What a Noise Criterion Curve Is For

Why this matters

Two offices can measure within a decibel of each other on an overall A-weighted reading, rate at the identical criterion number, and get opposite reactions from the people sitting in them. That happened in the pair of rooms below: 44.0 and 43.2 dB, A-weighted sound pressure level re 20 micropascals, overall, equivalent-continuous over 15 minutes, both rating NC-40 under the tangency rule, one accepted without comment and one rejected in writing. A criterion curve exists because acceptability follows the spectrum's shape, and its real payoff is that the failing band names the piece of equipment to go after. An overall number tells you a room is loud. A band-by-band comparison tells you what to fix.

What an overall level cannot tell you

A-weighting was built to approximate the ear's sensitivity at low levels, and one of the things it does is discard low frequency hard. At the 63 Hz octave band the A-weighting adjustment is about minus 26 dB. A duct rumble that dominates the room can arrive at the meter almost fully discounted, so a rumbly space and a hissy space land at the same overall figure and feel nothing alike.

That is not a flaw in A-weighting, it is A-weighting doing its job for a different question. The occupational question ("is this exposure hazardous") is A-weighted, and 29 CFR 1910.95 sets the federal floor there with an 85 dB A-weighted eight-hour time-weighted average action level and a 90 dB permissible exposure limit at the same averaging. The habitability question ("is this background acceptable in an office") is a spectrum question, and it needs octave bands.

What a criterion curve actually is

A criterion curve family is a set of published per-octave-band limits, each labelled with a number. The measured background is plotted against them band by band, and the room's rating is the lowest curve that no measured band exceeds. That last clause is the whole method and it is called the tangency rule: one band touching a curve sets the rating for the entire room, no matter how far under the other seven bands sit.

The two families in common use:

• NC, published over the 63 through 8000 Hz octave bands, rated by tangency. Simple, widely specified, and silent about shape once it has produced its number.
• RC, published over 16 through 4000 Hz, rated as the arithmetic average of the 500, 1000 and 2000 Hz octave band levels, plus a letter describing shape: neutral, rumble or hiss. It exists because the tangency number alone kept passing rooms occupants hated.

Both live in the ASHRAE Handbook, HVAC Applications, in the sound and vibration control chapter. That is design guidance, not code: it binds when a specification, a contract or an owner's project requirements document names it, in the edition named, and not otherwise. Typical published design guidance runs in the region of NC-25 to NC-30 for a conference room, NC-30 to NC-35 for a private office, and NC-35 to NC-40 for an open plan, and the numbers you commission against are the ones in the specification rather than the ones in anyone's memory.

The gate, stated once

A room meets a criterion of NC-40 when no octave band from 63 through 8000 Hz exceeds the published NC-40 limit for that band, with the measurement taken as unweighted sound pressure level re 20 micropascals, per octave band, equivalent-continuous over a stated duration, at the occupant position with the systems in their normal occupied state.

Read the curve values out of the Handbook table when a rating is contractual rather than from memory. The NC-40 limits used below are 64, 56, 50, 45, 41, 39, 38 and 37 dB at 63, 125, 250, 500, 1000, 2000, 4000 and 8000 Hz.

Two rooms, one gate

Both are private offices on the same floor, same air handler, measured at the seated position, unweighted sound pressure level re 20 micropascals, per octave band, equivalent-continuous over 15 minutes, systems in occupied state.

Room A against the gate. Every band is under the NC-40 limit. The closest is 125 Hz at 1 dB under, so 125 Hz is the tangent band and the rating is NC-40.

Room B against the gate. Every band is under. The closest are 2000 Hz and 4000 Hz, both 3 dB under, so those are the tangent bands and the rating is also NC-40.

Overall A-weighted, for comparison. Applying the A-weighting adjustment per band and summing on energy rather than arithmetically: Room A is 44.0 dB, A-weighted, overall; Room B is 43.2 dB. Eight tenths of a decibel apart, which is below what anyone can hear as a difference in level.

Same gate, same rating, same overall level. Room A generated a written complaint about a low throb that people said they felt in their chest. Room B generated nothing.

The second reading, and where the letter comes from

RC takes the same eight measured numbers and asks a different question.

Rating. Arithmetic average of the 500, 1000 and 2000 Hz octave band levels, unweighted.

• Room A: (39 + 34 + 30) / 3 = 34.3, so RC 34.
• Room B: (39 + 37 + 36) / 3 = 37.3, so RC 37.

Reference line. A line through the rating value at 1000 Hz, sloping at minus 5 dB per octave, holding the rating constant. For Room A that is 54 at 63 Hz, 49 at 125, 44 at 250, 39 at 500, 34 at 1000, 29 at 2000, 24 at 4000. For Room B, 57, 52, 47, 42, 37, 32, 27 across the same bands.

Deviation, Room A low bands. 63 Hz measured 62 against 54, plus 8. 125 Hz measured 55 against 49, plus 6. 250 Hz measured 46 against 44, plus 2. Two bands more than 5 dB over the reference in the low region, which is the rumble condition. High bands: 2000 Hz plus 1, 4000 Hz plus 1, nothing. RC 34(R).

Deviation, Room B low bands. 63 Hz minus 12, 125 Hz minus 10, 250 Hz minus 7. Nothing. High bands: 2000 Hz measured 36 against 32, plus 4; 4000 Hz measured 35 against 27, plus 8. Over the reference by more than 3 dB in the high region, which is the hiss condition. RC 37(H).

Room B has the higher RC rating number and the acceptable room. That is not a contradiction: the number describes the mid-band level and the letter describes the shape, and in this pair the shape is what the occupants were reacting to. A shaped, mildly hissy background at 43 dB, A-weighted, overall reads as ordinary building noise and does useful masking work. A low-frequency throb at the same overall level reads as a fault.

Where the deviation thresholds come from. The 5 dB low-region and 3 dB high-region deviations are the RC method's own, and they belong to whichever edition of the Handbook chapter your specification names. State the edition alongside the rating when you hand a number to anyone.

Sibling-rule check. Every level above carries quantity, weighting, reference, bandwidth and time basis, including in the table headers where plain speech would drop them. The band values are unweighted, the overall values are A-weighted, and the two are never compared directly or averaged together. The A-weighted overall figures were summed on energy, matching the level-addition article. 29 CFR 1910.95 is cited as the occupational floor and explicitly not as a habitability criterion. The masking observation about Room B matches the background-noise article's direction: raising a well-shaped background lowers intelligibility of intruding speech, which is a gain in an office and a loss in a training room. No rounding runs toward the flattering answer; the RC ratings were taken to the integer below the computed average in both rooms, which is the stricter reading for Room A's rumble finding.

What each shape points at

The letter is a work order.

• Rumble points at low-frequency energy that survived the duct run: fan-generated tones, a plenum too small, a duct that runs straight from the unit to the outlet, a lined section that stops absorbing below 250 Hz because the lining is too thin, or a structure-borne path around the isolation. The fix is upstream and it is almost never at the diffuser.
• Hiss points at velocity at or just before the outlet: a balancing damper too close to the diffuser, a neck too small for the flow, a flexible duct pulled into a bend. The fix is usually within a few feet of the outlet and it is usually cheap.
• Neutral, and still too loud is the honest case where the whole background needs to come down and the answer is airflow, path length or a bigger piece of equipment.

Getting above the ceiling to trace any of this means a portable ladder set per 29 CFR 1910.23 and eye protection per 29 CFR 1910.133, and in a building whose vintage is not documented, material above the grid is presumed asbestos-containing under 29 CFR 1910.1001(j) until sampled, which is an inhalation route rather than a contact one.

The failure mode to catch: accepting a room on a single A-weighted overall reading because it came in under the specification number. It passes, it gets signed, and the complaint arrives three weeks later with nothing in the record that could have predicted it. Take the bands the first time. The measurement costs the same.

References

• ASHRAE Handbook, HVAC Applications, sound and vibration control chapter, in the edition the specification names, which owns the NC and RC curve values, the tangency rule, the RC deviation thresholds and the room design guidance; it binds contractually rather than as code.
• 29 CFR 1910.95 for the occupational noise floor including the action level and permissible exposure limit; 29 CFR 1910.23, 1910.133 and 1910.1001(j) for ladder, eye and presumed asbestos-containing material controls above a ceiling.
• ANSI S1.4 and IEC 61672 for sound level meter and octave band filter performance classes, in the edition the specification names, since a rating is only as good as the class of instrument that produced it.
• See related: What Background Noise Does That Is Actually Useful; What Speech Intelligibility Depends On; What Tonal Noise Is and Why It Is Judged Differently.