Frequency is the first specification decision on any ultrasonic machine — and the one most buyers let the salesman make for them. It sets the size and violence of every cavitation bubble in the tank, which means it decides what cleans, what survives, and what disappoints. Ten minutes here saves years of the wrong tank.
A cavitation bubble grows during the low-pressure half of each sound cycle. Lower frequencies give it longer to grow, so it gets bigger — and a bigger cavity collapses with far more energy. Higher frequencies give bubbles less time, so they stay small, collapse gently, and form in vastly greater numbers, reaching into finer detail.
Low frequency hits harder in fewer places; high frequency hits softer in millions more places. Everything else about frequency selection is this sentence applied to your part.
A bubble in water has one size at which it resonates with the sound field, and that is the size the field drives hardest. It is set by the frequency alone: the resonant radius in metres is about 3.3 divided by the frequency in hertz. Nothing about the tank, the chemistry or the power changes it. That single relationship is why the bands behave the way they do.
| Frequency | Resonant radius | Diameter | What that size suits |
|---|---|---|---|
| 25 kHz | 130 µm | 265 µm | Heavy soil, castings, carbon |
| 28 kHz | 115 µm | 235 µm | Heavy soil, weld scale, foundry |
| 40 kHz | 80 µm | 165 µm | General workshop and precision |
| 68 kHz | 50 µm | 95 µm | Fine detail, small passages |
| 80 kHz | 40 µm | 80 µm | Sensitive substrates, fine finish |
| 130 kHz | 25 µm | 50 µm | Micro-features, polished surfaces |
Two things follow. The first is reach: a 25 kHz bubble is a quarter of a millimetre across and bridges straight over a 0.2 mm slot, while a 130 kHz bubble is small enough to sit inside it. The second is force. Collapse energy scales with the volume of the cavity, so the radius dropping five times takes the energy per collapse down by about a hundred. High frequency makes that back in numbers, not in force — many more bubbles, each doing far less.
A real tank holds a spread of bubble sizes, not one. The resonant radius is the peak of that spread and the useful number to design around. It is not a promise that every cavity in your tank is that size.
| Band | Character | Built for | Wrong for |
|---|---|---|---|
| 20–28 kHz | Large, aggressive bubbles | Coked carbon, heavy grease, castings, mining and engine work | Polished, plated and delicate surfaces |
| 40 kHz | The industrial all-rounder | General precision — machined parts, mixed workshop loads | Extremes at either end |
| 68–80 kHz | Fine, dense field | Fine detail, small passages, sensitive substrates | Heavy soils — it will bore you |
| 120–130 kHz | Gentlest, densest | Optics-adjacent, polished dies, micro-features | Anything a workshop calls dirty |
Frequency decides how loud the machine is, and that decides where in the building it can stand. It is the axis nobody asks about until the tank is installed and the shop cannot hear itself work.
| Band | What it sounds like | What that means |
|---|---|---|
| 20–28 kHz | The loudest. A hard, rattling hiss you feel as much as hear | Plan for a lid, a door, or distance. Check it against your noise assessment |
| 40 kHz | Quieter, but still plainly audible across a workshop | Fine in a workshop, less so in a room people sit in |
| 68–80 kHz | Quieter again — a softer, higher hiss | Usable in a lab or an inspection room |
| 120–130 kHz | The drive tone is above hearing. What you hear is the cavitation itself | The quietest option, but not silent |
High-frequency machines are sometimes sold as silent. They are not. The drive tone moves above the range most people hear, but cavitation collapse is broadband — it puts energy right across the audible range, and a loaded basket rattling in a steel tank makes its own noise. High frequency is quieter. It is not quiet.
The workplace exposure standard for noise is 85 dB(A) averaged over eight hours, with a peak of 140 dB(C) at any moment. Every 3 dB above 85 halves the time a worker can be exposed — 88 dB(A) for four hours, 91 dB(A) for two. A 28 kHz tank running open in a small room is worth measuring rather than guessing at.
Three things bring it down without changing frequency: a lid that actually seals, standing the tank away from hard parallel walls, and not running it empty. An unloaded tank is louder than a loaded one.
What is the soil? Baked and bonded → low band. Films and fines → high band. Mixed → keep reading.
What is the substrate? Robust ferrous castings shrug off 25 kHz all day. Soft aluminium, brass and coatings prefer 40+ — low frequency can mark soft surfaces over long exposures.
What is the geometry? Fine passages and dense detail favour the smaller bubbles of higher bands, which follow contours the big bubbles bridge over.
Published cavitation-erosion work confirms what workshops learn expensively: low-frequency exposure erodes soft substrates measurably over time — erosion-test foils and soft aluminium coupons show it within minutes — while the same energy at 80 kHz leaves them untouched. The practical translation is the exposure budget: aggressive frequencies buy speed on robust parts and spend surface on soft ones. When a load mixes both, sequence the bands (№10) instead of splitting the difference — a compromise frequency compromises everything.
Plenty of real loads sit across the trade: heavy soil and delicate features, castings and finished faces in one basket. Dual-frequency platforms answer with two resonant systems in one tank — run the aggressive band for the bulk, switch to the fine band for the finish, without moving the part. Our dual-piezo BLT architecture drives each frequency at its own true resonance — no off-resonance compromise. The dual-frequency article (№10) goes deep; the short version is that it exists precisely for mixed reality.
"Higher frequency = better machine." Frequency is a fit, not a grade. A 130 kHz tank on mining hydraulics is a very expensive way to do nothing; 25 kHz on polished mould cavities is vandalism with a power cord. The best frequency is the one your soil and substrate vote for.
The damage conversation deserves numbers rather than nerves. Low-frequency cavitation can mark soft substrates — but the mechanism is cumulative exposure, not instant harm. Short aggressive passes followed by fine-band finishing keep total low-band exposure under the threshold while still breaking heavy soil: the sequencing strategy that dual-frequency platforms automate. Hard ferrous parts, by contrast, tolerate the aggressive bands essentially indefinitely — which is why 25 kHz remains the rebuild industry standard.
Bubble diameter scales roughly inversely with frequency — the aggressive band collapses cavities several times larger, each carrying far more energy per implosion.Standard acoustic cavitation relationships; see also №1 in this series.
A general rebuild shop asks for "the right frequency" for a workload of cast-iron heads, aluminium housings and the occasional rack of injectors. There is no single right answer — there are three, and the exercise shows how the decision actually runs.
The iron heads carry baked carbon and need demolition: 25–28 kHz, hot alkaline chemistry, no apology. The aluminium housings carry oils and light varnish on machined faces that will be measured afterwards: 40 kHz covers them comfortably, and if the shop later takes on polished or anodised work, 68–80 kHz earns its place. The injectors have micro-drillings and lapped seats: high band only, ever. Three duties, three bands — and the purchasing decision comes down to this: either two tanks at different frequencies (the usual answer, and cheaper than it sounds when one is a small high-frequency bench unit), or one dual-frequency platform where floor space or budget forces a single vessel.
What the shop should not do is buy one 28 kHz tank and feed everything through it. The iron will be delighted; the aluminium will develop the matte frosting of early cavitation erosion within weeks, and the injectors should never go near it. Nor should they buy a "safe" 40 kHz compromise and wonder why head carbon takes three cycles. Frequency mismatch is invisible on the quote and expensive on the floor.
Frequency selection from a blog gets you to a shortlist. The trial gets you to a number — your parts, two or three candidate bands, results side by side. That comparison costs a box of parts and settles the biggest specification on the machine.
The trial runs candidate bands side by side and lets the results choose. Bring the shortlist; leave with the number.