This is the most technical article on this site, and also the most important one if you are trying to understand why measurement figures and people's experience so often fail to match.
What A-weighting is
A sound level meter by itself registers all frequencies equally. The human ear does not. At quiet levels the ear is least sensitive to very low and very high frequencies and most sensitive around 1–4 kHz.
To bring a measurement closer to human perception, a weighting filter is applied. The most common is A-weighting, and the result is written as dB(A).
How much A-weighting removes
| Frequency | A-weighting correction |
|---|---|
| 20 Hz | about −50 dB |
| 31.5 Hz | about −39 dB |
| 50 Hz | about −30 dB |
| 125 Hz | about −16 dB |
| 315 Hz | about −6.6 dB |
| 1000 Hz | 0 dB |
The figures are approximate, but the order of magnitude is exactly this. It means that a 50 Hz signal physically as strong as a 1000 Hz signal will appear about 30 decibels quieter in an A-weighted measurement.
Why this is a problem specifically for a hum
A-weighting was developed and calibrated to model perception of quiet, broadband sounds. It works well for traffic noise or general urban background.
It works less well in two cases, and industrial hum falls into both:
- When the sound is narrowband and tonal. A continuous tone is perceived as far more disturbing than broadband noise at the same level.
- When the sound carries substantial low-frequency energy. Low frequencies produce not only an auditory sensation but also vibration and a sense of pressure, which A-weighting does not model at all.
Why this changes conclusions
Noise reports often contain a conclusion in this form: "in the low-frequency range the noise level is considerably lower and no dominance can be established".
Where such a conclusion is drawn from A-weighted third-octave data, it can be self-confirming. Take a simple example with round numbers.
| Band | A-weighted | A-correction | Approx. linear |
|---|---|---|---|
| 50 Hz | 47 dB(A) | −30 | ~77 dB |
| 315 Hz | 71 dB(A) | −6.6 | ~78 dB |
In the A-weighted data the 50 Hz band looks 24 decibels quieter than the 315 Hz band — and from that follows the conclusion that low frequencies do not dominate. Converted to linear values the difference all but disappears: physically, the two bands carry roughly equal energy.
An important caveat
This example is an indicative illustration, not an acoustic assessment. Points like this are worth putting to a competent authority as questions rather than assertions — and it is never worth challenging a laboratory's accreditation or competence. The objection concerns methodology, not the accuracy of the measurement.
The significance is practical: if the conclusion that low frequencies are absent is methodologically questionable, then so is the conclusion that follows from it — that there is no tonal hum problem.
The appropriate method for assessing low frequencies
International practice for low-frequency noise assessment generally uses:
- Z-weighting (linear) — no frequency correction at all;
- C-weighting — far less low-frequency roll-off than A;
- the dB(C) − dB(A) difference — a widely used indicator: a difference above roughly 15–20 dB points to significant low-frequency content;
- third-octave band analysis compared against low-frequency criterion curves.
Three questions worth asking
If your situation involves a hum and the report concludes that low frequencies do not dominate, three questions are worth putting to the competent authority:
- whether it is methodologically sound to draw a conclusion about low-frequency dominance, or its absence, from A-weighted data;
- whether Z- or C-weighted levels should be used for assessing low-frequency noise;
- whether the laboratory could be asked for the Z-weighted third-octave data from the same measurement.
The third is the practically important one: modern instruments normally record all weightings simultaneously, so it requires no new fieldwork — only the same data in a different form.
What this site documents
This article is published by the Jūrkalnes and Irlavas neighbourhood community. We document the industrial noise produced by the outdoor Nestro filter at SIA Livonia Print, Jūrkalnes iela 15/25, Riga. The installation has run 24 hours a day since 2021, and its hum reaches around 12 streets. The residents’ request to the company is a single one: fit an acoustic enclosure around the filter and sound insulation on the air ducts.
