Troubleshooting Power Chart Gaps with Good-Quality Data

Power Chart line gaps can occur even when every returned sample has good quality. With pens.display.breakLine enabled, a long interval between stored samples can be rendered as a break; quality describes the samples that exist, not the unsampled time between them. Measure timestamp spacing first, then correct the history sampling policy or disable line breaking only when connecting sparse points is operationally honest.

How should the symptoms be read?

Look at the trend first. In the reported case, the Power Chart showed gaps while a regular Time Series Chart connected the same period, and the inspected history contained no bad-quality points. That combination separates three different conditions engineers often treat as one:

• Bad sample: A stored value has bad quality.
• Missing sample: No row exists for part of the requested time range.
• Display break: The chart chooses not to draw a segment across a long timestamp interval.

A query can return only good-quality rows and still contain a 20-hour interval with little or no recorded information. The absence of a bad-quality row does not fill that interval or prove that values were sampled continuously. Compare the timestamps of adjacent rows instead of checking quality alone.

The difference between the two chart components is also diagnostic. If one component connects sparse points and the Power Chart breaks them, the historian may be returning the same values to both while each chart applies a different rule for drawing between them.

What does the signal chain reveal?

The OPC tag is a floating-point process value. Its out-of-the-box historical tag group evaluates on a 10-second schedule, but that schedule is not the same as a promise to store one row every 10 seconds. The history subsystem applies the configured deadband and sampling rules before a value reaches the chart.

The controller or historian cannot reconstruct changes it never stored, and the chart cannot distinguish a truly steady process from unrecorded movement using only two distant values. Tuning the chart does not fix sampling, just as tuning does not fix wiring.

Which checks identify the actual cause?

Start with raw history, not the rendered line. Export or inspect timestamps across the visible break and calculate each adjacent interval. Several points in the reported long gap were approximately 24 hours apart and difficult to see at the selected chart scale. That finding changes the diagnosis from “bad data” to “sparse data.”

1. Query the same tag and exact time range used by the Power Chart.
2. List timestamp, value, and quality for every returned row around both edges of the gap.
3. Find the largest difference between adjacent timestamps. Do not infer continuity from good quality.
4. Compare that interval with the history configuration active when those rows were recorded.
5. Check Historical Deadband, Deadband Mode, Deadband Style, Sample Mode, Historical Tag Group, and Max Time Between Samples.
6. Compare the Power Chart with another chart using the identical tag, time range, and query mode.
7. Temporarily toggle pens.display.breakLine. If the values remain unchanged but the line becomes continuous, the toggle changed presentation rather than recovering data.

The observed configuration included Include Metadata set to false, Deadband Style set to Auto, Deadband Mode set to Absolute, Historical Deadband set to 100, Sample Mode set to Tag Group, and the default 10-second historical tag group. The displayed settings also showed Max Time Between Samples at 2 hours, but that value may have been changed after the affected data was collected. Configuration changes do not rewrite old timestamps.

How do you correct the history and display settings?

Choose the recording policy from the process requirement, not from the desired appearance of one chart. Decide how much value deviation may be discarded and how long operators may go without a stored confirmation point.

1. Record the current settings before changing them.
2. Compare the absolute deadband of 100 with the tag’s normal range, engineering units, noise, and smallest meaningful change. Reduce it if legitimate movement is being suppressed.
3. If the requirement is to record meaningful value changes directly, test Discrete deadband behavior instead of relying on Analog compression. Analog handling is not simply “store a point whenever the value differs by 100.”
4. Set Max Time Between Samples to the longest acceptable interval between stored confirmation points. The reported current value was 2 hours; historical points spaced near 24 hours indicate that older data must be evaluated against the setting active at that time.
5. Keep the 10-second tag-group rate only if that evaluation frequency matches the process. Increasing evaluation frequency alone will not force every evaluation to be stored when deadband rules suppress writes.
6. Leave pens.display.breakLine enabled when operators must see loss of temporal coverage. Disable it only if a straight line between sparse observations is an accepted visualization.

Apply one change at a time to a controlled test period. Changing deadband style, threshold, maximum interval, and display behavior together hides which setting fixed the symptom.

How do you verify the correction?

Verification must follow the signal chain from source to storage to display. Observe a period that includes steady operation and a real process change, then inspect the raw history rows.

1. Confirm that the live OPC value and quality update as expected.
2. Verify that stored timestamps never exceed the newly selected maximum interval during steady operation.
3. Drive or observe a legitimate value change and confirm that the history policy retains enough points to represent it.
4. Reload the same time range in both chart components.
5. Confirm that any remaining break corresponds to a measurable timestamp discontinuity, not merely a change in line style.

New settings validate only data collected after the change. Old gaps remain sparse because a display property cannot recreate measurements that were not stored.

Which pitfalls recur with compressed history?

• Equating good quality with complete coverage: Quality belongs to individual samples; continuity depends on timestamp density.
• Reading the tag-group rate as a storage rate: A 10-second evaluation opportunity can still produce widely separated stored rows after deadband processing.
• Using a large absolute deadband without engineering units: A threshold of 100 may be appropriate or excessive; the process range and required resolution decide.
• Treating Analog and Discrete as interchangeable: Select the style based on whether shape compression or change-oriented recording matches the requirement.
• Judging old data by current settings: A present 2-hour maximum does not explain rows collected while another value was active.
• Hiding the gap: Disabling pens.display.breakLine connects endpoints but does not prove what occurred between them.

FAQ