Why Adding Absorption Does Not Quiet the Neighbour

Why this matters

This is an expensive category error and it is committed by competent people. A tenant complains about a machine two rooms away, somebody lines the machine room with absorptive panels, and the tenant hears exactly what they heard before. The panels were not defective and the install was not wrong. Absorption in the source room only ever reduces the reverberant level in that room, so the most it can do for the neighbour is bounded by the ratio of room constants, and it touches the airborne share alone. Whatever arrives through the structure never entered the room air, so no quantity of panel sees it. What follows is one job where that was worked out afterwards, with the arithmetic that would have taken twenty minutes beforehand.

The complaint and the first purchase

A plant room 30 ft by 24 ft by 12 ft under an occupied tenant suite. The tenant reported a steady drone, worst in the early evening when their own space went quiet. Measured in the tenant suite: 46.5 dB re 20 micropascals, A-weighted, overall, equivalent-continuous over 15 minutes, against a background of 34 dB in the same units with the plant shut down.

The first response was 500 ft2 of absorptive panel in the plant room, which took the room from an average absorption coefficient of about 0.08 to about 0.25 in the mid bands, both figures from published coefficients measured to ASTM C423 with the mounting per ASTM E795 in the editions the specification cited. The panels were specified with a surface burning characteristics class suitable for the occupancy, tested to ASTM E84 in the edition the adopted code references, and set clear of every sprinkler head to keep NFPA 13 coverage intact in the edition the authority having jurisdiction adopted, which binds the owner and reaches the contractor through the permit. A sibling card covers the rest of the install hazards.

After: 44.8 dB at the same tenant position, same time basis, same background conditions. The tenant reported no change, which is exactly what a 1.7 dB reduction in a steady broadband sound produces, since roughly 3 dB is the smallest change most people report.

The measurement that should have come first

A differential test in the tenant suite, run afterwards, split the received level into two paths. A loudspeaker in the plant room, normalized to the machine's source-room level band by band, reproduced the airborne share alone; subtracting that from the total on an energy basis gave the rest. The method belongs to a sibling how-to. Wear hearing protection while either the machine or the loudspeaker is running in the plant room and station nobody there for the duration; 29 CFR 1910.95 sets the general-industry permissible exposure at 90 dBA over 8 hours on slow response and triggers a hearing conservation program at an 85 dBA 8-hour time-weighted average.

• Airborne through the boundary: 43.0 dB.
• Structure-borne, through the slab and back out of the tenant's ceiling: 44.0 dB.
• Check: 10^{4.30 = 19,950 plus 10}4.40 = 25,120 gives 45,070, which is 46.5 dB. The two shares add back to the measured total.

The structure-borne path was the larger of the two before anyone bought anything. Absorption cannot see it. Energy leaves the machine through its feet, its piping and its conduit, travels in the slab, and radiates back into the tenant's room from their ceiling. It never passes through the plant room's air, so covering the plant room's surfaces is not on that path at any point.

Two shares landing a decibel apart is not a coincidence in this arrangement, and it is worth recognizing as a pattern. A hard-mounted machine with a fan discharging into the room it stands in feeds both paths from the same shaft, so unless something in the installation is unusual, neither share is normally an order of magnitude clear of the other. That is the worst case for a single-path fix: whichever path you close first, the other one is sitting a decibel behind it waiting to cap your result, which is precisely what the gate below reports.

And the complaint pattern was telling you this already. The tenant reported it worst in the early evening, when their own space went quiet. Nothing about the machine changes at that hour. What changes is the masking their own activity was providing, which raised their threshold for noticing a steady drone during the day. That is a background-level effect, a sibling card owns it, and it is worth reading correctly because a complaint that appears on a schedule tempts everyone to go looking for a schedule in the equipment.

The arithmetic, run in the order it should have been

Room constants. The plant room's surfaces total 1,440 + 720 + 576 = 2,736 ft2. At an average absorption coefficient of 0.08, A = 218.9 sabins and R = 218.9 / 0.92 = 237.9. At 0.25, A = 684 sabins and R = 684 / 0.75 = 912.

The bound on what absorption can do to the airborne share. 10 log10 (912 / 237.9) = 5.8 dB. That is the entire budget, and it applies only to the airborne path.

Qualifier, printed: the wall has to be in the reverberant field. The 5.8 dB is a reduction in the reverberant level. If the machine sits close enough to the common boundary that the boundary is driven by the direct field, that portion of the boundary gets nothing. Critical distance after treatment, with the machine floor-mounted so the directivity factor is 2: square root of (2 x 912 / 50.27) = 6.0 ft. The machine stands 9 ft from the boundary, past that distance, so the reverberant field does drive it and the full 5.8 dB applies here. At 4 ft it would not have.

Applied to the airborne share only. 43.0 - 5.8 = 37.2 dB. Structure-borne is unchanged at 44.0 dB.

New total. 10^3.72 = 5,248 plus 25,120 gives 30,368, which is 44.8 dB. That matches what the meter said after the panels went in, which is the confirmation that the model is describing this building and not a textbook.

Net: 1.7 dB, from a treatment that tripled the room's absorption.

Ceiling check, printed. Before the panels, the airborne share was 19,950 against everything else at 25,120. The flanking card's gate puts the ceiling on perfecting the airborne path at 10 log10 (1 + 19,950 / 25,120) = 2.5 dB. The 1.7 dB achieved sits under it, as it must. Twenty minutes with that gate would have said, before the purchase order, that no absorption treatment in that room could return more than two and a half decibels to the tenant.

Sibling-rule check, printed. Absorption applied to the reverberant term only, with the direct field and the structure-borne path untouched, per the absorption-versus-blocking card: yes. Paths combined on energy rather than by subtracting decibels: yes. Every level carries quantity, weighting, reference, bandwidth and time basis: yes. The result quoted against the gate that caps it: yes. Rounding taken away from the flattering direction throughout.

What actually closed it

The machine was retrofitted onto isolators selected against its disturbing frequency, which is a sibling card's subject and not re-derived here. Before any of that work: shut the machine down, isolate it at its disconnect, lock and tag it, relieve stored energy in any accumulator, spring or pressurized line, and support the machine's weight independently before a single mount is unbolted, per 29 CFR 1910.147, because a machine coming off its mounts is a crush hazard rather than an acoustics problem. Every rigid pipe and conduit connection had to be broken and re-made flexible in the same visit, or the isolators would have been short-circuited and the money wasted a second time.

That work dropped the structure-borne share from 44.0 dB to 34.0 dB. With the panels already in place:

• Airborne 37.2 dB and structure-borne 34.0 dB give 5,248 plus 2,510 = 7,758, which is 38.9 dB.
• The isolator work therefore returned 5.9 dB on top of what the panels had done, and the total from the original complaint is 7.6 dB, which is 46.5 minus 38.9.

The part worth keeping. Run the same two fixes in the other order and the endpoint is identical, because energy addition does not care about sequence, but what you learn is completely different. Isolators first, panels not yet installed: 19,950 plus 2,510 = 22,460, which is 43.5 dB, a gain of 3.0 dB. Then the panels take it to 38.9 dB, a further 4.6 dB. The same panels that looked worthless when they went in first are worth 4.6 dB when they go in second, and the two sequences sum to the same total: 1.7 plus 5.9 is 7.6, and 3.0 plus 4.6 is 7.6.

So absorption was never useless in this building. It was masked. A path that dominates hides every improvement made to the paths under it, which is the whole reason the gate is worth running before the purchase rather than after.

What changes the answer

A source room where the boundary is in the direct field. Machine close to the wall, wall gets the direct field, absorption returns little regardless of the room constant. Compute the critical distance and compare it to the machine's setback before promising anything.

A source room that is already absorptive. The bound is a ratio, so a room at 0.25 going to 0.40 returns 10 log10 ((0.40/0.60) / (0.25/0.75)) = 3.0 dB, half of what the first treatment gave. Each further doubling of total absorption costs more surface than the last and returns the same 3.0 dB.

A purely airborne complaint. Where a differential test shows the structure-borne share is well below the airborne share, source-room absorption returns close to its full bound and is a legitimate, cheap fix. It is not that absorption never helps a neighbour. It is that it is capped by a ratio and blind to a path, and both have to be checked.

Absorption in the receiving room. It lowers the level the tenant hears without changing anything that crossed the boundary, and in a live-sounding tenant space it is sometimes the fastest partial relief available. Say what it is when you sell it: comfort in their room, not less sound arriving.

References

• ASTM C423 and ASTM E795 for absorption coefficients and mounting, and ASTM E84 for surface burning characteristics of interior finish, in the editions the specification or adopted code cites; consensus standards bind through that reference rather than on their own
• NFPA 13, in the edition adopted by the authority having jurisdiction, for sprinkler coverage and obstruction
• 29 CFR 1910.147 for control of hazardous energy; 29 CFR 1910.95 for occupational noise exposure in general industry
• See related: What Absorption Does That Blocking Cannot; What Flanking Is and Why It Decides the Outcome; How to Find a Flanking Path Without Opening the Wall; Why a Rigid Connection Defeats an Isolator