🔊

HVAC Noise Analyzer

Diagnose heating, ventilation and air-conditioning noise from your microphone. Watch the live spectrum across the bands where fans, ducts and air handlers misbehave, match a peak to a fan’s blade-passage frequency, find duct resonances from a duct length, and compare octave bands against the published NC (Noise Criteria) curves — with a plain-English hint for rumble, whoosh and whine.

This is an uncalibrated estimate from a consumer microphone — not a certified sound-level meter and not valid as compliance, complaint or legal evidence. The spectral shape and the fan blade-passage / duct math are reliable (and the BPF and duct formulas are exact). The NC-curve comparison needs SPL calibration to be quantitative — without it, it is a relative shape comparison only. Auto gain control and noise suppression must be off (this tool requests raw audio); a reading is meaningless otherwise. Phone/laptop mics roll off the deep bass and generally cannot capture true infrasound (<20 Hz). Nothing is recorded or uploaded.

Microphone

Idle — press Start. Audio is analyzed live and never recorded.

Consent: pressing Start asks your browser for microphone access. The mic feed is analyzed in real time only — never recorded, saved or uploaded.

Live spectrum (20 Hz – 5 kHz)

Broadband level
Strongest peak
Character

Broadband level is in dBFS (relative to digital full scale). Set the calibration offset below to also see an approximate dB SPL.

NC (Noise Criteria) comparison

In relative mode the octave bands are shifted to the loudest band so you can read the shape against the NC family. In SPL mode each band uses your calibration offset to estimate true dB SPL — only meaningful once calibrated.

Start the mic to compare octave bands to the NC curves.
Not calibrated — the NC comparison is relative only. To calibrate: read the level on a real sound-level meter or a calibrated phone app, type it in “known dB SPL”, and press Calibrate. The offset is shared across every noise tool on this site.

Fan blade-passage frequency (BPF)

BPF = (RPM ÷ 60) × number of blades. A spectrum peak at the BPF (or a harmonic) points the noise straight at that fan. Math is exact.

Duct resonance helper

A duct rings like a pipe. Quarter-wave (one open + one closed end): f = c ÷ (4L). Half-wave (both ends open, or closed): f = c ÷ (2L). Math is exact for the idealised duct.

How It Works

HVAC noise has a handful of fingerprints. A fan pushes air with a set of blades, and every time a blade sweeps past a fixed point (the housing cut-off, a strut, a grille) it makes a pressure pulse. Repeated at the rate the blades pass, this produces a tone at the blade-passage frequency, BPF = (RPM ÷ 60) × blades, plus harmonics at 2×, 3× and so on. Find a sharp spectrum peak, compute the BPF for your fan, and if they line up you have identified the source — this part of the tool is exact arithmetic and is calibration-independent.

A duct behaves like an organ pipe: standing waves form at frequencies set by its length and end conditions. A length open at one end and closed at the other resonates at f = c ÷ (4L) and its odd harmonics; a length open (or closed) at both ends resonates at f = c ÷ (2L) and all harmonics, with c ≈ 343 m/s. If a peak in the spectrum matches a predicted duct resonance, that run of duct is amplifying the noise and may need lining, a plenum, or a length change.

The NC (Noise Criteria) curves are a family of standardised octave-band limits (here NC-15 through NC-65 at the published octave centres 63, 125, 250, 500, 1000, 2000, 4000 and 8000 Hz) used to rate steady background noise in occupied spaces. The NC rating of a sound is the lowest curve that is not exceeded in any band. This tool overlays your measured octave bands on the NC family. The honest catch: a true NC rating needs calibrated dB SPL. Without calibration the comparison is a shape check — useful for seeing whether your noise tilts toward low-frequency rumble or high-frequency hiss relative to the curves, but not a real NC number. Calibrate against a sound-level meter and the SPL mode becomes quantitative (still an estimate from a consumer mic).

Finally, the character hint reads the spectral balance: energy bunched in the low bands tends to be rumble (low-frequency fan imbalance or structure-borne vibration); a broad mid-band hiss is whoosh (high air velocity at grilles/diffusers); and a strong narrow tone up high is whine (blade or bearing). These are guidance, not a diagnosis.

Frequently Asked Questions

Can this give me a real NC rating or dB level for my building?
Not on its own. A consumer microphone is uncalibrated and the browser/OS may apply automatic gain, so the broadband level is in relative dBFS, not certified dB SPL. Until you enter a calibration offset (from a real sound-level meter), the NC comparison shows only the shape of your noise against the curves. Even after calibrating it is an estimate, not compliance-grade evidence — for code or complaints use a Type 1/2 sound-level meter.
How do I match a peak to a fan?
Enter the fan’s speed (RPM) and blade count. The tool computes the blade-passage frequency, BPF = (RPM ÷ 60) × blades, and its first few harmonics. If a sharp peak in the spectrum sits at the BPF or a harmonic, that fan is almost certainly the source. The BPF arithmetic is exact and does not depend on calibration, so this is one of the most reliable things the tool does.
What is the duct resonance helper for?
A duct acts like a pipe and rings at frequencies set by its length and ends. Enter the run length and pick the end condition: quarter-wave (one open, one closed) gives f = c ÷ (4L); half-wave (both open or both closed) gives f = c ÷ (2L), using c ≈ 343 m/s. If a measured peak matches a predicted resonance, that duct is amplifying noise and is a candidate for lining or redesign. The formulas are exact for an idealised straight duct; bends, fittings and temperature shift the real values.
Why must auto gain control and noise suppression be off?
Those features constantly change the gain and reshape the spectrum to favour speech, which corrupts every noise measurement — a steady fan would appear to fade, and tonal hum could be filtered out. This tool requests raw audio with autoGainControl, noiseSuppression and echoCancellation all disabled. If your OS still forces them on, the reading is not trustworthy.
Can it detect very low rumble or infrasound below 20 Hz?
Only partly. The spectrum focuses on the bands where HVAC problems usually live, but phone and laptop microphones roll off steeply in the deep bass and generally cannot capture true infrasound (<20 Hz). If you suspect sub-20 Hz rumble or pressure pulsation, treat the low end of the display with caution and consider dedicated low-frequency instrumentation. The Low Frequency Noise Detector is a better starting point for that range.
Is my audio recorded or sent anywhere?
No. The microphone signal is analyzed in real time in your browser to draw the spectrum and compute levels, and is never recorded, saved or transmitted. All calculation stays on your device. The microphone is released when you press Stop or close the tab. The calibration offset is stored locally (in your browser) only.