Radar and ultrasonic level in difficult service

Why is my ultrasonic level sensor giving erratic readings when there's foam or a turbulent surface?

Foam absorbs the sound pulse. A turbulent surface tilts the rest of it away from the transducer. What comes back is too weak and too inconsistent to lock onto.

Ultrasonic works like a mirror. It needs a specular reflection off a flat surface, and a wave tilts that mirror so the echo goes somewhere other than back at the transducer. Light, dry foam is mostly air separated by thin liquid films. It absorbs. Dense, wet foam fails the other way — it reflects well enough that the device tracks the top of the blanket, which reads high and steady rather than erratic.

Endress+Hauser puts decibels on this for the Prosonic T FMU30. Add up the attenuation that applies to your process, then read the remaining range off the range diagram for your sensor. A calm surface costs 0 dB. Waves cost 5 to 10 dB. Strong turbulence, and the example E+H gives is stirrers, costs 10 to 20 dB. Against foaming the table gives no figure at all and says to contact your Endress+Hauser sales representative. That is itself the answer. Other rows stack on top: heavy dust 5 to 20 dB, a large filling curtain in the detection range 10 to 40 dB, a 60 °C difference between sensor and product surface another 10 to 15 dB. There is not much to spend. The 1½-in. FMU30 sensor reaches 5 m in liquids and the 2-in. sensor 8 m, both with an 11° emitting angle.

Read the failure before you fix it. A reading that wanders continuously is a weak echo. A reading that snaps between two fixed numbers is a false echo, usually a mapping or mounting problem — though a device alternating between the top of a foam blanket and the liquid beneath does the same thing.

• Move the sensor off the fill stream and away from the agitator. Re-run the false-echo learn with the process running, not on a still empty tank.
• Put the sensor in a stilling well. A vertical pipe with a vent hole above maximum level and an opening below minimum level mechanically excludes foam and flattens the surface.
• Check the transducer face for condensate or product film first. Same symptom.

If the foam is always there, ultrasonic is the wrong instrument. No configuration change recovers energy the foam absorbed. Guided wave radar with a coaxial probe excludes the foam mechanically, and Emerson's own summary of the 5300 is that it handles vapor, dust, turbulence and foam well. Differential pressure barely sees foam either. Foam weighs almost nothing.

Why does my radar level transmitter reading jump around when the agitator starts?

The agitator trades a flat reflecting surface for a sloped, moving, gas-entrained one, and it adds two more reflectors the device can track instead of the product: the blades and the vortex wall.

A tilted surface throws the beam away from the antenna. The surface echo drops; everything else stays where it was. How much else the device sees comes down to beam width. Rosemount publishes a beam angle per antenna for the 5408, which runs FMCW at 24.05 to 27.0 GHz: 22° for a 1½-in. cone, 18° for 2-in., 14° for 3-in., 10° for 4-in., and 4.5° for the 8-in. parabolic. At 33 ft (10 m) Emerson's own beam-width table gives footprints of 3.9 m, 3.2 m, 2.5 m, 1.8 m and 0.8 m. A wide beam over an agitated vessel collects competing reflections a narrow one never sees.

• Check mounting position first. Emerson is explicit about the 5408: not in the center of the tank, not close to or above the inlet stream, and a minimum of 8 in. (200 mm) from the tank wall. The recommended position is ½ of the tank radius for liquids and ⅔ of the tank radius for solids.
• Re-map false echoes with the agitator running, at the level where the problem shows. A map taken on a still tank describes a tank that does not exist during operation.
• Raise damping only after the echo curve is correct. Damping averages a number; it does not create a missing echo. If the surface genuinely swings ±100 mm, a 20 s filter gives you an average, not the level.
• Fit a still pipe or chamber. Emerson recommends one outright for tanks where there is excessive foaming or turbulence. For a cone antenna in a still pipe the published gap between antenna and pipe is a maximum of 0.2 in. (5 mm).

For a guided wave probe in the same vessel the risk is mechanical, not electrical. The 5300 manual's mounting rules: not close to inlet pipes, not close to agitators, avoid heating coils, and do not let the nozzle extend into the tank. On agitators specifically, Emerson recommends a probe tie-down if the probe can move to within 30 cm of an agitator or any other object during operation. It also says to avoid physical contact between probes and agitators, and to avoid applications with strong fluid movement unless the probe is anchored. A probe that tends to sway in turbulence gets anchored to the tank bottom.

How do I stop condensation forming on the face of an ultrasonic level transducer?

Wiping the face and raising gain buys time, nothing more. Keep the transducer face above the vapor dew point, or move the measurement out of the wet vapor space entirely.

A film of condensate on the face is an acoustic impedance mismatch. It damps the transducer, lengthens the ringdown, and weakens the transmitted pulse and the returning echo both. The symptom is a signal that degrades slowly, recovers when someone wipes the face, and degrades again.

• Insulate the nozzle and sensor body so the face tracks vapor temperature instead of ambient. Endress+Hauser sells a weather protection cover for the FMU30 and recommends it outdoors for the same reason — its installation notes say to protect the device against direct sun or rain, both of which drive the face away from vapor temperature.
• Heat trace the nozzle at low wattage to hold the face a few degrees above dew point.
• Bleed a small continuous flow of dry instrument air or nitrogen into the nozzle, directed across the face rather than at it.
• Mount the sensor plumb so droplets shed. Out of plumb also aims the beam off the surface, which costs you twice.

Grease, petroleum jelly, and hydrophobic sprays are not the permanent fix. They change the acoustic coupling, they are gone within weeks, and they hide the real cause.

In a permanently saturated vapor space — a sump, a digester, a sour water tank — condensation wins. Ultrasonic loses. Guided wave radar tolerates a wet probe. On non-contact radar the tolerance comes from the antenna, not the frequency: Emerson's antenna selection notes for the 5408 describe the process seal antenna, with all-PTFE wetted parts, as suitable for applications with heavy condensation and build-up, where the plain cone antenna is simply the default for most applications. A purge connection helps more than either. Higher frequency does not solve it. An 80 GHz device gets its narrow beam from a small aperture, and a small aperture is no less sensitive to a wet or coated face. Radar costs more and is still the correct answer here.

Why is my Rosemount 5300 guided wave radar reading empty when the tank is full?

The transmitter has lost the surface echo near the top of the probe and fallen through to the probe-end echo, which it reports as empty.

This is documented behavior, not a fault. The 5300 manual lists "sudden level jump detected" and gives a surface level inside the transition zone as one of its causes, noting that in the transition zones the level may jump to Full Tank or End of Probe. The current datasheet calls the same regions blind zones. Measurements may not be possible inside them, and accuracy there exceeds ±1.18 in. (30 mm). On a single lead probe the upper blind zone is 3.9 in. (10 cm) on water and 3.5 in. (9 cm) on oil, with reduced accuracy through the first 9.8 in. (25 cm). Install the probe in a nozzle and Emerson states the nozzle height shall be added to the specified upper blind zone. A level sitting in the nozzle is a level the device cannot see. Hence Emerson's own instruction: set the 4 and 20 mA points outside these zones.

• Upper Null Zone. The manual's wording is that measurements are not performed within the Upper Null Zone. If someone masked a rough nozzle with a 500 mm UNZ, everything above that point is invisible by design. Setting the UNZ to zero restores measurement close to the flange.
• Wrong probe type configured. The troubleshooting chart lists this exact symptom — surface pulse detected, level reported as Full or Empty — and says to check whether the Full Tank / Empty Tank warning is active, then confirm the transmitter is configured with the correct probe type and that the reference pulse is below the reference amplitude threshold.
• Probe length and tank height parameters. Level is calculated from these. A probe cut in the field and never re-entered gives a consistent, confident, wrong answer.
• Amplitude thresholds. A surface threshold set too high ignores a real echo. Check it against the echo curve, not against a number someone remembers.

Pull the echo curve before you adjust anything. Confirm which pulses are present: reference, surface, probe end. Trim Near Zone is the correct tool for stationary disturbances at the top, and the manual is explicit that it is not suitable for occasional ones. It also carries a level requirement — Emerson says to make sure the product level is below the Near Zone region, which it defines as 0 to 1.6 ft (0 to 0.5 m) below the Upper Reference Point. Run it against a high level and you trim out a real echo. If the tank routinely floods the top 100 mm of probe, the transmitter is mounted too low for the range being asked of it. No setting fixes that.

What is blanking distance on an ultrasonic level sensor and why does it matter?

Blanking distance is the dead band directly below the transducer face, where the sensor is still ringing from its own transmit pulse. Any level that rises into it is not measured at all.

Published values are small and specific. The Rosemount 3100 series calls it dead zone and lists 12 in. (0.3 m) across the whole series — the 8 m Rosemount 3101 and the 11 m 3102 and 3105 alike. The Endress+Hauser Prosonic T FMU30 lists 0.25 m (0.8 ft) for the 1½-in. sensor at roughly 70 kHz, which reaches 5 m in liquids, and 0.35 m (1.1 ft) for the 2-in. sensor at roughly 50 kHz, which reaches 8 m. Lower-frequency, longer-range transducers ring longer. Blanking grows with range.

Endress+Hauser gives the mechanism in one line: level echoes within the blocking distance cannot be evaluated, because of the transient characteristics of the sensor. The arithmetic agrees. 0.3 m of blanking is about 1.7 ms of round trip at 343 m/s, the order of a transducer's ringdown time. Treat that as a sanity check, not a manufacturer statement.

It matters for three reasons, and people get the first one backwards. Blanking is measured from the transducer face, not from the tank flange. A nozzle does not add to it — a nozzle raises the face and carries the blind zone up inside itself. E+H says this directly: install the instrument high enough that the blocking distance is not undershot even at maximum fill, and use a pipe nozzle if you cannot maintain the blocking distance any other way. What limits you is the nozzle itself, since a long or rough one generates its own echoes. For the FMU30, E+H publishes maximum nozzle lengths of 80 mm on DN50 for the 1½-in. sensor, 240 mm on DN80, 300 mm on DN100, and 400 mm on DN150 and larger, and requires a smooth nozzle interior.

Second, the span cannot reach into the dead band. E+H states that the span may not extend into the blocking distance, so the 100% point — 20 mA on a direct-acting level output — has to sit below it, measured from the face. Third, a level entering the blanking distance does not read high. Depending on model and how loss-of-echo is configured it holds the last value or goes to alarm, which means a high-high trip built on that point does not exist. E+H's published answer is a configurable safety distance above the blocking distance, which makes the instrument output a warning or alarm before the level gets there. Configure it. And note the caution attached: if the blocking distance is undershot it may cause device malfunction.

If you need measurement in the top 300 mm of the vessel and cannot raise the transducer, ultrasonic cannot give it to you. Guided wave radar gets closer: the 5300 upper blind zone on a single lead probe is 3.9 in. (10 cm) on water. But there the nozzle genuinely does add, because Emerson states the nozzle height shall be added to it. A point level switch gets you the exact elevation you bolt it at.

What is the minimum dielectric constant a radar level transmitter needs to see the product?

There is no single number. Guided wave radar works down to 1.4 on a rigid single lead probe and 1.2 on a coaxial probe. Non-contact radar has no published floor at all. In every case the achievable minimum trades against distance and antenna size.

The physics is one equation. Reflected power at the surface is ((√εr − 1) / (√εr + 1))². That is 64% for water at εr 80, 3.8% for a crude at 2.2, 2.9% at 2.0, and 0.70% at 1.4. Emerson says the same plainly in its interface technical note: on a product such as oil with a dielectric of 2, less than 5% of the signal is reflected back to the transmitter, and the rest travels through.

Rosemount publishes the trade-off for the 5300 directly. A rigid or segmented rigid single lead probe in standard construction is rated to 1.4, or 1.25 if installed in a metallic bypass or stilling well; in HTHP construction those become 1.6 and 1.4. A coaxial or large coaxial probe goes to 1.2 in standard construction, 1.4 in MTMP, HP or C, and 2.0 in HTHP coaxial. A flexible single lead is rated 1.4 up to 49 ft (15 m), 1.8 up to 82 ft (25 m), 2.0 up to 115 ft (35 m), 3 up to 138 ft (42 m), 4 up to 151 ft (46 m), and 6 for the full 164 ft (50 m). Same instrument, same product. Different answer depending on how far you are asking it to look. Emerson notes that probe end projection improves the minimum measurable dielectric, and says to consult the factory for that.

Watch the length limits while you are choosing. The coaxial and large coaxial probes that buy you the 1.2 rating top out at 19 ft 8 in. (6 m) of measuring range. Rigid single lead runs 3 m, 6 m or 10 m depending on rod diameter. Only the flexible single lead reaches 50 m, and that is the probe with the worst dielectric rating at length.

For the non-contact 5408, Rosemount publishes range by antenna instead of a single minimum dielectric. The solids table is the clearest example. On light powder at a dielectric constant of 1.2 — the datasheet's example is plastic powder — a 1½-in. cone reaches 16 ft (5 m), a 2-in. cone 16 ft (5 m), a 3-in. cone 49 ft (15 m), a 4-in. cone 66 ft (20 m), and the 8-in. parabolic 115 ft (35 m). Step up to plastic pellets at 1.35 and the same antennas give 10 m, 10 m, 20 m, 30 m and 55 m. From the 2-in. cone to the parabolic, antenna choice is worth about seven times the range on the same low-dielectric product. Emerson calls these guidelines and lists what moves them: how the product piles up, silo diameter against angle of repose, internals, dust, condensation and antenna build-up.

There is no published minimum dielectric for a free-space 5408, and no honest rule of thumb either — Emerson's own table measures a dielectric of 1.2 at 35 m with the right antenna and gives up at 5 m with the wrong one. Quote the antenna-versus-range table for your model and product instead of a number someone remembers.

Working values: water 80, ethanol about 24, acetone about 21, toluene 2.4, crude oil 2.1 to 2.4, diesel about 2.1, gasoline about 2.0, liquid nitrogen about 1.45. For the LPG family, published radar-selection charts group non-conducting liquefied gases at 1.4 to 1.9 and put both propane and butane at about 1.5. These are chart values. They move with temperature, so use the number at process temperature when the margin is thin. Below about 1.4 at any real distance radar is a fight, and differential pressure, a displacer, or a coaxial probe in a chamber will serve you better.

What's the difference between guided wave radar and non-contact radar level transmitters?

Guided wave radar sends the pulse down a probe that touches the product. Non-contact radar radiates it through the vapor space from an antenna. The probe is why guided wave handles low dielectrics, foam, and narrow vessels. It is also why guided wave is wrong for anything that coats, crystallizes, or pulls hard on it.

Guided wave radar is time domain reflectometry. The pulse travels along a probe. On a coaxial probe almost all of the energy stays inside the outer tube, so a weak reflector still returns a usable echo. A single lead radiates more of its field into the vessel — higher minimum dielectric, and it needs clearance from nearby metal. The 5300 ratings are that confinement written down: 1.2 on a coaxial probe against 1.4 on a rigid single lead in open service, and 1.25 on that same single lead once it sits inside a metallic bypass or stilling well. The metal around the probe is doing the work either way. Guided wave also has no beam to steer, so it works in a small chamber and beside internals. Emerson's own note on the 5408 is that guided wave radar is generally the better fit for chambers.

Non-contact radar has nothing in the tank. Nothing to coat, nothing to break, nothing whose length limits the range: a Rosemount 5408 reaches 130 ft (40 m) in its standard performance class, or 492 ft (150 m) with the extended range option. The price is a beam, and the beam has to see the surface. Antenna size sets that: 22° on a 1½-in. cone down to 4.5° on an 8-in. parabolic.

The guided wave limits are mechanical, and they are published. Maximum recommended viscosity on a 5300 is 500 cP on a coaxial probe, 1500 cP on large coaxial, and 8000 cP on single lead. On contamination and build-up, Emerson allows build-up on single lead, allows thin build-up but no bridging on large coaxial, and does not recommend coaxial at all. Range is the other limit: coaxial and large coaxial stop at 6 m. In agitated service the probe needs a tie-down if it can move to within 30 cm of an agitator or other object.

• Choose guided wave for low dielectric, interfaces, chambers and bridles, small or obstructed vessels, and foam you want to exclude mechanically.
• Choose non-contact for coating, crystallizing, or abrasive products, agitation strong enough to move a probe, tall vessels, and sanitary service. Emerson's process seal antenna, all-PTFE wetted, is the one it names for heavy condensation and build-up.

Neither one is automatic on a boiler drum bridle in steam. That is guided wave with vapor compensation configured — Emerson sells the 5301 with Dynamic Vapor Compensation for exactly this — or differential pressure with the wet leg maintained. Which one depends on what your site can actually keep in calibration.

Radar vs ultrasonic level sensor — which one should I use for my tank?

Use ultrasonic only in a vented, near-ambient tank with a clean vapor space and a calm surface. Everything else is radar.

The split is not accuracy. It is what the pulse travels through. Ultrasonic measures time of flight at the speed of sound, and the speed of sound belongs to the vapor, not to the instrument. Air is 343 m/s at 20 °C and 331 m/s at 0 °C — about 0.6 m/s per °C, or roughly 0.18% of the reading per °C near room temperature. On a 10 m distance an uncompensated 10 °C error is about 180 mm. Nearly every ultrasonic transmitter compensates with a temperature sensor at the transducer. That sensor reads vapor at the top of the tank, not the average along the path. Low-cost OEM sensors often have no compensation at all, and those carry the full error.

Vapor composition is worse, and nothing compensates for it. At 0 °C sound travels about 259 m/s in carbon dioxide and about 430 m/s in methane, against 331 m/s in air. Those are textbook values, not manufacturer-published, but the direction and the size of the problem are not in doubt. Endress+Hauser draws the line by vapor pressure. Below 50 mbar (1 psi) at 20 °C, ultrasonic measures with very high accuracy, and E+H names the media that qualify: water, aqueous solutions, water-solid solutions, dilute acids such as hydrochloric and sulfuric, dilute bases such as caustic soda, oils, greases, slurries and pastes. High vapor pressures and outgassing media are named as the failure case, with ethanol, acetone and ammonia given as the examples, and there the instruction is to contact the vendor.

Radar cares about dielectric constant. Not density, not temperature, not the speed of sound — and it works in vacuum, where ultrasonic has nothing to travel through at all. The exception is a dense vapor space, high-pressure steam or heavy hydrocarbon vapor, where the vapor's own dielectric slows the microwave pulse and the device needs vapor compensation configured. Instrument accuracy under reference conditions is ±2 mm on a 5408 in its standard performance class and ±1 mm in ultra accuracy, against ±0.25% of measured distance above 1 m on a Rosemount 3102 or 3105 ultrasonic, which is 20 mm at 8 m. The entry-level 3101 is ±0.5%, twice that. Range is 40 m on a 5408, or 150 m with the extended range option, against 11 m on the 3102 and 3105 and 8 m on the 3101.

Ultrasonic is still the right instrument on an open sump, a lift station, a day tank of water, or open channel flow over a flume or weir — clean water, ambient temperature, no foam, no vapor, short range. On a 3 m water sump with a calm surface there is no reason to spend radar money.

How do you measure an oil-water interface level with guided wave radar?

Configure the device for level and interface, enter the upper product dielectric constant, then confirm the oil layer is thick enough and low enough in dielectric for the pulse to reach the water. Skip that last step and the interface number is arithmetic, not measurement.

The oil returns a weak echo — under 5% at a dielectric of 2, in Emerson's own words — and passes the rest. The water beneath returns about 64% at a dielectric of 80. The device times both pulses, and uses the configured upper product dielectric to correct for the pulse being slowed by √εr through the oil. Get that constant wrong and the interface reading is wrong in proportion. Emerson's troubleshooting chart lists an incorrect upper product dielectric constant as a direct cause of an incorrect interface level reading, alongside an incorrect interface threshold.

Rosemount's published criteria for the 5300 are the whole application review:

• Upper product dielectric must be known and must not vary, and it must be lower than the lower product's. The typical target Emerson names is an upper product below 3 and a lower product above 20.
• The difference between the two dielectric constants must be greater than 6 on a 5300. Emerson's interface technical note gives greater than 10 for the 3300 and the 3308.
• Maximum upper product dielectric is 7 for single lead probes and 10 for coaxial. Note the conflict in Emerson's own library: the 2014 interface technical note gives 8 for single lead, 10 for coaxial and 7 for rigid twin lead. The current datasheet supersedes it. Where two vendor documents disagree, the one matching your device revision wins.
• Minimum detectable upper product thickness on the current 5300 datasheet is 1 in. (2.5 cm) on large coaxial, 2.4 in. (6 cm) on single lead, 2.8 in. (7 cm) on standard, MTMP, HP and C coaxial, and 8 in. (20 cm) on HTHP coaxial. The 5.1 in. (0.13 m) figure still quoted in the field comes from the 2014 technical note and is superseded. Worth knowing: it is the difference between an application that passes review and one that does not.
• Maximum upper product thickness comes off the manufacturer's chart. Emerson's worked example for a flexible single lead 5300: with an upper product dielectric of 2 over a lower product of 20, the maximum upper product thickness is 62 ft (19 m), and the maximum distance to the interface is 164 ft (50 m) minus that thickness.

Pick the model to match the installation: a 5301, which Emerson describes as liquid level or interface with a fully submerged probe, or a 5302 for liquid level and interface together.

In a chamber, watch for a trapped gas pocket. The device times that pocket as if it were oil. The pulse moves faster in gas than in oil, so the distance is under-read. With the upper dielectric set to 2 the error across the pocket is 29%: 40 cm of trapped gas is timed as 28 cm, and everything below it — the interface included — reads about 12 cm high. That figure is arithmetic from velocity = c/√εr, not a quoted vendor statement, but the direction is not in question. Vent the chamber. And remember that a stagnant bridle separates on its own schedule. The interface in the pipe is not the interface in the vessel unless there is real flow through both connections.

Will guided wave radar still work if there's a thick emulsion layer between the oil and water?

It will keep reporting a number. Past about 2 in. (50 mm) of emulsion, that number is often the top of the rag layer rather than the oil-water interface.

Emerson's own guideline is that emulsion layers smaller than 2 in. (50 mm) are quite manageable and will not prevent the measurement of the interface. Above that, the company's wording is that the results vary with the fluid mix and that in many cases the interface is measured at the top of the emulsion layer.

Guided wave radar needs a step change in dielectric to reflect from. A rag layer is a gradient, climbing from about 2 to 80 across its thickness. No single impedance step, so the energy spreads across the layer, and part of it is absorbed by the conductive water carried in it. Emerson's staged illustrations show the progression: a well-mixed, uniform emulsion returns no interface at all; a mix with no distinct interface leaves the device distinguishing only the upper product level; both level and interface echoes reappear once a bottom product layer starts to form. Separate further and you get three layers, with the emulsion itself showing as a visible echo between them.

Solids make it worse. Emerson's description is that emulsion can form from fine solids combined with emulsified oil and water, that stable liquid emulsion plus solid particles trigger rapid emulsion layer growth, and that the more particles are present, the larger the emulsion.

A stilling well is worth fitting, and not only for turbulence. Emerson says a stilling well provides more robust measurements because of the higher degree of liquid separation inside the well, and calls that helpful where an emulsion layer is present. Its own installation note recommends 4 to 8 in. pipe with multiple holes to ensure flow-through. It is a separation aid, not a solvent. A stagnant chamber can hold a rag layer after the vessel has cleared, so tie the chamber into real flow at both connections and compare against tap samples before you trust it.

If the measurement you actually need is where the rag layer sits and how thick it is — a desalter, a wash tank, a heater treater — no two-echo device gives you that. Guided wave radar is not the instrument. A profiling gauge is: a multi-point capacitance or admittance profiler, or a gamma density profiler, both of which report a profile from top to bottom. That is a different instrument from a different vendor, and on a desalter it is the right one.

When this page runs out

This page is reference material and it stops where your echo curve starts. It cannot tell you which pulse your transmitter is tracking, and that is the one thing that decides most of these calls. Every vendor number above is quoted from a named document: the Rosemount 5408 product data sheet 00813-0100-4408, the Rosemount 5300 product data sheet 00813-0100-4530 and its reference manual 00809-0100-4530, Emerson's guided wave radar interface measurement technical note 00840-0400-4530, the Rosemount 3100 series product data sheet 00813-0100-4840, and Endress+Hauser technical information TI00440F for the Prosonic T FMU30. Revisions move, and as the interface figures above show, sometimes a long way. Check yours. If the problem needs a waveform read against a specific firmware revision and the factory's application database, the manufacturer's application support group is the right call — most do this at no charge, in warranty or out, but confirm that rather than assume it. If what you need is a rag layer profile rather than a single interface number, that is a profiling gauge vendor. If the requirement is a custody transfer tank gauging system to API standards, that is a certified tank gauging provider, not a process instrumentation shop, and Bridges Industrial will say so.