Radar Level Transmitter Accuracy: What the Datasheet Number Does Not Tell You

A quoted radar accuracy such as plus or minus 2 mm applies to the specified model, measuring range and reference conditions. It is not automatically the uncertainty of installed level, volume or mass. Define the variable being accepted, identify the main contributions and compare the measurement with a suitable reference before deciding that an instrument needs adjustment or replacement.

Which terms need to be kept separate?

Control, inventory and protective functions may need different combinations of accuracy, repeatability and response time. A high-level alarm also needs enough margin for measurement error and the level rise during detection and action; repeatability alone is insufficient.

What do the reference conditions actually say?

Read the manufacturer's accuracy statement and exclusions for the exact configuration. Relevant conditions may include distance, ambient temperature, reference plane, target and signal quality. Do not assume all manufacturers use a water tank at 20 degrees Celsius and a distance above one metre.

Check whether the stated limit covers the whole usable range or whether near-end and low-signal regions have additional conditions. Identify separate analogue-output, temperature and long-term-stability specifications. A datasheet limit is neither a universal minimum error nor proof that any field deviation is acceptable.

Which installation contributions belong in the budget?

Deposits, competing echoes and a wrong range can produce faults or systematic errors, not merely small random uncertainty components. Investigate and correct identifiable problems before combining an uncertainty budget. Do not hide an incorrect target by adding a large allowance.

Can the components simply be added by root sum of squares?

Only with a justified measurement model. Convert components to standard uncertainties, apply the appropriate sensitivity coefficients and consider correlations. Datasheet plus-or-minus limits cannot all be treated as independent standard deviations without knowing their basis.

For a simplified level model, suppose three independent contributions already expressed as standard uncertainties are 2 mm, 3 mm and 4 mm, with sensitivity coefficients of magnitude one. The combined standard uncertainty is sqrt(2^2 + 3^2 + 4^2), or about 5.4 mm. Using a coverage factor of 2 gives an expanded uncertainty of about 10.8 mm. These are assumed teaching values, not a predicted result for a particular tank.

The coverage factor and interpretation must be stated and justified; a factor of 2 is not an unconditional guarantee of a particular coverage probability. A worst-case sum of limits is a different calculation. Do not call either calculation a universal field-accuracy figure for radar.

What does temperature change?

Temperature can affect electronics, mechanical dimensions and the microwave propagation conditions in the gas space. The importance depends on the model, vessel and vapour. Dense vapour or demanding steam conditions can require specific treatment rather than being dismissed as negligible.

Mechanical expansion can move the reference. With an assumed expansion coefficient of 12 micrometres per metre per kelvin, a 20 m steel length changes by about 2.4 mm for a uniform 10 K rise. This example does not represent every alloy, temperature profile or tank restraint.

Ultrasonic instruments have different sensitivity to the acoustic path, but no universal tenfold accuracy ratio applies. Compare actual compensated systems and their installation conditions instead of using a fixed technology multiplier.

How should the field comparison be made?

Use a reference method that is safe, appropriate to the process and sufficiently characterised for the acceptance decision. Manual gauging may be suitable in some tanks, but opening a hazardous vessel or using a tape is not an obligatory test method.

Compare the same variable, datum and time. Where practical, include representative low, middle and high operating points and the relevant filling or agitation states. Record repeat readings and reference uncertainty; a single comparison at one level does not demonstrate the full range.

A loop-current check tests the output path, not the radar's identification of the surface. An echo curve helps explain target tracking, but does not independently prove absolute accuracy. Use both types of evidence for the questions they can actually answer.

When is adjustment appropriate?

First check reference height, range, units, mounting, deposits and echo selection. A constant discrepancy may indicate a datum error, but it can also come from the reference method or another systematic effect. A changing discrepancy is a diagnostic clue, not a unique diagnosis.

A manufacturer-supported trim can be appropriate after the cause is understood and the procedure is authorised. Record the as-found result, adjustment and as-left verification. Calibration establishes a measurement relationship; it is not simply another name for changing a setting.

Send METRAVON the model, reference readings, tank drawing, configuration and operating trends for a performance review. State whether the required result is distance, level, volume or mass and provide the acceptance rule.

FAQ

Is a plus-or-minus-2-mm specification realistic in service?

It may be attainable under suitable conditions, but installed performance must be assessed for the actual model, reference, surface and output chain. Do not replace that assessment with a universal field allowance.

Is repeatability enough for a high-level alarm?

No. Setpoint accuracy, uncertainty, response time, fault handling and process rise rate also matter. Design the complete alarm or protective function.

Does a higher frequency guarantee better accuracy?

No. It may improve beam clearance for a particular antenna, but compare actual specifications and installed performance for the chosen models.

Can I combine every datasheet limit using RSS?

Not directly. Establish standard uncertainties, sensitivity coefficients and relevant correlations first. A limit, a standard deviation and an expanded uncertainty are different quantities.

Does a constant offset always mean the datum is wrong?

No. Check the datum first, but also examine reference-method bias, configuration and output scaling before choosing a correction.

Does an unchanged echo curve prove accuracy?

No. It supports a stability and target-tracking assessment. Absolute accuracy still needs an appropriate reference and a defined comparison or calibration procedure.

Technical references

• NIST: combining uncertainty components

Related applications

How to Specify a Radar Level Transmitter: The 14 Fields That Decide the Quote | Level Measurement Audit: How to Review an Installation in One Shift

Discuss the application with METRAVON