Why a Double Wall Beats a Heavy Wall and When It Does Not

Why this matters

A double-leaf partition can outperform a single wall several times its weight, which is why almost every high-performing partition built today is one. It can also perform worse than a single wall of the same total mass, in one specific octave band, and if that band is the one your complaint sits in you have built the wrong wall for more money. The behaviour is not subtle and it is not random. A double leaf is a resonator, it has a resonant frequency you can compute before anything is framed, and it is worse than a single wall at and below that frequency. This is a story about a wall that was built correctly to a specification and was still the wrong wall.

The mechanism: two masses and a spring

Two leaves with air between them are a mass, a spring and a second mass. The air in the cavity is the spring. Like any mass-spring-mass system it has a resonance, and its behaviour splits into three regions around it.

• Below the resonance, the spring is stiff relative to the motion and the two leaves move essentially together, so the assembly performs roughly like a single leaf of the combined surface mass. You paid for a cavity and got nothing for it.
• At the resonance, the system is driven at its natural frequency and transmission loss dips, and the dip can put the double wall BELOW a single wall of the same total mass.
• Above the resonance, the leaves decouple and transmission loss rises far more steeply than the 6 dB per octave a single limp panel gives. This is the entire reason double walls exist.

The mass-air-mass resonance in Hz is approximately 60 multiplied by the square root of the sum of the two surface masses divided by the product of the two surface masses and the cavity depth, with surface masses in kg/m2 and cavity depth in metres. That constant of 60 is for an empty cavity. With the cavity filled with absorption the compression becomes closer to isothermal and the constant is commonly taken nearer 43, which puts the resonance lower, which is one more reason to fill the cavity.

Read the relationship for its directions, which are what you use on site. Adding mass to both leaves lowers the resonance. Deepening the cavity lowers the resonance. Adding mass to only one leaf lowers it much less than adding to both, because the product term is dominated by the lighter leaf. And nothing in the relationship mentions the studs, which is the omission the second half of this article is about.

The complaint

A shop finished a new office alongside its machine room, built to a partition specification that had worked well for them in another building: one layer of 5/8 in gypsum board each side, on a single row of steel studs with a 3-5/8 in cavity, batt fill, sealed at the perimeter. The office was unusable during production. The complaint was described as a rumble, and the shop's own comparison point was an older office behind a single heavy masonry wall, which was quieter in the same conditions.

Band level differences between the machine room and the office, unweighted, Leq over 5 minutes, measured at fixed positions in each room. These are level differences between two rooms, not the wall's transmission loss, and the two are different quantities, which the transmission-loss card covers.

What was ruled out

A leak. Opening the office door changed the level difference above 500 Hz by several decibels and changed the 63 Hz band by under a decibel. A gap is a high-frequency-dominant loss, so a gap large enough to cause a 63 Hz problem would have wrecked the top of the range as well, and it had not.

Broad flanking. A flanking path around a partition generally drags the whole curve down together. This curve is not uniformly low: it is strong at 2 kHz and collapses in one band. That shape points at the wall behaving as a wall, badly, rather than at sound getting around it.

Insufficient mass. This was the shop's own first theory and it is the one the numbers kill. Look at the 31.5 Hz row: 22 dB, which is 8 dB BETTER than the 63 Hz row underneath it. A mass-controlled panel's performance rises with frequency and never dips. A local minimum in the curve is a resonance, and the only resonance a sealed cavity wall has in that region is its mass-air-mass resonance.

The number that named it

Surface mass of 5/8 in gypsum board is about 2.2 lb/sq ft, which is 10.7 kg/m2, and the cavity is 3-5/8 in, which is 0.092 m. With the cavity filled, the resonance is 43 multiplied by the square root of 21.4 divided by the product of 10.7, 10.7 and 0.092, which is 43 multiplied by 1.43, or about 61 Hz. Run the same arithmetic with the empty-cavity constant of 60 and it lands at about 86 Hz.

The 63 Hz octave band spans roughly 45 to 89 Hz, so both figures fall inside it. Whichever constant you argue for, that wall's resonance is in the complaint band. It was designed to have its weakest point exactly where this machine room is loudest.

The second bridge, which was on no drawing

  machine room side cavity office side

   leaf one leaf two
   |                                            |
   |             absorptive fill                |
   |                                            |
   |========= rigid stud, a bridge =============|
   |                                            |
   ============= shared floor track =============
                 also a bridge

Both boards were screwed to the same row of studs, and both leaves landed in a common floor track. A rigid connection between the leaves short-circuits the air spring: above the resonance, where the double wall was supposed to be earning its steep slope, energy takes the stud instead of the cavity.

The evidence for it is in the same table. From 125 Hz to 2 kHz the level difference rose from 28 to 47 dB, which is 19 dB over four octaves, or about 4.75 dB per octave. A decoupled double leaf should climb far more steeply than a single leaf's 6 dB per octave in that region. This wall was climbing more slowly than a single leaf. That is a bridge, not a material shortfall.

The rebuild and what it bought

The wall was rebuilt with a 6 in cavity, two layers of board each side, framing separated so no member touched both leaves, cavity filled, and the perimeter and floor junction detailed so the leaves are structurally independent.

Recompute the resonance with the same relationship and the filled-cavity constant: two leaves at 21.4 kg/m2 each and a cavity of 0.152 m give 43 multiplied by the square root of 42.8 divided by the product of 21.4, 21.4 and 0.152, which is about 34 Hz. That is now below the complaint band, in the 31.5 Hz octave.

The 63 Hz band gained 13 dB and the office became usable. From 63 to 250 Hz the rebuilt curve climbs 23 dB over two octaves, about 11.5 dB per octave, which is the steep decoupled slope the first wall never produced. And the 31.5 Hz band got 2 dB WORSE, which is not a defect in the work: the resonance moved down into that band, exactly as the arithmetic said it would. The weak point was not eliminated, it was relocated below the frequencies this machine room produces.

The demolition and rebuild carried hazards that had nothing to do with acoustics. Before cutting into either face, the circuits serving anything concealed in that wall were locked out and tagged and verified dead, which is 29 CFR 1926.417 for lockout and tagging of circuits in construction work, with 29 CFR 1910.333(b)(2) as the general-industry counterpart if your shop falls under that Part instead. Sanding and cutting joint compound and board produces respirable dust that can contain crystalline silica, so the work was done with tool-mounted dust collection and respiratory protection under a program meeting 29 CFR 1910.134, and 29 CFR 1926.1153 is the construction respirable crystalline silica standard that governs the control methods. In a building whose age was not documented, the existing joint compound was treated as presumed asbestos-containing and sampled before demolition rather than after, under 29 CFR 1926.1101 for construction work including renovation.

Sibling-rule check against the cards this one sits beside. Every level difference is named as a level difference between two rooms rather than as a transmission loss, and no room-side correction is applied to one figure and not another, since all readings are at the same fixed positions. The single-leaf 6 dB per octave figure is quoted at constant surface mass, as its owning card states it. The leak test uses the high-frequency-dominant behaviour of gaps that the transmission-loss card establishes. The resonance constants of 60 and 43 are both stated in the general section before the example uses them, and the example uses only the filled-cavity value because the cavity is filled in both builds.

What would change the recommendation

A mid-frequency complaint. Had the dominant content been at 500 Hz rather than 63 Hz, the original wall's resonance would have been two and a half octaves below the problem and entirely irrelevant. The finding would have been the bridging alone, and separating the framing without deepening the cavity would have fixed it.

No room to deepen the cavity. Where the cavity depth is fixed by the building, the resonance can only be moved by adding mass to both leaves, and it moves as the square root: doubling the surface mass of both leaves divides the resonance by about 1.4, not by 2. Getting a full octave requires roughly four times the mass on both faces, which usually runs into a structural limit before an acoustic one.

A lightweight floor structure. Both walls above were built on a slab. On a timber or lightweight framed floor, structure-borne flanking under the partition can dominate below 125 Hz regardless of what the wall itself does, and the level differences would then be reporting the floor, not the partition. The tell is that the curve fails to improve when the wall is improved.

References

• ASTM E90, laboratory measurement of airborne sound transmission loss, in the edition stated on the report for any partition specification you are asked to match; ISO 10140 is the international counterpart
• 29 CFR 1926.417, lockout and tagging of circuits for construction work, with 29 CFR 1910.333(b)(2) as the general-industry counterpart
• 29 CFR 1926.1153, respirable crystalline silica in construction, and 29 CFR 1910.134, respiratory protection
• 29 CFR 1926.1101, asbestos in construction work including renovation, for presumed asbestos-containing joint compound
• See related: What the Mass Law Predicts and Where It Fails; What Transmission Loss Actually Measures; Why Low-Frequency Noise Is the Hard Problem