Noise Lab
Drop a data center anywhere in Europe and see how far its hum carries.
Octave-band ISO 9613-2 propagation, computed live in your browser —
drag the building, move the sliders, watch the plume breathe.
How this model works — every assumption, honestly
The physics
Sound power from each piece of plant is propagated to ~15,000 receptor points in
8 octave bands (63 Hz–8 kHz) using ISO 9613-2, the method used in real
environmental noise assessments: spherical spreading (−6 dB per doubling of distance),
frequency-dependent air absorption per the full ISO 9613-1 formula (your temperature
and humidity sliders feed straight into it), the standard's per-band ground effect, and
single-edge diffraction over the optional perimeter wall (capped at 20 dB, with the
standard's meteorological correction). Band levels are A-weighted to human hearing and
energy-summed. Receptors are at 1.5 m ear height.
The sound sources
The facility is a square building (~700 m² per MW of IT load) populated from typical
vendor sound data: air-cooled chillers ≈ 99 dB(A) sound power per ~1.4 MW unit on the
roof, cooling-tower cells ≈ 97 dB(A) in an east yard, dry coolers ≈ 93 dB(A), rooftop
air handling ≈ 94 dB(A) per 3 MW, grid transformers ≈ 90 dB(A) with their 100/125 Hz
hum, and — when testing — enclosed, silenced 2.5 MW standby gensets at ≈ 108 dB(A) each
in a west compound. Each type carries a realistic octave spectrum. Night mode runs cooling
at part load (≈ −3 dB, fan affinity laws); the mitigation package applies −8 dB to cooling
plant and −10 dB to gensets — the realistic span between default and best-practice design.
The metrics
Day is full-load daytime LAeq (plus gensets when a test runs). Night is
Lnight: part-load cooling (fan affinity laws, ≈ −3 dB), no generator tests. Lden is
the EU Environmental Noise Directive's 24-hour composite (evening +5 dB, night +10 dB
weighting), combined at source level — exactly equivalent, since propagation is
level-independent. The rating level toggle adds the BS 4142 / TA Lärm-style +5 dB
character penalty wherever tonal plant (transformer hum, genset firing) contributes within
6 dB of the total — that's the number many regulators actually compare against limits.
The cyan dashed contour marks where the plant drops to the typical background level
of your chosen area character: beyond it, you'd struggle to pick the site out of the
soundscape.
Weather & wind
Temperature and humidity feed the ISO 9613-1 absorption directly. Downwind worst case
is the standard's core method — wind carrying sound toward every receptor at once.
Long-term average applies the standard's §8 meteorological correction
(Cmet, C0 = 3 dB) directionally around your prevailing wind: receptors
downwind stay at the worst case, receptors upwind average up to 3 dB quieter — which is
why the plume goes egg-shaped. Stated accuracy of the method is ±3 dB (the optional
uncertainty bands show it on the 45/55 contours).
What it deliberately ignores
No terrain and no shielding or reflection by other buildings (real cities shield and
canyon) — the two genuine fidelity ceilings of this browser tool. Flat ground, a generic
north-aligned layout, and closed windows knock a further ~20–25 dB off indoors. This is a
screening tool to build intuition about siting, scale and mitigation — a real permit
needs a certified assessment with measured source data and 3-D geometry. For scale:
55 dB(A) is a common daytime residential limit in EU regulation, 40–45 dB(A) a common
night limit, and the WHO night guideline for annoyance-free sleep sits near 40 dB outside
your window.
Sources
ISO 9613-1/-2, IEC 61672, Directive 2002/49/EC, WHO Night Noise Guidelines for
Europe (2009), vendor sound data / ASHRAE Applications ch. 49 — the full citation
list and the model-vs-published cross-checks for both labs live in
REFERENCES.md.