Skip to content
YpsiDatacenter.org YpsiDatacenter.org

Up-to-date information on the proposed University of Michigan / Los Alamos Datacenter Project

YpsiDatacenter.org YpsiDatacenter.org

Up-to-date information on the proposed University of Michigan / Los Alamos Datacenter Project

  • Home
  • Resources
    • About YpsiDatacenter.org
    • What Is a Data Center?
    • Project Timeline
    • Who’s Who
    • What They Said, and What the Record Shows
    • How Loud Is a Data Center?
    • How Much Water Does a Data Center Use?
    • Why We Oppose This Project
    • Why the Township Can’t Zone This — and Won’t Tax It
    • Responsible Opposition
    • The Nuclear Question
    • In the News
    • Sources & Links
  • Community Toolkit
    • Data Center Database
  • Take Action
    • Legislation
      • Bills We are Tracking
      • Contact Your Representatives
    • Sign Up for Updates
    • Contact Your Representatives
    • Our FOIA Guide Has Moved — and Grown
    • FOIA Center
    • Request a Yard Sign
    • Events
    • How to Use Your Ninety Seconds
    • Volunteer
  • Policies
    • Terms & Acceptable Use
    • Privacy Policy
  • News
  • Home
  • Resources
    • About YpsiDatacenter.org
    • What Is a Data Center?
    • Project Timeline
    • Who’s Who
    • What They Said, and What the Record Shows
    • How Loud Is a Data Center?
    • How Much Water Does a Data Center Use?
    • Why We Oppose This Project
    • Why the Township Can’t Zone This — and Won’t Tax It
    • Responsible Opposition
    • The Nuclear Question
    • In the News
    • Sources & Links
  • Community Toolkit
    • Data Center Database
  • Take Action
    • Legislation
      • Bills We are Tracking
      • Contact Your Representatives
    • Sign Up for Updates
    • Contact Your Representatives
    • Our FOIA Guide Has Moved — and Grown
    • FOIA Center
    • Request a Yard Sign
    • Events
    • How to Use Your Ninety Seconds
    • Volunteer
  • Policies
    • Terms & Acceptable Use
    • Privacy Policy
  • News
Close

Search

Subscribe

Environmental Noise: How It’s Measured, and Why Data Centers Are Different

The short version

Noise is usually the first complaint neighbors have about a data center, and the hardest to pin down. A typical noise ordinance sets a decibel limit and stops there. But data center noise is different: it’s constant, it’s heavy in low-pitched sound, and it often carries a steady tone — a hum that a standard decibel reading can underestimate.

This page explains the basics in plain language: what the numbers mean, how a proper noise study is done, which standards professionals follow, and what other communities have done about data center noise in particular. You don’t need to become an acoustician. You need to know enough to ask the right questions and recognize a weak answer.

I’m not an acoustical engineer. This page is drawn from federal guidance and public records from communities that have dealt with data center noise, all linked below.

Decibels in two minutes

Sound is measured in decibels (dB). The scale is logarithmic, which makes it behave in ways that surprise people:

  • +3 dB means twice the sound energy. Two identical machines running side by side are about 3 dB louder than one.
  • +10 dB sounds roughly twice as loud to most people.
  • Sources add up. One data center might meet a limit on its own; three next to each other might not. Ordinances that judge each project in isolation can miss this.

A-weighting and C-weighting

Our ears are less sensitive to very low and very high pitches, so sound meters apply a filter, called a “weighting,” to mimic hearing.

  • A-weighting (dBA) is the standard in almost every noise ordinance. It heavily discounts low-pitched sound.
  • C-weighting (dBC) is much flatter and keeps more of the low-pitched sound.

That difference is the heart of the data center problem. A facility’s low hum can measure modestly in dBA while being clearly audible — and annoying — inside nearby homes. When a dBC reading is much higher than the dBA reading at the same spot, that gap is a sign of significant low-frequency sound.

For even more detail, professionals measure in octave bands — breaking the sound into ranges of pitch — which shows exactly where the energy is and reveals tones.

The measurements you’ll see in a noise study

TermWhat it means
LeqThe average sound energy over a period of time — the most common measure of steady noise
LmaxThe loudest moment during a measurement
L90The level exceeded 90% of the time — used to describe the quiet background
L50The level exceeded half the time — the “typical” level
LdnA 24-hour average that adds a penalty to nighttime sound, because night noise disturbs more

Why data center noise is different

  • It never stops. Cooling fans and transformers run 24 hours a day, every day. There’s no quiet night. Some cooling designs — especially ones that save water by relying on air-cooled equipment — mean more fans running more of the time; see How Data Centers Stay Cool.
  • It’s low-pitched. Fans moving huge volumes of air, chillers and transformers produce low-frequency sound that A-weighting underweights and that travels farther and passes through walls and windows more easily.
  • It’s often tonal. A steady hum at one pitch is more noticeable and more annoying than broadband noise at the same level. Prince William County’s consultant found a pronounced 90 Hz tone at one data center complex near homes.3
  • Generators add spikes. Backup generators are tested on a schedule, adding loud periods on top of the constant hum.

As one account quoted in Prince William County’s materials put it, what makes data center noise an irritant is less its loudness than its constancy.4

What a proper noise study looks like

A credible study — whether a developer submits it or a community commissions one — should include:

  1. Existing (“ambient”) conditions first. Measurements at the nearest homes before construction, over several days and nights, so there’s a baseline to compare against.
  2. The right locations. At or near the homes most likely to be affected, not only at the edge of the developer’s property.
  3. Nights, not just days. Background sound drops at night, which is when a constant hum stands out most.
  4. Both dBA and dBC, plus octave bands, so low-frequency sound and tones are visible.
  5. A prediction model of the finished facility, using the actual equipment’s sound data, showing expected levels at each home.
  6. Proper equipment and conditions. A professional-grade, calibrated sound level meter, checked before and after each session, with a windscreen, and measurements avoided in rain or strong wind. Times, weather and anything unusual should be recorded.
  7. A post-construction study once the facility is running, to confirm the predictions were right.

Watch for studies that measure only in dBA, only during the day, only at the property line, or only for a few hours. Those can make a facility look quieter than it will be where people live.

The standards professionals follow

These are the formal standards behind environmental noise measurement. Most cost $100 or more each and aren’t freely available, but it’s useful to know their names: an ordinance can require measurements “in accordance with” them, and a qualified acoustical consultant will own them.

StandardWhat it covers
ANSI/ASA S12.9 (Parts 1–4)The main U.S. standard for describing and measuring environmental sound, including short-term and long-term measurements
ISO 1996-1 and 1996-2The international standard for describing, measuring and assessing environmental noise, including adjustments for tonal, impulsive and low-frequency sound
ANSI S12.18Outdoor measurement of sound pressure levels
IEC 61672 (and ANSI S1.4)Performance requirements for sound level meters — a “Class 1” or “Type 1” meter is the professional grade
ISO 9613-2How sound travels outdoors — the basis for most prediction models

The best free stand-in is the Federal Highway Administration’s Noise Measurement Handbook (2018), which covers planning, measurement methods, terminology, instruments and reporting, and is based on current national and international standards.1 It’s written for highway projects, but its guidance on how to measure properly applies broadly.

What a strong noise standard looks like — and the evidence behind it

Here’s the honest starting point: there’s no single official “data center noise standard.” No federal agency sets one, and the professional standards describe how to measure sound, not what level a community should allow. But there is solid, research-based evidence for each kind of noise a data center makes. A defensible ordinance is built from those pieces, then calibrated to what your own community sounds like at night.

1. Steady, round-the-clock noise

The health benchmark is the World Health Organization’s Guidelines for Community Noise (1999). It recommends no more than 30 dBA inside bedrooms for steady, continuous noise, and no more than 45 dBA outdoors at night so people can sleep with their windows open — a figure that assumes an open window cuts about 15 dB. During the day, it reports that few people are seriously annoyed below 55 dBA, and few are moderately annoyed below 50 dBA.12 Those are the numbers with a documented health basis.

2. Low-frequency hum: knowing when to look closer

The same WHO guidelines give a simple, sourced trigger: when the C-weighted level is more than 10 dB higher than the A-weighted level, there’s significant low-frequency content, and a frequency analysis should be done.12 Health Canada uses the same 10 dB indicator, as described in a case study in Canadian Acoustics.13 An ordinance can use that gap to require octave-band data whenever it appears.

3. Low-frequency hum: deciding whether it’s a problem

Two government methods were developed specifically to judge whether low-frequency sound inside a home is disturbing. Both look at the range from 10 Hz to 160 Hz, and both are measured indoors, where people actually experience it:

  • The UK’s NANR45, prepared for the Department for Environment, Food and Rural Affairs by the University of Salford (2005, revised 2011), gives a criterion curve in one-third-octave bands for investigating low-frequency noise complaints.14
  • The Danish Environmental Protection Agency’s method evaluates the A-weighted level across the one-third-octave bands from 10 Hz to 160 Hz, measured in dwellings, and recommends a limit of 85 dB(G) for infrasound — sound below 20 Hz.15

4. Enforceable limits by frequency at the property line

Octave-band limits sound exotic, but there’s a working American example that’s nearly fifty years old. New Jersey’s noise regulations, N.J.A.C. 7:29, have applied to stationary commercial and industrial sources since 1974.16 At residential property lines, local ordinances based on the state’s model — Cranbury Township’s, for example — limit continuous sound to 65 dBA during the day and 50 dBA at night — and separately limits each octave band, so a source can’t pass on its overall level while producing a loud low-pitched hum.17 The daytime octave-band limits, from 31.5 Hz up to 8,000 Hz, are 96, 82, 74, 67, 63, 60, 57, 55 and 53 dB.18

It’s already been applied to a data center: a sound modeling study for a QTS data center project in East Windsor, New Jersey evaluated the design against those state limits, and notes that octave-band limits are rare, used by only a few states.19

5. Tones

A steady tone is more annoying than broadband sound at the same level, so standard professional practice adds a penalty — typically several decibels — when a tone is present. The methods for detecting tones and setting the penalty are in the international standard ISO 1996 and the UK’s BS 4142, both of which a qualified acoustical consultant will have. An ordinance can require tonal assessment “in accordance with” them.

6. Generator testing and other short, loud events

Backup generators and other intermittent equipment need their own rules, separate from the steady-noise limit: a maximum-level (Lmax) limit, fixed testing hours on weekdays during the day, and a clear definition of “emergency” so routine or money-making generator use isn’t excused. Prince William County’s experience shows why the details matter — its adopted ordinance names generators in its tonal-sound definition and then exempts generator exercising.11

7. Everything together, calibrated locally

Limits should apply to the combined sound of all equipment and all phases of a project, so it can’t comply one building at a time. And no outside source can tell a community what its own nights sound like. Every serious assessment method compares new sound against the existing background, so the step that turns these benchmarks into the right local numbers is a multi-night background survey at representative homes.

At a glance

Kind of noiseBenchmarkSource
Steady noise, night30 dBA inside bedrooms; 45 dBA outdoors at nightWHO, 1999
Steady noise, dayBelow 50–55 dBA outdoorsWHO, 1999
Low-frequency content (trigger)dBC more than 10 dB above dBA → require frequency analysisWHO, 1999; Health Canada
Low-frequency disturbance (indoors)One-third-octave criteria, 10–160 HzUK NANR45; Danish EPA
Infrasound (indoors)85 dB(G)Danish EPA
Outdoor limits by frequencyA-weighted limits plus octave-band limits, 31.5–8,000 HzNew Jersey, N.J.A.C. 7:29
TonesPenalty added when a tone is presentISO 1996; BS 4142
Generators and short eventsLmax limit; daytime weekday testing; narrow emergency definitionCommon ordinance practice
All sources combinedLimits apply to the total, calibrated to local backgroundStandard assessment practice

These benchmarks don’t replace a qualified acoustical engineer. They’re the evidence to ask for — and the questions to bring — when an ordinance is drafted or a noise study is reviewed.

What other communities have done

Prince William County, Virginia has been dealing with data centers next to homes for years; residents of its Great Oak neighborhood near Manassas have suffered extensively from low-frequency data center noise. The county formed a Data Center Ordinance Advisory Group, and its consultant found that A-weighting didn’t effectively capture data center noise.2 Working with the advisory group, the county’s consultants proposed adding dBC limits and octave-band measurement to catch low-frequency sound.25 Its public presentations remain the best plain-language explanation available of why data center noise needs different rules.

What the county actually adopted was a compromise. On October 28, 2025, the Board of County Supervisors voted 5–2 for a modified ordinance that relies on C-weighting rather than the octave-band approach the advisory group and its consultants had developed.8 The county code now sets a C-weighted limit at the receiving property of 73 dB(C) during the day and 68 dB(C) at night.9 According to the Prince William Times, that version was about 5 decibels weaker than the consultant-backed proposal, and in its first six months it produced no noise violations for data centers — leading the board in 2026 to revive the stricter proposal.10 One analysis also points out that the adopted ordinance names backup generators in its definition of steady tonal sound but exempts generator exercising from that standard.11

The lesson for other communities: adding a C-weighted limit is a real improvement, but the number chosen, the measurement method and the exemptions decide whether it actually protects anyone.

Fairfax County, Virginia adopted data center zoning standards, effective September 11, 2024, that require a noise study showing compliance with the county’s noise ordinance before site plan approval, and a post-construction noise study.6 A publicly posted study for one proposed facility shows what that work looks like, including ambient measurements and nighttime background levels.

Wixom, Michigan has proposed a data center noise ordinance that adds a dBC limit to control low-frequency sound, and its public presentation compares limits used by other communities.7 It’s a useful in-state reference for Michigan townships.

If you’re living with the noise

  • Keep a noise log — date, time, how long, what you heard, where you were, and how it affected you (sleep, conversation, windows closed).
  • Record it. Short videos with audio, with the time noted, help document what you’re hearing, especially tones.
  • Phone apps can help, with limits. Free sound meter apps are useful for spotting patterns, but phone microphones are poor at low frequencies and won’t hold up for enforcement. Treat them as a diary, not evidence.
  • Report it to your township or city, in writing, every time.
  • Ask for the studies. The developer’s noise study, and any compliance testing, are often public records. See the FOIA Crash Course.
  • Organize with neighbors. Many logs from many households carry far more weight than one.

Good questions to ask about a proposed facility

  1. Will you submit a noise study, and does it include existing nighttime conditions at the nearest homes?
  2. Does it report dBC and octave-band levels, not just dBA?
  3. What are the predicted levels at each nearby home, day and night, with all equipment and any future phases running?
  4. Are any tones predicted, and how will they be controlled?
  5. When and how often will backup generators be tested?
  6. Will you conduct a post-construction study, and what happens if it shows the limits are exceeded?
  7. Does our local ordinance measure low-frequency sound at all?

Example: Ypsilanti Township’s noise rules

Ypsilanti Township’s zoning ordinance (Section 1400) sets all of its noise limits in A-weighted decibels and requires measurements to follow “the latest American National Standards Institute (ANSI) specifications.” It has no C-weighted or octave-band limit, addresses tones only through a general “unreasonable” test, and has no separate row for industrial uses. That’s typical of Michigan ordinances — and exactly the gap Prince William County found. See How Loud Is a Data Center?


Sources and further reading

  1. Federal Highway Administration, Noise Measurement Handbook (FHWA-HEP-18-065, 2018), and its companion Noise Measurement Field Guide (FHWA-HEP-18-066) with step-by-step checklists.
  2. Prince William County, Virginia, Noise Ordinance Update presentation to the Data Center Ordinance Advisory Group, March 18, 2025.
  3. Prince William County, Great Oak status and proposed ordinance levels, March 19, 2025.
  4. As quoted in Prince William County’s March 19, 2025 presentation, from the Prince William Times, February 27, 2023.
  5. “Prince William outlines noise ordinance update with data centers a priority,” InsideNoVa; and Prince William County, Noise Ordinance Update presentation, January 8, 2025.
  6. Environmental noise study for a proposed data center in Chantilly, Fairfax County, describing Fairfax’s data center use standards, published by FFXnow.
  7. City of Wixom, Michigan, Proposed Data Center Noise Ordinance presentation.
  8. “Prince William Supervisors greenlight new noise ordinance,” Bristow Beat, November 9, 2025.
  9. Prince William County Code of Ordinances, Chapter 14, Noise, § 14-4 (maximum permissible sound levels).
  10. “Prince William supervisors backtrack on data center noise ordinance,” Prince William Times, 2026.
  11. “A noise limit on data centers is only as good as its measurement,” Planetizen.
  12. World Health Organization, Guidelines for Community Noise (1999), full text.
  13. “Community noise investigation of a ‘low level thrumming’ attributed to…,” Canadian Acoustics, describing Health Canada’s 10 dB C-minus-A criterion.
  14. A. Moorhouse, D. Waddington and M. Adams, Proposed criteria for the assessment of low frequency noise disturbance, NANR45, prepared for DEFRA (2005; Revision 1, 2011); companion procedure for assessing low frequency noise complaints.
  15. Danish Environmental Protection Agency, Laboratory Evaluation of Annoyance of Low Frequency Noise (2002), describing the Danish assessment method and infrasound limit.
  16. New Jersey Administrative Code, N.J.A.C. 7:29, Noise Control, via the New Jersey State Library.
  17. Cranbury Township, New Jersey, noise control ordinance, based on the state model.
  18. West Windsor Township Planning Board, Penns Neck acoustical evaluation (May 2023), Table 1, summarizing the N.J.A.C. 7:29 octave-band limits.
  19. SLR, sound modeling technical memo for the QTS data center project, East Windsor Township, New Jersey.
Start Here
What is a Data Center?
Just the Facts
Project Links
Who's Who

Recent Posts

  • The Letter That Spoke for Ypsilanti Township
  • Not Yet: What the Regents Are Being Asked to Approve Thursday
  • “LANL Has Made No Commitment of Funds”
  • Four Years of Regents Minutes
  • How the Words Changed

Archives

  • November 2026
  • October 2026
  • September 2026
  • August 2026

Categories

  • Documents & Records
  • Legislation & Advocacy
  • Public Meetings
  • Responses
  • Uncategorized

Privacy Policy • Terms & Acceptable Use

Copyright 2026 — YpsiDatacenter.org. All rights reserved.

Get action alerts

One small, specific task at a time — a call or an email that takes five minutes. Unsubscribe anytime.