Noise Monitoring in Data Centers: How to Measure, Document, and Comply with OSHA 1910.95

Most data centers have never conducted a formal noise survey. The assumption that “it’s a tech environment, not a factory” has left a compliance gap that OSHA does not recognize. If workers are exposed at or above 85 dBA TWA, the monitoring requirement applies regardless of industry classification.

Soundtrace integrates noise monitoring data directly into each worker’s unified cloud profile, linking measured TWA to audiometric history, HPD selection, and STS determinations in a single record.

This guide covers the practical mechanics of noise monitoring in data center facilities: which instruments to use, where and when to measure, how to document results, and what triggers re-monitoring under OSHA 1910.95.

The CDC estimates that 22 million U.S. workers are exposed to hazardous noise annually. As data center capacity is projected to nearly double over the next five years, with the sector adding an estimated 97 GW of capacity between 2025 and 2030, the number of noise-exposed data center workers is growing faster than the compliance infrastructure to protect them.

When Noise Monitoring Is Required

OSHA 1910.95(d)(1) requires employers to develop and implement a monitoring program whenever information indicates that any employee’s exposure may equal or exceed 85 dBA TWA. In a data center, the “information” that triggers this requirement is straightforward: server rooms, generator test bays, and mechanical plant areas routinely produce noise above 85 dBA.

Re-Monitoring Triggers

Under 1910.95(d)(3), re-monitoring is required whenever a change in production, process, equipment, or controls may have increased noise exposures. For the data-center-specific list, see the next section. The general OSHA doctrine is covered in OSHA Noise Re-Monitoring Requirements: Change-in-Process.

Data-Center-Specific Re-Monitoring Triggers

1910.95(d)(3) is written generically because it has to cover everything from steel mills to dental offices. The translation for a modern data center is more specific, and missing one of these triggers is the single most common reason a data center’s monitoring records go stale.

Best practice is to bake re-monitoring into change management. If the change ticket affects power density, cooling topology, generators, or hall layout, it should require sign-off from EHS with either fresh dosimetry or a documented justification for why existing data still represents the worker’s exposure.

Commissioning vs. Steady-State Monitoring

Commissioning is not steady-state, and a single set of monitoring records cannot legally or technically cover both. During Level 4 functional performance testing and Level 5 integrated systems testing, generators, chillers, CDUs, and load banks run at full or stepped load for hours. The acoustic environment in the building is not what operations staff will experience after handover, and the people in the building are not the people who will be there long term.

Two distinct monitoring efforts are required:

• Commissioning monitoring. Conducted during fit-out and the Cx test sequence. Construction trades are covered under 29 CFR 1926.52; operations and QA staff present during Cx are covered under 1910.95. Document the test scenario for every dosimetry sample (e.g., "Generator A load-step test, 0→100% in 25% increments, 4-hour duration") so the record is reproducible.
• Steady-state monitoring. Conducted after handover, with IT load installed and operations staff working their normal shifts. This is the dataset that drives HCP enrollment for the operating workforce. Schedule it after the first 30–60 days of operation so the equipment is in its real-world duty cycle, not its initial burn-in.

Treating commissioning data as steady-state typically over-states TWAs (and over-enrolls workers in HCP). Treating steady-state data as commissioning data under-states peak exposures and under-protects the test crews. They are different programs with different records.

Generator and Load-Bank Test Monitoring

Generator and load-bank testing is the most predictable high-noise event in a data center. Monthly no-load generator exercises, quarterly load-bank tests, and annual full-facility transfer tests are scheduled, repeatable, and run by a known crew. That predictability is the opportunity: monitor a representative test for each scenario, document it once, and you have a defensible record for every subsequent test of the same configuration.

Monitoring approach:

• Area survey at the generator yard. Measure dBA at the generator face, at the operator position, at the exhaust silencer outlet, and at the property line. Repeat at 100% load with the load bank online. Property-line measurement is for community-noise records, not OSHA, but it lives in the same file.
• Personal dosimetry on every crew member. A test day TWA of 88–96 dBA is common for crew who spend the full test in the yard or switchgear room. Peaks of 100–105 dBA at the generator face are normal and require HPD regardless of TWA.
• Document the test scenario. Number of generators online, load step sequence, duration, weather, and whether silencer access doors were open. These variables move dBA by 5–10 dB and the record must reflect what was actually tested.
• Mandate dual HPD during the test. Treat generator test days as scheduled high-noise tasks. See HPD selection: Hearing Protection for Data Center Workers.

Colocation Tenant Equipment Monitoring

In a colocation facility, the noise environment in any given hall is the sum of the host’s mechanical systems and every tenant’s rack-level equipment. The host owns the shared environment; tenants own what they put in their cages. Monitoring responsibility splits along the same line, but the practical handoffs need to be written down.

Host obligations:

• Conduct and maintain area surveys for every operating hall, generator yard, mechanical room, and common space.
• Publish current room dBA in the tenant onboarding packet so tenants can make defensible HPD and HCP decisions for their workers.
• Conduct personal dosimetry for host facilities, mechanical, electrical, and security staff who work in shared zones.
• Re-monitor when host-controlled equipment changes (mechanical changes, generator additions, containment retrofits).

Tenant obligations:

• Conduct personal dosimetry for tenant deployment and rack staff during representative shifts inside the tenant cage.
• Notify the host in writing when planning equipment changeouts that materially change rack-level noise (most commonly, GPU pod deployments).
• Maintain HPD and audiometric records for tenant workers regardless of who owns the hall.

The multi-employer doctrine that governs this split is the same one covered in Contractors & Subcontractors: Hearing Conservation OSHA Compliance and Who Is Responsible for Vendor Employee Hearing Conservation. The clearer the contractual handoff in the master service agreement, the less likely either party will be cited for the other’s gap.

Two Methods: Area Surveys vs. Personal Dosimetry

Data center noise monitoring requires both area surveys and personal dosimetry, used for different purposes. See the full breakdown: Area Monitoring vs. Personal Noise Monitoring.

Area Noise Surveys

Area surveys use calibrated Type 2 (or better) sound level meters positioned at worker ear height throughout the facility. The purpose is to create a facility noise map that identifies zones at or near the action level.

In a data center, measure at minimum:

• Center of each server room hot aisle and cold aisle
• In front of CRAC/CRAH units
• Generator room or generator pad (during operation and during testing)
• Mechanical plant areas (chillers, cooling towers, UPS rooms)
• Network operations center (NOC) or monitoring rooms adjacent to server halls
• Loading docks and receiving areas
• Office spaces adjacent to data halls

Record the measurement location, date, time, instrument serial number and calibration date, and the measured dBA (slow response, A-weighted).

Personal Dosimetry

For HCP enrollment decisions, personal dosimetry provides the legally defensible data. The dosimeter is worn by the worker with the microphone clipped within 30 cm of the ear during a full representative shift including all high-noise tasks.

Data center roles that typically require dosimetry:

Server room technicians spend substantial time in the loudest zones and should be prioritized. Their exposure profile includes hot aisle work, rack-level troubleshooting, and cable management tasks directly adjacent to fan exhaust.

Maintenance and HVAC workers access mechanical areas that may exceed server room levels. Generator testing periods can push their TWA significantly above baseline.

Security and facilities staff who patrol server rooms on regular rounds accumulate exposure intermittently throughout the shift.

Construction workers during buildout are subject to separate standards (1926.52 for construction) but the monitoring obligation still applies.

Documentation That Survives an OSHA Inspection

Every noise monitoring event must generate a record that includes:

• Date, time, and duration of the measurement
• Location (specific enough to identify the zone)
• Employee name or job title for personal dosimetry
• Instrument make, model, serial number, and calibration date
• Measurement settings (exchange rate, criterion level, response time)
• Pre-use and post-use calibration readings
• Measured TWA, dose percentage, or area dBA reading
• Name of person conducting the monitoring

OSHA requires noise exposure records to be retained for at least two years. Best practice for data center operators is to retain records indefinitely, given the long latency of noise-induced hearing loss and the potential for workers’ compensation claims years after exposure.

For the EHS director inheriting a data center hearing conservation program with missing monitoring records, the gap is not just a compliance problem. It is a legal exposure problem. Every missing record is a period of unquantified risk, a shift where someone may have been overexposed without documentation, without protection, without anyone measuring what was happening.

How Soundtrace Simplifies Data Center Noise Monitoring

Soundtrace integrates noise monitoring data directly into each worker’s unified cloud profile, linking measured TWA to audiometric history, HPD selection, and STS determinations in a single record. For data center operators with distributed facilities:

• Real-time noise dashboards show current exposure data across all sites
• Automatic flagging when workers exceed the action level or PEL
• Integration with audiometric testing schedules for immediate follow-up
• Digital records with timestamp, location, and instrument metadata
• Multi-site compliance visibility from a centralized dashboard

Learn how Soundtrace noise monitoring works →

Frequently Asked Questions