25/07/2026
Submission to the NSW Legislative Council Public Accountability and Works Committee Inquiry into Data Centres
Thanks to Ray Williams MP assistance, the Secretariat this week accepted this submission for consideration. The Inquiry is due to report 30 Sept.
The focus is on specific issues around the acoustic and setback impacts, measurement and approval standards.
A bit of detail to set-up so jump to the 9 Recommendations for a quicker read. These were sense checked by an acoustic engineer.
“The Problems of Residential Proximity
Data centres are now being approved next to residential streets in NSW using an acoustic assessment framework written for general industry and road traffic, not for continuous, low-frequency, tonal mechanical plant running 24 hours a day, 7 days a week, for the life of the facility.
Using the Rouse Hill approval to illustrate the gap precisely: The site: 44 emergency generators, 42 chillers and 34 cooling towers, sits directly opposite two-storey homes at 141–167 Mile End Road. The noise impact assessment that supported approval was found compliant against the NSW EPA's *Noise Policy for Industry* (NPI 2017) and the Hills Shire Council Development Control Plan 2012.
However that compliance finding rests on:
- an A-weighted, outdoor, façade-level metric (L_Aeq,15min) that structurally discounts the low-frequency energy (63–250 Hz) where generator, chiller and cooling-tower noise is concentrated; the report's data shows the lead generator at 110 dB at 63 Hz, falling to 97 dB(A) once A-weighted, a roughly 13 dB "disappearing act" performed by the metric, not by the equipment;
- an event-based sleep-arousal test (L_Fmax) borrowed from road/rail noise methodology, applied to a source that has the opposite hazard profile; continuous, not episodic;
- no tonality correction actually applied, despite generators, chillers and cooling towers being intrinsically tonal sources, and despite the NPI itself providing a mechanism (Table C1 modifying factors) to capture exactly this;
- "indicative" plant data, with final equipment selection deferred to post-approval detailed design, meaning the acoustic case that secured the DA is not the acoustic case that will actually be built and used; and
- no independent, government-commissioned verification once the facility is running at full load, and no minimum physical separation from residential land as a backstop to modelling.
None of this reflects an error by the consultant, who applied current policy correctly. It reflects a policy framework — the NPI 2017, and the equivalent provisions of the Hills DCP 2012 — that predates the current wave of hyperscale, generator-and-chiller-dense data centres in Sydney's growth corridors, and that has not been updated to match the technology it now regularly assesses.
Jurisdictions further into this build-out; the UK, EU member states, and multiple US states and counties, are visibly moving to close the same gaps: mandatory low-frequency assessment, tonal-noise corrections applied by default, independent post-construction verification, and minimum physical setback distances treated as a non-negotiable safeguard rather than relying solely on acoustic modelling of equipment that has not yet been selected. As the state instrument directly responsible for the NPI and for local planning controls such as the Hills DCP, the NSW Government is best placed to close this gap before the current pipeline of DA approvals locks it in at scale across Sydney's growth areas.
---
9 Recommendations
1. Make tonal and low-frequency corrections automatic for this class of equipment
The NPI (2017) already provides for modifying factors; the Table C1 correction to be applied where a noise source is tonal, impulsive, or has low-frequency character. In practice, as in the Rouse Hill assessment, this is treated as discretionary ("where applicable") and is routinely not applied to generators, chillers, cooling towers or transformers, even though these are intrinsically tonal and low-frequency-dominant by design. Recommend that the EPA amend the NPI so that generators, chillers, cooling towers, transformers and comparable rotating/reciprocating mechanical plant automatically attract the relevant Table C1 modifying correction, with the burden on the applicant to demonstrate a source is *not* tonal or low-frequency-dominant, rather than the reverse. This should be the Committee's priority recommendation, as it requires only an amendment to existing EPA policy, not new legislation.
2. Require a dedicated low-frequency noise (LFN) assessment
In addition to the standard A-weighted NPI test, require a low-frequency-specific assessment (e.g. against UK NANR45:2005 reference curves, or ANSI S12.2 RNC/NCB curves) for any development with 24/7 generator, chiller, cooling-tower or transformer plant within a defined distance of a residential receiver.
3. Require assessment of realistic internal (in-room) conditions, not façade-only
Compliance should not be assessed solely at the outdoor property boundary or façade. Low-frequency sound transmits through glazing and light residential construction with far less attenuation than mid/high-frequency sound, and NSW housing commonly relies on partially open windows for ventilation. Require indoor assessment
(or a defensible façade-to-indoor transmission calculation using real, not assumed, window conditions) wherever a low-frequency or tonal exceedance risk is identified.
4. Require cumulative, precinct-level assessment in data centre growth corridors
Individual DA-by-DA assessment cannot capture the combined effect of multiple data centres approved in the same corridor, a live issue given the scale of the current NSW data centre pipeline (with a reported $100 billion of projects under assessment). The EU's Environmental Noise Directive requires strategic, area-wide noise mapping for major industrial sites in agglomeration zones; NSW should require an equivalent cumulative assessment wherever more than one data centre is operating, approved, or proposed within a defined radius of the same residential receivers, coordinated at Council or Department of Planning level rather than left to individual applicants.
5. Require disclosure and binding conditions for maintenance and testing schedules
Routine generator maintenance/testing and periodic full-load testing should not be
assessed only as a predicted noise level. The frequency, duration, time-of-day and
advance notice to affected residents should be disclosed at DA stage and fixed as an
enforceable condition of consent, a practice already used in the UK, where generator
testing is now routinely confined to defined weekday daytime windows.
6. Post-construction verification should be independent of the applicant
Require verification testing at full design load within a fixed period after occupation (60–90 days) and have this commissioned and paid for by the applicant, and conducted or supervised by the EPA, the consenting Council, or a consultant appointed and instructed directly by the EPA/Council, not the applicant's own acoustic consultant
reporting back on their own predictions. Results should be published, with automatic mitigation obligations (not further "review") triggered by any exceedance.
7. Set a minimum statutory setback of 250–450 m between data centre mechanical plant and residential zone boundaries
Acoustic modelling of "indicative" plant, as at Rouse Hill, is only ever as good as the equipment finally installed. A fixed minimum separation between generators/chillers/cooling towers and residential land is a non-negotiable backstop that does not depend on that modelling being right, and is now standard practice in
jurisdictions building the most hyperscale capacity; see the international examples and sources at Appendix A. A 250–450 m minimum for generator/chiller/cooling-tower plant from residential zone boundaries would bring NSW within the emerging international
range rather than well below it.
8. Require confirmed plant make and model at DA stage, not "indicative" data
At Rouse Hill, as is standard practice, the sound power data underpinning the compliance finding is explicitly described as "indicative," with final equipment selection deferred to detailed design after approval. Recommend that confirmed manufacturer
make/model and full octave-band sound power data for all major mechanical plant (generators, chillers, cooling towers, transformers) be a mandatory DA submission requirement, with any subsequent substitution of equipment during construction or operation requiring re-assessment against the same conditions before energisation.
9. Require compulsory assessment of emergency and full-load generator testing scenarios, not just business-as-usual operation
Current assessments typically model only steady-state cooling operation as the primary compliance scenario, with emergency/full-generator-load operation treated as a secondary case on the basis that it is infrequent. This under-represents residents' real exposure: generator load testing and emergency operation are measurably louder than routine cooling operation and are the scenarios most associated with complaint and sleep disturbance. Recommend that compliance be assessed, and conditioned, against the full-load emergency/generator-testing scenario as a mandatory case, on equal footing with business-as-usual cooling operation.”
Other submissions can be viewed at >>
The Parliament of New South Wales stands on the traditional lands of the Gadigal people. We pay our respects to their Elders past and present. We proudly acknowledge all the Aboriginal peoples of New South Wales and honour their enduring connection to the lands, waters and sky of the State.
