A cheap air ultrasonic sensor can work underwater through a wall, but only with its transducer face bonded rigidly to that wall and no air left in the path. An air gap reflects about 99.9% of the pulse at the water boundary, while a bonded 1/8 in [3 mm] acrylic window passes most of it through. Sound also runs 4.3 times faster in water, about 1480 m/s against 343 m/s in air, so every air-calibrated reading needs rescaling before you can trust it. The build below picks the acoustic face, bonds the sensor, seals the cable entry, rescales the timing and wet-tests the result. For the electronics, our 304 and 316L stainless bodies take whatever cutout the build needs.
Can an air ultrasonic sensor work underwater through a wall?
Yes, if you bond the transducer face rigidly to the wall or to a bonded window and leave no air anywhere in the path: sound crosses a bonded 1/8 in [3 mm] acrylic window with about 90% of its energy, while a dry air gap returns almost the whole pulse.
Nothing about the piezo disc needs to touch the water; it needs a rigid solid path and a rescaled clock. A typical air rangefinder pings at 40 kHz, which is a 37 mm [1.5 in] wavelength in water, so returns come from large, flat targets rather than thin edges. Water loading also damps the disc: the burst leaves weaker and the ring-down finishes sooner than in air. Expect a few metres of honest range, not the 4 m [13 ft] the module quotes in air. The 90% window figure is a thin-plate estimate at 40 kHz, water on both faces.
Why does the pulse need a rigid bond with no air gap?
Acoustic intensity crosses a boundary in proportion to how closely the two media match: air sits at about 0.000413 MRayl of impedance and water at 1.48 MRayl, so an air-to-water boundary reflects about 99.9% of the pulse and passes almost nothing.
Impedance is density times sound speed, and the reflected share at a flat boundary is the square of (Z2 − Z1)/(Z2 + Z1). Water at 20 C [68 F] carries sound at 1480 m/s with a density of 998 kg/m³, about 1.48 MRayl (1.48 × 10⁶ kg/m²s). Air at the same temperature sits near 0.000413 MRayl, almost 3600 times lower, so the boundary turns back almost the entire pulse and passes 0.1%.
Rigid epoxy cures to roughly 3 MRayl and acrylic sits near 3.2 MRayl, within a factor of two of water. That is why a hard bond passes most of the pulse where a soft one fails. A soft silicone pad works as a lossy spring: it absorbs and scatters instead of coupling, so the couplant you choose wants a hard, brittle cure. Treat every void as an air interface, because one bubble in the bond line reflects the whole pulse. Filled epoxies let you tune the number: a published transducer-backing recipe runs epoxy-to-powder weight ratios anywhere from 4:1 to 50:1.
What do you need before you start?
Gather a rigid-cure two-part epoxy, a 1/8 in [3 mm] acrylic disc, degreaser, clamping tape, a submersion-rated cable entry, and a bucket or tank deep enough to test at your real depth before the first dive.
- Two-part rigid epoxy with a hard, brittle cure; skip flexible or rubber-toughened formulas.
- An acrylic disc that overlaps the hole by 0.4 in [10 mm] or more all round; for a 0.63 in [16 mm] transducer over a 0.47 in [12 mm] hole, a 1.6 in [40 mm] disc sits comfortably.
- Isopropyl alcohol and a lint-free cloth for the transducer face and the bonding spot.
- Tape or a light clamp to hold contact through the cure.
- A submersion-rated cable entry: underwater cable entry with IP68 connectors, glands or potted penetrators compares the three routes.
- An O-ring retainer as mechanical backup, sized for 20-30% squeeze on the cord; O-ring squeeze for a sealed enclosure face gives the compression numbers.
- A tank deep enough for the working depth and the controller where you will rescale readings.
How do you bond the sensor to an acrylic window?
Cut the hole, epoxy the acrylic disc over it from the inside, then epoxy the bare transducer face to the center of the disc with a bond line under about 0.04 in [1 mm], clamped until the epoxy's full cure, commonly 24 h at room temperature.
- Have the hole cut with the housing where you can: the cutout designer places a 0.47 in [12 mm] hole on any face of a catalog or custom body, and custom stainless enclosures with factory cutouts compares cutting a standard box against a fully custom one. A machined edge bonds cleaner than a field-drilled one.
- Degrease both sides of the hole and the transducer face with isopropyl alcohol and let everything dry.
- Lay a continuous rigid-epoxy fillet around the rim, press the disc over the hole from the inside, and work the fillet until no silvered voids show. The fillet is the seal and the structure in one.
- Strip the metal grill or membrane off the transducer so the bare face stands proud, and degrease it again.
- Spread a thin epoxy film, press the face onto the disc center over the hole, and hold the bond line under 0.04 in [1 mm]; squeeze-out should ring the joint.
- Clamp lightly and keep the face flat against the disc. Give the full cure the epoxy datasheet calls for, commonly 24 h at 20-25 C [68-77 F].
- Bring the cable in through the submersion entry, leave a drip loop outside, and pot or clamp the strands; bare conductors wick water into the box along the strand gaps, as water wicking through cable strands explains.
How do you fix the timing for water?
Multiply every reading by 4.3, or set the speed constant to 1480 m/s where your controller exposes one: sound runs about 1480 m/s in fresh water at 20 C [68 F] against 343 m/s in air, so an air-calibrated module reports about 0.23 m for every true metre.
Work the numbers for a target 1.00 m [3.3 ft] out:
- travel time in water: 2 × 1.00 m ÷ 1480 m/s = 1.35 ms;
- the module converts with its air constant: 343 m/s × 1.35 ms ÷ 2 = 0.23 m displayed;
- correction factor: 1480 ÷ 343 = 4.3.
The dead zone grows by the same factor. Take a module that blanks 1.17 ms after transmit; its air dead zone is 0.20 m [0.7 ft], from 2 × 0.20 m ÷ 343 m/s. In water that blanking hides 1480 m/s × 1.17 ms ÷ 2 ≈ 0.87 m [2.9 ft]. Keep obstacles outside that radius, or shorten the blanking in your own code. Echoes rattling inside the 3 mm window return in about 2.2 µs, since 2 × 3 mm ÷ 2730 m/s = 2.2 µs, far inside any blanking period; the window cannot false-trigger the receiver on its own.
Which checks prove the bond and the seal when you finish?
Soak the housing at working depth for 30 minutes, dry the outside, and open it onto a dry paper towel; then check timing against a flat plate at 1.00 m, where the raw reading sits near 0.23 m and near 1.00 m after the 4.3 correction.
- Leak: soak for 30 minutes at working depth, where every 10 m [33 ft] of water adds about 1 bar [14.5 psi]; dry the housing, weigh it, or open it onto a dry paper towel. Any weight gain is a leak.
- Bond: hold the disc up to a light; a silvered ring or speck in the epoxy is a void, and the joint comes off and goes again.
- Timing: ping a 4 in [100 mm] flat plate at 0.50, 1.00 and 2.00 m. Raw readings land near 0.12, 0.23 and 0.46 m, since 0.50 m × 343 ÷ 1480 = 0.12 m, and land on the tape-measure values after the 4.3 factor.
- Receiver: if the module returns its minimum at every range, it triggers on the direct path across the disc; lengthen the blanking in your controller and ignore everything inside it.
- Electrical: run every wet test from a battery or an isolated low-voltage supply and keep mains out of the tank.
Which mistakes cost you the pulse or the seal?
Four failures account for nearly every dead build: air left in the coupling path, a soft rubbery bond, a void in the epoxy, and a cable entry that leaks or wicks; each one shows up in a bench test before launch.
- Grill or membrane left on the transducer: the trapped air layer restores the full air-gap reflection. Strip it before bonding.
- Silicone pad or soft urethane in place of epoxy: the couplant absorbs the burst. Use the hard, brittle cure.
- A void in the fillet or bond line: one bubble turns back the whole pulse. Work the epoxy and inspect it against a light before the soak.
- Bonding straight to a stainless wall: a 0.08 in [2 mm] stainless wall passes only about 36% of a 40 kHz pulse and a 0.12 in [3 mm] wall about 20%, on a thin-plate estimate at 40 kHz with water on both sides. Keep the acrylic window as the acoustic face and the stainless box as the structure.
- Fasteners through the wall below the waterline: every penetration is a leak path, and screws through a control panel wall covers the sealed-hardware practice.
- Dry strands at the cable entry: water follows them inside; pot or seal the conductor ends.
- A part-filled oil pocket, if you try that route: air above the oil blanks the pulse, so fill the pocket completely and mount it face outward.
When is a water-designed transducer the better buy?
Buy the water-designed transducer when depth, duty cycle or accuracy outrank cost: past about 10 m [33 ft] of depth or on continuous survey duty, the bonded-window build gives up; for shallow, intermittent prototypes it is the cheap and honest path.
| Case | Build that works | What holds underwater |
|---|---|---|
| Shallow prototype, short bursts | Bonded 1/8 in [3 mm] acrylic window, readings × 4.3 | The epoxy joint, proven by the soak test |
| Continuous duty past about 10 m [33 ft] | A transducer designed for water, matched face and timing | The transducer's own pressure design |
| Housing for the electronics | 316L stainless body with a factory-cut window hole and a submersion-rated cable entry | The penetrator joint and the body itself |
Our 304 and 316L wall-mount and HMI bodies are rated NEMA 4X / IP66 by the manufacturer; they are not UL listed. NEMA 250 defines Type 4X as hose-down with corrosion resistance, and Type 6 and 6P as temporary and prolonged submersion, so underwater the joint you build and test is the rating that counts, not the type number on the label. Order the body with the window hole already in place: cutouts, glands and HMI windows shows what the drawing needs, and the finished box is delivered in 10 business days, counted for custom work from your approval of the drawing.
Questions
Will a sealed window still work if the sensor sits in dry air behind it?
An air layer reflects about 99.9% of the pulse at the air-to-water boundary, and even a film a fraction of a millimetre thick does the same. A sealed housing with the sensor in dry air is silent underwater. The transducer face must bond rigidly to the wall or window, wetted over its full face by rigid epoxy with zero voids.
Can the transducer sit in oil instead of being bonded to the window?
Yes, and oil also pressure-compensates the pocket. Mineral oil carries sound at about 1.25 MRayl of impedance against water's 1.48 MRayl, so an oil-to-water boundary reflects under 1% of the energy; commercial sonar housings are oil-filled for this reason. Fill the pocket completely, because any air above the oil blanks the pulse, and keep the outer window thin and rigid.
Does sea water or cold water change the 4.3 factor?
Slightly. Sea water at 35 parts per thousand runs about 1500 m/s, so the multiplier moves from 4.3 to about 4.4, and temperature adds roughly 2.4 m/s per °C near room temperature, about 0.16% per °C. Calibrate the final factor against a flat plate at a known distance in the water you dive in.
What keeps the cable entry watertight at depth?
Use a potted penetrator or a gland whose datasheet states a submersion depth, because an IP68 stamp means immersion on conditions the manufacturer defines, not a depth you can assume. Match the clamping range to the cable diameter, leave a drip loop outside the box, and pot or seal the strand ends against wicking.
