Fan Noise or 50 Hz Hum? A Real Recording Diagnosis

The recording and what I did not know

This started with a short phone recording labelled only as an electric fan. I did not have reliable notes for microphone model, distance, fan speed, room dimensions, or phone processing. Rather than invent a controlled setup after the fact, I kept those gaps visible: this is a field sample, not a laboratory measurement.

The supplied file is 14.28 seconds of mono AAC audio in an M4A container at 44.1 kHz. I decoded it once to PCM WAV for repeatable analysis. I did not normalize it, raise the gain, denoise it, or filter the low end. The M4A is the original listening file; the WAV is the decoded analysis copy. Decoding does not restore information already discarded by AAC.

What the analyzer measured

The whole-file pass measured −23.21 LUFS integrated, −20.87 dBFS RMS, and a −0.07 dBFS sample peak. The quietest measured window was still −29.69 dBFS, which fits an almost continuous mechanical source rather than a recording with a clean silent baseline.

The spectrogram is more useful than a single spectrum here. Low-frequency energy changes over the recording, while the average spectrum does not present one clean fundamental with a stable ladder of harmonics.

Why I did not label it mains hum

The automatic screening pass raised several 50 and 60 Hz candidates near the start. That is useful as a list of places to inspect, but it is not a verdict. A rotating fan, air movement, room modes, phone high-pass behaviour, and genuine electrical hum can all put energy in the same low-frequency neighbourhood.

I checked the candidates against the full spectrogram and the average low-frequency spectrum. The suspected fundamental was only about 1 dB above its local background in the whole-file view, and the supporting harmonics were not stable enough across time. My conclusion is therefore deliberately narrower: broad, changing low-frequency fan noise is present; stable mains hum is not established by this recording.

How to repeat the check

Use the original M4A if you want to reproduce the browser workflow, or the decoded WAV if you want the exact PCM used for the figures. Small decoder differences should not change the practical conclusion, but they can move the last decimal place of a peak result.

1. 01

   Load the complete file

   Open the original recording and allow the complete timeline to finish before judging any hum candidate.

   

2. 02

   Focus the low-frequency analysis

   Use a 4096-point FFT and inspect the bass range around 20–250 Hz.

   

3. 03

   Compare time around the candidate

   Open the spectrogram and compare the suspected timestamp with the seconds immediately before and after it.

   

4. 04

   Check the harmonic evidence

   Look for a persistent fundamental plus harmonics at exact multiples, not just one bright low-frequency patch.

   

5. 05

   Listen before naming the cause

   At a safe level, decide whether the sound behaves like a steady electrical tone, rotating modulation, airflow, or a mixture.

   

Limits of this result

This recording cannot support calibrated sound-pressure claims. The phone microphone, automatic gain control, AAC encoding, unknown placement, and unknown fan setting all affect the result. A better follow-up would record the same fan at several distances and speeds, add ten seconds with the fan switched off, and capture a second track from a known audio interface.

Loudness and True Peak results at 44.1 kHz are engineering estimates under the analyzer's stated method, not a compliance certificate. The useful result here is comparative and diagnostic: it shows what is in this file and how strong the evidence is for each interpretation.