setting, change one documented parameter at a time where the approved procedure permits. Save the before-and-after records and compare them with the tag’s agreed requirements. Select settings from those results and the equipment documentation. Use this commissioning recommendation with the applicable equipment procedure.
Review bandwidth, sampling and filter response together before accepting a smoother trend as evidence that the measurement is correct.
Transmitter settings also need their own definitions. In ABB’s 266 FOUNDATION Fieldbus manual, PV_FTIME filters the process variable PV, while the raw FIELD_VAL is separate. Its low-cut function changes calculated values below a selected cutoff to zero. Those are different operations, and these parameter names describe this product rather than every transmitter (ABB, 2021).
For the tag under review, ask which variable the collector actually receives. Then use an authorized test to compare the response before and after the configured function. Record the result instead of relying on a setting’s name. If wiring remains suspect, review the installation against the approved design and the shielding and grounding guide before making changes.
A value needs context about its validity. IEC 61850-7-3:2010 distinguishes good, questionable and invalid quality and includes detail indicators such as oldData. These are IEC 61850 semantics; do not assume another protocol uses the same labels or encodings (IEC 61850-7-3:2010).
The distinction matters when commissioning a frozen value. Under the standard’s invalid state, the client must not use the value; questionable status leaves a judgment to the client. ABB’s FOUNDATION Fieldbus manual separately describes a status byte that communicates variable quality. The formats differ, so inspect the implemented mapping rather than translating names by intuition (IEC 61850-7-3:2010, ABB).
As a commissioning test, introduce a permitted test or fault condition and follow its representation across every boundary. Inspect both the displayed trend and the retrieved record. Ask whether a last value remains visible, how its quality is represented, and how the application distinguishes it from a current valid measurement. Document the observed behavior and compare it with the agreed requirement.
Treat time as another explicit test. Schneider’s MasterPact MTZ IEC 61850 guide documents fractional seconds and time-quality fields including ClockFailure and ClockNotSynchronized. Their separate representation supports a narrow inference: timestamp precision alone is not evidence of clock health (Schneider Electric).
Record where timestamps originate and compare them at the controller, collector and archive during the test. Include buffered replay if the installed path supports it. Do not infer successful source-time preservation from a finely resolved timestamp on the final display. This is a recommended acceptance check, not a claim that every historian has the same timestamp contract.
In AVEVA’s PI System presentation, exception reporting is an interface deadband that reduces events sent toward the Data Archive snapshot. Archive compression uses a swinging-door algorithm to reduce stored events within a configured deviation. These are PI-specific mechanisms, not interchangeable names for transmitter damping (AVEVA, 2023).
The presentation also distinguishes collection products: PI Connectors do not perform exception testing, and AVEVA Adapters can apply deadband filtering while ignoring tag exception settings. Check the actual collector before assuming a tag attribute controls the whole path (AVEVA, 2023).
| Decision | What to inspect | Acceptance question |
|---|---|---|
| Acquisition filtering | Input hardware and transmitter settings | Does the measured response meet the tag’s purpose? |
| Event reduction | Actual collector/interface configuration | Which observations are forwarded? |
| Archive compression | Historian parameters and retrieved records | Does retained history meet the agreed tolerance? |
Use the table as a review sequence. Do not select a storage setting to compensate for an acquisition problem that has not been diagnosed.
One PI example shows why parameter names need care. AVEVA compares a binary tag scanned once per second for 48 hours: 172,800 events with compression disabled versus 15 under its illustrated minimal-compression configuration. The example uses compression enabled, zero compression deviation, no exception and an eight-hour compression maximum. It is a specific demonstration, not a storage-saving prediction for an analog process tag (AVEVA, 2023).
The same presentation explicitly distinguishes compression disabled from compression enabled with zero deviation. Do not use “zero” as shorthand for “all observations stored.” It also states that out-of-order data bypasses compression, so include the installed replay behavior in a PI commissioning review (AVEVA, 2023).
Compare the information retained at each boundary with the tag’s purpose before accepting the final archived trend.
Use a staged test record for the selected tag. The following sequence is an engineering recommendation drawn from the documented mechanisms above; it is not a quoted standard procedure. ISA’s catalog identifies loop-check and acceptance-test standards, but a catalog entry does not supply their procedures (ISA standards catalog).
ABB’s 266 manual illustrates why simulation deserves a specific restoration check: software simulation is gated by a hardware switch. That device-specific procedure belongs in the applicable instrument work instruction; it must not become a generic instruction for other equipment (ABB, 2021).
Make the test record useful to the next engineer. Include the raw or reference trace where available, the retrieved history, configuration versions and acceptance decision. If the tag crosses an edge system, include that boundary in the gateway-versus-direct connection review. Do not leave it outside the commissioning scope merely because a historian can connect to it.
Treat them as separate decisions. In the PI example above, exception reporting reduces transmitted events and archive compression reduces stored events. Review acquisition first, then test whether each configured reduction preserves the information required by the tag (AVEVA, 2023).
This guide provides no universal cutoff. Start with the measurement task and the actual module and transmitter documentation. Test the selected configuration against the changes the tag must reveal, including its response during the permitted commissioning input.
Do not use resolution as the acceptance test. Schneider’s IEC 61850 implementation documents clock-health flags separately from fractional seconds. Check the implemented quality information and test timestamp preservation through collection and archived retrieval (Schneider Electric).
Keep the tag-path map, settings, acceptance criteria and observed results together. Compare known inputs and permitted abnormal states at acquisition, collection and retrieval. Label any untested behavior explicitly so the next review does not mistake an assumption for a successful check.
Choose one troublesome tag for this review. Trace its acquisition settings, quality and time representation, event reduction and archived retrieval. Finish with a recorded acceptance decision tied to the tag’s purpose, rather than a screenshot of a smoother line.