Why a Wall Never Performs Like Its Rating

Why this matters

A shop builds a demising wall to a detail stamped STC 50 and the tenant still hears the conversation next door. Nobody cheated. The submittal number was true about a specimen in a laboratory; the number the tenant is living with came out of a different test, of a different quantity, in a building that supplies every mechanism the laboratory was built to remove. That gap is not scatter and it is not a tolerance. It is a short list of named mechanisms, most of them paths and joints rather than materials, and most of them either measurable in the field or preventable at submittal. Treating it as a fudge factor is how a shop adds a fourth layer of board to a wall that was never the limiting element.

What the laboratory number is a number about

Sound transmission loss is measured in the laboratory to ASTM E90, in the edition the specifying document names, and it binds you only because a contract, a specification or a code section referenced it. Two reverberation rooms are built structurally separate from each other so that the only meaningful path between them is through the test opening. The specimen fills that opening, its perimeter is sealed into a test frame, and transmission loss is reported band by band, in one-third octave bands, unweighted, as a ratio of incident to transmitted sound power.

ASTM E413, again in the edition cited, then fits a reference contour to those bands and reports a single number, the sound transmission class. The contour spans 125 Hz to 4000 Hz, allows up to 8 dB of deficiency in any one band and 32 dB total, and returns the contour value at 500 Hz. Two consequences follow immediately: the rating is silent about anything below the 125 Hz band, and it can hide one deep dip inside an otherwise good assembly. Both belong to sibling articles rather than this one.

So the lab number is a statement about a sealed specimen, in a facility engineered so nothing goes around it, described by a contour that discards detail on purpose. Everything below is what that sentence excludes.

The field number is a different quantity, not a degraded one

Field work to ASTM E336, in the adopted edition, measures the space-averaged level difference between two rooms. That raw difference is noise reduction, and noise reduction is not transmission loss: it depends on how absorptive the receiving room is, because a dead room holds less of whatever arrives.

TL = NR + 10 log10 (S / A)

S is the area of the common partition and A is the total absorption of the receiving room, in the same units. This is a re-basing, not a penalty added to a rating. Read it in the direction that catches people: a heavily furnished, carpeted receiving room raises A, which makes the raw noise reduction look better than the partition is. Two suites with identical construction can report different noise reduction because one of them has furniture. A sibling card owns the fact that receiving-room absorption lowers the received level without changing what crosses the boundary; this is the same physics seen from the measurement side.

Normalize the field data and apply the E413 contour to it and you get an apparent sound transmission class. The word apparent is the whole point: it includes every path between the two rooms, not just the one through the wall.

What the laboratory left out, one item at a time

• Every path that does not go through the specimen. The facility is built to suppress them. Your building is built to carry them: slab, deck, corridor, plenum, ducts, continuous mullions. This is the dominant term in most disappointing field results and a sibling card owns it.
• The perimeter. The specimen is sealed into a frame. The field wall lands on a deflection track, a fluted deck, a floor with a tolerance, and a mullion pocket. A joint filled for a fire rating is filled to a fire test, not to an acoustic one.
• Anything that opens or passes through. Doors, transfer ducts, receptacles, access panels, pipe sleeves. The specimen had none of them unless the report says it did.
• Workmanship on the seal line. An unsealed edge is an aperture, and apertures are summed on transmission coefficient rather than on decibels, which is why a small one costs far more than it looks.
• Field mounting. Stud spacing, gauge, screw pattern, insulation type and depth, and how the board was fastened are all part of the tested assembly. A substitution that looks equivalent on a takeoff can move the assembly off its report.
• Bands below 125 Hz and single deep dips. Excluded by the contour, not by the physics.

Four of those six are paths or joints, one is construction, and one is an artifact of the rating contour. That is the finding: the gap is usually geometry, not material.

Worked example: turning one field reading into something comparable

A complaint between two offices. All levels are re 20 micropascals, A-weighted, overall, equivalent-continuous over 60 seconds, unless a band is named. Wear hearing protection whenever you are the one generating the test level in the source room, and station nobody there while it runs; 29 CFR 1910.95, Table G-16, sets the general-industry permissible exposure at 90 dBA over 8 hours measured A-weighted on slow response, with the hearing conservation program triggered at an 85 dBA 8-hour time-weighted average, and the construction counterpart is 29 CFR 1926.52. Do not prop, wedge or gasket a rated door assembly to take a reading; a rated opening protective has to self-close and latch, under NFPA 80 in the edition your authority having jurisdiction adopted, which binds the building owner and reaches you through the permit.

• Common partition: 12 ft by 10 ft, so S = 120 ft2.
• Source room, space-averaged: 84 dB.
• Receiving room, space-averaged: 46 dB. Noise reduction NR = 38 dB.
• Receiving-room background with the source off: 34 dB.
• Receiving-room total absorption, from a measured decay: A = 200 sabins (ft2).

Correction one, background. The signal sits 12 dB above background, so the correction is 10 log10 (10^{4.6 - 10}3.4) = 45.7 dB, a 0.3 dB shift. Between 3 and 10 dB of margin the correction is applied; below 3 dB the result stops being a value and becomes a one-sided bound written with a single inequality sign. At the 12 dB margin here the sibling procedure would decline the correction outright; it is computed and printed anyway because the number is small enough to carry. Carry 45.7 dB, so NR = 84 - 45.7 = 38.3 dB.

Correction two, the re-basing. TL = 38.3 + 10 log10 (120 / 200) = 38.3 - 2.2 = 36.1 dB. The absorptive receiving room was flattering the raw reading by 2.2 dB. Note what changed: NR is a level difference between two rooms, TL is a property of the partition. This is not a term added to NR, it is NR expressed on a different basis.

The comparison that is now legitimate. Lab STC 50 against a field figure near 36 on the E413 basis. Roughly 14 points, and none of it is measurement noise: the background correction moved 0.3 dB and the absorption normalization moved 2.2 dB, both in the direction that made the wall look worse once corrected.

Where it went. An unsealed head-of-wall run was found above the ceiling. Before going up: set the ladder to 29 CFR 1910 Subpart D in general industry or 1926 Subpart X in construction, leave thermal system insulation and surfacing material of unknown vintage undisturbed because it is presumed to contain asbestos until sampled under 29 CFR 1910.1001 in general industry and 29 CFR 1926.1101 for construction work, and do not cut, compress or add material to any fire-rated joint system, which voids its listing. This run was outside the rated joint. Sealing it and re-testing moved the field figure to 43. That single seal returned about 7 dB, which is more than a whole extra layer of board would have bought on an assembly already in the high 40s. The remaining gap did not respond to further work on the wall plane, which is the signature of a path that bypasses it.

Sibling-rule check, printed. Composite transmission summed on area-weighted transmission coefficient, per the small-hole card: yes, the seal was treated as an aperture, not as a decibel subtraction. Every level carries quantity, weighting, reference, bandwidth and time basis: yes. The remaining gap attributed to a path rather than to material, per the flanking card: yes, and it is stated as unresolved rather than as a quantified leak, because no differential test was run on it.

What would move this the other way

The lab number is close to the field number in exactly one situation: a small, simple, fully sealed partition between two rooms that share no structure, no plenum and no services, tested with a controlled source. Laboratory mock-ups of guest-room walls sometimes land within a couple of points. A masonry wall in a building with a topped slab and no plenum path can too. The moment there is a continuous ceiling grid, a shared deck, a duct crossing, or a door in the wall, the field figure belongs to whichever of those is weakest, not to the wall type on the drawing.

Outside North America the same logic runs on ISO 10140 for laboratory measurement and ISO 16283 for field measurement, with single numbers per ISO 717. The quantities are defined differently enough that an ISO weighted value and an ASTM class are not interchangeable, so do not convert one to the other on a submittal.

How to verify you got this right

Ask three questions of any isolation number before you quote it.

1. Which test, which edition, which mounting? If the answer is "the manufacturer says STC 50", you have a marketing figure until you see the report. A tested assembly's report names the mounting; a system built differently is a different assembly.
2. Is this a partition property or a room-pair result? Transmission loss and apparent class are partition-side. Noise reduction is room-pair and moves with furniture. Say which one your number is in the same sentence as the number.
3. Did anything bypass the wall? If you cannot answer, your field number is an upper bound on the wall's own performance and nothing more.

The failure mode to watch for is the fourth layer of board. When a wall comes in low and the response is more mass, and the retest returns a fraction of what the mass law predicted, the limiting element was never the wall plane. Stop adding and go find the path.

References

• ASTM E90 (laboratory airborne transmission loss), ASTM E336 (field measurement between rooms) and ASTM E413 (single-number classification), each in the edition the project specification or the adopted code cites; these are consensus standards and bind through that reference, not on their own
• 29 CFR 1910.95, Table G-16, occupational noise exposure in general industry; 29 CFR 1926.52 for construction work; 29 CFR 1910 Subpart D and 29 CFR 1926 Subpart X for ladders; 29 CFR 1910.1001 and 29 CFR 1926.1101 for presumed asbestos-containing materials
• NFPA 80, in the edition adopted by the authority having jurisdiction, for rated opening protectives
• See related: What Flanking Is and Why It Decides the Outcome; What a Small Hole Does to a Large Wall; What an STC Rating Covers and What It Misses; What Transmission Loss Actually Measures