What Flanking Is, and Why It Decides the Outcome

Why this matters

Two buildings, the same complaint, the same partition upgrade, the same crew. One tenant reports the problem solved and one reports no change at all. The difference is not workmanship and it is not the product. It is that in one building the wall was carrying most of the energy and in the other it was not. Flanking is every path between two spaces that does not go through the element you are being paid to improve, and because paths combine on energy rather than on decibels, the quietest possible partition cannot get you below whatever the flanking paths are already delivering. This card gives you the one piece of arithmetic that tells you, before you spend an hour, the most that any partition work can possibly return.

Paths add as energy, and that changes everything

Transmission loss is a ratio expressed logarithmically, so it does not sum. What sums is the transmission coefficient, the fraction of incident sound power that gets through. Convert each path to its coefficient, add the coefficients weighted by the area they act over, and convert back:

T_total = sum over paths of (area x transmission coefficient)
apparent TL = 10 log10 (total area / T_total)

Everything that follows is a consequence of that one line. Two equal paths give 3 dB worse than either alone. A path ten times weaker adds 0.4 dB and is irrelevant. A path ten times stronger owns the result and the other one might as well not exist.

Where the paths actually run

  SOURCE ROOM        |        RECEIVING ROOM
                     |
  plenum over ceiling|  path 1: over the wall
  ...................|.........................
  --------------------------------------------
     the wall itself (the element you were sold)
  --------------------------------------------
  ...................|.........................
  floor slab and deck|  path 2: through structure
                     |
  duct ==============|========  path 3: in the duct
  sleeve ------------|--------  path 4: at the sleeve

Path 1 is the one that catches most shops: a demising wall stopped at the ceiling line with a continuous plenum above it. Path 2 is structure-borne, and it does not care about air seals at all. Paths 3 and 4 are services, and a sibling card covers why an element that must open or pass something through is usually the weakest of the set.

Two properties make flanking hard to reason about without measuring. It attenuates far less per unit distance in structure than in air, so the receiving room is not always adjacent to the source. And it is frequency-selective: a plenum path can be the whole story at 250 Hz and irrelevant at 2000 Hz, which is why band data beats a single number here.

Path 1 is the one path in that sketch that has its own laboratory rating, and knowing it exists saves arguments. Ceiling attenuation class is measured to ASTM E1414 in the edition the specification cites, and it describes sound going up through a suspended ceiling, across a plenum, and back down on the other side of a partition that stops at the ceiling line. It is a rating of that route only. It is not a substitute for a wall rating, it does not transfer to a wall run deck to deck, and it is void the moment a light fixture, a return-air boot or an unsealed grille breaks the ceiling membrane. When a submittal answers a demising-wall question with a ceiling number, somebody has quoted a rating for one path as if it covered all of them.

The gate: how much any single-path fix can possibly return

State it once and use it everywhere. For a named receiving position, on a named octave band or on a named single-number rating, the ceiling on perfecting one path is:

ceiling in dB = 10 log10 (1 + T_path / T_everything_else)

You cannot reach the ceiling, only approach it, because perfecting a path means driving its coefficient to zero.

Read the first row and understand why partition upgrades disappoint. If the flanking paths already transmit ten times what the wall does, the entire budget for the wall buys four tenths of a decibel. For reference throughout this group: a change of about 3 dB is the smallest most people report as a change in a steady broadband sound, and about 10 dB reads as roughly half as loud. A pure tone is noticed at smaller changes, which a sibling card owns.

One gate, two buildings

Same complaint, same proposed fix: replace a wall whose field performance is 50 dB with one that performs at 60 dB, a 10 dB improvement in the element. All figures are apparent transmission loss on the same partition area, measured to ASTM E336 and classified per ASTM E413 in the editions the specification cites, and levels are re 20 micropascals, A-weighted, overall, equivalent-continuous over 60 seconds.

Building A. A differential test puts the dominant flanking path, a continuous plenum over the wall, at 45 dB.

• Coefficients: wall 10^{-5.0 = 1.00e-5, flanking 10}-4.5 = 3.16e-5.
• Total 4.16e-5, so apparent TL today = 43.8 dB.
• Ceiling check, printed: 10 log10 (1 + 1.00e-5 / 3.16e-5) = 10 log10 (1.32) = 1.2 dB. That is the most the wall work can ever return here, and it was knowable before anyone priced it.
• After the upgrade: 1.00e-6 + 3.16e-5 = 3.26e-5, apparent TL = 44.9 dB. Gained 1.1 dB, which is under the 1.2 dB ceiling as it must be, and below what anyone would report as a change.

Building B. Same wall, but a topped slab, no plenum path, and the services offset. The strongest remaining path measures 58 dB.

• Coefficients: wall 1.00e-5, everything else 10^-5.8 = 1.58e-6.
• Total 1.16e-5, so apparent TL today = 49.4 dB.
• Ceiling check, printed: 10 log10 (1 + 1.00e-5 / 1.58e-6) = 10 log10 (7.31) = 8.6 dB.
• After the upgrade: 1.00e-6 + 1.58e-6 = 2.58e-6, apparent TL = 55.9 dB. Gained 6.5 dB, under the 8.6 dB ceiling, and comfortably past the 3 dB most people notice.

The same product, installed the same way, returned 1.1 dB and 6.5 dB. Nothing about the wall explains the difference. The measure that separated the two jobs was the strength of the strongest path that is not the wall, and it took one differential test in each building to get it.

Sibling-rule check, printed. Paths summed on area-weighted transmission coefficient rather than by subtracting decibels, per the small-hole card: yes. Every level and every rating carries its quantity, weighting, reference, bandwidth and time basis: yes. Field figures called apparent transmission loss rather than compared directly to a laboratory class, per the lab-versus-field card: yes. No rounding taken in the flattering direction: the gains are quoted below their own ceilings.

What this does not tell you

The gate tells you how much room there is. It does not tell you which path is which, and it will not tell you whether the flanking is airborne or structure-borne. Those are different fixes and one of them cannot be sealed at all: adding mass, sealant and gaskets to a structure-borne path returns nothing, because the energy is travelling in the slab and radiating on the far side.

The gate is also band-specific. Running it on a single-number rating is legitimate for a first pass, but a plenum path that dominates at 250 Hz and vanishes at 2000 Hz will make the single number understate the fix at high frequency and overstate it at low. When the complaint is a hum or a bass line, run the gate on the band that carries the complaint.

How to catch a flanking problem before it costs you

• Before quoting a partition upgrade, ask what is continuous between the two spaces. Ceiling grid, plenum, floor deck, topping, mullion, shaft, duct, conduit, pipe. If anything is continuous, the wall type is not the answer on its own.
• Get one differential measurement. Change one path, hold everything else constant, and read the difference. The procedure is a sibling how-to; the point here is that the gate is worthless without a number for the other paths.
• Refuse the retest trap. If a wall upgrade returns much less than the mass-law prediction for the added mass, that is not a bad product. It is the gate. Stop adding layers and go find the path.
• Do not close a flanking path with something that breaks another system. A barrier laid in a ceiling plenum can obstruct sprinkler coverage, which is governed by NFPA 13 in the edition your authority having jurisdiction adopted and which binds the building owner through the permit, and it can block a return-air path the mechanical system depends on. Confirm both before you install anything above a ceiling, and restore any rated assembly you disturb with its own listed system rather than with acoustic sealant.
• Say what you are selling. Quote the wall as a wall, and say in writing that the outcome at the receiver depends on paths outside your scope until they are measured. A shop that promises a receiver-side result on a partition-side scope owns the difference.

References

• ASTM E336 (field measurement of airborne sound attenuation between rooms) and ASTM E413 (single-number classification), in the editions the project specification or the 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 obstruction and coverage
• See related: Why a Wall Never Performs Like Its Rating; How to Find a Flanking Path Without Opening the Wall; What a Small Hole Does to a Large Wall; Why a Door or a Duct Is Usually the Weakest Element