Radar Level Echo Diagnostics: Five-Layer Fault Tree

A jumpy or stuck radar reading rarely announces which layer it came from, and the expensive mistake is to start changing gain, mapping and empty-distance settings before the fault is localised. Radar level echo diagnostics works most reliably as a decision tree that walks the fault from the sensor outward through mounting, process, configuration and finally the PLC, changing nothing until the layer is named. This order matters because a process-foam problem "fixed" with a configuration change will return on the next batch.

Localise Before You Tune

The echo trace shows every reflection the instrument sees, but it does not label them. Before any setting is touched, the trace should be read against what the installation actually contains: the liquid surface, the nozzle, the tank wall, any internal pipework and the bottom. A strong, stable peak at the right height is a valid surface; a peak that appears only during filling, or that sits at a fixed height regardless of level, belongs to something else. Naming the layer first turns a confusing trace into a short list of suspects.

The discipline is to hold a test point at each layer rather than tuning blindly. Record the reading, the trace and the process state at each step so that if the fault moves you can see which action caused it. Configuration changes made without a held point are the reason many sites end up with a transmitter that "was adjusted until it worked" and nobody can reproduce the reasoning.

The Five Layers of a Radar Fault

Every fault in radar level echo diagnostics sits in one of five layers. The table maps common symptoms to the layer that owns them, which is the first branch of the decision tree. A symptom in the wrong layer is the most common misdiagnosis: a process foam treated as a sensor fault, or a PLC scaling error treated as a bad echo.

Layer ownership is assigned from the symptom pattern, not from a single reading. Confirm with a held test point at each layer before changing settings.

Walking the Sensor and Mounting Branches

The sensor branch is checked first because it is cheapest to exclude. A contaminated antenna, condensation on the lens or a failed transmitter module reduces signal quality and produces exactly the kind of weak, noisy trace that looks like a process problem. Cleaning the antenna or swapping the electronics removes the whole branch from the tree in minutes, and should always be done before mapping is invoked.

The mounting branch covers everything between the antenna and the liquid: nozzle length, beam obstruction from ladders or agitators, and tilt. A reading pinned at full scale while the tank is known empty almost always lives here, because the beam is hitting a near obstruction or the empty-distance setting places the surface inside the dead zone. This branch is solved by geometric and mapping windows, not by gain, which is why the tree visits it before configuration.

The Process and Configuration Branches

The process branch is where foam, vapour, agitation, coating and interface confusion live. These faults are real but intermittent, which is why they are so often misread as sensor drift. A value that jumps only during filling, or a trace whose second peak grows with product buildup, belongs here. The decision tree treats process faults by changing the physical or duty condition, or by accepting a verified limited range, rather than by chasing the echo with gain.

The configuration branch is entered only after sensor, mounting and process are cleared. Empty distance, gain, false-echo mapping and material factor all sit here, and a wrong value here produces a correct-looking trace with a wrong number. The trap is applying a configuration fix to a process fault: the reading may improve for one batch and fail for the next, because the underlying cause was never addressed.

A Worked Window Check on the Echo

The decision tree can be made quantitative by checking whether a detected peak falls inside the valid measurement window. This worked example uses assumed values and is a calculation method, not METRAVON field data.

Illustrative calculation, assumed site values:

• Nozzle dead zone (blocking distance): 0.3 m

• Maximum measurable range: 15 m

• Echo curve display range: 18 m

• Measurement window: 0.3 m to 15 m

• Detected peak height from antenna: 16.2 m

• Second peak height: 9.4 m

The display runs to 18 m, further than the 15 m measuring window, so a peak can be visible on the trace and still be outside span. The 16.2 m peak lies above the 15 m window and cannot be the surface. The 9.4 m peak sits inside the window and is the candidate surface — a candidate, not a confirmed one. Cross-check it against a manual gauge reading before accepting it. Boundary condition: if the real level were 16 m the tank would exceed the ordered range and the reading would be invalid regardless of echo quality, so verify the span against the actual fill height before trusting any peak.

Echo Features to Ignore and When the Tree Ends

A frequent error in radar level echo diagnostics is comparing signal-quality or signal-amplitude numbers between different manufacturers, on the assumption that a higher number means a better sensor. Those figures are defined inside each vendor's own firmware and scaling, so a 60% quality on one brand is not comparable to a 60% on another. Using them as a cross-vendor ranking produces false conclusions and should be avoided in any selection or fault comparison.

The tree says to stop tuning when the trace is clean and correct but the number is still wrong, because that fault lives in the PLC or scaling, not in the radar. At that point the work moves to the controller: check the 4–20 mA scaling, the engineering-unit conversion, the rounding and the stale-value flag. Continuing to adjust the transmitter only hides a configuration error downstream and makes the next troubleshooter start from a distorted baseline. The tree also stops when every layer is cleared yet a stable, repeatable offset remains; that is a datum difference to record and accept within budget, not a fault to map away.

FAQ

Q1. Why not just raise the gain to fix a weak echo?

Because gain masks the layer, it does not fix it. Radar level echo diagnostics must localise the fault to sensor, mounting or process first, then correct the cause rather than the symptom. A weak echo from a dirty antenna or heavy vapour will briefly improve, then fail again under the next condition.

Q2. My reading is pinned at full scale on an empty tank. Where is the fault?

Almost always in the mounting or configuration layer: the beam hits a near obstruction, or the empty-distance setting places the surface inside the dead zone. Check nozzle length and beam clearance before touching gain, because those are the usual owners of that symptom.

Q3. Can I compare signal quality between two brands?

No. Signal-quality and amplitude figures are defined inside each vendor's firmware and are not standardised, so a 60% on one brand is not equal to 60% on another. Compare instruments only on measured performance in the same tank, never on displayed percentages.

Q4. The trace looks fine but the PLC value is wrong. What now?

That fault lives in the configuration or PLC layer, not the radar. Check the 4–20 mA scaling, the engineering-unit conversion and the stale-value flag. Stop adjusting the transmitter, because the error is downstream of the sensor.

Q5. When is foam a process fault rather than a sensor fault?

When the value jumps only during filling or agitation and the trace shows a second peak growing with product, the cause is surface condition, not the antenna. Solve it by duty changes or by accepting a verified limited range, not by gain or mapping.

Q6. What if every layer is clear but the offset remains?

Then it is a datum or reference-plane difference, not a fault to tune away. Record it and accept it if it falls within the agreed error budget. Forcing agreement by mapping beyond the physical offset makes the reading wrong at other levels.

SUMMARY

Radar level echo diagnostics should localise a fault through five layers, sensor, mounting, process, configuration and PLC, before any setting changes; in a worked window check a 16.2 m peak outside a 0.3–15 m span is rejected while a 9.4 m peak is the candidate surface. Do not rank sensors by cross-vendor signal-quality numbers, and move to the PLC only when the trace is clean.