Transformer Noise and Sound Levels: How to Specify a dry type transformer for data center, Hospital, and Residential Projects

The commissioning engineer put a sound level meter on the mezzanine floor and watched the number settle at 62 dB(A). The drawing said 55. Nothing was broken. The unit had passed its factory sound test exactly as designed - at no load, in a test bay, at a fixed distance from the enclosure. Nobody had asked what the same transformer would sound like two floors above an open-plan office, at 70 percent load, at two in the morning.

That gap between the paper number and the number people actually hear is where most acoustic conversations with transformer buyers start. It is also why a dry type transformer for data center project now receives an acoustic clause in the specification, and why pad-mounted units feeding residential blocks are being measured by people holding sound meters instead of reading datasheets.

What Is Actually Making the Noise?

Two physically different mechanisms produce everything you hear from a transformer, and they behave differently across the load range.

The first is magnetostriction. Each time the core laminations are magnetised, the steel changes dimension by a few parts per million. Excitation reverses twice per cycle, so the resulting vibration sits at twice the line frequency - 120 Hz on a 60 Hz system - with harmonics stacked above it at 240 Hz, 360 Hz and beyond. This is the steady hum that never stops. It is present the moment the unit is energised, even with the secondary open.

The second is load noise. Current in the windings creates electromagnetic forces that push and pull the coils, the leads and, at larger ratings, the enclosure or tank walls. Load noise scales roughly with the square of the current, which makes it almost inaudible at 20 percent load and a genuine problem at 90 percent.

Here is the part most buyers miss: the sound level printed on a quotation is almost always the no-load figure. It is the number a factory can measure repeatably in a test bay, and the number the standard test codes were written around. It is not the number a neighbour will complain about.

The Standards Behind the Number on the Drawing

A sound level figure is meaningless without three things attached to it: the measurement distance, the loading condition, and the test method. Those three items live in published standards, not in sales literature.

For liquid-immersed distribution and power transformers, the test procedure - including audible sound level measurement - sits in IEEE C57.12.90. Dry-type units are covered by the parallel test code IEEE C57.12.91, and IEC 60076-10 defines the international method. The published average sound levels that engineers actually compare against come from NEMA ST 20 for dry-type units and NEMA TR 1, Table 0-2, for liquid-immersed units.

Two practical consequences follow:

• A number quoted at 0.3 m is not the same quantity as a number quoted at 1 m. Depending on the enclosure and the measurement surface, the two figures can sit several decibels apart - easily enough for a submittal that looks compliant on paper to fail a site measurement.
• A no-load figure is not a full-load figure. For a facility running at 60 to 80 percent load all night, the winding component is a real contributor to whatever the acoustic consultant measures.

The decibel scale is unforgiving in both directions. Three decibels is a doubling of sound energy, and roughly ten decibels is what a listener experiences as twice as loud. That is why 3 dB is worth arguing for at specification stage, and why a 10 dB misunderstanding becomes a project problem.

For orientation, a 1,000 kVA dry-type unit is commonly specified in the 55–60 dB(A) band, and each doubling of kVA adds roughly 2.5–3 dB on top. Liquid-immersed units of comparable rating usually land a few decibels lower, because the oil and the steel tank damp core vibration before it reaches the air.

No-Load vs Load Noise: Why a Dry Type Transformer for Data Center Sounds Louder Than Its Datasheet

When a project carries a hard acoustic limit, the useful question is not "what is the sound level?" It is "at what load, at what distance, measured how?"

No-load sound level is set almost entirely by core design. A manufacturer who wants a lower number has only a few honest levers: reduce the design flux density, use step-lap mitered joints instead of butt joints, clamp the core more tightly, and decouple the core-and-coil assembly from the enclosure with properly rated isolation pads. Reducing flux density is the most effective single change, and it is also the most expensive, because it requires more core steel. That is where a low bid and a low noise transformer stop being the same product.

Windings do not care about voltage; they care about current - and about its waveform. Load noise rises with the square of current, and faster than that when the current is rich in harmonics. A harmonic-rated or K-rated design changes the conductor geometry and the bracing to survive that duty, and it changes the acoustic signature at the same time. In our test bay, the difference between a standard dry-type unit and a harmonic-rated unit of the same kVA, both at rated current, is audible without instruments.

Cooling adds a third layer that specifiers routinely forget. Fans and pumps can add 5–10 dB(A) on top of the no-load figure when they stage in, which is why the fan-on rating and the fan-off load capability matter as much as the sound level on the nameplate. Specify only the no-load number, and a unit can pass the test bay and still be too loud in the room on a July afternoon.

When a contractor in Texas sends us a submittal for a 2,500 kVA dry-type unit sitting on a mezzanine two floors above office space, the first thing our engineering group checks is not the kVA. It is the sound level on the drawing, the distance to the nearest occupied room, and whether the cooling stages will be running at the load the building will actually carry.

What to Put in the Specification

The acoustic clause that prevents most of these conversations fits in fewer than ten lines. Instead of a single line reading "sound level: 55 dB", specify these items:

1. A maximum, not an average. Published average sound levels are statistical figures; a site needs a guaranteed ceiling.
2. The method, in full. "NEMA ST 20, measured at 1 m, at rated voltage and rated frequency, no load" - distance, standard, condition.
3. A test report in the FAT package. A certified sound level test delivered with the routine test documentation, not promised in an email.
4. A design flux density cap, evaluated at rated voltage and frequency, with the calculation submitted for review.
5. Step-lap core construction, and encapsulated (cast resin or VPI) windings for dry-type units rather than open-wound coils.
6. Fan-off load capability. "No fans below 50 percent load" is a different acoustic product from "fans above 25 percent".
7. The mounting detail. Isolation pads sized for operating weight, plus flexible conduit so vibration has no path into the structure.
8. The receptor. Nearest occupied room, nearest property line, night-time limit, and the local ordinance number where one exists.

None of those items cost anything to write. All of them are difficult to retrofit after the unit is sitting on the pad. If you are working through a specification now and want a second opinion on the acoustic clause, send the drawing and the siting plan to Ryancustomerinquiry@redq.cc - our engineering group reviews sound level targets against load profile and room geometry before quoting, not after.

The Siting Half of the Problem

Sound dissipates slowly and predictably. In free field, every doubling of distance removes about 6 dB: a unit measuring 60 dB(A) at 1 m is roughly 54 at 2 m, 48 at 4 m, 42 at 8 m and 36 at 16 m. That arithmetic is the first thing an acoustic consultant runs, and it is the first thing a transformer buyer should run, because reflections, adjacent equipment and building surfaces all push the real measurement above the calculation.

The regulatory pressure is arriving from outside the transformer industry. The U.S. Environmental Protection Agency has identified 55 dB outdoors and 45 dB indoors as the levels needed to prevent activity interference and annoyance, per the EPA noise guideline. Residents near new data centers are now measuring sites against those numbers: in California, data center noise entered the state regulatory debate in September 2026, and the Los Angeles Times published reader complaints on 2 September 2026 pointing to cooling and power-train noise well above that guideline. Utilities have been having the same argument about pad-mounted units on residential feeders for years - only the vocabulary is newer.

Three moves solve most of it, and none of them are exotic:

• Distance. Get the receptor out of the near field, or move the unit away from the receptor. It is the cheapest acoustic treatment in existence.
• Orientation. Radiators, fans and ventilation openings radiate very differently. Turning a unit 90 degrees costs nothing.
• Barriers. A masonry wall of adequate mass per unit area between the unit and the nearest window does more than any argument at handover.

For pad-mounted equipment on residential feeders the decision usually comes down to the enclosure itself: a heavier tank, stiffer panels, and a lower-flux core. As a pad mounted transformer manufacturer shipping into North American networks, we quote those options explicitly rather than leaving them to chance, because the alternative is a complaint file that outlives the maintenance contract.

How We Specify Sound Level on Custom Units

When a project carries a hard acoustic limit, our engineering group treats the sound level as a design input rather than a label. That means designing the core at a lower flux density, machining step-lap joints, controlling clamping pressure, and measuring the assembled unit in the test bay before it leaves the plant - the unit that will actually ship, not a prototype. A low noise transformer is the result of those decisions, not a catalogue option bolted on at the end.

The manufacturing base behind that work covers 120,000 m² in Haian, Jiangsu, with 37 patents and more than 180 sets of production and test equipment, operating since 2007. Units are designed and tested to UL, CSA, IEEE, IEC, DEKRA and CNAS requirements depending on the destination market, and since 2023 the company has operated as an Eaton joint-venture partner.

Two recent projects show how the acoustic question resolves in practice. A Canadian mining project took 13 cast-resin units rated 1,500 kVA, 6.6/0.6 kV, in IP44 enclosures certified to IEC 60076, in an environment where the enclosure has to do the acoustic work - Canada Golden Mining project. A U.S. site took 30 pad-mounted units at 3,250 kVA, 34.5/0.415 kV, certified per IEEE C57 by UL - USA pad mounted project. A Kenyan utility installation of eight 1,600 kVA oil-immersed units followed the same process - Kenya utility project. Different load profiles, same starting point: the sound level was specified, measured and documented before shipment.

If an acoustic requirement is tight, ask for the sound test report with the routine test package. A supplier who can produce it will not hesitate. A supplier who cannot will explain why the site should not worry about it - and that answer is itself a specification problem. Most dry type transformer for data center enquiries we receive already carry a sound level target; the useful conversation is about how it will be verified.

About the Author

This guide was prepared by the transformer engineering team at Ryan Electric (Jiangsu Ryan Electric Co., Ltd.), which designs and manufactures cast resin and VPI dry-type transformers, oil-immersed power transformers and pad-mounted units for data center, utility, mining and industrial projects across North America, the Middle East, Southeast Asia and Africa. Questions on sound level specifications, routine test reports or certification requirements are welcome at Ryancustomerinquiry@redq.cc.