Data center hum
Our increasingly digital world relies on massive Data Centers (or Data Centres), but their rapid proliferation often comes with a hidden, yet pervasive, cost: noise. Persistent, 24/7 low-frequency hums, frequently overlooked by standard noise specifications and regulations, lead to significant and expensive challenges for both operators and the surrounding communities.
This article explores why traditional regulatory and ordinance approaches both to data center noise and other industrial sources fail so spectacularly, often resulting in multi-million-dollar mistakes and continuing community frustration.
We highlight a crucial disconnect in how data center noise is measured and understood and the consequent impact on local communities of the inadequate noise specifications that are so often included in regulations and local noise ordinances for planning requirements.
We also provide details of a more effective, cost-efficient, and environmentally friendly alternative that can be used to update those noise specifications.
Environmental noise regulation
Remote Diagnosis
Email us a video from your smartphone for free evaluation of best practice
The Costly Misunderstandings re Data Center Noise
A stark example illustrates the widespread failure to properly address data center noise:
A "$5 million mistake plus $232,000 in fines for lack of a 5-minute noise diagnosis."
This incident underscores a deeper, common misunderstanding of the problem. Data center noise, primarily generated by their massive, 24/7 cooling and power systems, is a global issue. Communities worldwide are experiencing the detrimental effects of this constant drone on property values, quality of life, and with health consequences second only to air pollution, contributing to substantially increased rates of dementia, strokes, and heart attack deaths.
The core of the issue lies in the inadequacy of standard noise measurements. Conventional “A” weighting filters in sound level meters often filter out the very low-frequency hum that causes the most complaints.
This means the standard measurements used for planning and assessment can literally be "deaf" to the problem, like looking at a rainbow with the colour red removed...
Case Study: The Atlas Power Data Center – A Monument to Misdiagnosis
The Atlas Power data center in Williston, North Dakota, serves as a prime example of this widespread misunderstanding. Faced with a deluge of noise complaints, the company pursued what seemed like a logical, traditional solution: building a massive noise barrier. This "jaw-dropping" $5 million structure was intended to be the definitive answer.
It wasn’t. It was a spectacular failure.
Despite the monumental investment, the hum persisted, and complaints continued. The failure was completely predictable to anyone with a good understanding of acoustics. The wall, a $5 million eyesore, became a testament to solving the wrong problem.
Why Traditional Noise Barriers Fail Against Data Center Hum
The ineffectiveness of such a colossal wall against data center noise lies in the lack of an accurate diagnosis of the cause of the noise complaints and understanding of the fundamental principles of acoustics.
Low-frequency sounds, like the deep hum from data centers, possess long wavelengths, similar to ocean breakers. Conversely, higher-frequency sounds, typically measured and regulated using the conventional dB(A) scale, have short wavelengths, like ripples. Attempting to block a low-frequency hum with a barrier is akin to trying to stop an ocean breaker with a postcard; the sound energy simply flows around it.
This $5 million investment effectively bought a "King Canute noise barrier" – a structure that could not work.
The "Smoking Gun": narrow band frequency analysis
The key to understanding and solving data center noise problems lies in a precise diagnostic tool: narrow band frequency analysis. As demonstrated by my simple analysis of an audio clip from a local TV news story about the complaints, the problem at the Atlas Power data center was obviously not the overall noise level. Instead, it was a group of powerful low-frequency tones based on a dominant spike at 118 Hz originating from the cooling fans. A simple, rapid remote analysis that instantly pinpointed the real culprit.
Narrow band frequency analysis acts like a "surgeon’s scalpel" compared to the "crude conventional dB(A) bludgeon."
It doesn’t just detect the volume, it tells you the exact pitch and source of the sound – in this case, 6-bladed cooling fans running at 1180 rpm.
Performing this simple, five-minute diagnosis would have saved millions.
The Smart Solution: preventing noise at source
If a $5 million wall is the wrong answer, what's the right one? The optimal solution is not only more effective but also significantly cheaper and much greener: prevent the noise at source. Rather than attempting to block sound after it's generated, the elegant and innovative engineering fix involves using a computer model to understand airflow and then designing small retrofit aerodynamic components that eliminate the hum. Compare this precise acoustic surgery with the construction of a colossal castle around the site.
The difference in outcomes is dramatic. The wall is incredibly expensive, ineffective against the hum, an eyesore, and can even hinder cooling airflow. In contrast, the aerodynamic fix is low-cost, eliminates the problem hum, is invisible, and remarkably, can even make the data center more efficient by improving fan efficiency by up to 20%.
Data center cooling fan: 22.7% reduced fan power for the same cooling + 98% reduction in noise
As Arthur C. Clarke famously stated -
"Any sufficiently advanced technology is indistinguishable from magic!"
This innovative approach means a project to cut data center noise can potentially pay for itself through energy savings - or even generate a profit.
There is, however, an elephant in this room: the generally abysmal lack of knowledge of modern engineering noise control techniques, even amongst noise consultants. The industry tends to rely on expensive palliatives such as acoustic enclosures, silencers and, as above, noise barriers.
Problem Prevention Through Smarter Noise Regulations
This story extends far beyond a single data center in North Dakota; it’s a vital lesson for the entire data center industry, crypto-farms, and other large installations, as well as the communities that host them. To prevent devastating impacts on local residents and avoid future multi-million dollar mistakes, the regulatory framework must evolve.
This necessitates introducing smarter data center noise ordinances and regulations that look beyond simplistic overall decibel levels and specifically address tonality, particularly annoying low-frequency hums.
More nuanced noise ordinances and sound studies don't have to be complex. Although often castigated by acoustic consultants, entirely adequate tonal analysis is even readily available on smartphones for free. Such studies should be proactively carried out by designers and developers before construction begins. The emphasis should shift from reactive measures after widespread complaints, negative publicity, and health concerns, to precise and proactive planning.
As an aid to implementing smarter noise specifications and acoustic modelling, we developed a template plant noise data sheet for suppliers that captures tonal data in addition to the usual noise data. In almost every case, we end up analysing smartphone recordings from similar plant installed elsewhere. We always include tonal components in the data center noise models we develop to meet noise specifications to ensure that the site will not generate noise complaints from the local community.
Ultimately, the example of the wasted $5 million on an "emperor’s new clothes" solution isn't just about acoustics. It's about active listening, diagnosis using the correct noise analyses, asking the right questions, and prioritizing smart design over traditional, high-cost, brute-force noise control palliatives.
As our digital world continues to expand, a critical question remains: how can we ensure that the multitude of massive monuments to the digital age we build are not only useful but also good, quiet neighbours in the communities we all call home?
The answer lies in smarter, more precise approaches to data center noise regulation and management. To that end, we have been involved with regulators in a number of countries to help them update noise ordinances and specifications to include the low frequency and tonal noise elements that are so often the cause of complaints, even when “standard” noise requirements have been met.
Data Center Noise Regulation, Ordinances and Control: technical support
We have provided technical support to improve noise evaluation and control for environmental noise regulators and other organisations in number of countries on the following topics:
- noise evaluation and analysis
- updating noise guidance
- modern engineering noise control options
- template supplier noise data templates for site modelling
- online webinars
The following are some of the online resources that we have developed:
- Data center (or "centre") and crypto mining noise control using modern engineering technology
- Tonal (hum) noise evaluation best practice
- Tonalysis - how to add hum detection to dB(A)
- How to use free smartphone apps for hum analysis
- Chiller and cooler plant noise control
- Diesel generator noise control
- Greener quieter - more sustainable noise control
Online Training
- How to evaluate and resolve environmental noise complaints quickly
- Fan noise control masterclass
- Environmental noise control best practices
We also provide bespoke workshops covering related topics for regulators and other organisations.
