What an NRC Rating Actually Tells You

Why this matters

Noise reduction coefficient is the number every absorptive product leads with and the number most likely to be quoted back at you by a customer who read a spec sheet. It is a real measured quantity and it is genuinely useful for one narrow comparison. It is also an average of four bands, rounded, taken under a mounting condition that is part of the result, in a test that says nothing whatever about blocking. Two products that both read 0.75 can perform four decibels apart in the band a customer is actually complaining about. This card is mostly about what the number leaves out, because that is where the jobs go wrong.

What the number is, exactly

Absorption coefficients are measured to ASTM C423, in the edition the specification cites, by putting a specimen of known area in a reverberation room and reading how much faster the room decays with the specimen in it than without. The result is reported per one-third-octave band as a coefficient: the fraction of incident energy the surface does not return.

Noise reduction coefficient is then the arithmetic average of the coefficients at the 250, 500, 1000 and 2000 Hz octave bands, rounded to the nearest 0.05. That is the whole definition. Four bands, straight average, coarse rounding.

The same standard also reports the sound absorption average, which is the mean of the twelve one-third-octave bands from 200 Hz to 2500 Hz rounded to the nearest 0.01. It is a better number: finer resolution and finer rounding. It is still centred on the same part of the spectrum.

What it leaves out on purpose

• The 125 Hz band and the 4000 Hz band. Not averaged, not implied, not represented.
• The shape inside the four bands it does average. A coefficient set that rises steeply and one that is nearly flat return the same average.
• Precision. Rounding to 0.05 means a labelled 0.75 covers a true four-band average anywhere from 0.725 to 0.775. Two products carrying the same label can differ by that much before either label moves, and that is before any of the omissions above.
• Anything about transmission. Absorption and blocking are different quantities acting in different places, which a sibling card owns. A product can absorb 95 percent of what hits it and stop almost nothing from passing through, and many good absorbers are exactly that.
• The room it will actually live in. The test room is engineered to hold a diffuse field. A panel on one wall of a rectangular room with five hard surfaces is concentrated absorption in a field that is not diffuse, and concentrated absorption returns less than the arithmetic predicts.

The omission that bites: the missing low band

Almost every complaint that gets described as a rumble, a hum, a bass line or a drone lives at or below the 125 Hz octave band. NRC does not contain that band. So the number a shop uses to compare products is silent about the frequency range where the hard problems are, and it is silent in a way that systematically favours thin material, because thin porous absorbers do well in the bands NRC averages and poorly below them.

Porous absorption works by air moving through the material and losing energy to friction. Air particle velocity is zero at a rigid surface and reaches its maximum a quarter wavelength away from it. At 125 Hz in air a quarter wavelength is on the order of two feet. A one-inch panel tight against a hard wall sits almost entirely in the low-velocity region at that frequency and absorbs very little of it, no matter what its NRC says. The same panel with a deep airspace behind it is a materially different product at 125 Hz and the same product in the bands NRC averages.

Mounting is part of the number

ASTM C423 results are reported with the mounting used, standardized in ASTM E795 in the cited edition. Type A is the specimen laid directly against a rigid surface; other types specify furring or a stated airspace behind the specimen. The same material on the same day, tested in two mountings, returns two different coefficient sets and can return two different NRC values.

That makes the mounting a specification item, not a footnote. A product's published data sheet showing an impressive low-frequency figure and a data sheet showing a poor one may both be true and may both be that product, tested two ways. If the submittal does not name the mounting, the number is not yet a number.

Coefficients above 1.00, and what that tells you

C423 normalizes the measured absorption by the specimen's geometric area, and a finite specimen has edges that diffract sound into it from outside its footprint. So reported coefficients routinely exceed 1.00. Nothing absorbs more than all of the energy that hits it; the number simply is not literally a fraction for a finite specimen.

The practical consequence matters more than the curiosity. A panel with four exposed edges in the field behaves more like the tested specimen than the same panel butted into a continuous run does, so a wall covered edge to edge tends to deliver slightly less per square foot than a scattered layout. Do not size a job at the published coefficient and expect to find the last few percent.

Worked example: two products, same NRC, 4 dB apart

A conference room 20 ft by 14 ft by 9 ft. Surfaces total 560 + 360 + 252 = 1,172 ft2. The complaint is a low rumble from a rooftop unit, which lands in the 125 Hz octave band, so that is the band the arithmetic runs on. The existing wall and ceiling surfaces read 0.08 at 125 Hz. Plan: 400 ft2 of panel.

Say the two candidate data sheets read as follows, both mounted Type A, and note both are illustrative sets used to make the arithmetic legible rather than quoted from a product.

NRC, computed. Product A: (0.45 + 0.80 + 0.90 + 0.85) / 4 = 0.750, so NRC 0.75. Product B: (0.75 + 0.80 + 0.75 + 0.72) / 4 = 0.755, which rounds to NRC 0.75. Identical labels.

Correction, printed: added sabins are the difference, not the product coefficient. Panel replaces existing surface, so added absorption is 400 x (panel coefficient minus 0.08), and total surface area stays 1,172 ft2.

Baseline at 125 Hz. A1 = 1,172 x 0.08 = 93.8 sabins. Room constant R1 = 93.8 / 0.92 = 101.9.

Product A. Added 400 x (0.15 - 0.08) = 28 sabins. A2 = 121.8, average coefficient 0.104, R2 = 121.8 / 0.896 = 135.9. Reverberant-field gain 10 log10 (135.9 / 101.9) = 1.3 dB.

Product B. Added 400 x (0.55 - 0.08) = 188 sabins. A2 = 281.8, average coefficient 0.240, R2 = 281.8 / 0.760 = 370.9. Reverberant-field gain 10 log10 (370.9 / 101.9) = 5.6 dB.

The finding. Same NRC on both data sheets. 4.3 dB apart at the band the customer is complaining about, which is the difference between a change nobody reports and a change everybody does, since roughly 3 dB is the smallest change most people notice in a steady broadband sound.

Bound, printed. Product B's post-treatment average coefficient of 0.240 is past the point where a diffuse field can be assumed comfortably, and 400 ft2 of panel on one or two surfaces of a 1,172 ft2 room is concentrated rather than distributed absorption. Both push the real result below the model, so 5.6 dB is an upper estimate, written as a bound with one inequality rather than as a value. Product A's 1.3 dB is inside the model's comfortable range and stands as a value.

Sibling-rule check, printed. Absorption applied only to the reverberant term, per the absorption-versus-blocking card: yes, and no claim is made about anything crossing a boundary. Added sabins computed as the difference between panel and covered surface: yes. Coefficients carried per octave band with the band named, rather than as a single average: yes. The result reported as a bound where the model's validity condition was strained, with one inequality sign: yes. No rounding taken in the flattering direction.

What to ask for instead

Ask for the full octave-band coefficient set with the mounting named, and the report it came from. Then run the arithmetic on the band your complaint actually lives in. NRC's legitimate use is a fast first sort between products for a broadband speech-range problem in an ordinary room, and it is fine at that. It is not a design number and it has never been one.

Ask for the surface burning characteristics. Whatever you specify goes on a wall or ceiling as interior finish, and it has to meet the flame spread and smoke developed class the adopted building or fire code requires for that occupancy and location, tested to ASTM E84 in the cited edition. Do not substitute an untested material for a tested one because their NRC values match; an open-cell foam with a good absorption number and no test report is a fire load with no rating at all.

Ask what happens below 125 Hz if the complaint is low. Porous material with an airspace, a thicker assembly, or a tuned device are different answers, and the honest one is often that the fix belongs at the source rather than in the room.

References

• ASTM C423 (sound absorption by the reverberation room method) and ASTM E795 (mounting of test specimens), in the editions the specification cites; consensus standards bind through that reference rather than on their own
• ASTM E84, surface burning characteristics of building materials, in the edition the adopted building or fire code references
• See related: What Absorption Does That Blocking Cannot; What Reverberation Time Controls and What It Does Not; Why Low Frequency Noise Is the Hard Problem; Why a Single Number Hides the Fault