# Laboratory Flanking Transmission (ISO 10848)

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

Canonical: https://docs.jmrp.io/phonometry/buildings/insulation/flanking-lab/
Language: en
Alternate: https://docs.jmrp.io/phonometry/es/buildings/insulation/flanking-lab/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: 2026-11-16T21:36:14Z
Standards: ISO 10848-1:2006; ISO 10848-1:2017; ISO 10848-2:2006; ISO 10848-3:2006; ISO 10848-4:2010; ISO 140-9:1985; ASTM E413-22
Verified: 5 conformance checks (ISO 10848-1:2006; ISO 140-9:1985; ASTM E413-22) https://docs.jmrp.io/phonometry/reference/conformance/#room-building-acoustics
License: MIT (https://github.com/jmrplens/phonometry/blob/main/LICENSE)
Summary: Flanking transmission measured in the laboratory by ISO 10848: Kij, Dn,f, Ln,f, the Part 4 modal-overlap checks and the ceiling plenum path with Dn,c.

---

EN 12354 predicts a building from its junctions, and ISO 10848 is how a
junction gets its number. Two elements rigidly joined along a common edge are
built into a laboratory rig whose rooms are structurally isolated from each
other and from the surrounding structure, so the junction under test is the
only path between them; shakers and tapping machines drive one element, and the
velocity level differences across the junction condense into the
**vibration reduction index** $K_{ij}$, together with the overall flanking
descriptors $D_\mathrm{n,f}$ (airborne) and $L_\mathrm{n,f}$ (impact). This guide covers
that measurement chain: the direction-averaged velocity level difference,
the equivalent absorption lengths, the SEA validity checks of Part 4 and
the three accredited-report-style fiches. The empirical junction values a prediction
falls back on when no measurement exists live in
[Predicting Sound Insulation (EN 12354)](https://docs.jmrp.io/phonometry/buildings/design/insulation-prediction/);
the wave-theory $K_{ij}$ of ideal plate junctions is derived in
[Bending-wave transmission at plate junctions](https://docs.jmrp.io/phonometry/vibration/structural/junction-transmission/).

## The vibration reduction index and the flanking descriptors

ISO 10848:2006/2010 is the laboratory method that **measures** the junction
**vibration reduction index** $K_{ij}$ that the [EN 12354 prediction](https://docs.jmrp.io/phonometry/buildings/design/insulation-prediction/) takes
as an input, together with the overall flanking descriptors $D_\mathrm{n,f}$
(airborne) and $L_\mathrm{n,f}$ (impact). It is the measurement counterpart of the
empirical `junction_vibration_reduction()` of that prediction.

**Vibration reduction index (Formula (13)).**
$K_{ij} = \overline{D}_{\mathrm{v},ij} + 10\log_{10}\!\big(l_{ij} / \sqrt{a_i a_j}\big)$ dB, from
the direction-averaged velocity level difference
$\overline{D}_{\mathrm{v},ij} = \tfrac{1}{2}(D_{\mathrm{v},ij} + D_{\mathrm{v},ji})$ (Formula (11), which
makes $K_{ij}$ symmetric), the common-edge junction length $l_{ij}$ and the
**equivalent absorption lengths** $a_j = 2.2\pi^2 S_j /(T_{\mathrm{s},j} c_0)\sqrt{f_\mathrm{ref}/f}$
(Formula (12), $f_\mathrm{ref} = 1000$ Hz). For lightweight well-damped elements
$a_j = S_j / l_0$ ($l_0 = 1$ m) and Formula (13) reduces to the simplified
Formula (14), which needs no $T_\mathrm{s}$ at all.

![ISO 10848 junction measurement rigs for an L-junction and a T-junction of 140 to 200 mm concrete plates: a shaker or hammer excites element i, accelerometers on elements i and j read the velocity level difference Dv,ij, and the junction length of at least 2.3 m runs along the highlighted corner line](https://docs.jmrp.io/phonometry/media/diagram_junction_rig.svg)

Every symbol in Formula (13) is a length or an area in that drawing. The
highlighted corner line is $l_{ij}$, the common edge of the two elements; $S_i$
and $S_j$ are the areas of the two plates, which enter through their equivalent
absorption lengths $a_i$ and $a_j$; and $\overline{D}_{\mathrm{v},ij}$ is what the pair
of accelerometers reads, once with element $i$ driven and once with element $j$
driven. The $10\log_{10}(l_{ij}/\sqrt{a_i a_j})$ term is what makes the index a
property of the *junction* rather than of the two particular plates that were
built into the rig, which is the entire reason EN 12354 can consume it.

### The structural reverberation time, and where it comes from

$T_{\mathrm{s},j}$ is the second most influential input on this page and the one most
often assumed. It is **not** the room's reverberation time: it is the decay of
*bending-wave* energy on the element itself, measured with accelerometers, and
it is what carries the element's total loss factor into the absorption length,
$\eta_\text{total} = 2.2/(f T_\mathrm{s})$ (ISO 10848-1:2017, Formula (19)). Because
$a_j \propto 1/T_{\mathrm{s},j}$ and $K_{ij}$ carries $-10\log_{10}\sqrt{a_i a_j}$, a
$T_\mathrm{s}$ wrong by a factor of two moves $K_{ij}$ by 1.5 dB in each element, and it
moves it in the direction that makes a junction look better than it is when
$T_\mathrm{s}$ is overestimated.

Measure it as ISO 10848-1:2017 Clause 7.3 prescribes: point excitation by an
electrodynamic shaker with an MLS or swept-sine signal (the preferred
laboratory method, per ISO 18233) or by an impact hammer, the integrated
impulse response of ISO 3382-2 with backward integration, acceleration rather
than velocity so no integration smears the decay, at least three excitation
positions and three measurement positions per excitation, and the minimum
separations of Clause 7.3.3 (0.5 m from the element boundaries, 1 m between
the excitation and the measurement positions, 0.5 m between measurement
positions). Evaluate over 5 dB to 15 dB from a decay that starts at least
35 dB above the background (Clause 7.3.4); heavy elements coupled to other
heavy elements give a short straight decay followed by a curve as energy
returns from the rest of the structure, so the short evaluation range is not a
convenience but a requirement of the physics. Clause 7.3.5 sets the floor
below which a forward analysis stops being reliable: $T_\mathrm{s} > 70/f$ and
$T_\mathrm{s} > 2T_\text{det}$, with $T_\text{det}$ the reverberation time of the
averaging detector.

Expect a few tenths of a second on a heavy concrete plate built into a
building, falling with frequency, and much less on a well-damped lightweight
element, which is why the lightweight route of Formula (14) drops $T_\mathrm{s}$
entirely. The loss-factor identity is the sanity check on a measured value: at
500 Hz a $T_\mathrm{s}$ of 0.35 s gives $\eta = 2.2/(500 \times 0.35) = 0.013$, a
plausible total loss factor for an in-situ heavy plate, while an $\eta$ coming
out above about 0.1 or below 0.005 says the decay was misread rather than that
the plate is unusual.

**Overall descriptors.** $D_\mathrm{n,f} = L_1 - L_2 - 10\log_{10}(A/A_0)$ (Formula (4),
airborne) and $L_\mathrm{n,f} = L_2 + 10\log_{10}(A/A_0)$ (Formula (5), tapping machine),
$A_0 = 10\ \text{m}^2$; their $D_\mathrm{n,f,w}$ / $L_\mathrm{n,f,w}$ single numbers reuse the
ISO 717 rating engines. The single-number $\overline{K}_{ij}$ is the arithmetic
mean over 200–1250 Hz for one-third-octave bands, or over 125–1000 Hz for
octave bands (Annex A).

![ISO 10848 junction vibration reduction index Kij rising across one-third-octave bands from 100 Hz to 5000 Hz for a rigid T-junction of two heavy walls, with the single-number mean Kij over 200-1250 Hz drawn as a dashed line](https://docs.jmrp.io/phonometry/media/flanking_transmission.svg)

*$K_{ij}$ climbs steadily with frequency for a rigid junction of heavy walls,
so no single band represents it. The dashed line is the arithmetic mean over
200–1250 Hz, the range Annex A prescribes, and it is that mean (not the curve)
that an EN 12354 prediction consumes.*

**Show the code for this figure**

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

freqs = [100, 125, 160, 200, 250, 315, 400, 500, 630,
         800, 1000, 1250, 1600, 2000, 2500, 3150, 4000, 5000]
# Direction-averaged velocity level difference of a rigid T-junction (dB):
dv = np.array([4.5, 4.8, 5.2, 5.6, 6.0, 6.5, 7.0, 7.6, 8.1, 8.7,
               9.2, 9.8, 10.3, 10.9, 11.4, 11.9, 12.3, 12.7])
res = building.vibration_reduction_index(
    dv, junction_length=4.0, area_i=12.0, area_j=10.0, frequencies=freqs,
    structural_reverberation_time_i=0.35, structural_reverberation_time_j=0.40,
)
print(res.single_number)   # mean Kij over 200-1250 Hz (Annex A)
res.plot()
plt.show()
```

**Validity.** $K_{ij}$ rests on a statistical-energy-analysis simplification:
`check_strong_coupling()` checks the Formula (15) inequality and, for the
heavy junctions of Part 4, `modal_density()`, `band_mode_count()` and
`modal_overlap_factor()` (Formulae (5)/(4)/(6)) quantify where the mode count
is too low for $K_{ij}$ to be reliable. Pass the per-band modal overlap factor
to `vibration_reduction_index(..., modal_overlap=M)`: bands with $M < 0.25$
are flagged in `result.bracketed` and excluded from the single-number
$\overline{K}_{ij}$, as Part 4 Clause 9 requires. Because ISO 10848 contains no
worked numeric example, conformance is anchored on closed-form identities
(simplified $K_{ij}$, $a_j$ at $f_\mathrm{ref}$, $\eta$).

![ISO 10848-1 Formula (15) across one-third-octave bands from 100 Hz to 5000 Hz for a 140 mm concrete wall excited and the 200 mm wall it meets: the one-way velocity level difference Dv,ij rising from 4.2 dB to 12.6 dB against the 6.1 dB bound, with the strong-coupling side under the bound shaded and the five lowest bands, where the inequality does not hold, marked](https://docs.jmrp.io/phonometry/media/strong_coupling_check.svg)

*Formula (15) asks a 140 mm concrete wall excited ($m_i$ = 322 kg/m², $f_{\mathrm{c}i}$ = 133 Hz) and the 200 mm wall it meets ($m_j$ = 460 kg/m², $f_{\mathrm{c}j}$ = 93 Hz) for a one-way $D_{\mathrm{v},ij}$ of at least 6.1 dB. The measured difference falls short of it in the five lowest bands: there the two walls are strongly coupled, they cannot be taken as separate SEA subsystems, and the measured $K_{ij}$ may not be relevant (4.3.3). From 315 Hz up the inequality holds, so the coupling is weak enough for the SEA picture $K_{ij}$ rests on. The bound moves when $i$ and $j$ swap, so each direction is judged on its own.*

**Show the code for this figure**

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

freqs = [100, 125, 160, 200, 250, 315, 400, 500, 630,
         800, 1000, 1250, 1600, 2000, 2500, 3150, 4000, 5000]
# One-way velocity level difference, the 140 mm wall i excited (dB):
dv_ij = [4.2, 4.7, 5.1, 5.5, 5.9, 6.4, 6.8, 7.4, 7.9,
         8.5, 9.0, 9.6, 10.1, 10.7, 11.2, 11.7, 12.1, 12.6]
check = building.check_strong_coupling(
    dv_ij, mass_i=322.0, mass_j=460.0,
    critical_frequency_i=133.0, critical_frequency_j=93.0, frequencies=freqs,
)
print(round(check.min_velocity_level_difference, 1))   # 6.1 dB
print(int(check.holds.sum()), check.passes)            # 13 False
check.plot()
plt.show()
```

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

freqs = [200, 250, 315, 400, 500, 630, 800, 1000, 1250]
lij, s_i, s_j = 4.0, 12.0, 10.0     # junction length (m), element areas (m^2)
ts = np.linspace(0.30, 0.10, 9)     # structural reverberation time Ts (s)
dv_ij = [5.6, 6.0, 6.5, 7.0, 7.6, 8.1, 8.7, 9.2, 9.8]    # element i excited (dB)
dv_ji = [6.4, 6.8, 7.3, 7.8, 8.4, 8.9, 9.5, 10.0, 10.6]  # element j excited (dB)

# Kij from both excitation directions (symmetric via the direction average):
dbar = building.direction_averaged_level_difference(dv_ij, dv_ji)
res = building.vibration_reduction_index(dbar, lij, s_i, s_j, frequencies=freqs,
                                structural_reverberation_time_i=ts,
                                structural_reverberation_time_j=ts)
res.k_ij           # Kij per band (Formula (13))
res.single_number  # mean Kij over 200-1250 Hz, or None without the band set
res.octave_bands() # Kij in octave bands (its single number averages 125-1000 Hz)

# Overall airborne flanking descriptor and a Part-4 modal-overlap validity check:
dnf = building.normalized_flanking_level_difference(np.full(9, 75.0), np.full(9, 42.0),
                                           absorption_area=np.full(9, 12.0))
m = building.modal_overlap_factor(s_i, critical_frequency=85.0,
                         structural_reverberation_time=ts)
res_m = building.vibration_reduction_index(dbar, lij, s_i, s_j, frequencies=freqs,
                                  modal_overlap=m)   # M < 0.25 bands bracketed
res_m.bracketed    # per-band flags; bracketed bands leave the single number

# With 16 one-third-octave (or 5 octave) bands, dnf.plot() draws Dn,f vs the
# shifted ISO 717-1 reference with Dn,f,w annotated (needs matplotlib):
```

![Normalized flanking level difference Dn,f per one-third-octave band against the shifted ISO 717-1 reference curve, with the unfavourable deviations shaded and the Dn,f,w rating annotated](https://docs.jmrp.io/phonometry/media/flanking_level_difference.svg)

*The overall flanking descriptor $D_\mathrm{n,f}$ is an airborne quantity, so its
single number $D_\mathrm{n,f,w}$ comes from the unchanged ISO 717-1 engine; it
drops straight into the EN 12354-1 model as the flanking-path datum of the
tested junction (the impact counterpart $L_\mathrm{n,f}$ rates per ISO 717-2 the
same way).*

**Show the code for this figure**

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

# A lightweight junction in the laboratory: source-room level, receiving-room
# level over the flanking path, and the receiving-room absorption area.
l1 = np.full(16, 80.0)
dnf_target = np.array([48, 49, 50, 51, 52, 54, 55, 57,
                       58, 59, 60, 61, 62, 63, 64, 65], dtype=float)
dnf = building.normalized_flanking_level_difference(
    l1, l1 - dnf_target, absorption_area=np.full(16, 10.0)
)

# One line — Dn,f vs the shifted ISO 717-1 reference:
dnf.plot()
plt.show()

# By hand, from the rating the result carries:
w = dnf.rating
fig, ax = plt.subplots()
ax.semilogx(w.frequencies, dnf.d_n_f, "o-", label="Dn,f (flanking)")
ax.semilogx(w.frequencies, w.shifted_reference, "s--",
            label="shifted reference")
ax.fill_between(w.frequencies, w.measured, w.shifted_reference,
                where=w.measured < w.shifted_reference, interpolate=True,
                alpha=0.3, label="unfavourable deviations")
ax.set_xlabel("Frequency [Hz]")
ax.set_ylabel("Normalized flanking level difference [dB]")
ax.set_title(f"Dn,f,w = {w.rating} dB  (C={w.c:+d}; Ctr={w.ctr:+d})")
ax.legend()
plt.show()
```

## Making the measurement (ISO 10848-1:2017, Clauses 7 and 9)

The formulae above are indifferent to how their inputs were obtained. The
standard is not, and one of its requirements decides whether the number means
anything at all.

**Shielding comes first (Clause 9).** A junction of three or four elements has
two elements facing each room, and a receiving-room level fed by *all* of them
is not attributable to any one path. So the elements not under test are
shielded, successively, and the rule differs by quantity: measuring $D_\mathrm{n,f}$
for transmission through $i$ and $j$, shield the other element **in both
rooms**; measuring $L_\mathrm{n,f}$, shield the other element **in the receiving
room**; measuring $K_{ij}$ with structure-borne excitation, **no shielding is
needed**, because the excitation is applied to one element rather than to a
sound field. Radiated sound measured by intensity to ISO 15186-1 also needs no
receiving-room shielding, because the enclosing surface does the same job.

The shield has to be good enough to be worth building. Where the three or four
elements of the junction are identical, a $\Delta R$ of 10 dB in every band
suffices. In general, excite element 1 and then element 2 with the same kind of
excitation, read the average velocity level of the receiving element $j$ in
each case, and require
$\Delta R_\text{min} = 10 - L_{\mathrm{v},1j} + L_{\mathrm{v},2j}$ when
$L_{\mathrm{v},1j} - L_{\mathrm{v},2j} \le 10$ dB, and nothing at all when the difference already
exceeds 10 dB (Formulae (23) and (24)). Heavy shielding on a light element can
change the very transmission being measured and is to be avoided. **A $K_{ij}$
or $D_\mathrm{n,f}$ quoted without a statement of how the other elements were
shielded, and of how that shielding was verified, is a room-pair number with a
path label attached to it.** And it flows straight into an EN 12354
prediction, where the error is unbounded in direction.

**Accelerometers (Clause 7.2.3).** Mounted directly on the surface of the test
element, stiff in the direction normal to it. Bees or petroleum wax is
convenient, but a weak fixing costs high-frequency accuracy; where the surface
texture defeats wax, cement a small metal washer to it and wax or stud onto
that (ISO 5348). The accelerometer mass must be small enough not to load the
plate, which Formula (16) makes explicit:
$m_\text{acc} < 1/(2\pi f Y_\mathrm{dp})$, with the driving-point mobility of a
thin isotropic plate estimated by Formula (17) as
$Y_\mathrm{dp} = 1/(8\sqrt{B\rho_\mathrm{s}}) = 1/(2.3\,\rho_\mathrm{s} c_\mathrm{L} h)$. The inequality
bites at the **top** of the band range, so check it at the highest band, not
the lowest.

**Excitation (Clause 7.2.4).** Steady-state, from an electrodynamic shaker
driven with noise or from a tapping machine (adapted for walls, with the
impacts mechanically controlled to repeat), or transient, from single or
manually repeated hammer blows, in which case $D_{\mathrm{v},ij}$ must be measured on
both elements **simultaneously**. Manual multiple hits are spread over 1 m² to
2 m² at 1 Hz to 2 Hz over 20 s to 30 s. In every band the receiving element's
velocity level must exceed the background by at least 10 dB; below that, the
ISO 10140-4 correction applies and must not exceed 1.3 dB. Take care that the
source's own airborne noise, or radiation from the excited element, does not
excite the others. That is the same failure the shielding exists to prevent,
arriving by a different route.

**Positions (Clause 7.2.5).** At least **four** excitation positions on a
Type A (heavy, homogeneous) source element and **six** on a Type B one, with at
least **three** measurement positions per excitation position on each plate,
randomly distributed over the surface rather than gridded. The minimum
separations are: 0.5 m from an excitation position to the element boundaries,
1.0 m between excitation positions (0.7 m for Type B), 1.0 m from an excitation
position to the junction under test, 1.0 m from an excitation position to its
associated measurement positions, 0.25 m from a measurement position to the
element boundaries and 0.5 m between the measurement positions of one
excitation position. On inhomogeneous elements (hollow brick, beam-and-block)
the positions must also be spread over the sub-elements, and on framed floors
the tapping-machine axis sits at 45° to the joists. On lightweight junctions
the operator should stay off the floor unless their weight and damping can be
shown not to matter.

![Plan of a source element and the receiving element unfolded beside it across the junction under test: four excitation positions scattered over the source element, the three measurement positions of the first one on each plate, and the six minimum distances of Clause 7.2.5 drawn on the spans they rule, 0.5 m from an excitation position to the element boundary, 1.0 m between excitation positions, 1.0 m to the junction, 1.0 m to its measurement positions, 0.5 m between those and 0.25 m from one of them to the boundary; an inset shows a tapping machine at 45° to the joists of a framed floor](https://docs.jmrp.io/phonometry/media/diagram_junction_positions.svg)

The plan lays those numbers out on one element. Every distance on it is a
minimum, and none of the marks is on a grid: a layout passes when each span is
at least as long as the one drawn for it, whatever the positions are.

**Both directions.** $K_{ij}$ is symmetric only because the standard makes it
so: run the whole procedure with element $i$ excited and again with element $j$
excited, and average the two level differences before Formula (13), which is
exactly what `direction_averaged_level_difference` does in the snippet above.

## Suspended ceilings: the plenum flanking path (ISO 140-9, Vigran 9.2.3)

Two offices separated by a partition that stops at the suspended ceiling share
one continuous plenum above it. Sound leaves the source room through the
ceiling tiles, travels sideways over the partition and comes back down through
the tiles of the receiving room. That path is often the weakest link in an
open-plan fit-out, and it is not what a partition's $R_\mathrm{w}$ describes.

![Two panels. A: section of the ISO 140-9 test facility, two rooms of at least 50 m³ under one continuous suspended ceiling, 2.8 m ± 0.2 m high, the dividing wall ending under the ceiling with a bevelled head of at most 100 mm, a plenum 650 mm to 760 mm deep with absorbent lining on its end walls, and the measured path drawn up through the ceiling, along the plenum and down again; beside it a plan of the plenum, 4.5 m ± 0.5 m wide, with one side wall and both end walls lined, and what Clauses 5.1.2 to 5.1.5 ask. B: the symbols of Vigran's one-dimensional model on the same section, the ceiling lengths LS and LR, the source-room height HS, the plenum depth h and the reduction indices RS and RR](https://docs.jmrp.io/phonometry/media/diagram_ceiling_plenum_facility.svg)

The drawing keeps apart two things that share that geometry. Panel A is the
laboratory facility in which ISO 140-9 measures the path, with what its
Clauses 5.1.2 to 5.1.5 fix: an interior 4.5 m ± 0.5 m wide and 2.8 m ± 0.2 m
high under the suspended ceiling, two rooms of at least 50 m³ whose volumes
differ by at least 10 %, a dividing wall that ends under the ceiling with a
head bevelled to at most 100 mm, a plenum 650 mm to 760 mm deep, and an
absorbent lining on one side wall and both end walls of the plenum. Panel B
carries the symbols of the one-dimensional model below, which is Vigran's and
no part of the standard: its worked example, with a plenum 0.43 m deep, is not
a geometry the facility could have.

![Two panels: the ceiling-plenum flanking reduction index Rcl per octave band for two plenum depths against the sum of the two ceiling reduction indices, and the normalized ceiling attenuation of an accredited ASTM E1414 test report with the fitted ASTM E413 contour and its shaded deficiencies giving CAC 34](https://docs.jmrp.io/phonometry/media/ceiling_plenum_flanking.svg)

*Left: the plenum charges a fixed penalty against the sum of the two ceilings
(about 10 dB for this geometry), and doubling its depth pays back only 3 dB of
it, which is why the tile is the lever and the void is not. Right: the same
quantity as an accredited laboratory measures it, with the ASTM E413 contour
fitted to the measured $D_\mathrm{n,c}$ and its deficiencies shaded; the class is read
off the shifted contour at 500 Hz.*

**Show the code for this figure**

```python
import matplotlib.pyplot as plt
import numpy as np

from phonometry import building

fig, (ax_path, ax_cac) = plt.subplots(1, 2, figsize=(13.0, 5.6))

# Left: Vigran Section 9.2.3 geometry, LS = LR = 4.75 m, 9.5 mm plasterboard
# ceiling, reflecting plenum sidewalls, at two plenum depths.
freqs = np.array([63.0, 125.0, 250.0, 500.0, 1000.0, 2000.0, 4000.0])
ceiling = np.array([17.0, 21.0, 25.0, 29.0, 32.0, 30.0, 38.0])
x = np.arange(freqs.size)
for depth, style in ((0.43, "-o"), (0.86, "--o")):
    res = building.plenum_flanking_reduction_index(
        ceiling, ceiling, ceiling_length=4.75, plenum_height_m=depth,
        frequencies=freqs,
    )
    ax_path.plot(x, res.reduction_index, style, label=f"Rcl, plenum h = {depth:g} m")
ax_path.plot(x, 2.0 * ceiling, ":s", label="RS + RR (two ceilings)")
ax_path.set_xticks(x, ["63", "125", "250", "500", "1k", "2k", "4k"])
ax_path.set(xlabel="Frequency [Hz]", ylabel="Sound reduction index [dB]")
ax_path.legend()

# Right: an accredited ASTM E1414 test report rated per ASTM E413.
dnc = np.array([14.4, 18.6, 21.7, 24.1, 23.4, 30.3, 33.7, 35.2,
                41.6, 44.2, 42.1, 36.8, 35.7, 36.0, 36.9, 37.9])
cac = building.ceiling_attenuation_class(dnc)
xc = np.arange(dnc.size)
ax_cac.fill_between(xc, cac.measured, cac.shifted_reference,
                    where=cac.measured < cac.shifted_reference,
                    interpolate=True, alpha=0.25, label="deficiencies")
ax_cac.plot(xc, cac.measured, "-o", label="Dn,c (measured)")
ax_cac.plot(xc, cac.shifted_reference, "--s", label="ASTM E413 contour, fitted")
ax_cac.set(xlabel="Frequency [Hz]",
           ylabel="Normalized ceiling attenuation Dn,c [dB]",
           title=f"CAC = {cac.rating} dB")
ax_cac.legend()
plt.show()
```

### The one-dimensional model

Mechel's one-dimensional variant, as presented
by Vigran in Section 9.2.3, treats the plenum as a duct lined on one side. The
ceiling on each side has a transmission factor $\tau_\mathrm{S} = \tau_\mathrm{S,pl}\tau_\mathrm{S,a}$
(plates times any plenum absorber, Eq. (9.14)); the injected power splits,
a fraction $s_\mathrm{S}$ heading for the partition, and decays as $\exp(-mx)$ with the
power attenuation coefficient $m = 2\,\mathrm{Re}\{\Gamma\} = -2\,\mathrm{Im}\{k'\}$
(Eqs. (9.15) and (9.16)). Integrating over the ceiling length on both sides
gives Eq. (9.18), whose receiving side carries the leakage back into the room,
$m'_\mathrm{R} = m_\mathrm{R} + s_\mathrm{R}\tau_\mathrm{R}/h$ (Eq. (9.17)). Vigran prints the *unprimed* $m_\mathrm{R}$ in
that expression's denominator; that is a misprint, recorded in
[the errata register](https://docs.jmrp.io/phonometry/reference/errata/), and the derived $m'_\mathrm{R}$ is
used here. Read literally, the printed form is non-monotonic in the plenum
damping (adding absorber would predict a worse path than none) and unbounded as
$m_\mathrm{R} \to 0$. For a plenum with little
attenuation and $s_\mathrm{S} = s_\mathrm{R} = 0.5$ that collapses to the compact form that
makes the geometry visible (Eqs. (9.19) and (9.20)):

$$
R_\mathrm{cl} = R_\mathrm{S} + R_\mathrm{R} - 10\log_{10}\!\left[\frac{\varepsilon^2 L_\mathrm{R}}{4h}\right],
$$

with $\varepsilon = 1$ for totally absorbing plenum sidewalls and
$\varepsilon = 2$ for totally reflecting ones. A deep plenum helps, a long room
hurts, and doubling the tile insulation helps twice over because $R_\mathrm{S}$ and
$R_\mathrm{R}$ both appear. Referred to the partition area instead of the ceiling,
$R_\mathrm{cl,p} = R_\mathrm{cl} + 10\log_{10}(H_\mathrm{S}/L_\mathrm{S})$ (Eq. (9.13)), which is what lets the
ceiling path be added to the direct path as transmission factors.

```python
from phonometry import building
# `plenum_flanking_reduction_index` is the import of the figure block above.

freqs = [63.0, 125.0, 250.0, 500.0, 1000.0, 2000.0, 4000.0]
ceiling = [17.0, 21.0, 25.0, 29.0, 32.0, 30.0, 38.0]   # 9.5 mm plasterboard

# Vigran's example geometry: LS = LR = 4.75 m, plenum 0.43 m, reflecting walls.
res = building.plenum_flanking_reduction_index(
    ceiling, ceiling, ceiling_length=4.75, plenum_height_m=0.43, frequencies=freqs
)
print(round(res.geometry_term, 1))    # 10.4 dB charged against RS + RR
res.plot()                            # Rcl against the two ceilings

# A lined plenum attenuates the sideways path (Eq. (9.18) instead of (9.20)):
damped = building.plenum_flanking_reduction_index(
    ceiling, ceiling, ceiling_length=4.75, plenum_height_m=0.43,
    attenuation_source_per_m=[0.5] * 7, attenuation_receiving_per_m=[0.5] * 7,
)

# Referred to the partition instead of the ceiling (Eq. (9.13)):
print(building.partition_referenced_reduction_index(res.reduction_index, 2.7, 4.75))
```

### The measured quantity: normalized ceiling attenuation

A ceiling is not rated by $R_\mathrm{cl}$ but by the
**normalized ceiling attenuation** $D_\mathrm{n,c} = D - 10\log_{10}(A/A_0)$ (ISO 140-9:1985
clause 3.3), with $A$ the receiving-room equivalent absorption area and the
reference $A_0 = 10\ \text{m}^2$. The facility (panel A of the drawing that
opens this section, Clauses 5.1.2 to 5.1.5) has two rooms of at least
50 m³ whose volumes differ by at least 10 %, a dividing wall tapered to at most
100 mm at the top, and a plenum 650 mm to 760 mm deep with one sidewall and
both end walls lined; the standard prints the required lining absorption
$\alpha_\mathrm{s} \ge 0.65$ at 125 Hz and $\ge 0.80$ from 250 Hz to 4000 Hz, and
requires $\alpha < 0.10$ on the other sidewall and on the plenum ceiling.
The North American counterpart, ASTM E1414, uses $A_0 = 12\ \text{m}^2$, so an
ASTM value runs about $10\log_{10}(12/10) = 0.79$ dB higher for the same rooms.

ISO rates $D_\mathrm{n,c}$ with the ISO 717-1 curve, giving $D_\mathrm{n,c,w}$; ASTM E1414
rates it through ASTM E413 as the **ceiling attenuation class**. E413 rounds the
data to the nearest integer (clause 5.2), raises its reference contour in 1 dB
steps while the sum of the deficiencies stays at or below 32 dB and no single
deficiency exceeds 8 dB (clauses 5.3 and 5.4), and reads the rating off the
shifted contour at 500 Hz (clause 5.5).

```python
from phonometry import building
# `ceiling_attenuation_class` is the import of the figure block above.

# Dn,c from the measured pair: source- and receiving-room levels over the
# common plenum and the receiving room's absorption area, per band. ASTM E1414
# normalizes to A0 = 12 m2, ISO 140-9 to A0 = 10 m2.
l1_c = [80.0] * 16                       # source room, per band
l2_c = [45.0] * 16                       # receiving room, over the plenum path
absorption = [12.0] * 16                 # receiving-room A per band (m2)
astm = building.normalized_ceiling_attenuation(l1_c, l2_c, absorption, reference_area=12.0)
iso140_9 = building.normalized_ceiling_attenuation(l1_c, l2_c, absorption)
print(round(float(astm[0]), 2), round(float(iso140_9[0]), 2))   # 35.0 34.21

# A 28 mm perforated plaster acoustic tile, measured to ASTM E1414 (CAC 34).
dnc = [14.4, 18.6, 21.7, 24.1, 23.4, 30.3, 33.7, 35.2,
       41.6, 44.2, 42.1, 36.8, 35.7, 36.0, 36.9, 37.9]
res = building.ceiling_attenuation_class(dnc)
print(res.rating, res.deficiency_sum, res.max_deficiency)   # 34, 27.0, 7.0
res.plot()                                                  # Dn,c vs the contour

# The ISO single number of the same spectrum, shifted to A0 = 10 m2:
iso = [v - 0.79 for v in dnc]
print(building.weighted_rating(iso).rating)                          # Dn,c,w
```

## ISO 10848 flanking-transmission reports (`.report()`)

Each of the three results renders a one-page PDF fiche. `VibrationReductionResult.report()`
writes a **junction characterization** report of $K_{ij}$ (ISO 10848-1:2006):
the standard-basis line, an optional metadata header, the per-band $K_{ij}$
table beside the $K_{ij}(f)$ curve and a boxed single-number mean $K_{ij}$ over
the Annex A band range, with the count of averaged and bracketed bands. Bands
bracketed for poor modal overlap ($M < 0.25$, ISO 10848-4:2010 Clause 9) print
their value in brackets and are excluded from the mean; `verbose=True` adds a
column stating whether each band enters the mean.

`FlankingLevelDifferenceResult.report()` and `FlankingImpactLevelResult.report()`
write **measurement** reports of the overall descriptors $D_\mathrm{n,f}$ (airborne)
and $L_\mathrm{n,f}$ (impact, tapping machine), reusing the same two-panel insulation
layout: the per-band quantity beside the measured-versus-shifted-ISO 717-reference
curve and the boxed single number $D_\mathrm{n,f,w}$ (C; Ctr) (ISO 717-1) or
$L_\mathrm{n,f,w}$ (CI) (ISO 717-2). `verbose=True` annexes the ISO 717 evaluation per band (the value,
the shifted reference and the unfavourable deviation). A `requirement` supplied
on the `ReportMetadata` adds a verdict ($D_\mathrm{n,f,w}$ passes at or above it,
$L_\mathrm{n,f,w}$ at or below it), and `language="es"` renders every fiche in Spanish.
reportlab is required, and matplotlib too for the figure the fiche embeds (`pip
install "phonometry[report,plot]"`).

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

freqs = [100, 125, 160, 200, 250, 315, 400, 500, 630,
         800, 1000, 1250, 1600, 2000, 2500, 3150, 4000, 5000]
dv = np.array([4.5, 4.8, 5.2, 5.6, 6.0, 6.5, 7.0, 7.6, 8.1, 8.7,
               9.2, 9.8, 10.3, 10.9, 11.4, 11.9, 12.3, 12.7])
m = np.full(18, 1.0); m[:3] = 0.1                        # bracket the low bands
kij = building.vibration_reduction_index(
    dv, junction_length=4.0, area_i=12.0, area_j=10.0, frequencies=freqs,
    structural_reverberation_time_i=0.35, structural_reverberation_time_j=0.40,
    modal_overlap=m,
)
kij.report("Kij.pdf", metadata=ReportMetadata(specimen="Rigid T-junction"))

l1 = np.full(16, 80.0)
dnf = np.array([48, 49, 50, 51, 52, 54, 55, 57, 58, 59, 60, 61, 62, 63, 64, 65],
               dtype=float)
dres = building.normalized_flanking_level_difference(
    l1, l1 - dnf, absorption_area=np.full(16, 10.0)
)
dres.report("Dnf.pdf", metadata=ReportMetadata(requirement=55.0))   # Dn,f,w (C; Ctr)

recv = np.array([58, 57, 56, 55, 54, 52, 50, 48, 46, 44, 42, 40, 38, 36, 34, 32],
                dtype=float)
lres = building.normalized_flanking_impact_level(recv, absorption_area=np.full(16, 10.0))
lres.report("Lnf.pdf", metadata=ReportMetadata(requirement=55.0))   # Ln,f,w (CI)
```

The example fiches are regenerated with `make reports` and kept in the
repository. Click a preview to open the PDF:

**Report:** [ISO 10848-1 junction vibration reduction index report (PDF)](https://docs.jmrp.io/phonometry/media/iso10848_kij_example.pdf): One-page junction-characterization report of the vibration reduction index Kij: the metadata header, the per-band Kij table beside the Kij(f) curve, and the boxed single-number mean Kij over the Annex A band range with the count of averaged and bracketed bands (the three lowest bands bracketed for poor modal overlap).

Vibration reduction index fiche (VibrationReductionResult.report), mean Kij.

**Report:** [ISO 10848-2 airborne flanking level difference report (PDF)](https://docs.jmrp.io/phonometry/media/iso10848_dnf_example.pdf): One-page airborne flanking-transmission report of the normalized flanking level difference Dn,f: the metadata header, the one-third-octave Dn,f table beside the measured-versus-shifted-reference curve, the boxed Dn,f,w (C; Ctr) rated per ISO 717-1 and a PASS verdict against the 55 dB requirement.

Flanking level difference fiche (FlankingLevelDifferenceResult.report), Dn,f,w (C; Ctr).

**Report:** [ISO 10848-2 flanking impact level report (PDF)](https://docs.jmrp.io/phonometry/media/iso10848_lnf_example.pdf): One-page impact flanking-transmission report of the normalized flanking impact level Ln,f: the metadata header, the one-third-octave Ln,f table beside the measured-versus-shifted-reference curve, the boxed Ln,f,w (CI) rated per ISO 717-2 and a PASS verdict against the 55 dB requirement.

Flanking impact level fiche (FlankingImpactLevelResult.report), Ln,f,w (CI).

## What this guide covers

- **Covered.** ISO 10848-1:2006's vibration reduction index $K_{ij}$ (Formula (13)) from the
  direction-averaged velocity level difference (Formula (11)) and the equivalent
  absorption lengths (Formula (12)), with the simplified lightweight form
  (Formula (14)), the total loss factor and the Annex A single-number mean, via
  `building.direction_averaged_level_difference` and
  `building.vibration_reduction_index`; the overall descriptors $D_\mathrm{n,f}$
  (Formula (4)) and $L_\mathrm{n,f}$ (Formula (5)) with their ISO 717 single numbers,
  via `building.normalized_flanking_level_difference` and
  `building.normalized_flanking_impact_level`; the strong-coupling check
  (Formula (15)) and the ISO 10848-4:2010 modal-density, band-mode-count and
  modal-overlap validity checks that bracket unreliable bands
  (`building.check_strong_coupling`, `building.modal_density`,
  `building.band_mode_count`, `building.modal_overlap_factor`); and the three
  one-page fiches through `.report()`.

- **Not covered.** ISO 10848 Parts 2, 3 and 4 differ in which junction and specimen types they
  apply to; phonometry implements only the Part 1 $K_{ij}$/$D_\mathrm{n,f}$/$L_\mathrm{n,f}$
  formulae generically, plus the Part 4 modal-overlap validity check, not the
  facility-specific test setups the other parts describe. The acquisition of
  [Making the
  measurement](#making-the-measurement-iso-10848-12017-clauses-7-and-9) is
  documented here and enforced nowhere: nothing checks that the other elements
  were shielded, that the shield met $\Delta R_\text{min}$, that the position
  counts and separations were kept, that the accelerometer satisfied the
  mass-loading inequality, or that $T_\mathrm{s}$ was measured on the elements in their
  installed state rather than assumed. Because ISO 10848 contains no worked
  numeric example, conformance is anchored on closed-form identities (simplified
  $K_{ij}$, $a_j$ at $f_\mathrm{ref}$, $\eta$).

- **Covered.** The suspended-ceiling branch covers the normalized ceiling attenuation
  $D_\mathrm{n,c}$ of ISO 140-9:1985 clause 3.3, the ASTM E413-22 ceiling attenuation
  class that ASTM E1414 invokes, and the one-dimensional plenum model of Vigran
  Section 9.2.3, via `building.normalized_ceiling_attenuation`,
  `building.ceiling_attenuation_class`,
  `building.plenum_flanking_reduction_index` and
  `building.partition_referenced_reduction_index`. The plenum model has no
  published numeric output of its own, so it is anchored on its closed forms and
  on the structural properties a wrong reading breaks: monotonicity in the
  plenum damping, the bound $\tau_\mathrm{cl} \le 1$, and the convergence of Eq. (9.18)
  to Eq. (9.20); the measurement chain and the class are anchored on accredited
  ASTM E1414 laboratory reports. The plenum propagation constant $k'$ of a lined
  duct is an input here, not a prediction.

## See also

- [Predicting Sound Insulation (EN 12354)](https://docs.jmrp.io/phonometry/buildings/design/insulation-prediction/):
  the flanking model that consumes the measured $K_{ij}$, and its empirical
  Annex E junction values.
- [Bending-wave transmission at plate junctions](https://docs.jmrp.io/phonometry/vibration/structural/junction-transmission/):
  the wave-theory transmission coefficients behind an ideal junction's
  $K_{ij}$.
- [Laboratory Insulation Measurement](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-lab/):
  the ISO 10140 suite whose direct transmission these flanking paths bypass.
- [Insulation Ratings (ISO 717)](https://docs.jmrp.io/phonometry/buildings/insulation/insulation-ratings/):
  the reference-curve engines behind $D_\mathrm{n,f,w}$ and $L_\mathrm{n,f,w}$.
- [Theory](https://docs.jmrp.io/phonometry/reference/theory/rooms-buildings/): the flanking
  formulae in the EN 12354 derivation.
- [Room-to-room and open-plan acoustics](https://docs.jmrp.io/phonometry/buildings/rooms/open-plan-acoustics/): the open-plan context
  the ceiling flanking path usually appears in.
- API reference: [`building.measurement.flanking_transmission`](https://docs.jmrp.io/phonometry/reference/api/building/flanking-transmission/)
  and [`building.prediction.ceiling_plenum`](https://docs.jmrp.io/phonometry/reference/api/building/ceiling-plenum/).
- Theory: [Point mobilities and radiation efficiency](https://docs.jmrp.io/phonometry/reference/theory/vibration/#point-mobilities-and-radiation-efficiency-cremer-5-hopkins-29): the mobility and radiation-efficiency theory behind a flanking path.

## References

- **ASTM International. (2022).** [*Classification for Rating Sound Insulation*](https://www.astm.org/e0413-22.html) (ASTM E413-22).
  Table 1 is the reference contour and clause 5 the fitting rules that ASTM E1414 invokes to produce the ceiling attenuation class.
- **Hopkins, C. (2007).** *Sound insulation*. Butterworth-Heinemann. https://doi.org/10.4324/9780080550473
  The reference monograph for flanking transmission and the vibration reduction index, including the statistical-energy-analysis footing of the validity checks. ISBN 978-0-7506-6526-1.
- **International Organization for Standardization. (1985).** [*Acoustics — Measurement of sound insulation in buildings and of building elements — Part 9: Laboratory measurement of room-to-room airborne sound insulation of a suspended ceiling with a plenum above it*](https://www.iso.org/standard/3965.html) (ISO 140-9:1985).
  Clause 3.3 defines the normalized ceiling attenuation Dn,c with A0 = 10 m2, and the facility clauses give the plenum depth, the room volumes and the required lining absorption.
- **International Organization for Standardization. (2006).** [*Acoustics — Laboratory measurement of the flanking transmission of airborne and impact sound between adjoining rooms — Part 1: Frame document*](https://www.iso.org/standard/38284.html) (ISO 10848-1:2006).
  The frame document of the flanking-transmission measurement: definitions, the vibration reduction index Kij, the equivalent absorption length and the normalized flanking descriptors D_n,f / L_n,f that feed the EN 12354 prediction. Because ISO 10848 contains no worked numeric example, conformance is anchored on closed-form identities.
- **International Organization for Standardization. (2006).** [*Acoustics — Laboratory measurement of the flanking transmission of airborne and impact sound between adjoining rooms — Part 2: Application to light elements when the junction has a small influence*](https://www.iso.org/standard/35004.html) (ISO 10848-2:2006).
  Application to light elements whose junction has a small influence.
- **International Organization for Standardization. (2006).** [*Acoustics — Laboratory measurement of the flanking transmission of airborne and impact sound between adjoining rooms — Part 3: Application to light elements when the junction has a substantial influence*](https://www.iso.org/standard/35003.html) (ISO 10848-3:2006).
  Application to light elements whose junction has a substantial influence.
- **International Organization for Standardization. (2010).** [*Acoustics — Laboratory measurement of the flanking transmission of airborne and impact sound between adjoining rooms — Part 4: Application to junctions with at least one heavy element*](https://www.iso.org/standard/45360.html) (ISO 10848-4:2010).
  Application to junctions with at least one heavy element, with the modal-density, band-mode-count and modal-overlap validity checks (M < 0.25 bands bracketed, Clause 9).
- **International Organization for Standardization. (2017).** [*Acoustics — Laboratory and field measurement of flanking transmission for airborne, impact and building service equipment sound between adjoining rooms — Part 1: Frame document*](https://www.iso.org/standard/67226.html) (ISO 10848-1:2017).
  The current frame document, whose Clauses 7.2 to 7.4 and 9 carry the acquisition rules quoted here: accelerometer mounting and the mass-loading inequality (7.2.3), excitation (7.2.4), the excitation and measurement position counts and separations (7.2.5), the structural reverberation time (7.3) and the shielding requirement (Clause 9).
- **Vigran, T. E. (2008).** *Building acoustics*. CRC Press. https://doi.org/10.1201/9781482266016
  Section 9.2.3 presents Mechel's one-dimensional plenum model, Eqs. (9.13) to (9.20). ISBN 978-0-415-42853-8.
