Skip to content

Theory: Rooms and Buildings

Standards: ANSI/ASA S12.2 ISO 18233 ISO 3382 ISO 16283 ISO 717 ISO 354 Key references: Kuttruff 2016 Vorländer 2020 Allen & Berkley 1979 +11 more

Verified by 76 conformance checks against ANSI/ASA S12.2-2019, ISO 18233:2006, ISO 3382-1:2009, ISO 3382-2:2008, ISO 3382-3:2012, ISO 16283-1:2014, ISO 717-1:2020, ISO 354:2003, EN 12354-6:2003, ISO 12999-1:2020, ISO 16283-2:2020, ISO 16283-3:2016, ISO 717-2:2020, ISO 10140-2:2021, EN 12354-1:2000 and EN 12354-2:2000: 60 in Room & building acoustics, 9 in Building prediction & uncertainty, 3 in Room noise (ANSI S12.2-2019), 2 in Sound absorption in enclosed spaces (EN 12354-6), 1 in Enclosure and cabin insulation and 1 in Suspended ceilings in a reverberation room (EN 16487).

This page collects the theory behind rooms and buildings: impulse-response measurement and the room-acoustic parameters, the image-source and steady-state models of the sound field in a room, background-noise criteria, airborne and impact insulation with their single-number ratings and uncertainty, and flanking and absorption prediction. It is part of the theory reference; surface scattering and acoustic material characterisation live on Materials and Surfaces.

Each section keeps its own standard’s notation, and five letters carry more than one meaning across them:

Symbol Meaning Where
Equivalent absorption area, m² ( m² reference, specimen, , ) throughout
Power attenuation coefficient of air, 1/m image sources, ISO 354 and EN 12354-6
Number of repeated measurements ISO 12999-1 uncertainty only
/ Mass per unit area, kg/m² (EN 12354 writes , the panel-prediction literature ) flanking and panel prediction
Clarity, dB (subscripted 50 or 80) ISO 3382-1 parameters
Decay curvature, % ISO 3382-2 validity
, Spectrum adaptation terms, dB ISO 717 ratings
Room constant, m² steady-state room field
Pressure reflection factor of wall image sources
, Sound reduction index, dB ( weighted, apparent) insulation sections
Reverberation time, s ( is the centre time) throughout

ANSI/ASA S12.2-2019 rates steady background noise in rooms against families of octave-band curves (16 Hz – 8 kHz). The NC rating starts from the speech interference level (clause 3.2), which preselects the NC-(SIL) curve: if no band exceeds it, the spectrum is designated NC-(SIL) and the procedure ends there (clause 5.2.2). The two families otherwise differ in almost every respect:

NC RC Mark II
Curve family Table 1 curves NC-15 to NC-70, 16 Hz – 8 kHz A pure −5 dB/octave line keyed to its 1 kHz value, floored at dB at 16 and 31.5 Hz (Annex D)
How the rating is picked Tangency (clause 5.2.3): each band interpolated against the tabulated curves, the highest per-band index wins Arithmetic mean of the 500/1000/2000 Hz levels, rounded to an integer (D.4)
What is reported A continuous value plus the band that set it (e.g. NC-42.5 at 250 Hz, not snapped to a curve) An integer with a spectral-quality tag, e.g. RC-35(N)
Quality tag none Rumble “R” when any band ≤ 500 Hz exceeds the contour by more than 5 dB, hiss “H” when any band ≥ 1 kHz exceeds it by more than 3 dB, “RH” for both, else “N” (D.3)
Outside the family Flagged: >NC-70 with the band of maximum exceedance, or <NC-15, rather than a fabricated number —

The generated RC contours reproduce Table D.1 digit for digit, and feeding any Table 1 NC curve back returns its own tangency rating. NCB, RNC (Annex A) and the QAI (clause D.5) are deliberately out of scope.

Reading a rating. Neither number is a verdict on its own; the criterion comes from the use of the room, and Annex C (informative) is where the standard puts it. Its Table C.2 recommends, for example, NC 25-30 for a classroom or a small auditorium, 30-35 for a small private office, 40-45 for an open-plan area and 40-50 for public circulation. A rating in the low forties is unremarkable in an open office and a clear failure in a lecture room. The RC tag is a diagnosis rather than a grade: the reference contour is a straight −5 dB/octave line because that is the slope occupants describe as bland, so R means the spectrum carries more low-frequency energy than that slope allows and points at fan and duct-borne noise or a structure-borne path, while H points at the terminal device: diffuser velocity, turbulence at the outlet. Different tags, different remedies. And a rating computed from a spectrum whose lowest bands were never actually measured is not comparable with one where they were, because those bands set both the tag and, often, the tangency point.

See the Room Noise guide for usage.

Two panels for the same ventilation-dominated room spectrum. Left: the measured octave-band levels over the NC curve family, with a red diamond marking the tangent point at 250 Hz that sets the NC-42.5 rating. Right: the same spectrum over the reference RC-35 curve, with the low-frequency bands rising through the shaded rumble tolerance (plus 5 dB below 500 Hz) so the noise is classified RC-35(R), and the hiss tolerance (plus 3 dB at and above 1000 Hz) shaded for comparisonTwo panels for the same ventilation-dominated room spectrum. Left: the measured octave-band levels over the NC curve family, with a red diamond marking the tangent point at 250 Hz that sets the NC-42.5 rating. Right: the same spectrum over the reference RC-35 curve, with the low-frequency bands rising through the shaded rumble tolerance (plus 5 dB below 500 Hz) so the noise is classified RC-35(R), and the hiss tolerance (plus 3 dB at and above 1000 Hz) shaded for comparison

One ventilation-dominated spectrum, read by both procedures. The NC rating is set at 250 Hz (not at any of the four SIL bands) so the governing band is already telling the diagnostic story, and NC-42.5 would pass in an open-plan area and fail in a lecture room. The RC panel says the same thing in one letter: the bands below 500 Hz push through the +5 dB rumble tolerance, so the room is RC-35(R) and the remedy is on the low-frequency side, in the fan and the duct run, not at the diffuser.

Impulse response and room-acoustic parameters (ISO 18233, ISO 3382-1/-2/-3)

Section titled “Impulse response and room-acoustic parameters (ISO 18233, ISO 3382-1/-2/-3)”

Deterministic-excitation impulse response (ISO 18233)

Section titled “Deterministic-excitation impulse response (ISO 18233)”

A room/transmission path is modelled as linear time-invariant, so its impulse response carries everything. ISO 18233 replaces the classical noise-burst decay with a deterministic excitation that is deconvolved into , gaining 20–30 dB of effective signal-to-noise ratio. The exponential sine sweep (ESS, Annex B) has instantaneous frequency , so its phase is the closed-form integral of :

A constant time-per-octave makes the ESS spectrum pink (−3 dB/octave). Deconvolution is done by linear (non-circular, zero-padded) spectral division , the Tikhonov term (a fraction of ) preventing noise blow-up at the band edges. Since a low-to-high sweep places harmonic-distortion products at negative arrival times, they fall in the wrapped tail and are removed by keeping the causal part (Farina). The MLS method (Annex A) instead exploits that the circular autocorrelation of a maximum-length sequence of length is a periodic delta, so ; synchronous averaging of periods adds dB.

How long the record has to be, and what breaks first. The two rules that decide whether a deconvolved impulse response is usable are in clause 6.2, and they differ by excitation type. A repetitive signal such as an MLS must have a repetition period no shorter than the reverberation time of the room, (Formula 10), or the modes are not excited to steady state and the response wraps onto itself. A non-repetitive sweep may be of any suitable length, but it must be followed by silence, and the decay must be recorded over at least half the reverberation time (6.2.2.3); for a low-to-high sweep the silence needed is set by the reverberation time at the high frequencies, which is where padding has to be kept clear of the wrapped harmonic arrivals. A record that stops with the sweep folds the tail back onto the head and biases every and every clarity value derived from it, silently and always in the same direction.

The diagram below sizes both excitations for a room with s. For the sweep, Annex B (B.3.1) considers a length of two to four times the longest reverberation time, followed by a silent gap equal to that time, normally safe under moderate background noise, and doubling the sweep duration normally adds 3 dB of effective signal-to-noise ratio (B.6). For the MLS, the period of an order- sequence is : at 48 kHz, order 17 lasts 2.73 s and satisfies Formula 10, while order 15 lasts 0.68 s and does not. Annex A takes the sequence as replayed periodically (A.2), and the room answers periodically only once a whole period has run through it, so in practice the first period played is discarded and the second is kept; that step is measurement practice, which the standard does not state. The third lane is what the deconvolution of the sweep returns, the linear response at time zero and the harmonic packets ahead of it.

Three time lanes for a room with T = 1.2 s. The first shows the played sweep of 4.0 seconds, three and a third times T, followed by a silent gap of about one T and a bracketed record window of 5.2 seconds. The second shows a periodic MLS of order 17, period 2.73 seconds, with the first period marked as warm-up and discarded and the second kept, and below it a short order-15 period of 0.68 seconds marked as shorter than T so the tail folds onto the head. The third shows what the deconvolution returns: the linear impulse response and its tail at time zero with the causal part kept, the second, third and fourth harmonic packets at negative arrival times, and the deconvolution's own decaying noise tailThree time lanes for a room with T = 1.2 s. The first shows the played sweep of 4.0 seconds, three and a third times T, followed by a silent gap of about one T and a bracketed record window of 5.2 seconds. The second shows a periodic MLS of order 17, period 2.73 seconds, with the first period marked as warm-up and discarded and the second kept, and below it a short order-15 period of 0.68 seconds marked as shorter than T so the tail folds onto the head. The third shows what the deconvolution returns: the linear impulse response and its tail at time zero with the causal part kept, the second, third and fourth harmonic packets at negative arrival times, and the deconvolution's own decaying noise tail

The assumption that breaks first is time invariance, and clause 6.1 is explicit about it: moving the source or a microphone during the measurement is not acceptable, and neither air movement nor a change in the speed of sound through temperature is. The path length changes by a fraction of a wavelength between the start and the end of the excitation, and the deconvolution charges the mismatch to the noise floor. This is also what separates the two methods in practice: MLS rests on the exactness of the circular autocorrelation and degrades under both time variance and loudspeaker distortion, which it smears over the whole response, whereas the sweep pushes distortion products to negative times where they can be cut away. The sweep is therefore the default, and MLS is kept for cases where the excitation must be noise-like.

Schroeder backward integration (ISO 3382-1, 5.3.3)

Section titled “Schroeder backward integration (ISO 3382-1, 5.3.3)”

The band decay curve is the backward-integrated squared IR (Schroeder):

i.e. a reversed cumulative sum in discrete time. Backward integration cancels the random fluctuation of a single squared IR: for a purely exponential energy decay it gives , an exactly straight line . Background noise flattens , so integration is truncated at the crossing of the fitted decay line with the noise level and the missing tail is compensated by an exponential with the fitted rate; without that term the finite integral systematically underestimates .

The clip runs that reversed sum: the tail energy fills from the end of the record while the decay curve is drawn toward , and the T20 and T30 regression lines of the next subsection close it.

The tail energy of the squared impulse response fills from the end while the backward integral advances toward t = 0, and the Schroeder decay curve emerges on a companion axis, ending with the T20 and T30 regression lines.

Download the animation (WebM)

The tail energy of the squared impulse response fills from the end while the backward integral advances toward t = 0, and the Schroeder decay curve emerges on a companion axis, ending with the T20 and T30 regression lines.

Download the animation (WebM)

Squared impulse response with its Schroeder backward-integrated decay curve, and the EDT, T20 and T30 regression windows markedSquared impulse response with its Schroeder backward-integrated decay curve, and the EDT, T20 and T30 regression windows marked

A squared impulse response, its Schroeder backward integral and the EDT/T20/T30 regression windows of the next subsection.

Regression windows and validity (ISO 3382-2, Clause 6, Annex B/C)

Section titled “Regression windows and validity (ISO 3382-2, Clause 6, Annex B/C)”

Reverberation time is a least-squares fit over a window, extrapolated to 60 dB via (Annex C): EDT on 0 to −10 dB, T20 on −5 to −25 dB, T30 on −5 to −35 dB. A single-slope decay gives EDT = T20 = T30; a fast early / slow late double slope gives EDT < T30. Validity uses the dynamic-range rule of 5.3.3: the noise must sit at least 25 dB below the IR peak for EDT (evaluation span + 15 dB), tightened to 46 dB for T20 and 54 dB for T30 so the tail-compensation bias of a flagged-valid value stays within the 5 % JND. The curvature % (Annex B) flags a non-straight decay above 10 %.

The regression is only half of what makes a property of the room rather than of one point in it; the other half is spatial sampling, and ISO 3382-2 grades the result by it (clause 4.3, Table 1). The survey method needs at least one source position and two source-microphone combinations, one decay each, in octave bands only, for a nominal accuracy better than 10 %. The engineering method needs at least two source positions and six independent combinations with two decays each, and is the grade the sound-insulation standards call for (better than 5 % in octaves, 10 % in thirds). The precision method needs two source positions and twelve combinations with three decays each (2.5 % and 5 %). Independence is the operative word: microphone positions should be about half a wavelength apart, in practice a minimum of some 2 m, at least about a quarter wavelength (roughly 1 m) from any reflecting surface including the floor, never symmetric, and no closer to a source than , so that direct sound does not dominate the early decay. Positions packed closer than that are not the independent positions Table 1 counts, and a averaged over them carries no grade at all.

The diagram below draws those rules on a room of 10 m by 6 m, 3.5 m high, for which m³ and an expected of 0.6 s give m. It adds the two heights of ISO 3382-1 (clause 4.3): the acoustic centre of the source 1.5 m above the floor and, in rooms for speech and music, the microphones at 1.2 m, the ear height of a seated listener.

Room-acoustics measurement setup: a top-view room plan with two loudspeaker source positions and six microphone positions with the ISO 3382-1 spacing rules, and the ISO 3382-2 table of minimum positions for the survey, engineering and precision gradesRoom-acoustics measurement setup: a top-view room plan with two loudspeaker source positions and six microphone positions with the ISO 3382-1 spacing rules, and the ISO 3382-2 table of minimum positions for the survey, engineering and precision grades

Clarity, definition and centre time (ISO 3382-1, Annex A)

Section titled “Clarity, definition and centre time (ISO 3382-1, Annex A)”

Splitting the energy at an early/late boundary gives the early-to-late index and the definition ratio:

with ms (C50, speech) or 80 ms (C80, music), and the centre time . For a pure exponential decay these have closed forms and ; at s () they evaluate to C80 = 3.05 dB, C50 = −0.02 dB, D50 = 0.499 and Ts = 72.4 ms, the values the implementation reproduces. Table A.1 JNDs (EDT 5 %, C80 1 dB, D50 0.05, Ts 10 ms) bound how finely each is worth reporting.

ISO 3382 per-band parameters of a synthetic room impulse response: grouped EDT, T20 and T30 bars per octave band falling from about 1.4 s at 125 Hz to 0.7 s at 4 kHz, over a second panel where C50 and C80 rise with frequencyISO 3382 per-band parameters of a synthetic room impulse response: grouped EDT, T20 and T30 bars per octave band falling from about 1.4 s at 125 Hz to 0.7 s at 4 kHz, over a second panel where C50 and C80 rise with frequency

The closed forms above hold for a single exponential decay; a real room gives one set of values per band. The upper panel is the decay itself (EDT, T20 and T30 falling with frequency as air and surfaces absorb more), the lower panel the early/late split of the same impulse response, and C50 and C80 rise with frequency for the same reason the decay time falls: the later the energy, the more of it the room has already removed.

Open-plan spatial decay (ISO 3382-3, Clause 6)

Section titled “Open-plan spatial decay (ISO 3382-3, Clause 6)”

The spatial decay rate of A-weighted speech is the ordinary least-squares slope of against ( m) over the 2–16 m positions, rescaled to a per-doubling figure, and the nominal level is read off the same line at 4 m:

The distraction distance rD and privacy distance rP are the distances where a linear (not logarithmic) regression of STI against distance crosses 0.50 and 0.20; a fitted slope that does not fall (STI not decreasing with distance; a constant STI fits to a slope a few units in the last place either side of zero, read as no decrease) makes them undefined, realising the standard’s “can prove impossible to determine” note.

Open-plan spatial decay: A-weighted speech level and STI against source distance on a log axis, with the D2,S regression, the Lp,A,S,4m marker at 4 m and the rD and rP distance crossingsOpen-plan spatial decay: A-weighted speech level and STI against source distance on a log axis, with the D2,S regression, the Lp,A,S,4m marker at 4 m and the rD and rP distance crossings

Two regressions on two different axes, which is what makes this clause hard to hold in the head. The level line is fitted against and read twice, as the slope per doubling, and at m for . The STI line is fitted against itself, linearly, and read where it crosses 0.50 and 0.20 for the distraction and privacy distances. If that second fit comes out flat or rising, the two distances do not exist rather than being large.

Room sound field, predicted (Kuttruff, Vorländer, Bies)

Section titled “Room sound field, predicted (Kuttruff, Vorländer, Bies)”

Image-source room impulse response (Kuttruff 4.1, Vorländer 11)

Section titled “Image-source room impulse response (Kuttruff 4.1, Vorländer 11)”

A rectangular room reflects a point source in its walls; each reflection equals the free-field sound of a mirror image of the source. Mirroring the source position in a wall at distance with unit normal (, Vorländer Eq. 11.36) turns the source into a regular lattice of images, and the room impulse response is the sum of the direct sound and one delayed, attenuated impulse per image (Kuttruff Eqs. 4.4–4.5),

with the spherical spreading, the product of the wall pressure reflection factors (Vorländer Eq. 11.39; in energy) each raised to the number of reflections that image made off wall , and the air pressure attenuation ( the power attenuation coefficient of the table above, so the intensity falls as and the pressure as ). Along one axis the reflection counts of the image at lattice index and parity are and off the two walls (Allen & Berkley 1979), so the total order is ; a shoebox has audible images up to order (Kuttruff Eq. 9.23), and the reflection density grows as (Kuttruff Eq. 4.6).

Integrated from , that density is a count: the sphere of radius about the receiver holds images, one per room volume. The clip sweeps that sphere over the lattice of a room of 7 m by 5 m, 3 m high, and counts the arrivals against the law:

On the left, the 7 by 5 metre room sits at the centre of its grid of dashed mirror rooms, with the source as a star and the receiver as a triangle. A circle expands from the receiver at the speed of sound; as it reaches each image source that image lights up in its reflection-order colour and draws a straight line back to the receiver. On the right, the reflectogram fills in at the same instant: each image the circle reaches writes one stem at its arrival time and its level relative to the direct sound, under the dashed one-over-r spreading envelope. Below it a running count of arrivals climbs alongside the dashed analytic law, the two staying together as the count passes 44 at 30 milliseconds, 108 at 40, 202 at 50 and 344 at 60, where the law reads 348.

Download the animation (WebM)

The echogram is a lattice being swept, not a decaying signal. Because the expanding sphere gains volume as the cube of time while the images are spaced one per room volume, arrivals accumulate as the cube of time and their rate grows as its square: the early discrete reflections and the late diffuse tail are the same construction read at two densities.

On the left, the 7 by 5 metre room sits at the centre of its grid of dashed mirror rooms, with the source as a star and the receiver as a triangle. A circle expands from the receiver at the speed of sound; as it reaches each image source that image lights up in its reflection-order colour and draws a straight line back to the receiver. On the right, the reflectogram fills in at the same instant: each image the circle reaches writes one stem at its arrival time and its level relative to the direct sound, under the dashed one-over-r spreading envelope. Below it a running count of arrivals climbs alongside the dashed analytic law, the two staying together as the count passes 44 at 30 milliseconds, 108 at 40, 202 at 50 and 344 at 60, where the law reads 348.

Download the animation (WebM)

The echogram is a lattice being swept, not a decaying signal. Because the expanding sphere gains volume as the cube of time while the images are spaced one per room volume, arrivals accumulate as the cube of time and their rate grows as its square: the early discrete reflections and the late diffuse tail are the same construction read at two densities.

The initial decay rate of the specular reverberant energy recovers the Eyring reverberation time (Kuttruff Eq. 5.23), with here the total boundary area, because the mean reflection rate equals . The match is close only near cubic geometry; an elongated room sustains energy along its long axis, so the pure specular decay runs slower than Eyring’s diffuse-field estimate (the anisotropy the Fitzroy/Arau-Puchades models correct). The model is specular only (no diffraction or diffuse scattering) and exact only for real, angle-independent reflection factors.

Steady-state room field (Bies 6.4, Kuttruff 5.6)

Section titled “Steady-state room field (Bies 6.4, Kuttruff 5.6)”

A source of constant power sets up a steady level made of a direct field falling with distance and a diffuse reverberant field that is (approximately) uniform. With the room constant (Bies Eq. 6.44), the total boundary area and its mean absorption coefficient, and the directivity factor ,

the optional last term correcting a characteristic impedance away from 400 Pa·s/m. In air at 20 °C, 101.325 kPa and 50 % relative humidity it adds 0.13 dB:

from phonometry import fluids, room
air = fluids.air(
temperature_c=20.0, static_pressure_pa=101325.0, relative_humidity_percent=50.0
)
rho_c = air.characteristic_impedance # Pa·s/m
plain = room.steady_state_spl(90.0, 2.0, room_constant=50.0)
corrected = room.steady_state_spl(
90.0, 2.0, room_constant=50.0, characteristic_impedance=rho_c
)
print(round(rho_c, 1), round(corrected - plain, 2)) # 412.5 0.13

The critical distance is where the two fields cross; Kuttruff’s reverberation distance (Eq. 5.44, for ) uses the Sabine area instead of , the two coinciding for small . The Schroeder frequency (Kuttruff Eq. 3.44) roughly marks the modal-to-diffuse transition, a heuristic crossover rather than a sharp cutoff: well below it discrete room modes dominate and these diffuse-field relations grow unreliable, well above it the modes overlap and the relations hold. Borderline rooms still warrant a band-by-band check.

See the Image Sources and the Steady-State Room Field guide for usage.

Sound insulation and absorption, measured (ISO 16283, ISO 10140, ISO 717, ISO 354)

Section titled “Sound insulation and absorption, measured (ISO 16283, ISO 10140, ISO 717, ISO 354)”

Field insulation and weighted rating (ISO 16283-1, ISO 717-1)

Section titled “Field insulation and weighted rating (ISO 16283-1, ISO 717-1)”

Per one-third-octave band the level difference (energy-averaged over microphone positions, ) is normalised two ways: the standardized level difference with s (so when ), and the apparent sound reduction index with the Sabine absorption area , hence .

In a room below 25 m³ that default average is not the level ISO 16283-1 asks for in the three lowest bands. Clause 8 adds a corner measurement: the levels taken in the room corners are energy-averaged (Formula 12) and combined with the default average as (Formula 13), and that replaces the default level in the 50, 63 and 80 Hz bands only, because the default positions do not sample the modal field there. The library consumes levels and does not perform this combination, so it has to be applied before calling. It matters because the extended spectrum adaptation terms and are dominated by exactly those three bands: a small-room rating computed without the corner procedure is not comparable with one that used it.

The single-number rating (ISO 717-1, Clause 4.4) shifts the Table 3 reference curve in 1 dB steps toward the measured curve until the sum of unfavourable deviations is maximal but 32.0 dB (16 thirds) or 10.0 dB (5 octaves); the rating is the shifted reference at 500 Hz. The spectrum adaptation terms are and with (Table 4 spectra No. 1 pink noise, No. 2 urban traffic), each rounded to an integer. The ISO 717-1 Annex C worked example (, , , unfavourable sum 31.8 dB) is reproduced exactly.

The search is iterative, and the two positions that decide it are the ones a finished figure has already dropped: the last shift whose unfavourable sum still exceeded 32.0 dB, and the shift after the answer, which is allowed but gives a decibel away. The clip walks the Annex C spectrum of ISO 717-1 through them, and then the bare heavy floor of ISO 717-2 Annex C through the mirrored rule of the next subsection.

The reference curve steps in 1 dB increments toward the measured spectrum while the unfavourable deviations shade in band by band and their sum falls toward the 32.0 dB cap. The airborne engine stops at 31.8 dB for Rw = 30 dB, one step before the position that would waste the allowance; the impact engine then runs the same rule with the deviation's sign reversed, the curve stepping upward and stopping at 28.0 dB for Ln,w = 79 dB.

Download the animation (WebM)

The reference curve steps in 1 dB increments toward the measured spectrum while the unfavourable deviations shade in band by band and their sum falls toward the 32.0 dB cap. The airborne engine stops at 31.8 dB for Rw = 30 dB, one step before the position that would waste the allowance; the impact engine then runs the same rule with the deviation's sign reversed, the curve stepping upward and stopping at 28.0 dB for Ln,w = 79 dB.

Download the animation (WebM)

Both walks stop where the standard says, at the largest unfavourable sum that does not exceed 32.0 dB: 31.8 dB for dB and 28.0 dB for dB. The step before is rejected for overshooting the cap, and the step after is allowed but rates the element a decibel worse for no reason (a lower , a higher ), which is what “as large as possible” in ISO 717-1:2020, 4.4 and ISO 717-2:2020, 4.3.1 excludes.

Reading and . The weighted rating condenses a whole curve against one fixed reference curve, so it answers for the reference source and no other. The adaptation terms put the source dependence back by re-rating the same measured curve against two standard spectra: No. 1 for sources whose spectrum is flat-ish and mid-and-high dominated (speech, television, everyday activity) and No. 2 for sources dominated by low frequencies, such as urban road traffic, distant aircraft and bass-heavy music. Both terms are almost always negative, and the value is the number of decibels to add to to get the level difference that source will actually experience, which is why regulations state requirements as or and not as alone. That makes a diagnostic: near is ordinary for a homogeneous heavy wall, while values approaching or worse mark a construction whose insulation collapses in the lowest bands (typically a lightweight double leaf near its mass-spring-mass resonance, or a window). Two partitions with the same can differ by many decibels under traffic. The terms are computed over the same band range as the rating they accompany, so an adaptation term quoted over 100–3150 Hz and one extended to 50 Hz are not comparable quantities.

Measured one-third-octave sound reduction index with the shifted ISO 717-1 reference curve and the resulting weighted rating at 500 HzMeasured one-third-octave sound reduction index with the shifted ISO 717-1 reference curve and the resulting weighted rating at 500 Hz

A measured R spectrum against the shifted ISO 717-1 reference: the rating is the shifted reference read at 500 Hz.

Impact insulation (ISO 16283-2, ISO 717-2)

Section titled “Impact insulation (ISO 16283-2, ISO 717-2)”

Impact insulation swaps the airborne source for a standardized tapping machine and rates the receiving-room level, so the sign conventions flip. The standardized and normalized impact levels are (the reverberation term is subtracted, opposite to ) and with m² and . The ISO 717-2 rating shifts the Table 3 reference curve until is maximal but 32.0 dB (16 thirds) or 10.0 dB (5 octaves); the unfavourable deviation now counts where the measurement exceeds the reference (impact noise is worse when louder), the mirror image of ISO 717-1. The rating is the shifted reference at 500 Hz, reduced by a further 5 dB for octave bands, and the adaptation term is with the energetic sum over 100–2500 Hz (thirds) or 125–2000 Hz (octaves). The ISO 717-2 Annex C examples are reproduced exactly (thirds , ; octaves , ), via the same monotone shift search as ISO 717-1 run on the negated curves.

Measured one-third-octave normalized impact sound pressure level with the shifted ISO 717-2 reference curve and the resulting weighted rating read at 500 HzMeasured one-third-octave normalized impact sound pressure level with the shifted ISO 717-2 reference curve and the resulting weighted rating read at 500 Hz

The mirror image of the airborne figure above, drawn so the flip is visible rather than asserted. There the unfavourable deviations were counted where the measurement fell below the reference; here they are counted where it rises above it, because a louder receiving room is a worse floor. Everything else is the same procedure: the reference curve shifted in 1 dB steps until the unfavourable sum is as large as it can be without passing 32.0 dB, and the rating read off the shifted reference at 500 Hz.

Absorption in a reverberation room (ISO 354)

Section titled “Absorption in a reverberation room (ISO 354)”

Sound absorption (ISO 354) measures the equivalent absorption area from Sabine’s relation applied to a reverberation room empty and with the specimen: (the term is the air absorption, the power attenuation coefficient in 1/m), so the specimen area is and its coefficient . With the speed of sound from Eq. (6), (°C), and converted from an ISO 9613-1 attenuation coefficient by . Because diffraction and edge scattering intercept more than the flat sample area, is left unclamped and may exceed 1.0 (Clause 3.7 NOTE 2).

The two reverberation times are not free inputs. They are the output of a tightly specified arrangement, and means nothing without it. The room holds at least 150 m³ (200 m³ strongly recommended for a new one), its longest dimension stays below , and it carries enough fixed or rotating diffusion for the decays to be exponential (clause 6.1). The plane specimen covers 10 to 12 m² (the upper limit scaled by above 200 m³) and is rectangular with a width-to-length ratio between 0.7 and 1, sits at least 0.75 m from any room boundary and preferably has its edges non-parallel to the nearest room edge (6.2.1). Its mounting is part of the result (Annex B): rigid backing, air gap or framed edges change the answer, which is why a catalogue without its mounting code is not usable and why the same product gives different curves in different laboratories. Each is the average of at least 12 spatially independent decays, from at least 3 microphone and 2 source positions (7.1.4), with microphones at least 1.5 m apart, 2 m from any source and 1 m from any room surface and from the specimen (7.1.2), and source positions at least 3 m apart (7.1.3). Both measurements are made between 30 % and 90 % relative humidity and at 15 °C or more, at nearly the same climate (6.3.2).

The diagram below draws that arrangement for a 200 m³ room and a 10.8 m² specimen. It also shows what the paragraph leaves to the standard’s own text: the diffusers of Annex A, sheets of roughly 0.8 m² to 3 m² and about 5 kg/m²; the empty-room absorption area that must stay under the Table 1 limit (6.1.4), with measured without the frame of the specimen (6.2.1.3); each decay evaluated over 20 dB starting 5 dB below the initial level (7.4.1); and two of the Annex B mountings, Type A directly against the floor and Type E-400 with 400 mm between the exposed face and the surface behind it.

Plan of an ISO 354 reverberation room with non-parallel walls: the 10.8 square metre rectangular specimen on the floor with its edges deliberately not parallel to the room edges and a 0.75 metre clearance dimensioned to the nearest boundary, three microphone positions and two loudspeaker positions with their separation rules, suspended diffuser panels near the ceiling, an inset showing the empty room giving T1 and A1 above the room with the specimen giving T2 and A2, and a strip comparing the Type A and Type E-400 mountings in sectionPlan of an ISO 354 reverberation room with non-parallel walls: the 10.8 square metre rectangular specimen on the floor with its edges deliberately not parallel to the room edges and a 0.75 metre clearance dimensioned to the nearest boundary, three microphone positions and two loudspeaker positions with their separation rules, suspended diffuser panels near the ceiling, an inset showing the empty room giving T1 and A1 above the room with the specimen giving T2 and A2, and a strip comparing the Type A and Type E-400 mountings in section

That last condition is not housekeeping. The result is the difference of two large and nearly equal absorption areas, so everything that changes between the empty and the loaded measurement lands in it: a shift in temperature or humidity moves both and the air term , which is why the two decays belong in the same session with the climate recorded. And an above 1 is not an error but the signature of edge diffraction intercepting more energy than the flat projected area (largest for small specimens and at low frequency). The value belongs to a specimen of that size, aspect ratio and mounting and cannot be transplanted to a surface of another shape without judgement.

Laboratory vs field normalization (ISO 10140, ISO 16283)

Section titled “Laboratory vs field normalization (ISO 10140, ISO 16283)”

The field indices carry a prime because the receiving-room level they start from includes flanking transmission around the partition; the laboratory indices do not, because a qualified facility suppresses it. The formulas are therefore the same, and the prime is a statement about the specimen rather than about the arithmetic: the direct laboratory sound reduction index (ISO 10140-2) and the apparent field index (ISO 16283-1) are one closed form, evaluated with the facility’s known in the laboratory and with the receiving room’s measured in the field. What differs in what is normalised is not laboratory against field but airborne against impact: the airborne indices scale by the area-to-absorption ratio , the impact levels by .

Plan view of an ISO 10140 laboratory transmission suite: structurally decoupled source and receiving reverberation rooms of about 59 and 51 cubic metres, the test element mounted in the 10 square metre test opening between them, a corner loudspeaker in the source room and a continuously moving microphone with a sweep radius of at least 1 m in each roomPlan view of an ISO 10140 laboratory transmission suite: structurally decoupled source and receiving reverberation rooms of about 59 and 51 cubic metres, the test element mounted in the 10 square metre test opening between them, a corner loudspeaker in the source room and a continuously moving microphone with a sweep radius of at least 1 m in each room

What “a qualified facility suppresses it” means physically is drawn above: an ISO 10140-5 suite is two structurally decoupled reverberation rooms of roughly 50 to 60 m³ sharing a test opening of about 10 m², built so that transmission around the specimen sits far below transmission through it. The facility measures its own flanking limit, and a specimen may be reported only up to that limit, which is why a laboratory carries no prime, and why the same construction built into a building reads several decibels lower. The practical consequence is a diagnosis: when a field refuses to rise after the element is improved, the measurement has hit a flanking path, not the element. The impact pair is the normalized laboratory level (ISO 10140-3) versus the field (ISO 16283-2), both referenced to m². Before either is formed the receiving-room level is corrected for background noise by the energy subtraction for a 6–15 dB signal-to-background margin, capped at a fixed dB (the limit of measurement) at or below 6 dB and omitted at or above 15 dB (ISO 10140-4, Clause 4.3), the laboratory analogue of the 6/10 dB rule of ISO 16283-1. The façade extension (ISO 16283-3) replaces the source-room level by the level 2 m in front of the façade, , and adds a fixed angle-of-incidence correction to the element sound reduction index, dB for the 45° loudspeaker method () and dB for the all-angle road-traffic method (); all three carry the ISO 717-1 airborne single number.

Sound insulation and absorption, predicted (EN 12354-1/-2/-6, Bies, Cremer, Hopkins)

Section titled “Sound insulation and absorption, predicted (EN 12354-1/-2/-6, Bies, Cremer, Hopkins)”

Flanking transmission prediction (EN 12354-1/2)

Section titled “Flanking transmission prediction (EN 12354-1/2)”

The apparent field index is the energetic sum of the direct path and, for each flanking element across its junction with the separating element, the three paths , and (EN 12354-1, simplified single-number model, Formula 26):

The direct path Dd through the separating element and the three flanking paths Ff, Df and Fd across each junction between a flanking element and the separating elementThe direct path Dd through the separating element and the three flanking paths Ff, Df and Fd across each junction between a flanking element and the separating element

The four labels are only meaningful on this drawing. Upper case is the source side and lower case the receiving side, so goes straight through the separating element, runs along a flanking element on both sides of the junction, and and cross from one to the other. The junction they cross is where and the coupling length live in the formula below, and the separating element is where is measured. Each junction contributes three flanking terms, so a room with four flanking elements sums thirteen paths.

The direct path is (Formula 27), the separating-element laboratory index plus any lining improvement. Each flanking path (Formula 28a) is

with , the laboratory indices of the two elements meeting at the junction ( source side, receiving side), the combined lining improvement, the separating-element area, the junction coupling length and m the reference coupling length. is the junction vibration reduction index (Annex E), an empirical function of the mass ratio : for a rigid cross-junction (through) and (corner), read at 500 Hz, and floored at (Formula 29). Two linings combine as (Formulas 30/31). The impact counterpart (EN 12354-2, Formula 21) is the direct subtraction , with the bare-floor equivalent level (Annex B), the covering improvement (ISO 717-2) and the flanking correction from Table 1. The EN 12354-1 Annex H.3 ( dB) and EN 12354-2 Annex E.3 ( dB) worked examples are reproduced exactly.

Where the simplified model applies. It assumes homogeneous elements whose transmission is governed by their mass per unit area, joined rigidly, so that the junction attenuation can depend on the mass ratio alone; the Annex E regressions are empirical fits to exactly that family. They return no meaningful answer for framed lightweight constructions, for elastically decoupled junctions, or for elements with a strong internal loss mechanism. Those need the detailed band-by-band model or measured values. The floor exists to stop the regressions producing implausibly high junction attenuations for small elements, so a path that lands on the floor is a signal that the mass-ratio fit is at the edge of its range rather than a result to report as-is. Because the paths sum energetically, the apparent rating is pinned by whichever path is strongest: improving the separating element alone stops paying the moment one flanking path dominates, and identifying that path is the model’s chief practical value, well before the last decibel of the total. The 2 dB standard deviation the standard quotes (Clause 5) is the spread of the simplified model against measurement inside that intended population, not a general error bar.

Absorption in enclosed spaces (EN 12354-6)

Section titled “Absorption in enclosed spaces (EN 12354-6)”

EN 12354-6:2003 predicts the equivalent absorption area of a room from its parts (the normative Clause 4 model). The total (Formula 1) sums the surfaces, the objects and the air:

with the power attenuation coefficient of air (Formula 2; Table 1 tabulates it for six temperature/humidity climates over the octave bands 125 Hz – 8 kHz), the volume fraction occupied by objects (Formula 3), and a hard irregular object approximated by (Formula 4). The reverberation time follows from Sabine applied to the free volume (clause 4.4, Formula 5):

with m/s chosen so that is the familiar (clause 4.4 NOTE). The three Annex E worked cases are reproduced: the bare 29.75 m³ room gives m² and s at 1 kHz, and adding hard objects () raises to 5.03 m² and drops to 0.9 s. The informative Annex D method for irregular spaces and unevenly distributed absorption is out of scope.

Two panels for a 60 cubic metre office with a bare versus acoustically-treated ceiling: the equivalent absorption area per octave band, much higher with the acoustic ceiling, and the reverberation time falling from about five seconds at low frequency for the bare room to under one second with the acoustic ceilingTwo panels for a 60 cubic metre office with a bare versus acoustically-treated ceiling: the equivalent absorption area per octave band, much higher with the acoustic ceiling, and the reverberation time falling from about five seconds at low frequency for the bare room to under one second with the acoustic ceiling

What Formula 1 does band by band: the equivalent absorption area on the left and the reverberation time it implies through Formula 5 on the right, for the same room bare and treated. The Annex E case quoted above is the same arithmetic on a smaller room: from 2.26 to 5.03 m² and from 2.1 to 0.9 s at 1 kHz, and the figure shows why the two move in opposite directions and not proportionally.

See the Enclosed-Space Absorption guide for usage.

Predicted panel sound insulation (Bies 7.2, Hopkins 2.9/4.3.10, Cremer 5)

Section titled “Predicted panel sound insulation (Bies 7.2, Hopkins 2.9/4.3.10, Cremer 5)”

Where EN 12354 takes the element as a measurement, the sound reduction index of a panel can also be predicted from its physical properties. A limp panel follows the mass law (Bies Eq. 7.40), which rises 6 dB per octave and 6 dB per doubling of the surface mass (the of the EN 12354 subsections above); the field-incidence value subtracts 5.5 dB (one-third octave). That 5.5 dB is not derived: it is the result of averaging the normal-incidence mass law over incidence angles up to the limiting angle observed in practice rather than over the full hemisphere, so it is a fixed empirical offset and not a function of anything. Stiffness adds a coincidence dip at (Eq. 7.3), where the free bending wavelength matches the acoustic trace wavelength. Sharp’s method holds the mass law to , drops linearly in to the dip and rises again above with the loss factor (Eq. 7.44); the there is an empirical fit to measured coincidence dips, and the loss factor appears explicitly because the depth of the dip is controlled by damping rather than by mass. A double wall is a mass-spring-mass system with the cavity as the spring: below (Eq. 7.62) it follows the mass law of the combined mass, and above it the two leaves’ mass laws add plus the cavity term , saturating at +6 dB beyond (Eq. 7.64); a porous fill lowers . The 60 in is the numeric constant left once the stiffness of an air layer of thickness is substituted into the mass-spring-mass resonance, so it is fixed to air at ordinary conditions and changes for a gas-filled or evacuated cavity; and marks where half a wavelength fits across the cavity, above which the cavity stops behaving as a pure spring and standing waves in it limit what the construction can achieve. Small air paths cap any construction: the transmission coefficient of a straight slit (Gomperts, Hopkins Eq. 4.99, with resonances at ) or a circular hole (Wilson & Soroka, Eq. 4.102) combines with the wall in the area-weighted energy sum (Eq. 4.92), so a bare opening of relative area limits the composite to . The resonant transmission path and the double-wall radiation draw on the plate radiation efficiency and point mobilities of the vibration theory.

Four panels: the single-panel mass law with its coincidence dip, the double wall with the mass-spring-mass resonance and cavity gain, the plate radiation efficiency rising to unity above the critical frequency, and a composite wall whose 1 % open slit caps R at the open-area limitFour panels: the single-panel mass law with its coincidence dip, the double wall with the mass-spring-mass resonance and cavity gain, the plate radiation efficiency rising to unity above the critical frequency, and a composite wall whose 1 % open slit caps R at the open-area limit

The four behaviours of the paragraph above, one per panel. Top left, the mass law rising 6 dB per octave with Sharp’s coincidence dip cut into it at . Top right, the double wall: no better than the combined mass below , then the cavity term climbing until it saturates. Bottom left, the radiation efficiency that decides how much of the plate’s vibration becomes sound. Bottom right, the ceiling a leak imposes: a 1 % open area holds the composite at dB however good the wall is, which is the panel worth showing a client.

The dip of the top left panel is a matching of wavelengths, and it can be watched. At the free bending wavelength equals the trace of a wave at grazing incidence; for any smaller angle of incidence there is a frequency above at which the two match again (Bies 7.2.1). With a trace of at an angle from the normal, and a bending wavelength that falls as , that frequency is , which is at 45°. The clip sends a sustained plane wave at 45° onto a 10 mm steel plate, for which the library gives Hz, at half that frequency and at twice it:

Two side-by-side elastic FDTD fields of the same 10 mm steel plate lying in air while a sustained plane wave arrives at 45 degrees. At 602 Hz, half the coincidence frequency, the wave reflects almost totally and the faint transmitted level matches the oblique mass law. At 2410 Hz, twice the coincidence frequency, the acoustic trace matches the free bending wavelength and a 45-degree transmitted beam grows below the plate, holding the same level as the low-frequency panel where the mass law predicted 12 dB more insulation, the air below the plate drawn on both panels with an annotated 150-fold display gain (+44 dB) so the transmitted field is legible, with the 1205 Hz coincidence frequency of the library and both measured and mass-law levels annotated.

Download the animation (WebM)

Two side-by-side elastic FDTD fields of the same 10 mm steel plate lying in air while a sustained plane wave arrives at 45 degrees. At 602 Hz, half the coincidence frequency, the wave reflects almost totally and the faint transmitted level matches the oblique mass law. At 2410 Hz, twice the coincidence frequency, the acoustic trace matches the free bending wavelength and a 45-degree transmitted beam grows below the plate, holding the same level as the low-frequency panel where the mass law predicted 12 dB more insulation, the air below the plate drawn on both panels with an annotated 150-fold display gain (+44 dB) so the transmitted field is legible, with the 1205 Hz coincidence frequency of the library and both measured and mass-law levels annotated.

Download the animation (WebM)

At 602 Hz the plate lets through −47 dB, where the oblique mass law predicts −48 dB. At 2410 Hz it still lets through −47 dB, where the mass law said −60 dB: the 12 dB that two octaves should have added are gone. The air below the plate is drawn with a gain of 150 (+44 dB) on both panels so that the transmitted field shows.

To-scale cross-section of a 2 mm slit through a 100 mm wall: the hatched wall drawn in section with the narrow horizontal air gap at mid-height, an incident-sound arrow pointing at the gap from the left, the 100 mm wall depth and 2 mm slit width dimensioned, and circular transmitted wavefronts sketched spreading from the slit exit on the rightTo-scale cross-section of a 2 mm slit through a 100 mm wall: the hatched wall drawn in section with the narrow horizontal air gap at mid-height, an incident-sound arrow pointing at the gap from the left, the 100 mm wall depth and 2 mm slit width dimensioned, and circular transmitted wavefronts sketched spreading from the slit exit on the right

See the Predicting Panel Sound Insulation guide for usage.

ISO 12999-1 supplies the uncertainty of the quantities above from inter-laboratory (ISO 5725) reproducibility and repeatability rather than a GUM functional model. Three measurement situations fix the standard uncertainty : situation A (laboratory characterisation) uses the reproducibility standard deviation ; situation B (same location, different teams) the in-situ ; situation C (same location, operator and equipment, repeated) the repeatability . The per-band and single-number values are tabulated for airborne /// (Tables 2/3), impact / (Table 4 bands, situations B/C only; Table 5 ratings adding a situation-A estimate) and the covering reduction (Tables 6/7, situation A only). The expanded uncertainty is (Formula 2) with the coverage factor of Table 8 (at 95 %, two-sided, one-sided; a minimum is enforced). A two-sided interval reports a value (Formula 3); a one-sided factor declares conformity, requirement for a lower limit (Formula 5) or requirement for an upper limit (Formula 4). Uncorrelated components combine in quadrature (Formula C.2), independent measurements reduce to (Formula A.7), and the uncorrelated single-number uncertainty is the energy-weighted quadrature sum of the band uncertainties (Formula B.2).

See the Room Acoustics and Field Insulation Measurement (ISO 16283) guides for usage.