How to Decide Whether a Room Needs Absorption or Isolation

Why this matters

Absorption and isolation are two different jobs and the wrong one gets bought constantly, because both get sold under the word "soundproofing." Hanging panels in a room to stop the neighbour hearing through the wall does almost nothing for the neighbour, and the arithmetic below puts a number on "almost nothing": in the same room, on the same complaint, absorption buys the neighbour 1.6 dB while closing one gap buys around 10 dB per factor-of-ten reduction in the open area. That is not a small difference in effectiveness, it is a different category of work. The decision is settled by one question about paths, not by how loud anything is.

Before you take a reading or get above a ceiling

Set a portable ladder on a level surface at the angle and extension 29 CFR 1910.23 requires, and wear eye protection to 29 CFR 1910.133 before pushing a ceiling tile up, because the first thing out of a plenum is grit.

Do not disturb unidentified material above the ceiling or on structure. In a building of unknown vintage, thermal system insulation and surfacing material are presumed asbestos-containing under 29 CFR 1910.1001(j), which makes this an inhalation route that a glove and a pair of glasses do nothing about: leave it alone, have it sampled, and do not drill, cut or brush it.

Where a reading has to be taken beside running equipment, keep every guard in place as 29 CFR 1910.212 requires, and wear hearing protection whenever the space is at or above the 85 dB A-weighted eight-hour time-weighted average action level in 29 CFR 1910.95(c), which is the federal occupational floor and has nothing to say about whether an office is pleasant.

The two jobs, stated so they cannot be confused

Absorption removes energy from air already inside a volume. It converts acoustic energy to heat inside a porous or resonant material. It changes how long sound persists in a room and how much reverberant level builds up for a given source, and it does that only for listeners breathing the same air as the source.

Isolation stops energy crossing a boundary. It is a mass, stiffness, damping and airtightness problem in an assembly, plus every path that goes around that assembly. It changes what the neighbour gets.

A material can be excellent at one and useless at the other. A 2 inch glass fibre panel absorbs most of what hits it above about 500 Hz and has essentially no transmission loss, because it has almost no mass. A layer of dense board has real transmission loss and reflects nearly everything back into the room.

The gate: does the complainant share air with the source

Ask the complainant where they are standing when it bothers them, then ask what happens when the door between them and the source is closed.

• Complaint drops sharply with the door shut and the person is in a different volume: this is a transmission problem. Absorption in the source room is not the lever. Work the boundary and the paths around it.
• Complaint barely changes, or the complainant is in the same volume as the source: the reverberant field in that volume is the lever, and absorption is the work.
• Complaint changes with the door shut but not much, and the two rooms share a ceiling plenum, a duct, a floor slab or a common stud: you have a flanking path, which is still a transmission problem. The gap between what the wall is rated for and what the neighbour hears is usually a path, not a material.

What a level has to carry before it settles anything

Any number either side quotes has to name five things or it is not a measurement, and two people comparing numbers that carry fewer than five are not comparing anything:

• Quantity. Sound pressure level (referenced to 20 micropascals) or sound power level (referenced to 1 picowatt). Only pressure belongs to a place. Power belongs to the machine.
• Weighting. A, C, or unweighted (Z). A-weighting discards low frequency, which is exactly where a rumble complaint lives.
• Reference. 20 micropascals for pressure, 1 picowatt for power. Say it.
• Bandwidth. Overall, or per octave or one-third-octave band.
• Time basis. Fast, slow, equivalent-continuous over a stated duration, or peak.

So: "48 dB, A-weighted sound pressure level re 20 micropascals, overall, equivalent-continuous over 15 minutes." Not "48 decibels."

The room case, worked

A conference room, 20 ft by 14 ft by 9 ft high, 2,520 cubic feet. Surfaces: 280 sq ft ceiling, 280 sq ft floor, 612 sq ft of wall. People say meetings are exhausting and the far end of the table is hard to follow, and the complaint does not involve anyone outside the room. Same-volume complaint, so absorption is the lever.

Reverberation time by the Sabine relationship, which holds for a reasonably diffuse field at modest average absorption, holding room volume constant: RT = 0.049 x volume in cubic feet, divided by total absorption in sabins (square feet of absorption).

Current absorption at the 500 Hz octave band, using published coefficients from each product's own ASTM C423 report rather than memory:

• Painted gypsum ceiling, 280 sq ft at a coefficient in the published 0.03 to 0.10 range, taken at 0.05: 14 sabins
• Painted gypsum walls, 612 sq ft at 0.05: 30.6 sabins
• Carpet with pad on slab, 280 sq ft at a coefficient in the published 0.25 to 0.55 range, taken conservatively at 0.35: 98 sabins

Total 142.6 sabins. RT = 0.049 x 2,520 / 142.6 = 123.5 / 142.6 = 0.87 seconds at 500 Hz, empty.

Target 0.60 seconds. Required absorption = 123.5 / 0.60 = 205.8 sabins, so 63.2 sabins short.

Correction printed, mounting. A published NRC is the four-band average of the 250, 500, 1000 and 2000 Hz coefficients from ASTM C423 rounded to the nearest 0.05, and it already contains the mounting used in the test. A 2 inch panel published at NRC 0.95 on type A mounting (laid directly on the test surface) is being installed directly on the wall, so the mounting matches and this is not re-based. Furring the same panel out on standoffs would be a different mounting and a different report, not the same number plus a bonus.

Correction printed, net absorption. Absorption added is not the panel's coefficient, it is the panel's coefficient minus the coefficient of the surface it covers, because that surface stops absorbing on its own account: 0.95 - 0.05 = 0.90 net per square foot.

Area needed: 63.2 / 0.90 = 70.2 sq ft. Round up, never down, because under-treating is the failure that puts you back in the room: eight panels at 3 ft by 3 ft, 72 sq ft.

New absorption: 142.6 + (72 x 0.90) = 142.6 + 64.8 = 207.4 sabins. RT = 123.5 / 207.4 = 0.60 seconds.

Correction printed, model validity. Sabine degrades once average absorption gets high. Average coefficient after treatment = 207.4 / 1,172 sq ft of surface = 0.177, below about 0.20, so the model still applies. Past that, a different formulation is needed and the answer stops being a one-line calculation.

Correction printed, band. Every figure above is the 500 Hz octave band. It says nothing about 125 Hz, where a 2 inch porous panel absorbs a fraction of what it does at 1000 Hz. If the complaint had been a low rumble rather than speech clarity, this arithmetic would have been the wrong arithmetic.

The boundary case, same room, opposite answer

Same room, different complaint: the office next door hears conference calls through a demountable partition. Different volume, complaint drops when the door is shut, so this is transmission.

The partition is 10 ft by 9 ft, 90 sq ft, laboratory-rated STC 45 under ASTM E90 and classified under ASTM E413. That rating contains a laboratory mounting with no flanking, so it is used below only to size the gap's dominance, never as a prediction of the field result; a real field composite needs an apparent transmission loss measured under ASTM E336.

There is a 0.5 in by 96 in gap where the partition head meets a deflection track: 48 square inches, 0.333 sq ft.

Composite transmission loss combines the paths by area-weighted transmitted energy, not by averaging ratings. Treating the gap as no transmission loss at all:

• Panel term: 89.67 sq ft x 10 raised to (-45/10) = 0.00284
• Gap term: 0.333 sq ft x 1 = 0.333
• Sum 0.336, divided by 90 sq ft = 0.00373
• Composite transmission loss = -10 x log10(0.00373) = 24.3 dB

An STC 45 partition is delivering about 24 dB. The gap is 0.37 percent of the area and it owns 99 percent of the transmitted energy.

Now price absorption against it. Reverberant sound pressure level in the source room falls by 10 x log10(new absorption / old absorption), holding the source's sound power constant and with the receiver in the reverberant field rather than close to the talker. The eight panels give 10 x log10(207.4 / 142.6) = 10 x log10(1.455) = 1.6 dB.

So on the neighbour's complaint: 72 sq ft of panels, 1.6 dB. Reducing the open gap area by a factor of ten moves the leak-limited ceiling by 10 dB, because the composite ceiling goes as -10 x log10 of the open area fraction. Closing it entirely lifts the ceiling to whatever the flanking paths allow, which is a measured number under ASTM E336 in the field, not the lab STC.

Sibling-rule check, run against both cases above. Every level in this article carries quantity, weighting, reference, bandwidth and time basis. Every published coefficient carries its test method and mounting. The lab STC 45 is stated as containing its own no-flanking mounting and is not used as a field prediction. The absorption arithmetic is stated at one octave band and is not carried to another. The 1.6 dB reverberant reduction carries its held-constant terms (constant source sound power, receiver in the reverberant field). No rounding runs in the direction that makes the recommendation look better: the panel area rounded up, the carpet coefficient was taken at the low end of its published range, and the gap was treated as fully open.

Verifying you called it right

For an absorption call, the check is a repeat reverberation measurement to ISO 3382-2 at the same octave bands, or, without an analyzer, the simple field check: two people at opposite ends of the table, normal voice, before and after. If the room still rings, you targeted the wrong bands.

For an isolation call, the check is a field measurement to ASTM E336 between the two rooms, before and after, at the same source position. If the number barely moves after you seal a gap you were sure was the path, the energy is going somewhere else, and the plenum above the partition is where to look first.

The failure mode to catch early: a client who has already bought panels for a transmission complaint. Hang them and the complaint stands, you have spent the budget, and the credibility to ask for the real work is gone. Say which category the job is before anything is ordered.

References

• 29 CFR 1910.95 for the occupational noise floor including the 85 dB A-weighted eight-hour time-weighted average action level; 29 CFR 1910.212 for guarding, 1910.23 for ladders, 1910.133 for eye protection, 1910.1001(j) for presumed asbestos-containing material.
• ASTM C423 for absorption coefficients and NRC, including the mounting, and ASTM E795 for the mounting designations; the product's own report in the edition tested owns the number.
• ASTM E90 and ASTM E413 for laboratory airborne transmission loss and STC; ASTM E336 for field airborne measurement between rooms; ISO 3382-2 for reverberation time.
• See related: What Sound Actually Is, and Why That Decides Everything; Why Sound Power and Sound Pressure Are Not the Same Number; How to Add and Subtract Noise Levels Correctly.