Choosing an ultrasonic cleaning frequency is the first specification most buyers skip, and it’s the one that decides whether a cleaning process works or quietly ruins parts. Get the ultrasonic cleaner frequency wrong and you’ll either spend twenty minutes failing to shift machining oil, or you’ll pull a batch of aluminium out of the tank with a dull, pitted surface that wasn’t there when it went in. This is what the numbers mean on the shop floor, and how to pick one before you buy a tank.
What the frequency number is actually describing
An ultrasonic cleaner works by cavitation. The transducers drive pressure waves through the liquid; tiny bubbles form in the low-pressure phase and collapse in the high-pressure phase, and that collapse does the cleaning. Frequency sets how many of those cycles happen each second, which changes both the size of the bubbles and how much energy each one releases when it implodes.
Lower frequency means fewer, larger, more energetic bubbles. Higher frequency means many more, much smaller, far gentler ones. That single trade-off explains almost every frequency decision in industrial cleaning. Everything below follows from it.
The working ranges, and what each one is for
| Range | Bubble behaviour | Suits | Risk if misapplied |
|---|---|---|---|
| 20–28 kHz | Large, high-energy collapse | Heavy machining oil, carbon, castings, weld scale, large steel parts | Pitting on soft metals; erosion of plating and polished faces |
| 40 kHz | Balanced — the general-purpose range | Most machined parts, fasteners, bearings, tooling, mixed workloads | Slower than 28 kHz on baked-on carbon |
| 80–120 kHz | Small, dense, gentle | Optical components, coated surfaces, thin films, fine particulate | Won’t shift heavy grease or machining oil on its own |
| Megasonic, around 950 kHz | Sweeping action, minimal implosion energy | Sub-micron particles on wafers and optical surfaces | Not a cleaning step for any bulk soil |
Those ranges aren’t theoretical. They’re the frequencies we actually build at: our immersible transducer rods run at 20, 28 and 35 kHz; standard multi-tank production lines are built at 28/40 kHz; the three-frequency benchtop series reaches 80 and 120 kHz; and megasonic units operate near 950 kHz.
How to choose without guessing
Work through it in this order. Whoever runs the parts usually knows all four answers already.
- What is the soil? Machining oil, drawing compound, buffing wax and carbon need the energy of 28 kHz. Fingerprints, dust, polishing residue and fine particulate don’t.
- What is the surface? Bare steel and castings tolerate aggressive cavitation. Aluminium, plated finishes, polished surfaces, optical coatings and thin-walled parts don’t — this is where damage happens, and it’s usually discovered after a full batch has gone through.
- Is there a written residue specification, or is it visual? A particle count or cleanliness class normally pushes you toward higher frequency and more rinse stages. A visual standard rarely does.
- Does the part mix span both ends? If it does, a dual-frequency tank costs less than two machines and removes the temptation to run everything at one setting.
Frequency by part family: what we see in practice
Across the custom lines we’ve delivered — forty non-standard systems across six industries — the same patterns repeat. They’re a starting point, not a substitute for testing your own part.
- Machined steel and castings — 28 kHz. Drawing oil and swarf come off quickly, and the surface tolerates the energy.
- Aluminium and plated parts — 40 kHz, and watch cycle time. Aluminium is the material that most often comes out of a tank looking worse than it went in.
- Engine components with carbon deposits — 28 kHz, heated, with longer cycles. Carbon is one of the few soils where more energy genuinely helps.
- Bearings and precision assemblies — 40 kHz, then a dedicated rinse. Residual detergent matters more than the last trace of oil.
- Optical components, lenses, coated glass — 80 kHz or higher, low power. The coating decides everything.
- Wafers and sub-micron work — megasonic, after conventional cleaning has already removed the bulk soil.
- Medical instruments with hinges and lumens — 40 kHz with agitation, because geometry traps soil that frequency alone won’t reach.
Three mistakes that cost more than the machine
Running everything at the lowest available frequency. It’s the fastest setting for heavy soils, so it becomes the default. Then a batch of aluminium or plated parts goes through and comes out dull, pitted or etched. Cavitation damage looks like a chemistry problem, so the detergent usually gets blamed first — and changing it doesn’t help.
Expecting high frequency to do low-frequency work. 80 kHz won’t remove machining oil in any reasonable cycle time. When it fails, the common reaction is to raise temperature or concentration. Neither addresses the cause, and both cost money.
Skipping degassing. A freshly filled bath holds dissolved air that absorbs cavitation energy. Results in the first few minutes after filling aren’t what the machine can actually do. If you evaluate a frequency on that evidence, you’re deciding on bad data — and this is the single most common reason a new machine “doesn’t work as well as the demo”.
Frequency is one of four variables, not the whole process
Frequency, chemistry, temperature and time work as a set, and changing one shifts what the others need to be.
Raising temperature makes most detergents more active, but past a point it degrades cavitation itself — hotter is not linearly better, and there’s a peak beyond which cleaning gets worse. Pushing concentration above the formulated range wastes chemistry and, on some alloys, starts attacking the surface you’re trying to protect. Longer cycles don’t compensate for the wrong frequency; they just expose the part to the wrong conditions for longer.
When a process works, record all four together. That combination is the process. It’s also what makes results repeatable when a different operator runs the machine on a different shift — which is the actual problem in most plants, not the cleaning itself.
Where tank design changes the answer
One more variable that specification sheets rarely mention: how the part sits in the tank. Cavitation only cleans surfaces the liquid can reach with sound energy. A part resting on the tank floor isn’t cleaned where it touches. Parts stacked against each other aren’t cleaned where they meet. A component pressed against the tank wall can be both poorly cleaned and marked.
If cleaning results vary across a single basket, fixturing is the first thing to examine — before frequency, before chemistry. We’ve seen more processes fixed by changing how parts are held than by changing the machine.
If you’re not sure, don’t guess
Frequency selection is one of the few decisions where a sample test settles the question in a day, while a specification argument can run for weeks. Jietai Ultrasonic has been building cleaning equipment since 2003, from our own factory in Dongguan, and the most useful thing we do for a new customer usually isn’t a quotation.
Send us the actual part with its actual soil. We run it on our own machines and send it back with the frequency, chemistry, temperature and cycle time we used — and if a frequency damaged the surface, we tell you that too, because you need to know that before you buy a tank rather than after. Here’s how the free cleaning test works.
That’s more useful than any table, including this one, because it’s your part rather than a representative one.