What Tonal Noise Is and Why It Is Judged Differently

Why this matters

A hum complaint is the one that survives good work. You reduce the measured level, the customer agrees the meter moved, and the neighbour is angrier than before. That is not stubbornness, it is physics plus regulation working together: a tone is audible far below the level of the broadband noise around it, and most ordinances that mention tones apply a penalty on top of the measured level before comparing it with the limit. So a job can come back measurably quieter, still failing its limit, and worse to sit under than it was. This card is about recognising a tone, and about the difference between the number you measured and the number you are judged on.

What makes a tone a tone

A tone is energy concentrated in a narrow band relative to what sits either side of it. That is a comparison, not an absolute level, which is why it can only be found in band data and never in an overall level.

The field test is the same shape everywhere it appears: take third-octave bands, unweighted, dB re 20 uPa, Leq over a stated period, and compare each band against its two immediate neighbours. Where a band stands clearly above both of them, you have a tone candidate at that band's frequency. Octave resolution is too coarse for this test - an octave band is three third-octaves wide, so it reports the tone averaged in with two bands of ordinary broadband and the peak flattens out.

The margin that counts as a tone, and the size of any penalty, are set by the instrument you are being judged against, not by physics. A local noise ordinance sets both, and they differ from town to town. The published standards used to argue these cases set them differently again, and they generally require a larger margin at low frequencies than at mid frequencies, because the ear's own frequency resolution is coarser down there. Read the ordinance your site sits under before you quote a margin to anybody. There is no federal community noise limit and no federal tone test.

Why a tone is heard below the noise around it

The ear does not compare a tone with the overall level, it compares the tone with the noise energy in a narrow band around that tone. Broadband noise spreads its energy across the whole spectrum, so only a small slice of it lands in the tone's own neighbourhood and only that slice can hide the tone. A tone sitting 15 dB under the overall broadband level can therefore be plainly audible, and a tone that is inaudible in the daytime can arrive at night when the broadband residual falls and the tone does not.

This is the trap in a broadband control. Take 6 dB off the broadband and leave the tone alone and you have made the tone MORE audible, because you removed the only thing that was partly covering it. The meter says you improved the site. The complainant says you made it worse. Both are reporting honestly.

The penalty is applied to the rating, not to the measurement

When an ordinance says a tonal penalty applies, that penalty is added to the measured level to produce a different quantity: a rated level to be compared with the limit. It is an addition, not a correction. Nothing about the measurement was wrong, so the measured level stays on the record exactly as measured, and the rated level is reported beside it with the penalty named and the clause cited. Writing one number and calling it both is how a report gets torn up.

Penalties commonly sit around 5 dB where an ordinance carries one, but the trigger test, the size and whether it applies at all belong to the adopted ordinance text. Two consequences follow, and they are the reason this distinction is worth the paragraph. First, a penalty is a step change: eliminating the tone removes the whole penalty at once, so it can be worth several times what any realistic broadband reduction buys. Second, a penalty cannot be argued down with better broadband work, because it is not attached to the broadband at all.

The case: a hum that survived a good barrier

A shop had a condensing unit on a low roof beside a residential property line and a night-time hum complaint. Every reading below is LAeq,15min, dB re 20 uPa, overall unless a band is named, taken at the same property-line position with a class 1 meter calibrated before and after, between 11 pm and midnight. Residual with the unit off was 34 dB(A) at that position on each visit. Visit one sat 10.8 dB above it, so its correction was declined and said to be declined. Visits two and three sat 7.3 dB and 6.9 dB above it, inside the zone where a correction applies, so 0.9 dB and 1.0 dB were taken off before any comparison, and every figure below is on the corrected basis.

Roof access on every visit was under the building's fall-protection arrangements with the perimeter guarded, since the measurement position work and the later repair both put people within reach of an unprotected edge, which is what 29 CFR 1910.28 addresses for general industry walking-working surfaces.

Visit one. Measured 44.8 dB(A). The town's night-time limit as adopted is 45 dB(A) at the receiving property line, so on the face of it the site passed by 0.2 dB. Then the third-octave set, unweighted:

The 125 Hz band stands 13 dB above the band below it and 14 dB above the band above it. That is a tone, and its A-weighted contribution is 47 - 16.1 = 30.9 dB(A), which is 13.9 dB below the 44.8 dB(A) total and therefore contributes under 0.2 dB to it. The entire complaint was living inside a band that the overall level could not see. Its frequency, near 120 Hz on a 60 Hz supply, is twice line frequency, which is the one candidate frequency that does not move when equipment speed changes.

With the ordinance's tonal penalty of 5 dB applied, the rated level was 49.8 dB(A), failing the 45 dB(A) limit by 4.8 dB. The site that appeared to pass by 0.2 dB was actually failing by nearly 5.

Visit two, after the obvious control. The shop fitted an absorptive screen whose published insertion loss is given from 500 Hz upward. It worked: the content at 500 Hz and above fell about 6 dB, and the measured overall dropped from 44.8 to 41.3 dB(A), a genuine 3.5 dB improvement that any meter would confirm. The 125 Hz band was unchanged at 47 dB unweighted, because the screen has no published performance there and because a low-frequency tone was never going to be a screen's problem. Rated level with the penalty still applied: 46.3 dB(A), still failing by 1.3 dB. The complainant reported the hum as worse, which is what unmasking sounds like.

Visit three, the fix that mattered. With the unit de-energized, locked out and verified dead at the disconnect before any cabinet panel was touched, as 29 CFR 1910.333(b)(2) requires for work on electrical equipment, a cabinet panel was found free to vibrate at twice line frequency alongside the compressor. Before drilling the panel for a stiffener, the crew traced and cleared what sat behind it: a refrigerant line drilled into is a pressurized-system release and a burn, and a wiring bundle drilled into is a fault on re-energizing. Stiffening and damping that panel took the 125 Hz band from 47 to 33 dB unweighted, which put it level with the 160 Hz band and below the 100 Hz band, so the tone test no longer found a tone.

Measured overall fell from 41.3 to 40.9 dB(A), a change of 0.4 dB that nobody would notice on a meter. But the penalty came off, so the rated level went from 46.3 to 40.9 dB(A), a change of 5.4 dB, and the site passed by 4.1 dB. The complaint stopped that week.

Sibling-rule check. Every level carries quantity, weighting, reference, bandwidth and time basis, and the band levels are unweighted with their A-weighted contribution shown separately rather than mixed. The background gate is applied per figure rather than once for the visit, and all three overall levels are on the same corrected-or-declined basis, so no corrected figure is being compared against an uncorrected one. The penalty is presented as an addition producing a rated level, never as a correction to a measurement. The tone frequency is identified by the fixed twice-line-frequency field rather than assumed, consistent with the spectrum card. The 3.5 dB broadband gain is not credited with anything the tone owns.

The general lesson in those three numbers

A 3.5 dB measured improvement bought no change in outcome. A 0.4 dB measured improvement bought a 5.4 dB change in the rated level and ended the complaint. The tone was worth about thirteen times the broadband work in rated terms, and it cost less to fix.

That ratio is not a coincidence of this site. Whenever a penalty is in play, the tone is a step you either clear or do not, and steps are worth more than slopes. It also generalises past ordinances: even where nothing is being enforced and there is no penalty clause anywhere, a human complainant applies their own penalty to a tone, and they apply it at levels a meter would call inoffensive.

What would change the recommendation. If the ordinance carries no tonal provision, the rated level equals the measured level and the arithmetic above collapses to a plain 0.4 dB gain. The tone is then a customer-relations problem rather than a compliance one, and it is still usually the right thing to fix first, but the business case changes and you should say so. If the tone is generated inside a machine rather than by a radiating surface bolted to it, detuning it means changing speed, blade count or tooth count, which is a manufacturer conversation and not a field repair. And if the tone is at a frequency that moves with shaft speed rather than sitting fixed at twice line frequency, it is aerodynamic or mechanical, and the identification path runs through the spectrum worksheet rather than through the electrical side.

How to verify you got this right

Re-run the tone test, do not re-run the level. The question is whether the suspect band still stands above both neighbours by the margin the governing ordinance names, at the same position and the same time basis as the original set. A band that fell but still stands proud has not cleared the trigger and the penalty still applies.

Then check the one thing a level cannot tell you: ask the complainant. A tone that has been detuned rather than removed will still be reported, at a different pitch, and your band data will show it moved rather than gone.

References

• Your local noise ordinance as adopted and amended by the jurisdiction the receiving property sits in, which is the instrument that sets the limit, the tone test, the penalty and the measurement position; there is no federal community noise limit
• 29 CFR 1910.333(b)(2), for the de-energize, lock out and verify sequence required before working on the electrical equipment involved in the repair above
• 29 CFR 1910.28, walking-working surfaces, for fall protection during roof-level measurement and repair work in general industry
• See related: What Octave Bands Tell You That an Overall Level Cannot; How to Read a Spectrum Back to a Mechanical Cause