Precision Level Measurement: Accuracy, Uncertainty and Field Verification

Precision in level measurement is not a product name or a single catalogue accuracy number. It is the ability of the complete installed system to meet a defined uncertainty and repeatability requirement over the relevant process range. Begin with the decision supported by the measurement: inventory reconciliation, batching, pump control, alarm or custody transfer. State allowable error in engineering units and identify where it matters most. The uncertainty budget should include sensor performance, reference datum, mounting tolerance, tank geometry, dielectric or density variation, surface movement, temperature, signal conversion and reference-test uncertainty. A transmitter with a small laboratory error can still produce poor field results when installed in a long nozzle, on a flexible roof or against an incorrect tank datum. Selection, installation, PLC scaling and acceptance testing must therefore be reviewed as one measurement chain.

Define the real requirement

Express accuracy as an allowable level, volume or mass error across stated operating conditions. Avoid asking for the highest available percentage without linking it to a process decision.

Separate repeatability, resolution, hysteresis, drift and total uncertainty. A display with more digits does not create better measurement evidence.

Build an uncertainty budget

List sensor error, mounting and datum tolerance, process-property effects, tank-table uncertainty, analog or digital conversion and reference-method uncertainty.

Combine credible contributions using an agreed method and document assumptions. The dominant term may be vessel geometry or density rather than the transmitter.

Control the reference datum

Use one surveyed reference plane for physical level, transmitter distance, PLC scaling and tank volume calculations. Record nozzle and antenna offsets explicitly.

Roof movement, settlement or sensor relocation changes the relationship. Preserve drawings and as-built measurements so future technicians can reconstruct it.

Account for process conditions

Review foam, vapour, dust, turbulence, buildup, temperature, pressure, dielectric constant and density across normal and abnormal states.

Define which states are measurable and how invalid or low-confidence data will be flagged. Filtering should not hide a real rate-of-change or delay an alarm.

Verify the signal chain

Check raw sensor value, 4–20 mA or digital output, input scaling, engineering units, damping, fault current and displayed result at several simulated and physical points.

Keep sufficient resolution through each conversion. Incorrect zero, span or unit conversion can exceed the sensor’s own error by an order of magnitude.

Perform field acceptance

Compare against a traceable independent reference at representative low, middle and high levels where safely practical. Record both increasing and decreasing conditions.

Retain configuration, echo or diagnostic data, reference method and uncertainty. Revalidate after mounting, firmware, antenna, tank geometry or scaling changes.

Engineering checklist

Tie accuracy to a process decision.
Separate resolution from total uncertainty.
Use one controlled reference datum.
Include process and vessel effects.
Verify every signal conversion.
Retain traceable field acceptance records.

Frequently asked questions

Is catalogue accuracy the expected field accuracy?

No. Installed uncertainty also includes mounting, process, geometry, signal and reference contributions.

Does a digital output remove measurement error?

It can remove analog conversion error, but it does not correct the sensor, datum, process or geometry effects.

How many field test points are needed?

Use enough representative points to prove the stated requirement and risk, normally including low, middle and high conditions.

Document the decision and revalidation triggers

Issue an acceptance record that states the required uncertainty, validated range, reference datum, test method, environmental conditions, raw readings, corrections and remaining limitations. Identify changes that invalidate the result, including sensor relocation, nozzle work, tank-table revision, firmware replacement or PLC rescaling. A precision claim is defensible only when another engineer can reproduce the measurement chain and understand which operating states were excluded from the original verification.

Need a project-specific review? Send process data, drawings, photographs and acceptance criteria through our contact page.