An overall A-weighted level answers one question. How much acoustic energy arrives at the receiver, weighted roughly for the sensitivity of human hearing.

It does not answer the question residents near battery energy storage sites raise more often. What does it sound like. That is a question about character, and the most consequential aspect of character on this equipment is tonality.

What follows is not an argument that tonal equipment is non-compliant, nor that a compliant level is meaningless. It is an argument that the two questions are different and need different data.

What a tone actually is

Broadband noise distributes acoustic energy across a wide range of frequencies. Traffic, wind in vegetation, general mechanical noise and airflow all behave this way.

A tone concentrates energy in a narrow frequency band, producing a sound with an identifiable pitch. A transformer hum, a fan whine and a converter whistle are all tonal. Almost every mechanical source has some spectral structure, so what matters for assessment is whether the tone stands sufficiently above the surrounding frequencies to be heard as a distinct pitch.

That is why every objective tonality test operates on a comparison. It asks whether the level in one frequency band exceeds the levels in the bands either side of it, and by how much.

Why A-weighting does not describe character

A-weighting is a single frequency weighting curve applied to a spectrum before the levels are summed. It approximates the relative sensitivity of human hearing at low sound levels, which means it discounts low frequency content substantially.

Two consequences follow for battery energy storage sites.

The first is that a low frequency tone contributes less to the A-weighted total than its audibility would suggest. A transformer producing energy at 100 Hz and its even harmonics may be clearly audible at a dwelling while adding relatively little to the overall dB(A) figure, because the weighting has already reduced it.

The second is that two sources with identical A-weighted levels can be entirely different experiences. A broadband source at 33dB(A) is likely to blend into the residual environment. A tonal source at the same 33dB(A), with most of its energy concentrated in one or two narrow bands, may be plainly audible. The number is the same. The complaint is not.

The ear finds tones that broadband masking would hide

The reason a tone stands out is a property of hearing rather than of the source.

The auditory system does not analyse sound as a single wideband signal. It behaves more like a bank of overlapping filters, each responding to a limited range of frequencies. Masking happens within those filters. A background sound masks a target sound most efficiently when their energy falls in the same auditory filter.

Broadband background noise spreads its energy across many filters, so within any single filter the masking energy is only a fraction of the total. A tone places all of its energy inside one filter. The result is that a tone remains detectable at levels where an equal amount of broadband energy would be inaudible.

This is the physical reason tonality attracts a regulatory penalty. It is also why battery sites in quiet rural locations are particularly exposed. Where the residual environment is low and broadband, a tone has very little to hide behind.

Which battery storage equipment produces tones

Three source groups on a typical battery energy storage site produce tonal content, and they do so by different mechanisms.

Transformers produce the most predictable tonal content on the site. IEC 60076-10-1 attributes transformer hum principally to magnetostriction, the dimensional change in core steel as flux density varies, and describes it as low in frequency with fundamental frequencies of 100 Hz or 120 Hz at twice the supply frequency. The standard notes that the non linear behaviour of magnetostriction produces several even multiples of the exciting frequency, and that where a direct current bias is present the vibration pattern repeats at the fundamental power frequency rather than twice it, introducing odd harmonics and potentially raising the level significantly. On the Australian 50 Hz grid the expected series is 100 Hz, 200 Hz, 300 Hz and upward.

Power conversion equipment produces tonal content through electromagnetic forces acting on inductors, reactors and filter components. The frequencies fall into two groups, low order content at multiples of the fundamental and higher frequency content at and around the switching frequency. The specific frequencies depend on the converter topology, the modulation strategy and the operating point, so they are a property of the particular unit rather than of converters in general.

Fans produce a tone at the blade passage frequency, which ASHRAE guidance gives as fan speed in revolutions per minute multiplied by the number of blades, divided by 60, together with its harmonics. Every fan does this.

There is a compounding effect specific to this equipment type. Battery sites contain many identical units, so the tonal content of each unit lands at the same frequency across the whole site. Adding identical sources raises broadband and tonal energy by the same amount, ten times the logarithm of the number of sources, so the tonal character does not become less prominent as the site grows.

How Australian frameworks assess tonality

Every objective tonality test used in Australian environmental noise regulation works on one-third octave band data. Beyond that, they differ substantially.

A comparison presented to the Australian Acoustical Society in 2015 examined the approaches then applied across six jurisdictions. The differences it identified are considerable.

The comparison found that at low frequency more than 12dB can separate the level at which a tone would attract a 5dB penalty in Western Australia from the level at which the same tone would attract one in New South Wales. It also concluded that the approaches vary in how well they agree with the narrowband method in ISO 1996-2, with the variable penalty approaches showing better agreement than the fixed penalty ones.

Important qualification. That comparison describes the position in 2015 and several of these frameworks have since been revised. New South Wales replaced the 2000 Industrial Noise Policy with the Noise Policy for Industry in 2017, and the EPA has stated that the methods for determining annoying characteristics were updated to reflect current science and standards, with an additional assessment step added to the low frequency assessment. Victoria has moved from State Environment Protection Policy N-1 to the Environment Protection Act 2017 framework and EPA Publication 1826.4. The table above should be treated as an illustration of how much these approaches differ from one another, not as a current compliance reference. Confirm the operative test for the applicable jurisdiction and the specific conditions of consent before relying on it.

What has not changed is the magnitude of the consequence. The noise impact assessment for a battery energy storage project at Glen Innes in New South Wales applied a 5dB tonality correction, with a stated maximum of 10dB where two or more undesirable characteristics are present, against a night-time project trigger level of 35dB(A). A 5dB correction against a 35dB(A) trigger is not a rounding adjustment. It is a substantial part of the available design margin.

One-third octave against narrowband, and why resolution matters

A one-third octave band is a relatively wide slice of the spectrum. A tone at 100 Hz sits inside the 100 Hz third octave band, but so does everything else in that band.

Two failure modes follow. A narrow tone can be diluted when averaged across a band that also contains broadband energy, so a clearly audible tone may not trigger a third octave test. Conversely, a band can be elevated by energy that is not audibly tonal at all, so the test can flag something a listener would not identify as a pitch.

ISO 1996-2 addresses this by providing a reference narrowband method alongside simpler band based approaches. Australian objective tests are band based, which means the data resolution a project holds determines what it can demonstrate. Octave band data is enough to model propagation. One-third octave data is the minimum for an objective tonality test. Narrowband data is what settles an argument about whether a specific tone is present and at what frequency.

Two things this does not mean

A compliant overall level does not demonstrate the absence of a tonality issue. An assessment that reports an A-weighted level against a criterion and says nothing about spectrum has not tested for tonality. It has answered a different question.

Tonal equipment is not automatically non-compliant. Transformers and converters produce tonal content as a consequence of how they work. Where the level at the receiver is low enough that a correction still leaves the result within the criterion, the design is compliant. Tonality is a factor to be quantified, not a defect to be alleged.

The failure mode worth guarding against is neither of these. It is an assessment that stops at the overall level, and a project that discovers the correction after the layout and equipment selections are fixed.

What tonal content means for treatment

Character determines what works.

Broadband content responds well to conventional screening and absorption. Distance, barriers and absorptive surfaces all attenuate mid and high frequency energy efficiently.

Low frequency tonal content does not respond the same way. Barrier attenuation depends on the path difference relative to the wavelength, and at 100 Hz the wavelength is around 3.4 metres, so a screen has to be very large to achieve much. Absorption is similarly less efficient at those frequencies for practical material thicknesses. What works instead is mass, sealing, separation distance and layout, together with source side measures where the supplier can offer them.

The practical consequence is that the treatment strategy for a battery site cannot be selected from an overall dB(A) figure. It has to be selected from a spectrum, and the spectrum has to come from the supplier before the layout is committed. AcousTech has examined the difficulty of the low frequency case in more detail in designing attenuation for performance below 125 Hz.

Three questions that close the gap

A compliant dB(A) figure tells a project that the total energy arriving at a receiver is within a limit. It says nothing about whether that energy is concentrated in a narrow band that the human auditory system is unusually good at detecting, whether a modifying factor correction applies, or what a resident will describe when asked what they can hear.

Three questions close that gap, and none of them is expensive to ask. Does the project hold one-third octave sound power data for the equipment that produces tonal content, at the operating state that governs. Has the applicable tonality test for the relevant jurisdiction been applied to that data, using the current framework rather than a superseded one. And does the predicted result still sit within the criterion once any correction is added.

If the answer to all three is yes, the dB(A) figure means what it appears to mean. If any of them is unanswered, the figure is describing a quantity that was never the thing in question.