A Practical Guide to Ultrasonic Transducer Selection for Industrial NDT

---

Every ultrasonic inspection begins with getting sound into the part. That is the transducer's job. The right way to do it changes with the material acoustic properties, the part geometry, the surface condition, the object temperature, and the access window.

The right transducer is the one that matches those conditions: the coating that cannot be stripped, the steel that is still hot, the concrete that will not hold couplant, the weld that is accessible from one side only. The three core transducer families, piezoelectric, EMAT, and dry point contact, exist because no single generation mechanism handles all of these conditions at once. This guide covers how each family works and is built, the inspection methods that pair with each, and a practical framework for selecting a transducer for a specific inspection task.

Why transducer choice still defines an ultrasonic inspection

An ultrasonic inspection is defined first by its method. The opening questions are what is being inspected, which defects matter, and which technique finds them: pulse-echo, phased array, tomography, or guided wave. The method is the first decision. The transducer comes second, selected to deliver that method on the surface and conditions you actually have. That second choice is the subject of this article, because once the method is fixed, the transducer is what determines whether it works.

It helps to separate two things that often get treated as one. Ultrasonic testing is the method. The transducer is the accessory that makes the method work on a real surface. Point two probes built on different physical principles at the same defect, drive them with the same instrument, and they will not return the same result. The instrument only reads what the transducer gives it.

The transducer is also the only part of the measurement chain in contact with the test object. It works on the surface as it actually is, with its coating, temperature, curvature, and access window. The instrument and the settings can be changed during the work. If the transducer does not match the conditions, no setting recovers the inspection.

This is why there is no single best ultrasonic transducer. The core ultrasound generation methods covered here, piezoelectric, EMAT, and DPC, exist because no single mechanism wins on surface condition, temperature, sensitivity, and access at the same time. Each family is the strongest answer in a different part of that trade-off.

The sections that follow set out the inspection methods first, then take each transducer type in turn: what is the transducer type and operating frequency, how the families compare, and how to choose for the job in front of you.

How ultrasonic transducers work, the fundamentals

Industrial ultrasonic NDT works in a band that runs from about 20 kHz to 25 MHz. Where a job sits in that band is set mostly by the material. Concrete and other coarse, heterogeneous media force you to the low end, in the tens of kilohertz, because lower frequencies survive scattering. Thin metal sections and aerospace composites sit at the high end, from several megahertz up to 25 MHz, where short wavelengths resolve small features. The rule behind that spread is straightforward. Higher frequency gives finer resolution but less penetration, and the transducer is what sets the usable frequency.

Every transducer does the same job twice. It turns an electrical pulse into a mechanical wave inside the part, then turns the returning wave back into an electrical signal. What separates the four ultrasound generation methods is how they make that conversion. Piezoelectric transducers use an active ceramic in direct contact with the part. EMAT uses electromagnetic fields, through the Lorentz force and magnetostriction, to generate the wave inside conductive metal. DPC transducers press a hard, spring-loaded tip against the surface and couple mechanically. A fourth method, laser excitation, generates the wave thermally with a short light pulse; it appears in the comparison tables for completeness. The three families each get their own section below.

The wave mode matters as much as the mechanism. Four modes carry most NDT work. Longitudinal waves move particles along the direction of travel and are the fastest, about 5,920 m/s in carbon steel [1]. They handle thickness gauging and most straight-beam flaw detection. Shear waves move particles across the direction of travel at roughly half that speed, about 3,255 m/s in steel [1], and are the basis of angle-beam weld inspection. Surface, or Rayleigh, waves stay within about one wavelength of the surface and find surface-breaking cracks. Guided waves, the Lamb and shear-horizontal modes in plates and the torsional and longitudinal modes in pipes, travel along the structure itself and can screen long runs of pipe from a single position. Some transducers produce a mode naturally. Others need a wedge or a shaped coil to get there.

How a transducer launches the wave decides where the energy goes. Normal-beam excitation drives a localized beam for thickness measurement, while angle-beam and comb excitation set up guided waves that travel along the structure.
Image source: GW Ultrasonics

All of this depends on getting the wave across the boundary and into the part, and this is the largest physical obstacle. Acoustic impedance is a material's density times its sound velocity, and the gap between a solid and air is enormous. Steel sits around 45 × 10^6 Pa·s/m. Air sits near 415. At that mismatch, more than 99.9% of the energy reflects at a solid-air interface before any of it enters the part [2]. This is the coupling problem, and it runs through every section that follows. Each family is defined by how it solves or sidesteps that interface, whether with a liquid couplant, a dry contact tip, an electromagnetic field, or a laser. In the field today, most inspections still solve it the oldest way, with liquid couplant. The chart below shows how common each coupling approach is in practice, not how capable each one is.

Relative prevalence of coupling approaches in field ultrasonic NDT. Liquid gel and immersion coupling dominate. Dry contact arrays, EMAT, and air-coupled methods make up the smaller share.

Inspection methods and how they pair with each transducer type

An inspection is specified by its method before any probe is chosen. Engineers ask for a method by name, and each method runs on a particular set of transducers. This section sets out the methods. The transducer families that deliver them are covered in the sections that follow. The chart below places each family by operating frequency and the material it typically serves.

Operating frequency and typical application by transducer type, from 25 kHz to 10 MHz. Low-frequency air-coupled and DPC transducers cover composites and concrete, while metal testing sits higher up the band.

Pulse-echo is the default. One transducer sends the pulse and reads the echo, and it works on piezoelectric, EMAT, and DPC alike. It carries most thickness gauging, weld joint inspection, and corrosion mapping.

Pitch-catch and through-transmission split the job across two transducers, on the same face or on opposite faces. They earn their place on highly attenuating parts such as honeycomb composites, where a single-sided echo does not have the signal to spare.

Angle-beam testing tilts the wave to catch defects a straight beam would miss, the classic setup for welds. Piezo does it with a wedge that mode-converts the longitudinal wave to shear. EMAT does the same thing through coil geometry, with no wedge.

Phased array (PAUT) drives a multi-element array with timed delays to steer, focus, and sweep the beam electronically, with no moving parts. It is the dominant modern method for weld inspection and corrosion mapping. EMAT arrays can work this way too, though the technique is less mature there.

Time-of-Flight Diffraction (TOFD) places two longitudinal-wave probes either side of a weld and times the signals diffracted off crack tips. It sizes through-wall flaw height accurately and independently of orientation, with a near-surface dead zone that pulse-echo or PAUT has to cover.

Full Matrix Capture and the Total Focusing Method (FMC/TFM) fire each array element in turn while all elements receive, then focus computationally at every pixel rather than at one fixed depth. That gives more uniform resolution than physical-focus PAUT. It is standard with DPC arrays for concrete tomography and increasingly standard with PAUT for welds and additive manufacturing.

Ultrasonic tomography builds a three-dimensional image from a dense array and FMC/TFM reconstruction, the standard approach for concrete from one side, with immersion piezo versions used on production parts.

Guided wave testing clamps a ring of transducers around a pipe and launches torsional or longitudinal modes that travel along the wall. From a single position it screens tens of metres of pipe, around 30 m each way in typically corroded refinery line and further in cleaner pipe [3].

The matrix below shows which method runs on which transducer family at a glance.

Knowing which method runs on which transducer is half the picture. The other half is the transducers themselves: how each family generates sound, where it excels, and where it does not. The sections that follow take each family in turn, starting with the piezoelectric probe.

Piezoelectric transducers, the industry workhorse

The piezoelectric probe solves the coupling problem with a thin film of liquid. It is the industry default, and most ultrasonic inspection still starts with it.

How piezoelectric transducers work

A piezoelectric element converts electricity into sound and back. Apply a short voltage pulse and the element strains and launches a wave. When the echo returns, the same element turns that strain back into a voltage the instrument can read. That two-way conversion is the direct and inverse piezoelectric effect, and it is the whole basis of the probe.

The active material sets the starting performance. Most general-purpose probes use PZT ceramic, which is robust, low cost, and has a high piezoelectric coefficient. Dicing that ceramic and refilling the gaps with epoxy makes a 1-3 piezocomposite, which lowers the acoustic impedance toward that of the part, widens the bandwidth, and cuts cross-talk between array elements. Composite probes deliver 3 to 20 dB more sensitivity than standard ceramic probes, along with a shorter pulse, depending on the application [4]. Single-crystal PMN-PT pushes sensitivity and bandwidth further still, with measured gains of 5.2 dB in relative sensitivity and more than 27.7% in -6 dB bandwidth over a comparable PZT ceramic array [5]. It is also fragile, depolarizes under the high drive voltages heavy industrial work demands, and has a lower temperature ceiling, so it stays mostly in high-frequency, moderate-power arrays rather than general field use. For hot work, high-temperature materials such as bismuth titanate and lithium niobate, along with proprietary high-temperature ceramics, extend operation toward 550 °C.

Around the active element sits a stack that shapes the pulse. The same impedance gap from the last section applies here, so a quarter-wavelength matching layer at the front steps the wave into the part, and its impedance is engineered, often as a bespoke composite, rather than picked off a shelf. A heavy tungsten-loaded backing block at the rear damps the ring. Heavier damping shortens the pulse and sharpens axial resolution, lighter damping holds the ring and lifts sensitivity at the centre frequency, so the designer picks a point on that curve to suit the inspection. A wear face protects the front, and the housing and cable close out the build.

Inside a single-element piezoelectric contact probe. The active stack sets the performance: the piezoelectric element generates the wave, the matching layer improves transmission into the part, the wear face protects the front, and the backing block damps the ring to keep the pulse short.

Configurations

The same element appears in several forms. A single-element straight-beam probe is the default for thickness work. A dual-element probe splits transmit and receive onto separate crystals behind a delay line, which kills the near-surface ringdown and makes it the standard for corrosion surveys on rough or pitted metal. An angle-beam probe sits on a plastic wedge that mode-converts the longitudinal wave into a refracted shear wave in steel, the classical setup for weld inspection. A phased array packs 16, 32, 64, or 128 elements into one housing and steers and focuses the beam electronically, with no moving parts, and it is the foundation for Total Focusing Method imaging.

Where piezoelectric genuinely excels

On clean, ambient-temperature metal, piezoelectric gives the best absolute sensitivity of any family at the lowest cost per reading. It works across the widest range of materials, metals, plastics, composites, and glass, which no other family matches. The installed base is enormous, so trained operators, calibration blocks, and written procedures already exist almost everywhere, and that ecosystem is itself a real advantage. Phased array adds fast corrosion mapping and full weld coverage at code-required quality under standards such as ASTM E2700 and ISO 17640. For most inspections on accessible metal, piezoelectric is the benchmark the other families are measured against.

Limitations

The limitations trace back to the couplant and the surface under it. Gel drips, freezes, boils off, and contaminates clean parts. Real assets need grinding, brushing, or paint removal before a single reading. Standard probes fade above about 50 °C, and high-temperature delay-line setups that reach 500 °C demand brief-contact technique, repeated re-zeroing, and special couplants. Coarse-grained materials are the other classic problem. Cast austenitic stainless and dissimilar-metal welds scatter the beam so badly that the standard answer is a low-frequency longitudinal-wave dual-matrix array probe at 1.5 to 3.5 MHz. Those frictions, the couplant, the surface prep, and the temperature ceiling, are exactly the constraints the next family removes by generating the wave inside the metal itself.

How EMAT works

An EMAT has two parts, a radio-frequency coil and a static bias magnet, and the magnet can be a permanent magnet or an electromagnet. The coil sits close to the surface and drives a high-frequency current. That current induces eddy currents in the skin depth of the metal, and those eddy currents interact with the bias field through the Lorentz force to launch a wave directly inside the part. Nothing is pressed against the surface and no couplant carries the wave across. The metal itself becomes the source.

On ferromagnetic metal, carbon steel and most pipework, a second mechanism joins in. The alternating field makes the steel's magnetic domains change shape slightly, an effect called magnetostriction, and that adds to the wave. For a PPM-EMAT on mild steel, the magnetostrictive contribution has been modelled and measured at roughly 55% of the Lorentz contribution in the static case [6]. It is a major second source, not the dominant one, and it is why EMAT signals on steel are usually stronger than on aluminium, where only the Lorentz mechanism is available.

EMAT, generating sound inside the metal

EMAT removes the couplant from the equation. Instead of pushing a wave across the surface from outside, it generates the wave inside the metal itself, and that changes what an inspection can do.

EMAT makes the wave inside the metal. A coil induces eddy currents in the surface, and those currents interact with a bias field, from a permanent magnet through the Lorentz force. No couplant crosses the gap.

S3950 EMAT demonstration:

Why no couplant changes the operational envelope

Removing the couplant removes the biggest bottleneck in field inspection. EMAT reads through paint, light coatings, rust, scale, and oxide, so most of the surface prep that piezo demands goes away. It tolerates a small lift-off, typically zero to 3 mm in the field and up to about 10 mm in the lab, though the signal falls off quickly as the gap grows. The peak frequency also shifts about 100 kHz for every millimetre of lift-off on a Rayleigh-wave EMAT [7], which the instrument has to track.

Temperature is the other envelope change. Liquid couplant gives out around 150 °C. EMAT keeps working well past that because nothing touches the hot surface but the probe face and its standoff. Thickness measurements on low-carbon steel have been demonstrated at 450 °C with no cooling at all [8], and water-cooled designs push toward 1,000 °C. That puts live refinery lines, boiler water-walls, hot-rolled steel in the mill, and insulated in-service pipework within reach without a shutdown.

Coil geometries and wave mode selectivity

EMAT picks its wave mode from the shape of the coil and the direction of the bias field, not from a wedge. A spiral pancake coil under a normal field drives a radial Lorentz force and makes normal-incidence shear waves for thickness gauging and corrosion mapping. A racetrack coil under periodic permanent magnets selects Lamb modes, with the symmetric or antisymmetric mode set by where the wires sit relative to the magnets. A meander-line coil fixes the wavelength by its pitch and generates angled bulk waves, Rayleigh surface waves, or specific guided modes. The practical payoff is that shear-horizontal waves, which are hard to make with piezo in a scanning setup, are routine with EMAT. The same magnetostrictive effect drives a close cousin, the magnetostrictive transducer, which launches guided waves along a pipe to screen tens of metres from a single position.

S7392 EMAT demonstration:

Permanent magnets vs pulsed electromagnets

This is where the bias source becomes a field decision, not just a design detail. Permanent magnets, usually NdFeB or samarium-cobalt for higher temperature, made early EMAT practical. They also drag hard on steel, so the inspector cannot slide the probe smoothly, they collect shavings, scale, and debris on the face between readings, and the NdFeB grade weakens as the surface heats toward its Curie point.

A pulsed electromagnet turns the bias field on only for the microsecond of the measurement. Between pulses there is no field, so there is no magnetic drag during a scan, nothing pulls debris onto the face, and there is no permanent magnet to weaken when the steel is hot. ACS builds both. Permanent-magnet probes remain the right answer for point measurements and lower-cost work, and pulsed-electromagnet probes are the path for continuous scanning on hot or dirty steel.

Permanent magnet versus pulsed electromagnet. The permanent magnet is simple and proven but drags and collects debris. The pulsed electromagnet energises only during the measurement, so the probe slides freely and stays clean.

5e Limitations

EMAT's limitations are physical and well documented. Generating sound electromagnetically is less efficient than the piezoelectric effect. The one-way insertion loss is 40 dB or more under ASTM E1774-17, which means a lower signal-to-noise ratio than piezo on the same reflector. Higher pulser voltages, broadband matched receivers, and tone-burst excitation claw some of that back, but the floor is real. EMAT also works only on conductive metal, so plastics, composites, glass, and ceramics are out. There is a dead zone on the first echo, around 6 mm in steel, so thickness work reads between the second and third back-wall echoes. The probes cost more than piezo, and the permanent-magnet versions are heavier. For the field-by-field comparison of EMAT and piezoelectric on thickness gauging, see our article Ultrasonic Thickness Gauging. None of this changes the basic fact that EMAT works only because the part is metal. Step onto concrete and neither couplant nor electromagnetics helps, which is the problem the next family was built to solve.

Dry-point-contact (DPC) transducers and arrays

Neither of the families above works on concrete. It does not conduct, so EMAT has nothing to work with, and in many cases it is too rough and porous for liquid couplant to stay put. Inspecting concrete from one side needed a different idea, and dry point contact is it. This is also where ACS Group's own published research sits, so this section leans on it directly [9].

Why concrete demands a different transducer

Concrete is not a uniform solid. It is a matrix of cement paste and coarse aggregate, shot through with air voids, rebars, and ducts. High-frequency ultrasound scatters off all of that, and the signal-to-noise floor collapses before the wave reaches anything useful. The fix is to drop the frequency. Working in the low range, around 50 kHz for concrete arrays and broadly from tens of kilohertz upward, keeps the wavelength larger than the typical aggregate, so the wave travels through the matrix instead of bouncing off every stone [9].

Couplant is the second problem. A liquid film cannot hold on a rough, porous surface, because the gel drains into the pores and the contact turns inconsistent. The dry point contact answer is a hard, wear-resistant ceramic tip, spring-loaded, pressed onto the surface. The tip is much smaller than the wavelength, so it acts as a single point that bridges the surface roughness and injects energy through a small, high-pressure contact, with no liquid and no surface prep [9].

Why shear waves matter for concrete inspection

A useful property falls out of the dry-point-contact mechanism. It naturally generates shear waves as well as longitudinal ones [9]. That matters because cracks behave as directional reflectors, and shear waves read them cleanly. A shear wave diffracts around a crack tip without converting to other modes, so the return signal stays readable and the depth of a vertical crack can be measured from a single side. Longitudinal waves split into secondary modes on the same defect and muddy the picture. Clean shear-wave crack sizing from one side is a capability ordinary piezo on concrete cannot match.

From a single tip to an array

One tip at low frequency has coarse spatial resolution. The answer is an array of them. Each element rides on its own spring, so the array keeps contact across a rough or curved surface without losing elements. The array then runs Full Matrix Capture, where every element transmits in turn while all elements receive, and the recorded set is reconstructed with the Total Focusing Method, which focuses at every image point rather than at one fixed depth. In concrete, plain TFM is usually backed by SAFT, the synthetic aperture technique that holds up in the noisy, scattering conditions concrete creates [9]. Half-skip TFM adds a supplementary path that bounces off the back wall to size vertical surface-breaking cracks more reliably than direct imaging [10]. The result is a real-time three-dimensional image of the inside of the structure, built from one side, showing rebar, voids, ducts, delaminations, the back wall, and cracks. The highest-end ACS arrays carry up to 64 independent elements, confirmed by ACS. ACS has published the 3D-FMC tomography behind this imaging in more technical depth.

DPC array tomography in action and on the working surface. Left: the face carries a grid of independently spring-loaded dry-point-contact elements, while the display shows the reconstructed image of the concrete interior built by FMC and TFM. Right: the same class of array in the field, where the chalk grid marks scan positions and each placement reconstructs a slice of the interior from one side, with no couplant and no surface prep.

Where DPC tomography is the only practical option

Most civil structures give you one side to work from, and that is the case DPC tomography is built for. Bridge decks, tunnel liners, slabs, and walls all get inspected from the surface you can reach. It verifies grouting in the tendon ducts of post-tensioned structures, finds voids and honeycomb in load-bearing elements, and gauges slab thickness where there is no access to the far face. For the full concrete inspection workflow alongside the other methods, see our article NDT of Concrete in Practice.

Ultrasonic Pulse Velocity Testing (UPVT) versus Ultrasonic Pulse Echo Tomography (UPET)

Two methods get confused here, and they are not the same. Ultrasonic pulse velocity testing, or UPVT, sends a pulse from a transmitter on one face to a receiver on another and times it. The result is one velocity number, an integrated indicator of concrete quality and uniformity. Ultrasonic tomography, also called ultrasonic pulse-echo tomography, is an imaging method returning 2D / 3D representation of the inspection volume rather than a single number. They answer different questions, and ACS builds instruments for both. They are not substitutes for each other.

Limitations

The low frequency that makes concrete workable also caps the resolution. DPC tomography resolves features at the centimetre-to-decimetre scale, not the millimetre scale. The surface can be rough, but it still cannot be loose, flaking, or covered by decorative finishing materials. Penetration depends heavily on the concrete itself. Up to about 2 m is typical, and the highest-end ACS array reaches roughly 6 m in good-quality concrete with signal boosting, confirmed by ACS, while a heavily attenuating mix or / and multiple rebar layers cuts that down. DPC also still asks for a hand on the surface, one position at a time. The next section covers two families that drop even that, one by crossing the air gap without contact, the other by staying permanently in place.

Air-coupled and embedded transducers

Two more families drop contact with the surface entirely. One crosses the air gap without touching the part, and the other stays bonded in place for years. ACS builds both, and they cover jobs the contact families cannot.

Air-coupled transducers

Air-coupled transducers send the wave through the air itself, with no contact and no couplant. The obstacle is the same impedance mismatch from the fundamentals, except the wave now crosses two air-to-solid boundaries instead of one, so the total path loss runs to 100 to 150 dB [11]. Designs fight that with heavily impedance-matched piezocomposite elements, increasingly with micromachined elements, and with high-power tone-burst excitation and pulse compression to recover signal. Because air damps high frequencies hard, air-coupled work stays low, roughly 50 to 800 kHz [11], and most field setups use through-transmission, with a transmitter on one side and a receiver on the other. That needs access to both sides of the part, which suits panels moving down a production line but rules out one-sided field work. Single-sided air-coupled pulse-echo is possible but harder.

Its strength is parts that cannot take a liquid. Foam-cored sandwich panels, CFRP and GFRP composites, ceramics, wood, and food or packaging all inspect well with air coupling, and it picks up delaminations, disbonds, and impact damage in layered structures where immersion is not acceptable or the geometry is awkward.

Transducer Arrangements for Air-Coupled Ultrasonic Testing