Why an Open Office Is an Acoustics Problem, Not a Layout Problem

Why this matters

An open plan that people hate gets rearranged, usually more than once, and rearranging it moves the one variable that the space has already made nearly worthless. In the floor plate below, speech was falling off at 3.5 dB per doubling of distance, so pushing two desks from 4 m apart to 8 m apart bought 3.5 dB and cost a weekend. The quantity that predicts whether an open plan works is the rate at which speech level decays across the floor, and layout does not set it. The ceiling, the screens, the hard surfaces and the background level set it.

The complaint

A professional services tenant moved 46 people out of private offices into an open plan on one floor. Within a month the complaint was consistent and specific: people could follow conversations they were not part of, from three or four workstations away, and could not stop following them. Nobody complained about loudness. They complained about being unable to ignore what they could hear, which is a different symptom and points at a different cause.

Two things about that complaint are diagnostic before anyone measures. First, intelligible intruding speech is the most disruptive background there is, far worse per decibel than steady mechanical noise, because the brain processes language whether or not it was invited. Second, "I can understand them" is a signal-to-noise statement, and the signal is somebody else's voice.

Three explanations that got tried first

Layout. Two rearrangements. The second one pushed the loudest team to a corner and put more distance between facing pairs. Complaints did not move. That is consistent with the measurement that came later, and it is the observation that should have redirected the work.

The ceiling tile. The tile was assumed to be absorptive because it was a lay-in acoustic tile. The submittal put it at an NRC in the low range, which is a facing over a board with modest absorption above 500 Hz, tested under ASTM C423 on a type E-400 mounting (suspended 400 mm from the test surface, which is how a lay-in tile is actually used). That was a real finding, but the tenant treated it as the whole answer, and replacing tile alone would not have delivered.

Changing out lay-in tile means a portable ladder set per 29 CFR 1910.23 and eye protection per 29 CFR 1910.133, and in a building whose vintage is not documented, thermal system insulation and surfacing material above the grid are presumed asbestos-containing under 29 CFR 1910.1001(j). That is an inhalation route, so it gets sampling and a plan, not a pair of gloves.

The people. Voice-level policies, a quiet-hours schedule, a sign. This produced nothing measurable, which is expected: normal vocal effort is what the space has to survive, and asking 46 people to whisper is a control that decays to nothing in about two weeks.

The measurement that settled it

Open plan performance is measured to ISO 3382-3, in the edition the specification names, along a line of workstations. It reports four things, and each one carries its conventions:

• Spatial decay rate of speech, D2,S, in dB of A-weighted sound pressure level per doubling of distance. This is the slope of the fitted line, not a level.
• Speech level at 4 m, Lp,A,S,4m, in dB, A-weighted sound pressure level re 20 micropascals, overall, from a standardized speech source.
• Distraction distance, rD, the distance at which the Speech Transmission Index falls to 0.50.
• Privacy distance, rP, where it falls to 0.20.

Measured on this floor: D2,S of 3.5 dB per distance doubling, Lp,A,S,4m of 51 dB, and a distraction distance well past 10 m. The informative guidance in ISO 3382-3 puts a well-performing open plan above about 7 dB per doubling, below about 48 dB at 4 m, and inside about 5 m of distraction distance. Every one of the four was on the wrong side.

Background was 38 dB, A-weighted sound pressure level re 20 micropascals, overall, equivalent-continuous over 15 minutes, and uneven across the plate. A quiet floor sounds like a nice thing to have and is the opposite in an open plan: it is the denominator of every intelligibility ratio in the building.

None of these figures is an occupational exposure question. 29 CFR 1910.95 sets the federal floor at an 85 dB A-weighted eight-hour time-weighted average action level, and an office at 38 dB is nowhere near it. Nothing here is an OSHA matter; it is a fitness-for-purpose matter.

The three paths a voice takes to the next desk

Prose makes the reader assemble this, so here it is:

 ceiling tile (a mirror if it does not absorb)
 --------------------------------------------
      \                            /
       \   reflected path         /
        \                        /
 talker  \                      /   listener
   o  ---------- over the screen --------> o
   |          screen top                   |
   |  ____________|____________            |
   |                                       |
   \_____ around the screen edge __________/

A screen only interrupts the direct path. Raise the screen and the reflected path off the ceiling becomes the controlling one, so a taller screen under a reflective ceiling buys very little. That is why the ceiling and the screen are one decision, not two, and why treating either alone reads as a failed investment.

What was changed, and what each change bought

Three changes, in one phase, because doing them separately would have left the tenant complaining after each one.

Overhead absorption. Tile replaced with a product whose ASTM C423 report shows high absorption across the 500 to 4000 Hz octave bands on the E-400 mounting it is installed on. The mounting matters: a number tested tight to a hard surface does not describe a tile hung 400 mm below the deck, and using the wrong report is a re-basing, not a small error.

Screens at the right height. Screens between facing workstations at 65 in above finished floor, high enough to break the line of sight between seated heads, with an absorptive face toward the talker rather than a hard laminate that turns the screen into another mirror.

A commissioned masking system. Distributed loudspeakers, adjusted to 45 dB, A-weighted sound pressure level re 20 micropascals, overall, equivalent-continuous, uniform within a couple of decibels across the plate, with the per-octave-band spectrum set to a published masking contour rather than left at whatever the amplifier did.

Re-measured to the same ISO 3382-3 procedure along the same line: D2,S of 7.5 dB per distance doubling, Lp,A,S,4m of 47 dB.

What that does at a desk 8 m away. Speech level at 8 m is the 4 m value minus one doubling of decay.

• Before: 51 minus 3.5 = 47.5 dB, A-weighted, overall, against a 38 dB background. Signal-to-noise ratio of the intruding voice: +9.5 dB. Comfortably intelligible, which is exactly what people reported.
• After: 47 minus 7.5 = 39.5 dB against a 45 dB background. Signal-to-noise ratio: minus 5.5 dB. The words stop resolving.

Where each change landed. The signal-to-noise ratio of the intruding voice moved 15 dB, from +9.5 to minus 5.5. Eight decibels of that came from lowering the speech itself: 4 dB of source level plus 4 dB of extra decay over the single doubling from 4 m to 8 m, both owned by the ceiling and the screens. The remaining 7 dB came from raising the background. They are not independent levers on the same quantity: absorption and screens lower the numerator, masking raises the denominator, and the complaint is about the ratio. Doing only the masking would have raised the noise floor without lowering the speech, which produces a floor that is both noisy and still distracting, and that is the single most common way a masking install gets a bad name.

Distraction distance came back inside the ISO 3382-3 informative target. Complaints did not go to zero; they went to the ordinary level of an open plan, which is the honest outcome to promise.

Sibling-rule check. Every level above carries quantity, weighting, reference, bandwidth and time basis, including in the change descriptions where plain speech would have dropped them. The decay rate carries its unit of analysis (per doubling of distance, along a measured line, from a standardized source), and the 8 m figures apply exactly one doubling from the 4 m reference rather than interpolating. The absorption coefficients carry their ASTM C423 mounting, and the mounting change is called a re-basing rather than an adjustment. The direction of the background relationship matches the speech-intelligibility article: raising the background lowers intelligibility, which is a loss for the person you want to hear and a gain against the person you do not. 29 CFR 1910.95 is cited as the occupational floor and explicitly not as a comfort criterion.

What would have made a different call correct

A tonal source in the background. A masking spectrum with an audible tone or a whine is worse than no masking, because a tone draws attention rather than covering. Where the background already has a tonal component, fix the tone first; a sibling article covers why tonal noise is judged on different terms than broadband.

A ceiling that cannot be treated. Exposed structure with mechanical above it is a common architectural choice and it takes the largest absorptive surface out of play. Then the levers are horizontal baffles or clouds, hung to expose both faces, plus more aggressive screening, and the honest answer is that the target decay rate may not be reachable. Say that before the work rather than after.

A room where the wanted talker is also at a distance. A training area inside an open plan wants intelligibility, not privacy, and masking works against it. Zone the masking so it does not cover the space where people are supposed to hear each other, and accept that the boundary between the two zones is a compromise.

References

• ISO 3382-3 for open plan measurement, in the edition the specification names; it reports spatial decay rate, speech level at 4 m, distraction distance and privacy distance.
• ASTM C423 for absorption coefficients and the mounting used in the test, including the suspended mountings that describe a lay-in ceiling.
• 29 CFR 1910.95 for the occupational noise floor, 1910.23 for portable ladders, 1910.133 for eye protection, 1910.1001(j) for presumed asbestos-containing material above a ceiling.
• See related: What Speech Intelligibility Depends On; What Background Noise Does That Is Actually Useful; What Tonal Noise Is and Why It Is Judged Differently.