# Sound power from surface vibration (ISO/TS 7849)

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

Canonical: https://docs.jmrp.io/phonometry/devices/emission/vibration-sound-power/
Language: en
Alternate: https://docs.jmrp.io/phonometry/es/devices/emission/vibration-sound-power/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: 2026-07-11
Updated: 2026-11-16T21:36:13Z
Standards: ISO/TS 7849-1:2009; ISO/TS 7849-2:2009
Verified: 3 conformance checks (ISO/TS 7849-1:2009; ISO/TS 7849-2:2009) https://docs.jmrp.io/phonometry/reference/conformance/#room-building-acoustics
License: MIT (https://github.com/jmrplens/phonometry/blob/main/LICENSE)
Summary: Sound power from surface vibration by ISO/TS 7849: velocity level, radiation factor, the Part 1 limit and the Part 2 value.

---

The airborne sound power a machine radiates through the structure-borne
vibration of its outer surface can be estimated from the surface vibratory
velocity and a **radiation factor** $\varepsilon$ (the radiation efficiency),
without an
acoustic measurement. The radiated power is (ISO/TS 7849-1, Formula 6)

$$
P = Z_\mathrm{c} \, \langle v^2 \rangle \, S \, \varepsilon \quad [\mathrm{W}],
$$

with $Z_\mathrm{c}$ the characteristic impedance of air, $\langle v^2 \rangle$ the
mean-square vibratory velocity over the radiating area $S$. Expressed in levels
(velocity level re $v_0 = 5\times10^{-8}\ \text{m/s}$), each part writes that
relation in its own quantities. **Part 1** is A-weighted throughout: one
A-weighted surface velocity level $L_{v\mathrm{A}}$, the fixed $\varepsilon_\mathrm{A} = 1$, and a
single A-weighted number, the *upper limit* of the airborne sound power level
(Formula 12)

$$
L_{W\mathrm{A},\max} = L_{v\mathrm{A}} + 10\log_{10}\frac{S}{S_0}
      + 10\log_{10}\frac{Z_\mathrm{c,n}}{Z_{\mathrm{c},0}}.
$$

**Part 2** is the band-wise counterpart, with the radiation factor
$\varepsilon_j$ determined for each band (Formula 15)

$$
L_{W,j} = L_{v,j} + 10\log_{10}\frac{S}{S_0} + 10\log_{10}\varepsilon_j
      + 10\log_{10}\frac{Z_\mathrm{c,n}}{Z_{\mathrm{c},0}},
$$

where $S_0 = 1\ \text{m}^2$, the normalized impedance
$Z_\mathrm{c,n} = 411\ \text{N·s/m}^3$ and the reference
$Z_{\mathrm{c},0} = 400\ \text{N·s/m}^3$ give the fixed
$10\log_{10}(411/400) = 0.118\ \text{dB}$
term. This guide runs the band-wise form for both parts, so that the Part 1
upper limit and the Part 2 engineering value can be compared band by band; a
strict Part 1 survey is the single A-weighted number above, which the result
object carries as `sound_power_level_a`. This module feeds the structure-borne
source and building prediction standards (ISO 9611, EN 15657, EN 12354-5).

Before any levels, the radiator itself. The `radiation_efficiency` plate model
that supplies a predicted radiation factor retains its geometry, and
`sigma.plot_geometry()` draws the plate in its baffle to scale, here 1.5 m by
1.25 m and simply supported.

![To-scale front view of the plate of the radiation model: a grey 1.5 m by 1.25 m simply supported plate inside its hatched rigid baffle, both side lengths dimensioned and the boundary condition named in the title](https://docs.jmrp.io/phonometry/media/radiation_plate_geometry.svg)

*The radiator behind the radiation factor, to scale: the 1.5 m by 1.25 m
simply supported plate in its rigid baffle, whose area $S$ enters $L_W$
directly while its size decides how far $\varepsilon$ falls below one under
coincidence.*

**Show the code for this figure**

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

# The simply supported 1.5 x 1.25 m plate of the radiation model.
f = np.geomspace(50.0, 5000.0, 200)
sigma = vibration.radiation_efficiency(f, 1.5, 1.25, 2100.0)
sigma.plot_geometry()
plt.show()
```

## 1. The two parts

The two parts differ only in the radiation factor. **Part 1 (survey)** assumes
$\varepsilon = 1$ and yields the *upper limit* $L_{W,\max}$, needing only the
velocity level and the area. **Part 2 (engineering)** applies a frequency-band
radiation factor $\varepsilon_j$ determined (per ISO 9614) as
$\varepsilon_j = P_j/(Z_\mathrm{c,n}\,\langle v_j^2 \rangle\,S)$.

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

bands = np.array([250.0, 500.0, 1000.0, 2000.0])
lv = np.array([82.0, 85.0, 83.0, 79.0])          # mean velocity level per band [dB]

# Part 1 upper limit (epsilon = 1):
upper = emission.sound_power_from_vibration(lv, area=1.6, frequencies=bands)
print(round(upper.total_level, 1))               # 90.9  dB re 1 pW, unweighted
print(round(upper.sound_power_level_a, 1))       # 88.9  dB(A) re 1 pW

# Part 2 engineering value with a measured radiation factor:
eps = np.array([0.45, 0.75, 0.95, 1.00])
eng = emission.sound_power_from_vibration(lv, area=1.6, radiation_factor=eps, frequencies=bands)
print(np.round(eng.sound_power_level, 1))        # per-band L_W
print(round(eng.sound_power_level_a, 1))         # 88.4  dB(A) re 1 pW

eng.plot()   # the LW(f) spectrum, as in the figure above (needs matplotlib)
```

The two totals are not interchangeable: `total_level` is the unweighted band sum
$10\lg\sum_j 10^{0.1 L_{W,j}}$, while `sound_power_level_a` is the A-weighted
total that the fiche boxes and that a noise declaration quotes. They differ by
the A-weighting of the spectrum: 2.0 dB on this casing, several decibels more
on a low-frequency-heavy one.

Set the two parts side by side on one casing and the price of Part 1's
assumption is a picture rather than an argument:

![Radiated sound power level per octave band, comparing the ISO/TS 7849-1 upper limit (radiation factor of one) with the ISO/TS 7849-2 engineering value (measured radiation factor)](https://docs.jmrp.io/phonometry/media/vibration_sound_power.svg)

*Paired bars per band for a 1.6 m² casing: Part 1's upper limit against Part 2's
engineering value. The gap **is** $-10\log_{10}\varepsilon_j$: 7.0 dB at
125 Hz, 3.5 dB at 250 Hz, 1.2 dB at 500 Hz and nothing from 2 kHz up; so the
error Part 1 makes is largest in the low bands, exactly where a large thin
casing radiates worst.*

**Show the code for this figure**

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

# Surface velocity levels and a measured radiation factor per octave band.
bands = np.array([125.0, 250.0, 500.0, 1000.0, 2000.0, 4000.0])
lv = np.array([78.0, 82.0, 85.0, 83.0, 79.0, 74.0])
eps = np.array([0.20, 0.45, 0.75, 0.95, 1.00, 1.00])

lw_max = emission.radiated_sound_power_level(lv, 1.6)  # Part 1, eps = 1
lw_eng = emission.radiated_sound_power_level(lv, 1.6, radiation_factor=eps)  # Part 2

# The figure above: the two parts band by band.
x = np.arange(bands.size)
fig, ax = plt.subplots()
ax.bar(x - 0.2, lw_max, width=0.4, label="Part 1 upper limit ($\\varepsilon$ = 1)")
ax.bar(x + 0.2, lw_eng, width=0.4, label="Part 2 engineering ($\\varepsilon$ measured)")
ax.set_xticks(x, [f"{b:g}" for b in bands])
ax.set(xlabel="Frequency [Hz]", ylabel="Sound power level $L_W$ [dB re 1 pW]")
ax.legend()
plt.show()

# A single determination is a result object that draws its own spectrum:
res = emission.sound_power_from_vibration(lv, area=1.6, radiation_factor=eps,
                                          frequencies=bands)
res.plot()
plt.show()
```

## 2. Velocity level, calibration and the radiation factor

The velocity level is $L_v = 20\log_{10}(v/v_0)$ (Formula 3); a sinusoidal
calibration acceleration converts as
$L_v = 20\log_{10}\!\left(\hat{a}/(2\pi f\,v_0\sqrt{2})\right)$ (Formula 8). The
radiation factor comes from an independently measured power:

```python
from phonometry import emission

# The standard's worked calibration EXAMPLE: 9.81 m/s^2 at 100 Hz.
print(round(float(emission.velocity_level_from_acceleration(9.81, frequency=100.0)), 1))   # 106.9 dB

# Radiation factor from a measured power (ISO 9614): eps = P / (Zc <v^2> S).
eps = emission.radiation_factor(3.0e-4, area=2.0, mean_square_velocity=(1e-3)**2)
print(round(float(eps), 3))                                                # 0.365
```

Surface velocity levels from several positions are combined with the energetic
mean `mean_velocity_level` (Formula 10) or its area-weighted form (Formula 11),
and the correction `extraneous_velocity_correction` removes extraneous
vibration per Table 2.

Those positions are not free: the standard divides the radiating surface $S$
into $N$ equal cells and puts one accelerometer at the centre of each. Table 1
fixes the initial $N$ from the area alone: **5** positions below 1 m²,
**10** from 1 m² to 10 m², and $S/S_0$ positions above that ($S_0 = 1$ m²); so
the 4 m² surface drawn below takes ten. Increase $N$ when the spread of the
A-weighted velocity level over the positions, highest minus lowest in decibels,
exceeds twice the Table 1 count; a strongly non-uniform field can instead be
handled by redistributing the positions, which makes the partial areas $S_i$
unequal and so calls for the area-weighted Formula 11 rather than the plain
energetic mean. The Table 2 correction is a ladder on the same kind of spread:
with the machine stopped and the surrounding plant running, the mean extraneous
velocity level must sit at least 3 dB below the machine's, and
`extraneous_velocity_correction` returns 3 dB at a 3 dB difference, 2 dB at
4-5 dB, 1 dB at 6-9 dB and 0 dB from 10 dB up.

![ISO/TS 7849 surface-velocity method: a machine under test with its vibrating measurement surface of 2.5 by 1.6 metres divided into ten equal cells, an accelerometer at each cell centre, the radiated airborne sound leaving the surface, the Table 1 initial number of measurement positions (5 below 1 square metre, 10 from 1 to 10, S over S0 above) and the survey relation LWA = LvA + 10 lg(S/S0) + 10 lg epsilon](https://docs.jmrp.io/phonometry/media/diagram_vibration_sound_power.svg)

The Table 2 ladder is one call on a measured difference:

```python
# With the machine stopped and the surrounding plant running, the mean
# extraneous velocity level is 5.4 dB below the operating one.
k1a = emission.extraneous_velocity_correction(5.4)
print(k1a)                                        # 2.0 dB, subtracted from Lv
print(round(float(np.mean(lv)) - k1a, 1))         # 78.2 dB after the correction
```

### Mounting the transducer

Everything above assumes $L_v$ is the surface's own normal velocity. Whether it
is depends on the transducer and on how it is attached, and both have a
frequency limit that is easy to exceed without any warning in the data.

Use a piezoelectric accelerometer light enough not to mass-load the panel, and
only over the flat part of its response: the usable upper limit is about **one
third of the transducer's own mounted resonance**, which keeps the reading
within roughly 1 dB. A 30 kHz-resonance miniature accelerometer is therefore
good to about 10 kHz; a heavy 6 kHz one only to 2 kHz. Mount it per ISO 5348
with its sensitive axis **normal to the surface**, because the method needs the
normal component and nothing else, and fix the cable to the machine close to the
transducer so that cable motion does not drive the charge amplifier.

The attachment is the part that silently costs bandwidth:

| Attachment | Usable to | Notes |
| :--- | :--- | :--- |
| Stud (threaded) | the transducer's own limit | The reference mounting; needs a drilled and faced boss |
| Adhesive, or a thin layer of adhesive wax | about 10 kHz | Wax not on hot surfaces (ISO 5348 gives a thin layer of beeswax a maximum of 40 °C) |
| Clamping magnet | about 2.5 kHz | Smooth flat steel only; a 110 g magnet plus a 30 g transducer also caps the measurable acceleration, and paint or roughness destroys the adhering force |
| Hand-held probe | low bands only | For a quick survey, never for a determination |

That table has a direct consequence for the examples on this page, which run
octave bands out to **4 kHz**: a magnet mount loses the top band outright, and
loses it quietly (the level simply reads low rather than failing a check). A
scanning laser vibrometer avoids the mass loading and the mounting limit
altogether, provided the beam stays perpendicular to the surface, since it too
measures only the component along the beam.

## 3. When the radiation-factor assumption breaks

The whole method stands on one substitution: replacing the acoustic
measurement by $\varepsilon$. The Part 1 value $\varepsilon = 1$ is close to
the true radiation
factor only above the **critical (coincidence) frequency** of plate-like
parts, where bending waves travel faster than sound and the surface radiates
like a piston. Below coincidence, adjacent zones of the
plate move in antiphase and their radiation largely cancels: $\varepsilon$
drops far
below one and falls quickly with decreasing frequency, so the survey method
can overstate the low-frequency bands of a large thin casing by 10 dB and
more. The same cancellation makes small sources radiate poorly (the acoustic
short circuit around an unbaffled panel). Two further assumptions are easy
to violate in the field:

* **The measured vibration must be the machine's own.** Vibration fed in
  from neighbouring machinery inflates $\langle v^2 \rangle$; Table 2
  prescribes the
  source-off check and `extraneous_velocity_correction` applies it.
* **The surface must be the dominant radiator.** Airborne sound from
  openings, intakes or internal sources that bypasses the measured casing is
  invisible to a velocity survey; the method characterises the
  structure-borne part only.

### What the cancellation looks like

"Adjacent zones move in antiphase and their radiation largely cancels" is the
sentence the whole caveat rests on, and it is a statement about a phase
relationship between a travelling structural wave and the air on top of it
(which a spectrum cannot draw). The clip below drives a 10 mm steel plate in air
**along its whole length**, with a force on the plate rather than a wave
arriving at it, because that is the emission case this method is about, and
because a plate driven only over a patch radiates from the patch, which below
coincidence would be the only thing in the air and would teach the opposite of
the truth. It then runs the same scene twice, at $f_\mathrm{c}/2$ and at $2f_\mathrm{c}$, with
nothing differing between the panels but the drive frequency.

**Video:** [A steel plate driven below and above its coincidence frequency](https://docs.jmrp.io/phonometry/media/anim_elastic_radiation_efficiency.webm): A bending wave runs to the right along a steel plate with air above it. In the lower-frequency panel the plate ripples with a wavelength shorter than the sound wavelength and the pressure above it forms alternating lobes that cling to the plate and fade out within a few centimetres, so nothing leaves. In the higher-frequency panel the plate ripples with a longer wavelength and a plane beam of pressure departs at forty-five degrees along the whole driven span, growing as it goes.

Below coincidence the bending wavelength is **shorter** than the acoustic one.
Adjacent half-waves of the plate push and pull the same air in antiphase, and
what the air shows is a skin of alternating lobes that clings to the plate and
dies within a fraction of a wavelength. The pressure at the surface is of the
same order as in the other panel; it is simply a quarter-cycle out of step
with the velocity, so it carries no power away: that is what
$\varepsilon \ll 1$ *means*. Above coincidence the bending wave outruns sound,
the trace match is satisfied, and a plane beam leaves at
$\arcsin(c_0/c_\mathrm{B}) = 45°$.

Each panel is annotated with the closed forms its regime obeys, taken from the
plate's own constants: below coincidence, that $\lambda_\mathrm{B} = 0.40$ m is *shorter*
than the 0.57 m of sound in air, so $\sin\theta = \lambda/\lambda_\mathrm{B}$ has no
solution (there is no angle for the sound to leave at), and the skin that
stands in its place decays with a height of $1/\sqrt{k_\mathrm{B}^2 - k_0^2} = 0.091$ m;
above coincidence, that $\lambda_\mathrm{B} = 0.20$ m is *longer* than the 0.14 m in air,
so the trace match is satisfied and sends a beam out at 45°. Nothing is
measured off the scene itself. Two quantities were tried and both were spoiled
by the ends of the plate (a damping layer reflects a bending wave partially),
so the air carries a leftward beam on top of the rightward one, and the reading
is the difference of the two rather than either. Terminating a bending wave
well enough to read a number off it needs a much wider domain than a picture of
this can hold.

Two cautions on reading it. The scene is an effectively **infinite** plate in
air, so what it demonstrates is below-coincidence cancellation, not the
acoustic short circuit around an unbaffled panel: those are the same
cancellation at two different scales, and the second one needs a finite source
with a path around its edge. And the clip deliberately does not print a
radiation factor. A $\varepsilon$ read off a scene like this would be an
artefact of the drive: any finite driving aperture has a wavenumber spectrum
broad enough to put some of itself inside the radiating region, and below
coincidence that aperture radiation would be the only thing crossing an
intensity line, since the travelling subsonic wave contributes exactly
nothing. The number belongs to the closed-form curve in the next subsection;
the clip is here for the mechanism behind it.

### Estimating the radiation factor before you measure it

None of that has to be taken on trust: the library predicts $\varepsilon$ for a
plate, and it is the same `radiation_efficiency` model that drew the plate at
the top of this page. The **critical frequency** follows from the surface mass
and the bending stiffness (a 3 mm steel panel puts it near 4 kHz), and it is
where bending waves travel at the speed of sound and the plate begins to
radiate like a piston. Everything below it is the acoustic short circuit.

![The radiation efficiency sigma of the 1.5 by 1.25 metre plate on log-log axes from 50 Hz to 5 kHz, rising from about 0.003 at 63 Hz to a peak just above the critical frequency of 4073 Hz. The whole region below the sigma equals one line is shaded and labelled as the range where ISO/TS 7849-1 assumes epsilon equals one, with the radiation index marked at minus 23.4 dB at 125 Hz and minus 17.6 dB at 1 kHz](https://docs.jmrp.io/phonometry/media/radiation_efficiency_sigma.svg)

*The shaded area is the assumption Part 1 makes. For this plate it is worth
25 dB at 63 Hz and 14 dB at 2 kHz, and it closes only at coincidence: this is
the "10 dB and more" of the paragraph above, drawn.*

**Show the code for this figure**

```python
import matplotlib.pyplot as plt

# The same 1.5 x 1.25 m plate, now as a 3 mm steel panel so the coincidence
# frequency comes from the material rather than being assumed.
f = np.geomspace(50.0, 5000.0, 400)
stiffness = vibration.plate_bending_stiffness(2.0e11, 0.003, 0.30)
fc = vibration.coincidence_frequency(7800.0 * 0.003, stiffness)
print(round(fc, 0))                                   # 4073.0 Hz
sigma = vibration.radiation_efficiency(f, 1.5, 1.25, fc)
print(round(float(np.interp(1000.0, f, sigma.radiation_index)), 1))    # -17.6 dB

sigma.plot()          # sigma(f) with the sigma = 1 line and fc marked
plt.show()
```

A predicted $\varepsilon$ feeds straight back into the determination:
`sound_power_from_vibration(lv, area=…, radiation_factor=sigma.radiation_efficiency,
frequencies=bands)` turns a Part 1 survey into a first estimate of the Part 2
value. Take it for what it is: a *baffled flat plate with a known boundary
condition* is not a machine casing with ribs, apertures and a gearbox bolted to
it, but it is enough to know whether the Part 1 number is 1 dB or 15 dB
pessimistic before committing to an intensity measurement. The same
$\varepsilon$ appears as $\sigma$, the radiation efficiency, in
[Panel Sound Insulation](https://docs.jmrp.io/phonometry/buildings/design/panel-sound-insulation/):
the emission standards write it $\varepsilon$ and the building-prediction
standards write it $\sigma$, and it is the same quantity computed by the same
function.

Part 2 exists exactly for the radiation-factor problem: it replaces the
fixed $\varepsilon = 1$
with a band-by-band $\varepsilon_j$ determined from one reference measurement
of the
radiated power (ISO 9614 intensity), after which the velocity survey can be
repeated cheaply on nominally identical machines.

![The ISO/TS 7849-2 determination of the radiation factor: the same machine casing carrying its ten-cell accelerometer grid with the mean square velocity measured on the surface, enclosed by an ISO 9614 measurement surface offset 0.25 m from it over which a p-p probe traces a serpentine sweep, the band power P_j leaving that surface, and the relation epsilon_j equals P_j divided by the characteristic impedance times the mean square velocity times the area, taken on one machine in one operating mode](https://docs.jmrp.io/phonometry/media/diagram_radiation_factor.svg)

That pairing is the whole of Part 2, and it is where the transferability has to
be qualified. Clause 4.4.3 is explicit that $\varepsilon_j$ is determined not
only by the geometry and the mechanical properties of the structure but **also
by the distribution and manner of the excitation and by the internal loss
factor**, so a factor measured at idle does not carry to full load, and one
measured on a soft mount does not carry to a bolted installation: the operating
and mounting condition is part of the $\varepsilon$, exactly as it is part of
every other emission value in this section. Clause 4.2 adds the modal caveat:
in bands where only a very few modes are excited, $\varepsilon_j$ is set by
*which* modes happen to be excited rather than by the surface's average
radiating behaviour, so it is unstable from run to run and from machine to
machine. That makes the low bands the least transferable part of the spectrum,
and they are also where the Part 1 upper limit is worst. The practical rule:
reuse an $\varepsilon$ only across the same machine family, the same operating
mode and the same mounting, and re-verify it on a sample.

## 4. The measurement report (`.report()`)

A determination ends as a *document*. The `VibrationSoundPowerResult` exposes a
`.report()` method that writes a one-page PDF fiche laid out like a sound-power
test sheet: the standard-basis line naming the applied method (the ISO/TS
7849-1 survey method with a fixed radiation factor $\varepsilon = 1$, or the
ISO/TS
7849-2 engineering method with a determined radiation factor), an optional
metadata header (client, machine/source, test environment, instrumentation,
climate, date), a per-band table (nominal octave/one-third-octave frequency,
the surface vibratory velocity level $L_v$ and the band sound-power level
$L_W$), the sound-power spectrum $L_W(f)$ with a nominal band axis, and a boxed
A-weighted sound power level $L_{W\mathrm{A}}$ (dB re 1 pW) with the total $L_W$, the
radiating area $S$ and the applied method alongside. Those two come from the
result's `sound_power_level_a` (the declarable A-weighted total) and
`total_level` (the unweighted band sum); the table columns come from
`velocity_level`, `sound_power_level` and, with `verbose=True`,
`radiation_factor`.

The metadata is supplied through a `ReportMetadata`, whose applicable fields
here are the **machine/source description** (`specimen`), the **test
environment** (`test_room`), the **client**, the **instrumentation**, the
**temperature**, **relative humidity** and **ambient pressure**, the **date of
test** (`test_date`) and the footer identity (`laboratory`, `operator`,
`report_id`, `notes`); the radiating area $S$ comes from the result itself and
is printed in the result box, together with the sound-power relation
$L_W = L_v + 10\log_{10}(S/S_0) + 10\log_{10}\varepsilon + 10\log_{10}(Z_\mathrm{c,n}/Z_{\mathrm{c},0})$ in the
basis strip. Supplying `requirement` adds a PASS/FAIL verdict against a declared
A-weighted sound-power limit (a sound-power emission is a quantity where less is
better, so the source passes at or below the limit). `verbose=True` adds the
radiation factor $\varepsilon$ column to the table. `language="es"` renders the
Spanish fiche with comma decimals. Rendering needs reportlab and, for the figure the fiche embeds, matplotlib
(`pip install "phonometry[report,plot]"`).

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

freqs = np.array([125, 250, 500, 1000, 2000, 4000], float)
lv = np.array([78.0, 82, 85, 83, 79, 74])          # surface velocity level [dB]
eps = np.array([0.20, 0.45, 0.75, 0.95, 1.00, 1.00])  # measured radiation factor
res = emission.sound_power_from_vibration(
    lv, area=1.6, radiation_factor=eps, frequencies=freqs,
)

res.report(
    "vibration_sound_power.pdf",
    metadata=ReportMetadata(
        client="Example manufacturing plant",
        specimen="Gearbox casing (steel panel)",
        test_room="Machine hall (source vibration survey)",
        instrumentation="Piezoelectric accelerometer (ISO 16063-21 calibration), s/n 0042",
        laboratory="Phonometry reference example",
        report_id="EXAMPLE-7849",
        requirement=90.0,
    ),
)   # LWA = 88.7 dB(A) re 1 pW -> declared limit 90 dB(A): PASS
```

The example fiche is regenerated with `make reports` and kept rendered in the
repository; click the preview to open the PDF.

**Report:** [ISO/TS 7849 sound power from vibration example report (PDF)](https://docs.jmrp.io/phonometry/media/iso7849_vibration_power_example.pdf): One-page ISO/TS 7849-2 sound-power-from-vibration determination fiche: a header with the client, the machine/source, the machine-hall test environment and the accelerometer and climate, the octave-band table (125 Hz to 4 kHz) of surface vibratory velocity levels Lv and radiated band sound-power levels LW, the sound-power spectrum LW(f) with a nominal band axis, the boxed A-weighted sound power level LWA = 88.7 dB(A) re 1 pW with the total LW = 90.0 dB, the radiating area S = 1.60 m2 and the engineering method, and a PASS verdict against the declared 90 dB(A) limit, closed by a basis strip stating the LW = Lv + 10 lg(S/S0) + 10 lg(epsilon) + 10 lg(411/400) relation with its fixed impedance term and the radiation-factor model.

Sound power from vibration fiche (VibrationSoundPowerResult.report), an ISO/TS 7849-2 engineering-method determination with the measured radiation factor and the boxed LWA.

## What this guide covers

- **Covered.** ISO/TS 7849-1:2009 and ISO/TS 7849-2:2009 as far as they define the
  calculation of sound power from vibration: the radiated power
  $P = Z_\mathrm{c} \, \langle v^2 \rangle \, S \, \varepsilon$ (Formula 6), the velocity
  level and its calibration conversion (Formulae 3, 8) run by `velocity_level`
  and `velocity_level_from_acceleration`, the surface mean (Formulae 10/11) run
  by `mean_velocity_level`, the extraneous-vibration correction of Table 2 run
  by `extraneous_velocity_correction`, and the sound power level with the fixed
  impedance term (Formula 12, Part 1) or a measured per-band radiation factor
  (Formula 15, Part 2) run by `radiated_sound_power_level` and
  `sound_power_from_vibration`. `radiation_factor` implements Part 2's
  single-machine Formula 8, converting an independently measured sound power
  into $\varepsilon$.

- **Not covered.** The measurement clauses of both parts (instrumentation, source installation,
  environmental conditions: clauses 5 to 7) are not *implemented* as code, and
  neither are their measurement-uncertainty clauses and informative annexes:
  they are laboratory practice, and this guide states the rules of it that
  decide whether the inputs are valid at all: the Table 1 position counts, the
  Table 2 extraneous-vibration ladder and the transducer mounting limits above.
  Part 2's clause 8 also asks for the radiation factor of a machine batch or
  family, averaging $\varepsilon_j$ over several machines and its standard
  deviation (Formulae 9, 10); only the single-machine Formula 8 is implemented,
  so pass an already-averaged $\varepsilon$ for a family determination. The
  sound power that feeds `radiation_factor` must come from an ISO 9614 intensity
  *determination*, which is [Sound Power by Intensity
  Scanning](https://docs.jmrp.io/phonometry/devices/emission/sound-power-intensity/)
  (`sound_power_intensity`, or `sound_power_intensity_precision` for grade 1); the
  [sound intensity guide](https://docs.jmrp.io/phonometry/devices/emission/intensity/) covers
  the probe and the field indicators behind it but computes no power.

## See also

- [Sound Power](https://docs.jmrp.io/phonometry/devices/emission/sound-power/): the route chooser
  this method is the last entry of, what the accuracy grades promise, and
  the ISO 4871 noise-emission declaration a result feeds.
- [Sound Power by Intensity Scanning (ISO 9614)](https://docs.jmrp.io/phonometry/devices/emission/sound-power-intensity/):
  the determination that produces the $P_j$ behind a Part 2 radiation factor.
- [Sound Intensity (p-p)](https://docs.jmrp.io/phonometry/devices/emission/intensity/): the probe,
  its residual index and the field indicators that qualify that measurement.
- [Panel Sound Insulation](https://docs.jmrp.io/phonometry/buildings/design/panel-sound-insulation/):
  the same radiation-efficiency model, written $\sigma$ instead of
  $\varepsilon$, and the coincidence physics behind it.
- [Structure-borne source characterisation](https://docs.jmrp.io/phonometry/buildings/design/structure-borne-power/):
  ISO 9611, EN 15657 and EN 12354-5, the standards that consume this kind of
  result.
- [Mechanical mobility (ISO 7626)](https://docs.jmrp.io/phonometry/vibration/structural/mechanical-mobility/):
  the transducer practice behind a measured surface velocity.
- API reference: [`emission.vibration_sound_power`](https://docs.jmrp.io/phonometry/reference/api/power/vibration-sound-power/).
- Theory: [Point mobilities and radiation efficiency](https://docs.jmrp.io/phonometry/reference/theory/vibration/#point-mobilities-and-radiation-efficiency-cremer-5-hopkins-29): the radiation efficiency that turns a surface velocity into a radiated power.

## References

- **Cremer, L., Heckl, M., & Petersson, B. A. T. (2005).** *Structure-borne sound: Structural vibrations and sound radiation at audio frequencies* (3rd ed.). Springer. https://doi.org/10.1007/b137728
  The radiation-efficiency treatment behind section 3: coincidence, the cancellation below the critical frequency and the radiation of finite plates. ISBN 978-3-540-22696-3.
- **International Organization for Standardization. (1998).** *Mechanical vibration and shock — Mechanical mounting of accelerometers* (ISO 5348:1998).
  The mounting the page asks for: the characteristic responses of Figures 5 to 10, and the maximum of 40 °C Figure 10 gives a thin layer of beeswax. The edition read is its unchanged German adoption DIN ISO 5348:1999-07; since revised as ISO 5348:2021 (https://www.iso.org/standard/78160.html).
- **International Organization for Standardization. (2009).** [*Acoustics — Determination of airborne sound power levels emitted by machinery using vibration measurement — Part 1: Survey method using a fixed radiation factor*](https://www.iso.org/standard/40537.html) (ISO/TS 7849-1:2009).
  The upper-limit method with ε = 1: the radiated power P = Z_c⟨v²⟩Sε (Formula 6), the velocity level and its calibration (Formulae 3, 8), the mean over the surface (Formulae 10/11), the extraneous correction (Table 2) and the sound power level with the fixed impedance term (Formula 12). Conformance is anchored on the standard's own worked calibration example.
- **International Organization for Standardization. (2009).** [*Acoustics — Determination of airborne sound power levels emitted by machinery using vibration measurement — Part 2: Engineering method including determination of the adequate radiation factor*](https://www.iso.org/standard/40538.html) (ISO/TS 7849-2:2009).
  The engineering method with a measured band-wise radiation factor (Formula 4/8) and its sound power level (Formula 15). Conformance includes the exact round-trip between the radiation factor and L_W = 10 lg(P/P₀).
