# Façade Sound Insulation

> One page of the phonometry documentation, as Markdown. Index: https://docs.jmrp.io/phonometry/llms.txt

Canonical: https://docs.jmrp.io/phonometry/buildings/insulation/facade-insulation/
Language: en
Alternate: https://docs.jmrp.io/phonometry/es/buildings/insulation/facade-insulation/index.md
Version: phonometry 4.0.0
Install: pip install "phonometry[full]"
Author: José Manuel Requena Plens (https://orcid.org/0000-0003-1250-6212)
Published: 2026-07-28
Updated: 2026-11-16T21:36:13Z
Standards: ISO 16283-3:2016; EN 12354-3:2000; EN 12354-4:2000; ISO 717-1:2020
Verified: 22 conformance checks (ISO 16283-3:2016; EN 12354-3:2000; EN 12354-4:2000; ISO 717-1:2020) https://docs.jmrp.io/phonometry/reference/conformance/
License: MIT (https://github.com/jmrplens/phonometry/blob/main/LICENSE)
Summary: Façade insulation in Python, measured by ISO 16283-3 (loudspeaker and road-traffic methods) and predicted by EN 12354-3, with EN 12354-4 outdoor radiation.

---

A façade is the one element of a building that faces the noise everybody
complains about: the road, the railway, the flight path. It is also the
element whose insulation gets asked for twice, once predicted and once
measured, which is why both halves live on this page. ISO 16283-3 measures the
finished envelope, referencing the receiving-room level to the level 2 m in
front of the façade under a 45° loudspeaker or under the road traffic itself.
EN 12354-3 predicts the same standardized level difference from the sound
reduction indices of the wall, the glazing and the air inlet before any of them
is installed, and EN 12354-4 turns the envelope round to radiate an indoor
source outwards. A closing section puts the measured and the predicted
$D_{2\mathrm{m,n}T}$ side by side and shows what does, and does not, separate them. The
reference curves behind every single number live in
[Insulation Ratings (ISO 717)](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-ratings/); the
internal partitions of the same building in
[Field Insulation Measurement (ISO 16283)](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-field/)
and
[Predicting Sound Insulation (EN 12354)](https://docs.jmrp.io/phonometry/buildings/design/insulation-prediction/).

## How do I compute the façade sound insulation D2m,nT,w in Python?

Pass the level 2 m in front of the façade, the receiving-room level and the
receiving-room reverberation time to `building.facade_insulation()`, then rate
its `d_2m_nt` spectrum with `building.weighted_rating()`: the façade quantity
is airborne, so the ISO 717-1 engine applies unchanged (Annex F). Before the
façade exists, `building.facade_sound_reduction()` predicts the same quantity
from the element indices and returns the rating itself; for the EN 12354-3
Annex F example below, `fac.d_2m_nt_w` is 33 dB.

## Field measurement (ISO 16283-3)

A façade measurement asks a question an internal partition never does: **what
is outside, and where is it?** There is no source room, so the reference level
has to be established in a free field that a reflecting building is standing
in, and every quantity on this page inherits the geometry of that choice.
ISO 16283-3 resolves it by splitting the method twice over: a *global* result
that describes the whole envelope as built, against an *element* result that
describes one component; and a *loudspeaker* source that the operator controls,
against *road traffic* that is free but arrives from everywhere. Those two
splits give the four combinations of Table 1 (plus railway and aircraft in
Annex E), and they are chosen on different grounds, which
[Choosing and running the method](#choosing-and-running-the-method-clauses-9-and-10)
sets out. The quantities themselves follow from the choice: ISO 16283-3
references the receiving-room level $L_2$ to the level **2 m in front of the
façade** $L_{1,2\mathrm{m}}$, giving the level difference $D_{2\mathrm{m}}$ and, exactly as in
the airborne case, its standardized and normalized forms:

$$
D_{2\mathrm{m}} = L_{1,2\mathrm{m}} - L_2, \quad
D_{2\mathrm{m,n}T} = D_{2\mathrm{m}} + 10 \log_{10}\frac{T}{T_0}, \quad
D_{2\mathrm{m,n}} = D_{2\mathrm{m}} - 10 \log_{10}\frac{A}{A_0},
$$

with $T_0 = 0.5$ s, $A_0 = 10$ m² and $A = 0.16\ V/T$ (dwellings). When the
microphone sits **on the test element** (surface level $L_{1,\mathrm{s}}$) the *element*
method also yields an apparent sound reduction index, carrying a fixed
angle-of-incidence correction: $-1.5$ dB for the 45° loudspeaker method,
$-3$ dB for the all-angle road-traffic method:

$$
R'_{45°} = L_{1,\mathrm{s}} - L_2 + 10 \log_{10}\frac{S}{A} - 1.5, \qquad
R'_\mathrm{tr,s} = L_{1,\mathrm{s}} - L_2 + 10 \log_{10}\frac{S}{A} - 3.
$$

The façade quantity is airborne, so its single-number rating uses the
**ISO 717-1** reference curve through `weighted_rating` unchanged (Annex F).

Neither constant is arbitrary. A microphone fastened on the façade reads
incident *plus* reflected pressure, because the façade is the reflector; the
subtraction converts that surface level back to the level of the incident wave
alone, and how much has to come off depends on the geometry the sound arrives
with. ISO 16283-3 states the assumption behind each in its own definitions:
the $-1.5$ dB of $R'_{45°}$ "is based on the assumption that the sound is
incident from one angle only, 45°, and the sound field in the receiving room
approximates to a diffuse field" (3.12, Note 3), while the $-3$ dB of
$R'_\mathrm{tr,s}$ assumes the sound arrives "from all angles" (3.13, Note 2). So the
1.5 dB between the two methods is a statement about the *source*, not about the
wall, and it is why a loudspeaker result is labelled $D_{\mathrm{ls},2\mathrm{m,n}T}$ and a
traffic result $D_{\mathrm{tr},2\mathrm{m,n}T}$ rather than both being called the same thing.

The 2 m of the global method is a compromise of the same kind. Closer in, the
microphone reads a pressure dominated by the façade's own reflection and by
whatever detail of the elevation it happens to sit against; much further out,
it stops describing this façade and starts describing the street. Two metres is
far enough that the direct and reflected paths have settled into a stable sum
over most of the range and close enough to stay in the façade's field, but the
standard is candid about the cost, noting under Clause 9.6.1 that "systematic
errors will occur at low frequencies due to interference effects". The
practical consequence is the one to carry away: the reference level of a
$D_{2\mathrm{m,n}T}$ is measured in front of a reflecting surface and is therefore
higher than the free field the same source would produce with the building
removed, so a $D_{2\mathrm{m,n}T}$ and an internal $D_{\mathrm{n}T}$ are not the same kind of
number even when they are the same size.

![Section through a dwelling facade: a loudspeaker on the ground with the 45 degree plus or minus 5 degree incidence line to the centre of an 11.5 square metre test specimen, the slant distance r dimensioned at 5 m for the element method and 7 m for the global method and the perpendicular D beneath it, a microphone fastened flush on the specimen with its 10 mm parallel and 3 mm normal tolerances, and the global microphone 2.0 plus or minus 0.2 m from the facade plane at 1.5 m above the receiving-room floor](https://docs.jmrp.io/phonometry/media/diagram_facade_setup.svg)

### Choosing and running the method (Clauses 9 and 10)

**The instruments (Clause 4).** Everything that measures a sound pressure
level, the microphones with their cables, windscreens and recorders, meets
class 0 or 1 of IEC 61672-1 for random incidence, and the microphone used for
the surface measurements of the element methods has a diameter of **13 mm at
most** (Clause 4.1). The filters are class 0 or 1 of IEC 61260, and the
reverberation time equipment complies with ISO 3382-2. Check the whole chain
with a class 0 or 1 sound calibrator (IEC 60942) at the beginning and at the
end of every measurement session, and at least at the beginning and the end of
each measurement day, and note each reading in the field record. If two
consecutive checks differ by more than 0.5 dB, with no adjustment in between,
everything measured since the last satisfactory check is discarded
(Clause 4.2). Unless national regulations dictate otherwise, Clause 4.3
recommends having the calibrator calibrated every year at most, and the meter
and the filter set verified every two. The same classes and the same 0.5 dB
rule open the other two parts of ISO 16283, in
[Field Insulation Measurement](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-field/#taking-the-measurement-iso-16283-1-clauses-7-and-9).

**Global or element.** They answer different questions and ISO 16283-3
Table 1 says which is preferred for which. The **global** methods give
$D_{2\mathrm{m,n}T}$ (or $D_{2\mathrm{m,n}}$) and "quantify the airborne sound insulation of a
whole façade or a whole building in a specified situation"; that result
"cannot be compared with a sound reduction index obtained in a laboratory",
because it is a property of the building as built. The **element** methods give
$R'_{45°}$ or $R'_\mathrm{tr,s}$ and are the route when the aim is to characterise one
component or to compare against laboratory data. Method 1, element loudspeaker,
is the *preferred* way to estimate an element's apparent index; method 6, global
road traffic, is the *preferred* way to estimate the global insulation of a
façade exposed to road traffic.

**Loudspeaker or traffic.** A loudspeaker is controllable and always
admissible; traffic is free and is what the façade actually faces, but it comes
with conditions. Traffic varies, so indoor and outdoor levels **must be measured
simultaneously**; the average must include at least **50 vehicle pass-bys**; and
the receiving-room background must be at least 10 dB below the measured
equivalent level, because a background correction normally cannot be applied at
all under a varying source (Clause 10.2). The traffic must flow approximately
along a straight line within ±60° of the angle of sight from the façade, with
deviations up to ±15°, and the elevation angle from the point of closest
approach must be under ±40° (Clause 10.3.2). The standard also states the
ceiling on the element traffic method plainly: background noise usually limits
it to elements or façades with $R'_\mathrm{w} < 40$ dB (Clause 10.1 NOTE).

**Where the loudspeaker goes (Clauses 9.3 and 9.4).** Angle of incidence
45° ± 5°, preferably on the ground or as high above it as practice allows, at a
slant distance $r$ from the source to the centre of the specimen of at least
**5 m** ($D > 3.5$ m) for the element method and at least **7 m** ($D > 5$ m)
for the global method: the larger distance because the global method must
illuminate the whole façade evenly, not one element. Choose the position so the
level varies as little as possible over the specimen: the directivity
requirement is under 5 dB of local variation per band over a surface the size
and orientation of the specimen, relaxed to 10 dB (and stated in the report)
where one dimension exceeds 5 m. The source spectrum must be steady and
continuous, with the one-third-octave levels inside an octave differing by no
more than 6 dB at 125 Hz, 5 dB at 250 Hz and 4 dB above; and it must put the
receiving-room level at least 6 dB over the background.

**Where the microphones go.** *Element method* (Clause 9.5.1): the microphone is
fastened directly to the specimen, either with its axis parallel to the surface
(centre of the membrane ≤ 10 mm from it) or normal to it (≤ 3 mm), with a
hemispherical windscreen, securely fixed so it cannot move and so that the
fixing itself does not change what the specimen transmits. Use **3 to 10** fixed
positions, distributed evenly but asymmetrically and never in a regular grid.
The count is not chosen up front: begin with $n = 3$, and if any two positions
differ by more than $n$ dB in any band, add positions, up to 10. A specimen
mounted in a **recess** takes 10 positions outright, and a difference above
10 dB must be stated in the report. *Global method* (Clauses 9.6.1 and 10.4.1):
one microphone $(2.0 \pm 0.2)$ m from the plane of the façade, in front of the
middle of it, at a height of **1.5 m above the receiving-room floor**, or 1 m
in front of a balustrade or similar protrusion where one intervenes. A room
with more than one outside wall, or a very large one, needs a position on each
façade, and a very large façade needs several loudspeaker positions each
complying with Clause 9.4 in full.

None of this is checked by the functions below: `facade_insulation` and
`facade_sound_reduction` energy-average whatever positions they are given
(Clause 9.5.1, Formula (7)) and combine loudspeaker positions with Formula (8)
if you do it yourself.

```python
import numpy as np
from phonometry import building

# Outdoor level 2 m in front of the façade, receiving-room level and T per
# one-third-octave band; surface_level is the microphone on the test element.
l1_2m = np.full(16, 75.0)                              # L1,2m outdoors
l2 = np.full(16, 33.0)                                 # receiving-room L2
t2 = np.full(16, 0.5)                                  # receiving-room T (s)

fac = building.facade_insulation(l1_2m, l2, t2, volume=50.0, area=11.5,
                        surface_level=np.full(16, 78.0), method="loudspeaker")
print(round(float(fac.d_2m[0]), 1))                    # 42.0  D2m = L1,2m - L2
print(round(float(fac.d_2m_nt[0]), 1))                 # 42.0  (= D2m since T = T0)
print(round(float(fac.d_2m_n[0]), 1))                  # 40.0  normalized to A0 = 10 m^2
print(round(float(fac.r_prime[0]), 1))                 # 42.1  R'45deg (loudspeaker, -1.5 dB)

# The road-traffic element method carries the -3 dB all-angle correction instead
tr = building.facade_insulation(l1_2m, l2, t2, volume=50.0, area=11.5,
                       surface_level=np.full(16, 78.0), method="road_traffic")
print(round(float(tr.r_prime[0]), 1))                  # 40.6  R'tr,s (traffic, -3 dB)

# The façade quantity is airborne: rate D2m,nT with the ISO 717-1 engine
print(building.weighted_rating(fac.d_2m_nt).rating)             # 42  Dls,2m,nT,w

fac.plot()   # per-band D2m,nT with D2m, D2m,n and R' overlaid (needs matplotlib)
```

![Field facade insulation of a dwelling under the 45-degree loudspeaker method: the standardized D2m,nT, the raw D2m, the normalized D2m,n and the apparent reduction index R'45 per one-third-octave band, with the Dls,2m,nT,w rating annotated](https://docs.jmrp.io/phonometry/media/facade_field_insulation.svg)

*The four façade quantities of one measurement: the raw $D_{2\mathrm{m}}$, its
standardized and normalized forms, and the element $R'_{45°}$ carrying the
−1.5 dB angle-of-incidence correction. The rating box reads the single
number $D_{\mathrm{ls},2\mathrm{m,n}T,\mathrm{w}}$ obtained by feeding $D_{2\mathrm{m,n}T}$ to the ISO 717-1
engine.*

**Show the code for this figure**

```python
import matplotlib.pyplot as plt
import numpy as np
from phonometry import building

# A dwelling façade, 45° loudspeaker method: outdoor level 2 m in front,
# receiving-room level and T per one-third-octave band.
bands = np.array([100, 125, 160, 200, 250, 315, 400, 500,
                  630, 800, 1000, 1250, 1600, 2000, 2500, 3150], float)
l1_2m = np.array([76.0, 77.0, 78.0, 78.5, 79.0, 79.0, 79.0, 79.0,
                  78.5, 78.0, 77.5, 77.0, 76.5, 76.0, 75.0, 74.0])
d2m = np.array([24.0, 25.5, 27.0, 28.5, 30.0, 31.5, 33.0, 34.5,
                36.0, 37.0, 38.0, 38.5, 39.0, 39.0, 38.5, 38.0])
t2 = np.array([0.65, 0.62, 0.58, 0.55, 0.52, 0.50, 0.49, 0.48,
               0.47, 0.46, 0.45, 0.44, 0.43, 0.43, 0.42, 0.42])
fac = building.facade_insulation(l1_2m, l1_2m - d2m, t2, volume=32.0,
                                 area=10.8, surface_level=l1_2m + 3.0,
                                 method="loudspeaker", frequencies=bands)

# One line — D2m,nT with D2m, D2m,n and R' overlaid per band:
fac.plot()
plt.show()

# By hand, from the result's fields:
w = building.weighted_rating(fac.d_2m_nt)
fig, ax = plt.subplots()
ax.semilogx(bands, fac.d_2m_nt, "-s", label="D2m,nT (standardized)")
ax.semilogx(bands, fac.d_2m, "--o", label="D2m")
ax.semilogx(bands, fac.d_2m_n, ":", label="D2m,n (normalized)")
ax.semilogx(bands, fac.r_prime, "-.", label="R'45°")
ax.set_xlabel("Frequency [Hz]")
ax.set_ylabel("Level difference / reduction index [dB]")
ax.set_title(f"Dls,2m,nT,w = {w.rating} dB  (C={w.c:+d}; Ctr={w.ctr:+d})")
ax.legend()
plt.show()
```

`surface_level`, `area` and `volume` are all optional: with only `l1_2m`, `l2`
and `t2` the function returns `d_2m` and `d_2m_nt`; add `volume` for `d_2m_n`;
add `surface_level` **and** `area` **and** `volume` for `r_prime`. Positions are
energy-averaged with the surface-level formula (Clause 9.5.1); band levels are
assumed already corrected for background noise.

### `facade_insulation()` parameters

| Parameter | Type | Units | Range / default | Notes |
| :--- | :--- | :--- | :--- | :--- |
| `l1_2m` | 1D or 2D array | dB | one/band, or `(positions, bands)` | Level 2 m in front of the façade $L_{1,2\mathrm{m}}$ |
| `l2` | 1D or 2D array | dB | same band count | Receiving-room levels |
| `t2` | 1D array | s | > 0, one per band | Receiving-room reverberation time |
| `area` | float, optional | m² | > 0, with `surface_level`, `volume` | Test-element area $S$ (enables $R'$) |
| `volume` | float, optional | m³ | > 0 | Receiving-room $V$ (enables $D_{2\mathrm{m,n}}$; required for $R'$) |
| `surface_level` | 1D/2D array, optional | dB | same band count | Surface level $L_{1,\mathrm{s}}$ on the element (enables $R'$) |
| `method` | str | — | `'loudspeaker'` (−1.5 dB) / `'road_traffic'` (−3 dB) | Angle-of-incidence correction of $R'$ |
| `t0` | float | s | default `0.5` | Reference reverberation time $T_0$ |
| `frequencies` | 1D array, optional | Hz | — | Band centres carried on the result for plotting |

`facade_insulation()` returns a `FacadeInsulationResult` (`d_2m`, `d_2m_nt`,
`d_2m_n` or `None`, `r_prime` or `None`, `frequencies`); feed any 16-band façade
quantity to `weighted_rating` for its ISO 717-1 single number.

### ISO 16283-3 field façade report (`.report()`)

`FacadeInsulationResult.report(path)` writes the one-page ISO 16283-3 field
façade test report: the standard-basis line, an optional metadata header, the
one-third-octave table beside the measured-versus-shifted-reference curve, the
boxed field rating $D_{2\mathrm{m,n}T,\mathrm{w}}\ (C;\ C_\mathrm{tr})$, the engineering-method
statement, an optional requirement verdict (a level difference passes at or above it) and a
footer. `quantity="d_2m_nt"` (default) reports the standardized level
difference; `"d_2m_n"` the normalized one; `"r_prime"` the apparent sound
reduction index $R'_{45}$. `verbose=True`, `metadata`, `language="es"` and the
`phonometry[report]` extra behave exactly as in the fiches above.

```python
fac.report("D2mnTw_facade.pdf")                        # D2m,nT,w (C; Ctr)
fac.report("Rp45_facade.pdf", quantity="r_prime")      # R'45,w (C; Ctr)
```

**Report:** [Field façade ISO 16283-3 example report (PDF)](https://docs.jmrp.io/phonometry/media/iso16283_facade_example.pdf): One-page field façade sound insulation test report: the metadata header, the one-third-octave D2m,nT table beside the measured-versus-shifted-reference curve, the boxed D2m,nT,w (C; Ctr) field rating, the engineering-method statement and a PASS verdict.

Field façade fiche (FacadeInsulationResult.report), D2m,nT,w (C; Ctr).

## Prediction from the elements (EN 12354-3)

Parts 3 and 4 predict the two directions across the building envelope, both from
the same energy summation of the element **transmission factors**
$\tau = 10^{-R/10}$, area-weighted by $S_i/S$ (a small element or air path enters
through its element-normalized level difference $D_\mathrm{n,e}$ with the reference area
$A_0 = 10\ \text{m}^2$):

$$
R' = -10 \log_{10}\!\Big( \sum_i \tfrac{S_i}{S}\,10^{-R_i/10}
                          + \sum_k \tfrac{A_0}{S}\,10^{-D_{\mathrm{n,e},k}/10} \Big).
$$

**Part 3: outdoor → indoor.** From $R'$ (Formula 10) follow the loudspeaker- and
traffic-referenced indices $R_{45} = R'+1$ and $R_\mathrm{tr,s} = R'$, and the primary
output, the standardized level difference at 2 m (Formula 13)

$$
D_{2\mathrm{m,n}T} = R' + \Delta L_\mathrm{fs} + 10 \log_{10}\frac{V}{6\,T_0\,S}, \qquad T_0 = 0.5\ \text{s},
$$

with the façade-shape term $\Delta L_\mathrm{fs}$ (Annex C; 0 dB for a flat reflecting
façade; `facade_shape_level_difference` looks it up from the Figure C.2 table
for galleries, balconies and terraces, interpolating over the underside
absorption $\alpha_\mathrm{w}$). Single-number ratings reuse EN ISO 717-1
(`weighted_rating`).

![Per-element partial sound reduction indices and the resulting façade apparent reduction R' and standardized level difference D2m,nT for the EN 12354-3 Annex F worked example, the air inlet limiting the low bands](https://docs.jmrp.io/phonometry/media/facade_prediction.svg)

*A composite façade is governed by its worst element, and the drawn curves say
which one that is band by band: the air inlet, a fraction of a square metre,
holds the low bands down while the 6 m² wall is doing nothing to limit the
result. The two output curves sit far below every element curve because the
composite is an energy sum of transmitted power, not an average of indices.*

```python
from phonometry import building

# EN 12354-3 Annex F: an 11.3 m² façade (V = 50 m³, flat so ΔLfs = 0) of a double
# wall, a window, a small skylight and an acoustically-treated air inlet (a Dn,e
# element).
elements = [
    building.FacadeElement(name="wall",     area=6.0, r=[41, 46, 52, 58, 64]),   # octave 125-2000
    building.FacadeElement(name="window",   area=4.5, r=[23, 22, 30, 36, 37]),
    building.FacadeElement(name="skylight", area=0.5, r=[24, 27, 30, 33, 30]),
    building.FacadeElement(name="air inlet", dn_e=[28, 23, 25, 38, 44]),         # small element
]
fac = building.facade_sound_reduction(elements, area=11.3, volume=50.0,
                             frequencies=[125, 250, 500, 1000, 2000], fraction=1)
print(fac.r_tr_s_w, fac.c_tr, fac.d_2m_nt_w)   # 31 -3 33  (R'tr,s,w / Ctr / D2m,nT,w)

fac.plot()   # per-element partial indices with R' and D2m,nT overlaid (needs matplotlib)
```

**Show the code for this figure**

```python
import numpy as np
import matplotlib.pyplot as plt
from phonometry import building

# EN 12354-3 Annex F: an 11.3 m² façade (V = 50 m³, flat so ΔLfs = 0) of a double
# wall, a window, a small skylight and an acoustically-treated air inlet (a Dn,e
# element).
elements = [
    building.FacadeElement(name="wall",     area=6.0, r=[41, 46, 52, 58, 64]),   # octave 125-2000
    building.FacadeElement(name="window",   area=4.5, r=[23, 22, 30, 36, 37]),
    building.FacadeElement(name="skylight", area=0.5, r=[24, 27, 30, 33, 30]),
    building.FacadeElement(name="air inlet", dn_e=[28, 23, 25, 38, 44]),         # small element
]
fac = building.facade_sound_reduction(elements, area=11.3, volume=50.0,
                             frequencies=[125, 250, 500, 1000, 2000], fraction=1)

x = np.arange(5)
fig, ax = plt.subplots(figsize=(9, 5.5))
for name, rp in fac.element_r.items():
    ax.plot(x, rp, "--", alpha=0.6, marker=".", label=f"Rp — {name}")
ax.plot(x, fac.r_prime, "k-", lw=2.5, marker="o", label="R′ (façade)")
ax.plot(x, fac.d_2m_nt, lw=2, marker="s", label="D2m,nT")
ax.set_xticks(x); ax.set_xticklabels([125, 250, 500, 1000, 2000])
ax.set_xlabel("Frequency [Hz]"); ax.set_ylabel("Index / level difference [dB]")
ax.set_title("EN 12354-3 façade sound insulation (Annex F)")
ax.legend(ncol=2); ax.grid(alpha=0.4)
fig.tight_layout(); plt.show()
```

The composite is easier to reason about drawn as areas. The elevation below is
a **second, simpler façade** than the Annex F one above: a 6 m² masonry wall,
a 1.5 m² window and its 0.3 m² roller shutter box, chosen so the tiles are
legible, not a drawing of the prediction just computed.
`plot_facade_elements` tiles it with every element's drawn area equal to its
real area, and a prediction that retained its `elements` redraws *its own*
façade with `fac.plot_geometry()`.

![To-scale elevation of a composite facade: a hatched 6 m2 masonry wall, a 1.5 m2 window and a narrow 0.3 m2 roller shutter box drawn as tiles of a 3.95 m by 1.97 m facade whose drawn areas equal their real areas, each tile labelled with its area and the overall width and height dimensioned](https://docs.jmrp.io/phonometry/media/facade_elevation_geometry.svg)

*The areas the energy sum weighs, to scale: the window holds a quarter of the
wall's area but each square metre of it transmits a hundred times more, so the
small tiles decide $R'$.*

**Show the code for this figure**

```python
import matplotlib.pyplot as plt
from phonometry import building

# The prediction's composite drawn as areas: 6 m2 of wall, a 1.5 m2 window
# and its 0.3 m2 roller shutter box.
elements = [
    building.FacadeElement(name="Masonry wall", area=6.0, r=[50.0] * 5),
    building.FacadeElement(name="Window", area=1.5, r=[30.0] * 5),
    building.FacadeElement(name="Roller shutter box", area=0.3, r=[22.0] * 5),
]
building.plot_facade_elements(elements)
plt.show()

# A prediction retains its elements, so it redraws its own elevation:
#   fac = building.facade_sound_reduction(elements, area=7.8, volume=50.0)
#   fac.plot_geometry()
```

The façade prediction also writes a one-page **prediction** report through a
`report(path)` method, the same layout as the airborne and impact prediction
fiches. `FacadePredictionResult.report()` renders the façade-element table (each
element's weighted partial index $R_\mathrm{p,w}$) beside the per-element / $R'$ /
$D_{2\mathrm{m,n}T}$ plot, the boxed predicted $D_{2\mathrm{m,n}T,\mathrm{w}}$ (with $R'_\mathrm{tr,s,w}$ and
$C_\mathrm{tr}$), the prediction statement and, when a `requirement` is supplied, a
PASS/FAIL verdict
(the level difference passes at or above it). `verbose=True` annexes each
element's share of the transmitted sound energy, which singles out the limiting
element (the air inlet here, not the wall). The report needs the ISO 717-1
single-number ratings, so build the result on the 5 octave or 16 one-third-octave
bands. The applicable `ReportMetadata` fields describe the predicted situation:
`specimen` (the façade element set), `area` (the exposed façade area), the
receiving-room `receiving_volume`, the outdoor/traffic situation in `test_room`,
plus the calculator / laboratory identity fields (`client`, `manufacturer`,
`measurement_standard`, `laboratory`, `operator`, `report_id`, `test_date`), a
free-text façade-shape and model summary in `notes` and the target
$D_{2\mathrm{m,n}T,\mathrm{w}}$ in `requirement`. Metadata, `language="es"` and the `phonometry[report]` extra
behave as in the measurement fiches.

```python
from phonometry import building, ReportMetadata

# EN 12354-3 Annex F facade -> D2m,nT,w = 33 dB (R'tr,s,w = 31, Ctr = -3).
elements = [
    building.FacadeElement(name="Masonry wall", area=6.0, r=[41, 46, 52, 58, 64]),
    building.FacadeElement(name="Glazing",      area=4.5, r=[23, 22, 30, 36, 37]),
    building.FacadeElement(name="Roof light",   area=0.5, r=[24, 27, 30, 33, 30]),
    building.FacadeElement(name="Air inlet", dn_e=[28, 23, 25, 38, 44]),
]
fac = building.facade_sound_reduction(elements, area=11.3, volume=50.0,
                             frequencies=[125, 250, 500, 1000, 2000], fraction=1)
fac.report("D2mnT_prediction.pdf", metadata=ReportMetadata(
    specimen="Masonry wall + window + roof light + air inlet", area=11.3,
    receiving_volume=50.0, requirement=30.0,
    notes="Flat facade, ΔLfs = 0 dB (Annex C)."))          # D2m,nT,w = 33 dB
```

**Report:** [Predicted facade EN 12354-3 example report (PDF)](https://docs.jmrp.io/phonometry/media/iso12354_facade_prediction_example.pdf): One-page predicted facade sound insulation report (EN 12354-3 Annex F): the metadata header (facade element set, exposed area, receiving-room volume, traffic situation), the facade-element table (each element's weighted partial index Rp,w) beside the per-element partial-index and R' / D2m,nT chart, the boxed predicted D2m,nT,w = 33 dB (with R'tr,s,w = 31 and Ctr = -3), the prediction statement noting the model's \~2 dB standard deviation and a PASS verdict against the 30 dB requirement.

Predicted facade fiche (FacadePredictionResult.report), D2m,nT,w.

## Indoor sound radiated outdoors (EN 12354-4)

**Part 4: indoor → outdoor.** The sound power level radiated by a segment
(Formula 2) is $L_W = L_{p,\mathrm{in}} + C_\mathrm{d} - R' + 10 \log_{10}(S/S_0)$ with $S_0 = 1$ m²
and the inside-field diffusivity term $C_\mathrm{d}$ (Annex B; −6 dB ideal diffuse, −5 dB
average industrial). Openings are elements whose "R" is the silencer insertion
loss (a bare opening is 0 dB). The exterior level follows from the simplified
Annex E attenuation $A_\mathrm{tot}$ of a finite radiating side, $L_p = L_W - A_\mathrm{tot}$.

```python
from phonometry import building

# EN 12354-4 Annex G, side 1: a 10×20 m concrete wall segment with a 6×4 m
# industrial door, inside level Lp,in, Cd = -5 dB. The 40 dB cap on R' is an
# Annex G example footnote (field leaks), not part of Formula (2)/(3): pass it
# explicitly to reproduce Annex G; by default no cap is applied.
bands = [63, 125, 250, 500, 1000, 2000, 4000, 8000]
seg = building.radiated_sound_power(
    [building.FacadeElement(name="wall", area=176.0, r=[32, 36, 36, 33, 39, 49, 57, 63]),
     building.FacadeElement(name="door", area=24.0,  r=[21, 23, 28, 30, 30, 30, 30, 30])],
    lp_in=[70, 74, 76, 72, 70, 67, 62, 57], area=200.0, c_d=-5.0,
    r_prime_cap=40.0, octave_bands=bands)
print(round(seg.l_w[0], 1), round(seg.l_w[1], 1))     # 59.8 61.2  (LW at 63/125 Hz)

# The exterior level is computed for a whole SIDE, not for the segment above.
# Annex G Table G.8 sums side 1's three segments to LWA = 62.9 dB(A) over the
# full 10 x 60 m side; Table G.9 then reads the level 5 m in front of its
# centre. (`seg.l_w_dba` is 58.2 dB(A): one segment of that side, not the side.)
a_tot = building.outdoor_attenuation(width_m=60.0, height_m=10.0, distance=5.0)
print(round(a_tot, 1), round(building.outdoor_level(62.9, a_tot), 1))   # 26.3 36.6

seg.plot()   # radiated LW per octave with the A-weighted LWA line (needs matplotlib)
```

![Radiated sound power level per octave band of the EN 12354-4 Annex G wall segment with an industrial door, with the A-weighted single number drawn as a dashed line across the bars](https://docs.jmrp.io/phonometry/media/radiated_power_outdoor.svg)

*The Annex G side-1 segment: the wall dominates the area but the door's
weaker $R$ carries the radiated power, so the octave spectrum stays flat
where the wall alone would fall. The dashed line is the A-weighted single
number $L_{W\mathrm{A}}$ formed from the octave bands.*

**Show the code for this figure**

```python
import matplotlib.pyplot as plt
import numpy as np
from phonometry import building

# EN 12354-4 Annex G, side 1: a 10×20 m concrete wall segment with a 6×4 m
# industrial door, inside level Lp,in, Cd = -5 dB.
bands = [63, 125, 250, 500, 1000, 2000, 4000, 8000]
seg = building.radiated_sound_power(
    [building.FacadeElement(name="wall", area=176.0, r=[32, 36, 36, 33, 39, 49, 57, 63]),
     building.FacadeElement(name="door", area=24.0,  r=[21, 23, 28, 30, 30, 30, 30, 30])],
    lp_in=[70, 74, 76, 72, 70, 67, 62, 57], area=200.0, c_d=-5.0,
    r_prime_cap=40.0, octave_bands=bands)

# One line — LW per octave with the A-weighted LWA line:
seg.plot()
plt.show()

# By hand, from the result's fields:
x = np.arange(len(bands))
fig, ax = plt.subplots()
ax.bar(x, seg.l_w, label="radiated LW per octave")
ax.axhline(seg.l_w_dba, ls="--", color="tab:red",
           label=f"LWA = {seg.l_w_dba:.1f} dB(A)")
ax.set_xticks(x, [str(b) for b in bands])
ax.set_xlabel("Frequency [Hz]")
ax.set_ylabel("Radiated sound power level [dB re 1 pW]")
ax.set_title("EN 12354-4 radiated sound power (Annex G)")
ax.legend()
plt.show()
```

> **Worked-example note.** The 2000 worked examples carry small internal rounding
> inconsistencies at the higher octave bands (Part 3's printed $R'$ disagrees with
> its own per-element partial indices at 1 k/2 k; Part 4's $R'$ rows above 500 Hz
> disagree with its Table G.2 inputs). The implementation is faithful to the
> formulas: it reproduces the low bands, every single-number rating and the whole
> Annex E propagation exactly.

### `FacadeElement` / `facade_sound_reduction()` / `radiated_sound_power()` parameters

| Parameter | Type | Units | Range / default | Notes |
| :--- | :--- | :--- | :--- | :--- |
| `FacadeElement.area` | float | m² | > 0 for `r` / `insertion_loss` | Element area $S_i$ (ignored for `dn_e`) |
| `FacadeElement.r` / `dn_e` / `insertion_loss` | float or seq | dB | give exactly one | Area element $R_i$ / small-element $D_\mathrm{n,e}$ / opening insertion loss |
| `facade_sound_reduction(area)` | float | m² | > 0 | Total façade area $S$ |
| `facade_sound_reduction(volume)` | float | m³ | > 0 | Receiving-room volume $V$ (Formula 13) |
| `facade_sound_reduction(delta_l_fs)` | float | dB | default `0` | Façade-shape term $\Delta L_\mathrm{fs}$ (Annex C; look it up with `facade_shape_level_difference`) |
| `radiated_sound_power(lp_in)` | float or seq | dB | — | Inside level $L_{p,\mathrm{in}}$ per band |
| `radiated_sound_power(c_d)` | float | dB | default `-6` | Diffusivity term $C_\mathrm{d}$ (Annex B) |
| `radiated_sound_power(r_prime_cap)` | float | dB | default `None` (off) | Optional field cap on $R'$, an Annex G example footnote (it uses 40 dB), not part of Formula (2)/(3) |
| `radiated_sound_power(octave_bands)` | seq of int | Hz | default `None` | Octave centres matching the bands; enables the A-weighted $L_{W\mathrm{A}}$ |
| `facade_sound_reduction(frequencies)` | seq | Hz | default `None`; length = band count | Band centres carried on the result for plotting |
| `facade_sound_reduction(fraction)` | int or `None` | — | `1` (octaves) / `3` (one-third octaves) / `None` | Which band set the input is in; `None` infers it from the count, as on the [ratings page](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-ratings/) |
| `outdoor_attenuation(width, height, distance)` | float | m | > 0 | Finite radiating side and reception distance (Annex E) |
| `outdoor_level(l_w, attenuation)` | float or seq | dB | broadcast-compatible | Exterior $L_p$ from one or more sides (Formula E.1) |

`facade_sound_reduction()` returns a `FacadePredictionResult` (`r_prime`, `r_45`,
`r_tr_s`, `d_2m_nt`, `element_r`, and the `r_tr_s_w` / `d_2m_nt_w` / `c_tr` single
numbers); `radiated_sound_power()` a `RadiatedPowerResult` (`l_w`, `r_prime`,
`l_w_dba`). Both expose `.plot()`.

## Measurement against prediction

The two halves of this guide describe the same wall from opposite ends of its
life: EN 12354-3 before it exists, ISO 16283-3 once it does. They are worth
comparing because they end on the same quantity, the standardized level
difference at 2 m, and because the geometry term that carries the prediction
from an index to a level difference is the measurement's own normalisation
written another way. Start from the measured definition, replace $D_{2\mathrm{m}}$ by
the apparent index it implies and substitute the Sabine absorption area
$A = 0.16\ V/T$:

$$
D_{2\mathrm{m,n}T} = D_{2\mathrm{m}} + 10 \log_{10}\frac{T}{T_0}
          = R' + 10 \log_{10}\frac{A}{S} + 10 \log_{10}\frac{T}{T_0}
          = R' + 10 \log_{10}\frac{0.16\ V}{S\ T_0}.
$$

The reverberation time cancels: what survives is the room volume against the
façade area. EN 12354-3 Formula (13) writes that same term as
$10 \log_{10} V/(6\,T_0\,S)$, which is the Sabine constant carried as
$1/6 = 0.167$ instead of the $0.16$ of the measurement standard. The two
therefore differ by a fixed $10 \log_{10}(0.167/0.16) = 0.18$ dB in every
band, of any room, and by nothing else. (The 2017 revision of the
prediction standard harmonised its constant to 0.16, so the gap is
specific to the EN 12354-3:2000 edition the module implements.)

That is small enough to check numerically. Take the Annex F façade predicted
above, put it in a receiving room of 50 m³ with $T = T_0$, and reconstruct the
receiving-room level a measurement would have found:

```python
import numpy as np
from phonometry import building

# The Annex F façade predicted above, now "measured" in a receiving room of
# 50 m3 whose reverberation time is exactly T0 = 0.5 s.
elements = [
    building.FacadeElement(name="wall",     area=6.0, r=[41, 46, 52, 58, 64]),
    building.FacadeElement(name="window",   area=4.5, r=[23, 22, 30, 36, 37]),
    building.FacadeElement(name="skylight", area=0.5, r=[24, 27, 30, 33, 30]),
    building.FacadeElement(name="air inlet", dn_e=[28, 23, 25, 38, 44]),
]
volume, area = 50.0, 11.3
fac = building.facade_sound_reduction(elements, area=area, volume=volume,
                             frequencies=[125, 250, 500, 1000, 2000], fraction=1)

# What the field measurement would read: A = 0.16 V / T is the Sabine area of
# ISO 16283-3, and R' - 10 lg(S/A) is the level difference it implies.
t2 = np.full(5, 0.5)
l1_2m = np.full(5, 75.0)
l2 = l1_2m - (fac.r_prime - 10 * np.log10(area / (0.16 * volume / t2)))
meas = building.facade_insulation(l1_2m, l2, t2, volume=volume)

print(np.round(meas.d_2m_nt - fac.d_2m_nt, 2))   # [-0.18 -0.18 -0.18 -0.18 -0.18]
print(building.weighted_rating(meas.d_2m_nt).rating, fac.d_2m_nt_w)   # 33 33
```

The two curves run parallel to the second decimal and both rate at
$D_{2\mathrm{m,n}T,\mathrm{w}} = 33$ dB. Everything that separates a real prediction from a real
measurement therefore sits in the inputs, not in the formulas:

* **The element indices.** The prediction consumes laboratory $R$ values
  measured with flanking suppressed and with the element mounted as the
  laboratory mounts it. The built façade adds its perimeter seals, its
  roller-shutter box, the joint between frame and reveal and whatever the site
  did to them.
* **The shape term.** $\Delta L_\mathrm{fs}$ is 0 dB only for a flat reflecting
  façade. A balcony, a gallery or a terrace changes the field at the 2 m
  position, and a measurement takes that change as it is while a prediction
  has to look it up in Annex C.
* **The source.** ISO 16283-3 labels the loudspeaker result $D_{\mathrm{ls},2\mathrm{m,n}T}$
  precisely because a loudspeaker at 45° is not road traffic; its element
  index carries a −1.5 dB angle correction where the all-angle traffic method
  carries −3 dB. EN 12354-3 predicts the traffic-referenced $R_\mathrm{tr,s} = R'$
  directly and offers $R_{45} = R' + 1$ for the loudspeaker comparison.
* **The receiving room.** The prediction assumes the design volume and a
  diffuse receiving field. The measurement takes the room furnished, and the
  diffuse assumption behind $A = 0.16\ V/T$ weakens exactly in the small rooms
  and low bands where façade requirements bite.

The practical reading: size the glazing and the air inlets with the
prediction, accept the finished façade with the measurement, and do not treat
a couple of decibels between the two as an error in either. A prediction built
on laboratory indices and a field measurement of the finished envelope each
carry an uncertainty of that order on their own; the [ISO 12999-1 uncertainty section](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-field/#measurement-uncertainty-iso-12999-1) puts a number on the
measurement half. And when the two disagree by much more than that, the
per-element share of the transmitted energy in the prediction report is the
place to look first: on this façade it is the air inlet, not the wall, that
decides the result.

## What this guide covers

- **Covered.** ISO 16283-3:2016 (the field façade quantities $D_{2\mathrm{m}}$, $D_{2\mathrm{m,n}T}$,
  $D_{2\mathrm{m,n}}$, $R'_{45°}$ and $R'_\mathrm{tr,s}$ of the loudspeaker and road-traffic
  element methods, with the position averaging of Clause 9.5.1 and the Clause 14
  test report) via `building.facade_insulation`; EN 12354-3:2000 (the
  transmission-factor summation of Formula 10, the $R_{45}$ and $R_\mathrm{tr,s}$
  indices, the $D_{2\mathrm{m,n}T}$ of Formula 13 and the Annex C façade-shape term) via
  `building.facade_sound_reduction` and
  `building.facade_shape_level_difference`; and EN 12354-4:2000 (the radiated
  sound power of Formula 2 with the Annex B diffusivity term, and the Annex E
  attenuation of a finite radiating side) via `building.radiated_sound_power`,
  `building.outdoor_attenuation` and `building.outdoor_level`. The single-number
  ratings reuse the verified ISO 717-1 engine, and both the measurement and the
  prediction write their one-page fiche through `.report()`.

- **Not covered.** The band levels handed to `facade_insulation` are assumed already corrected
  for background noise, and ISO 16283-3's own procedural requirements (the ones
  set out under [Choosing and running the
  method](#choosing-and-running-the-method-clauses-9-and-10)) are documented
  but not checked: nothing verifies the 45° ± 5° incidence, the 5 m / 7 m slant
  distances, the 3→10 microphone escalation and its recess special case, the 50
  pass-bys and the simultaneity of a traffic measurement, or that the method
  chosen matches the question asked. The low-frequency corner procedure of
  Clause 7.3 (the façade counterpart of the [ISO 16283-1
  procedure](https://docs.jmrp.io/phonometry/buildings/insulation/low-frequency-procedure/),
  triggered by the same 25 m³ receiving-room threshold) is implemented and
  reached through `facade_insulation`'s `low_frequency=` argument, for the
  loudspeaker methods Clause 6 confines it to; the railway and aircraft methods
  of Annex E are not. The EN
  12354-3 and EN 12354-4 worked examples of the 2000 editions carry small
  internal rounding inconsistencies at the higher octave bands, noted where they
  appear: the implementation follows the formulas rather than the printed rows.

## See also

- [Field Insulation Measurement (ISO 16283)](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-field/):
  the airborne and impact field measurements of the internal partitions, and
  the ISO 12999-1 uncertainty that qualifies every field rating.
- [Predicting Sound Insulation (EN 12354)](https://docs.jmrp.io/phonometry/buildings/design/insulation-prediction/):
  the airborne and impact flanking models of EN 12354-1/2 that share the
  transmission-factor summation used here.
- [Insulation Ratings (ISO 717)](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-ratings/): the
  reference-curve engine behind the $D_{2\mathrm{m,n}T,\mathrm{w}}$, $R'_{45,\mathrm{w}}$ and
  $R'_\mathrm{tr,s,w}$ single numbers.
- [Sound Insulation Survey Method (ISO 10052)](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-survey/):
  the octave-band façade quantity of the quick survey method.
- [Predicting Panel Sound Insulation](https://docs.jmrp.io/phonometry/buildings/design/panel-sound-insulation/):
  the mass law, coincidence dip and aperture models behind the element $R$
  values the prediction consumes.
- [Outdoor Sound Propagation](https://docs.jmrp.io/phonometry/environment/propagation/outdoor-propagation/): the
  propagation from the source to the 2 m position in front of the façade, and
  onwards from the radiating side of EN 12354-4.
- [Theory](https://docs.jmrp.io/phonometry/reference/theory/rooms-buildings/): the reference-curve
  derivation behind the weighted single-number ratings.
- API reference: [`building.prediction.facade`](https://docs.jmrp.io/phonometry/reference/api/building/facade/) and [`building.measurement.insulation`](https://docs.jmrp.io/phonometry/reference/api/building/insulation/).

## References

- **European Committee for Standardization. (2000).** [*Building acoustics — Estimation of acoustic performance of buildings from the performance of elements — Part 3: Airborne sound insulation against outdoor sound*](https://knowledge.bsigroup.com/products/building-acoustics-estimation-of-acoustic-performance-in-buildings-from-the-performance-of-elements-airborne-sound-insulation-against-outdoor-sound) (EN 12354-3:2000).
  The façade sound insulation prediction (Annex F worked example). The linked catalogue record is the BSI Knowledge page for BS EN 12354-3:2000.
- **European Committee for Standardization. (2000).** [*Building acoustics — Estimation of acoustic performance of buildings from the performance of elements — Part 4: Transmission of indoor sound to the outside*](https://knowledge.bsigroup.com/products/building-acoustics-estimation-of-acoustic-performance-in-buildings-from-the-performance-of-elements-transmission-of-indoor-sound-to-the-outside) (EN 12354-4:2000).
  The outdoor-radiation prediction (Annex G worked example). The linked catalogue record is the BSI Knowledge page for BS EN 12354-4:2000.
- **Hopkins, C. (2007).** *Sound insulation*. Butterworth-Heinemann. https://doi.org/10.4324/9780080550473
  The transmission physics of the envelope behind both the measured and the predicted façade quantities. ISBN 978-0-7506-6526-1.
- **International Organization for Standardization. (2016).** [*Acoustics — Field measurement of sound insulation in buildings and of building elements — Part 3: Façade sound insulation*](https://www.iso.org/standard/59748.html) (ISO 16283-3:2016).
  The field façade element and global methods, with the angle-of-incidence corrections.
- **International Organization for Standardization. (2020).** [*Acoustics — Rating of sound insulation in buildings and of building elements — Part 1: Airborne sound insulation*](https://www.iso.org/standard/77435.html) (ISO 717-1:2020).
  The reference-curve rating that turns every façade quantity on this page into its single number (Annex F of ISO 16283-3, EN ISO 717-1 for the prediction).
