# Wind turbine sound at a dwelling: IEC TS 61400-11-2

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

Canonical: https://docs.jmrp.io/phonometry/environment/sources/wind-turbine-receptor/
Language: en
Alternate: https://docs.jmrp.io/phonometry/es/environment/sources/wind-turbine-receptor/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: 2026-10-01
Updated: 2026-11-16T21:36:14Z
Standards: IEC TS 61400-11-2:2024; ISO 9613-1:1993
Verified: 31 conformance checks (IEC TS 61400-11-2:2024; ISO 9613-1:1993) https://docs.jmrp.io/phonometry/reference/conformance/
License: MIT (https://github.com/jmrplens/phonometry/blob/main/LICENSE)
Summary: Wind turbine sound at a dwelling by IEC TS 61400-11-2:2024: wind shear, indoor low frequency, emergence, amplitude modulation and rating level.

---

[IEC 61400-11](https://docs.jmrp.io/phonometry/environment/sources/wind-turbine-noise/) measures what a
turbine emits, close to it and downwind. IEC TS 61400-11-2:2024 measures what
arrives at a dwelling several hundred metres away, where the wind that drives
the rotor also stirs the trees, raises the background and bends the sound on
its way. The measurement is long, weeks rather than hours, and everything is
sorted by wind speed and direction before it is averaged. This page covers the
calculations the TS writes down: the wind speed at the heights the bins need,
the bin averages with their uncertainty, the background correction, the low
frequency level indoors, the emergence, the amplitude modulation rating and the
rating level they feed.

## How the measurement goes

![Section from two wind turbines to a dwelling. The binning wind speed comes from the power curve or the nacelle anemometer, recalculated to 10 m, sampled at 1 Hz or faster and averaged over 10 s to 10 min; a mast or remote sensing device has one height within 40 % of the hub height and another at least 15 m from it; temperature and humidity are read 1 m to 1.5 m above the ground. The free-field microphone stands on a tripod inside a two-stage or oversize wind screen, five wavelengths of the lowest frequency of interest, typically 15 m to 20 m, from the dwelling. A façade microphone is the alternative only where the signal needs it, on a façade plane within 0.05 m over 1 m, its edges more than 1 m away and the turbines within 45 degrees of its normal, and never for tonal audibility, amplitude modulation or impulsivity. A wind rose shows twelve sectors 30 degrees wide from north, the first from 345 to 15 degrees, with speed bins 1 m/s wide centred on integers and the downwind sector within 45 degrees of the line from the turbines. A panel lists the amplitude modulation analysis: 100 ms L_Aeq in one-third octaves from 20 Hz to 10 kHz at least, 10 s blocks, 60 to a 10 min period, wind referred to 10 min averages, free field only](https://docs.jmrp.io/phonometry/media/diagram_wind_turbine_receptor.svg)

A free-field microphone with a two-stage or oversize wind screen on a tripod or
a pole is the preferred mounting (8.1.3). A microphone on the facade raises the
signal-to-noise ratio, if the facade is plane within ±0.05 m over 1 m round it
and the edges are more than 1 m away, but it cannot be used for tonal
audibility, amplitude modulation or impulsivity (8.1.3); a proxy position
chosen elsewhere is of limited use for them too (8.1.2). Wind speed and direction come from the
turbines themselves, from their power curves or nacelles, or from a mast, and
are brought to a 10 m reference height (9.3). Everything acoustic is then
binned in 1 m/s wind speed classes and 30° sectors (10.1). For amplitude
modulation the sound level meter logs 100 ms $L_\mathrm{Aeq}$ in one-third
octave bands from 20 Hz to 10 kHz at least, and wind is referenced to 10 min
averages (13.6.1).

## Wind speed at the heights the bins need

The power law of Equation (K.1) moves a wind speed between two heights with the
shear exponent $\alpha$, and Equation (K.2) inverts it:

$$
V_z = V_{z,\mathrm{ref}} \left(\frac{z}{z_\mathrm{ref}}\right)^{\alpha},
\qquad
\alpha = \frac{\ln(V_z / V_{z,\mathrm{ref}})}{\ln(z / z_\mathrm{ref})}
$$

Tables K.1 and K.2 evaluate both between 120 m and 10 m. The binning wind speed
of 9.3.2 is instead a hub height speed brought down to 10 m with the
logarithmic profile and the reference roughness length $z_{0\mathrm{ref}} =
0.05$ m of 3.31, which assumes neutral air; the power law through two measured
heights carries the stratification with it.

```python
from phonometry import environment

# Table K.1: 8 m/s at 120 m with alpha = 0.3 is 3.8 m/s at 10 m.
v10 = environment.power_law_wind_speed(
    8.0, height_m=10.0, reference_height_m=120.0, shear_exponent=0.3
)
# Table K.2: 5 m/s at 10 m under 8 m/s at 120 m is alpha = 0.19.
alpha = environment.wind_shear_exponent(
    8.0, 5.0, height_m=120.0, reference_height_m=10.0
)
# The binning wind speed of 9.3.2.1: 8 m/s at the hub, logarithmic, z0 = 0.05 m.
v_bin = environment.logarithmic_wind_speed(8.0, height_m=10.0, reference_height_m=120.0)
print(round(v10, 2), round(alpha, 2), round(v_bin, 2))   # 3.8 0.19 5.45

profile = environment.wind_shear_profile(4.5, 8.6, upper_height_m=120.0)
print(round(profile.shear_exponent, 2), profile.typical)   # 0.26 True
```

At a wind farm the binning wind speed is the mean of the turbines that make the
sound (9.3.2.3). The turbines are sorted by their predicted level at the
dwelling and the quietest ones are left out while the total drops by no more
than 1 dB; typically two or three remain.

```python
relevant = environment.sound_relevant_turbines(
    [33.5, 38.2, 29.4, 36.9, 25.1, 31.8, 27.0]
)
print(relevant.indices)                                  # (1, 3, 0)
print(round(relevant.total_level, 1), round(relevant.relevant_level, 1))  # 42.3 41.4
print(round(relevant.binning_wind_speed_m_s([6.1, 6.4, 5.8, 6.6, 5.5, 6.0, 5.7]), 2))  # 6.37
```

![Two panels. Left: height from 0 m to 150 m against wind speed, the power-law profile with a shear exponent of 0.26 drawn through two measured points, 4.5 m/s at 10 m and 8.6 m/s at 120 m, and the logarithmic profile with a roughness length of 0.05 m through the lower point, which bends away to lower speeds above it. Right: the predicted levels of seven turbines at a dwelling, loudest first, with the three loudest, turbines 2, 4 and 1, in the sound relevant colour and the other four grey, the total of 42.3 dB drawn as a line and the total less 1 dB as a dashed line just under it, and a line of diamonds over the bars for the energy sum of the loudest turbines taken together, 38.2 dB, 40.6 dB, 41.4 dB and on towards the total, which first reaches the dashed line at the third turbine](https://docs.jmrp.io/phonometry/media/wind_turbine_receptor_shear.svg)

*Left: the logarithmic profile of 9.3.2.1 is the neutral atmosphere; on a stable
night the measured shear is larger and the hub sees more wind than the 10 m
speed suggests. Right: leaving out the four quietest turbines drops the total
by 0.9 dB, the fifth would drop it by more, so three turbines set the binning
wind speed; the diamonds are the three loudest together reaching the dashed
line. The figure numbers the turbines from 1, so the `indices` (1, 3, 0) above
are turbines 2, 4 and 1 on it.*

**Show the code for this figure**

```python
import matplotlib.pyplot as plt

fig, (left, right) = plt.subplots(1, 2, figsize=(13.0, 5.4))
profile.plot(left)
relevant.plot(right)
plt.show()
```

## Bin averages and the background

Each bin averages its intervals by energy, Equation (1), or arithmetically for
a statistical level such as $L_{90}$, Equation (8). The type A uncertainty is
the spread about that average over $N(N-1)$, Equation (2) or (9); the type B
part is the root mean square of the per-interval uncertainties, Equations (3)
and (4), with examples in Table 3; Equation (5) combines them. The background,
measured with the turbines stopped and binned the same way, is subtracted bin by
bin under the rule of 11.7: logarithmically, Equation (6) with the uncertainty
of Equation (7), where the total is at least 3 dB above it; by a suggested 3 dB
where it is at least 0 dB but less than 3 dB above; and not at all where the
background is louder, when the turbine level cannot be determined.

```python
import numpy as np

rng = np.random.default_rng(611)
speeds = rng.uniform(2.6, 11.45, 420)              # binning wind speeds, m/s
turbine = 33.0 + 2.0 * np.minimum(speeds, 9.0)     # rated power at 9 m/s
background = 15.0 + 3.4 * speeds                   # wind in the trees
total = 10.0 * np.log10(10.0 ** (turbine / 10.0) + 10.0 ** (background / 10.0))
total_levels = np.round(total + rng.normal(0.0, 1.2, speeds.size), 1)
off_speeds = rng.uniform(2.6, 11.45, 240)           # turbines stopped
off_levels = np.round(15.0 + 3.4 * off_speeds + rng.normal(0.0, 1.5, 240), 1)

on = environment.bin_sound_levels(total_levels, speeds, type_b_uncertainty_db=0.5)
off = environment.bin_sound_levels(off_levels, off_speeds, type_b_uncertainty_db=0.5)
corrected = on.background_corrected(off)
print(corrected.wind_speeds_m_s)                   # [ 3.  4. ...  10. 11.]
print(np.round(corrected.level_differences, 1))  # 13.7 ... 3.4 2.1
print(np.round(corrected.turbine_levels, 1))     # 39.2 ... 50.4 52.2
print(corrected.regimes[-1].value)                  # three_db
```

![Two panels against wind speed from 3 m/s to 11 m/s. Left: 420 ten-minute total levels scattered from 36 dB to 58 dB and the energy average of each 1 m/s bin drawn on them as a square with its combined uncertainty of about half a decibel. Right: per bin, the total level rising from 39 dB to 55 dB, the background measured with the turbines stopped rising faster from 26 dB to 53 dB, and the turbine level after the background correction, which follows the total at low wind and falls below it as the background closes in; the 11 m/s bin, where the background is within 3 dB, is circled](https://docs.jmrp.io/phonometry/media/wind_turbine_receptor_bins.svg)

*Right: from 3 m/s to 10 m/s the total is 3.4 dB or more above the background and
the background is subtracted; at 11 m/s it is 2.1 dB above, and 11.7 suggests
taking 3 dB off the total instead. Its NOTE 1 warns that an excess over a limit
cannot then be established.*

**Show the code for this figure**

```python
fig, (left, right) = plt.subplots(1, 2, figsize=(13.0, 5.4))
on.plot(left)
corrected.plot(right)
plt.show()
```

A prediction from IEC 61400-11 sound power levels carries its own uncertainty
(10.3.4): Equation (10) weights each turbine's sound power uncertainty by its
share of the energy at the dwelling, and Equation (11) adds the 2 dB the TS
expects of ISO 9613-2 or Nord2000 downwind and 0.5 dB for the modelling.

```python
prediction = environment.predicted_receptor_level([36.9, 38.2, 33.5], 1.5)
print(round(prediction.level, 1), round(prediction.combined_uncertainty_db, 2))  # 41.4 2.55
```

## Low frequency sound indoors (Annex C)

Measuring low frequency sound inside a house for weeks is intrusive and rarely
clean, so Annex C predicts it from the sound power levels, band by band from
10 Hz to 200 Hz:

$$
L_{p,\mathrm{LF}} = L_{W,\mathrm{LF}} - 10\lg\left(l^2 + h^2\right) - 11~\mathrm{dB}
+ \Delta L_{\mathrm{g},\mathrm{LF}} - \Delta L_\mathrm{a} - \Delta L_\sigma
$$

with $l$ the horizontal distance and $h$ the hub height, 11 dB for
$10\lg(4\pi)$, the ground correction $\Delta L_{\mathrm{g},\mathrm{LF}}$ of Table C.2
for impedance class D, the air absorption $\Delta L_\mathrm{a} = \alpha
\sqrt{l^2 + h^2}$ and a facade insulation $\Delta L_\sigma$ such as the two
Danish rows of Table C.3. The turbines are summed per band, and the level is the
energy sum of the bands; the Danish limit of C.6.1 is 20 dB for the A-weighted
sum from 10 Hz to 160 Hz.

```python
power = 118.0 - np.arange(environment.LOW_FREQUENCY_BANDS_HZ.size, dtype=float)
low = environment.wind_turbine_low_frequency_level(
    np.vstack([power, power, power - 1.0]),      # three turbines, 10 Hz to 200 Hz
    distance=[650.0, 820.0, 1100.0],
    hub_height_m=120.0,
    facade_insulation=environment.LOW_FREQUENCY_FACADE_INSULATION[
        "Denmark brick or similar"
    ],
)
print(round(low.outdoor_level, 1), round(low.indoor_level, 1))   # 35.1 16.5
danish = environment.LOW_FREQUENCY_BANDS_HZ <= 160.0
print(round(10.0 * np.log10(np.sum(10.0 ** (low.indoor_levels[danish] / 10.0))), 1))  # 15.8
```

### Why Table C.2 sits under ISO 9613-1

Table C.2's caption says 70 % relative humidity and 10 °C, and the text under
it says the air absorption coefficient is calculated according to ISO 9613-1.
From 125 Hz to 200 Hz it is: 0.41, 0.59 and 0.80 dB/km are ISO 9613-1 at that
atmosphere, evaluated at the nominal band centres. ISO 9613-1's own Table 1 is
computed at the exact mid-band frequencies and prints 0.584 dB/km (0.58) at
160 Hz, whose exact mid-band is 158.5 Hz; the TS takes nominal centres here,
while its Table 7 follows the exact mid-bands. From 50 Hz to 100 Hz it is not. ISO 9613-1 at 70 % gives 0.08, 0.12,
0.19 and 0.28 dB/km at 50 Hz, 63 Hz, 80 Hz and 100 Hz; the table prints 0.07,
0.11, 0.17 and 0.26, which is 0.01 to 0.02 dB/km less.

Those lower cells are not a rounding accident. From 10 Hz to 100 Hz the row is,
cell for cell, the air attenuation of the Danish statutory order on wind turbine
noise (BEK nr. 135 of 2019, Table 1.4), which is set for **80 %** relative
humidity and 10 °C, takes the Nord2000 attenuation rates and ignores air
absorption below 25 Hz, as the DELTA report behind it recommends. Wetter air
absorbs less at these frequencies, and ISO 9613-1 at 80 % gives 0.07, 0.11, 0.17
and 0.25 dB/km there, within 0.01 dB/km of the printed cells. The table
therefore joins two atmospheres at 100 Hz: the Danish 80 % below, the caption's
70 % above. The order stops at 160 Hz with 0.38 and 0.55 dB/km at 125 Hz and
160 Hz, which the TS replaced with its own 70 % values.

![Two panels. Left: the A-weighted low frequency band levels of three turbines at a dwelling from 10 Hz to 200 Hz by Equation (C.1), outdoors rising from minus 11 dB at 10 Hz to 29 dB at 200 Hz with a total of 35.1 dB, and indoors behind the Danish brick facade of Table C.3, flattening around 6 dB to 8 dB above 40 Hz with a total of 16.5 dB. Right: each printed air attenuation cell of Table C.2 from 25 Hz to 200 Hz minus ISO 9613-1 at 10 degrees C, one line at 70 % and one at 80 % relative humidity, with a shaded strip of plus or minus 0.005 dB/km for the printed rounding; the 80 % line stays inside the strip or just above it up to 100 Hz and jumps to about 0.035 dB/km from 125 Hz, while the 70 % line dips to minus 0.02 dB/km between 50 Hz and 100 Hz and returns to the strip from 125 Hz](https://docs.jmrp.io/phonometry/media/wind_turbine_receptor_low_frequency.svg)

*Right: a cell and a computed value that differ by less than 0.005 dB/km print as
the same number. Up to 100 Hz the cells are the 80 % atmosphere's (the
0.006 dB/km at 100 Hz is the Nord2000 band rate of the Danish order, 0.26 where
ISO 9613-1 gives 0.254), and from 125 Hz the caption's 70 % one's.*

**Show the code for this figure**

```python
import warnings

from phonometry.environment.propagation.air_absorption import air_attenuation

bands = environment.LOW_FREQUENCY_BANDS_HZ
fig, (left, right) = plt.subplots(1, 2, figsize=(13.0, 5.4))
low.plot(left)
shown = bands >= 25.0
for humidity in (70.0, 80.0):
    with warnings.catch_warnings():       # ISO 9613-1 is advisory below 50 Hz
        warnings.simplefilter("ignore")
        iso = 1000.0 * np.asarray(
            air_attenuation(
                bands[shown], temperature_c=10.0, relative_humidity_percent=humidity
            )
        )
    printed = environment.LOW_FREQUENCY_AIR_ATTENUATION_DB_PER_KM[shown]
    right.semilogx(bands[shown], printed - iso, marker="o",
                   label=f"Table C.2 minus ISO 9613-1 at 10 °C, {humidity:.0f} %")
right.axhspan(-0.005, 0.005, alpha=0.2)
right.set(xlabel="Frequency [Hz]", ylabel="Difference in alpha [dB/km]")
right.legend()
plt.show()
```

The difference is at most 0.02 dB/km, a few hundredths of a decibel over the
distances Annex C is used for, but a reader who recomputes the row at the
caption's atmosphere cannot reproduce its lower half, so it is in the
[errata register](https://docs.jmrp.io/phonometry/reference/errata/). The library publishes the row as printed,
`LOW_FREQUENCY_AIR_ATTENUATION_DB_PER_KM`, and uses it by default, so its
results reproduce the TS and the Danish order; for ISO 9613-1 at another
atmosphere pass the coefficients yourself:

```python
with warnings.catch_warnings():
    warnings.simplefilter("ignore")
    iso_70 = 1000.0 * np.asarray(
        air_attenuation(bands, temperature_c=10.0, relative_humidity_percent=70.0)
    )
at_70 = environment.wind_turbine_low_frequency_level(
    np.vstack([power, power, power - 1.0]),
    distance=[650.0, 820.0, 1100.0],
    hub_height_m=120.0,
    air_attenuation_db_per_km=iso_70,
    facade_insulation=environment.LOW_FREQUENCY_FACADE_INSULATION[
        "Denmark brick or similar"
    ],
)
print(round(at_70.indoor_level - low.indoor_level, 3))   # -0.005
```

Table C.4, the reporting table, adds the distance term that Equation (C.1)
subtracts; the equation is the one implemented, and the table is in the
[errata register](https://docs.jmrp.io/phonometry/reference/errata/).

## Emergence (Annex J)

French regulation limits the emergence, the ambient sound criterion with the
turbines running less the background criterion with them stopped, per complete
wind speed class, Equation (J.1). It is not a limit on the turbine sound itself.

```python
emergence = environment.sound_emergence(
    [38.5, 41.0, 43.5], [33.0, 37.5, 41.5], wind_speeds_m_s=[5.0, 6.0, 7.0]
)
print(emergence.emergence)   # [5.5 3.5 2. ]
```

## Amplitude modulation (clause 13)

The level of wind turbine sound rises and falls once per blade passage, the
"swish" a neighbour describes, and more of it annoys more than its average
level says. Clause 13 rates it with the reference method of the Institute of
Acoustics Amplitude Modulation Working Group (13.2). The input is a series of
100 ms A-weighted levels summed over seven one-third octave bands, 50 Hz to
200 Hz, 100 Hz to 400 Hz or 200 Hz to 800 Hz (13.6.2.1.1), with three higher
bands in Annex G.3 for smaller turbines whose swish lies above 800 Hz.

```python
freqs = [50, 63, 80, 100, 125, 160, 200, 250, 315, 400, 500, 630, 800]
rng = np.random.default_rng(7)
logged = 40.0 + rng.normal(0.0, 1.0, (1000, len(freqs)))    # 100 s, unweighted bands
band_2 = environment.amplitude_modulation_band_levels(logged, frequencies=freqs, band=2, weighting="Z")
print(band_2.shape)   # (1000,)
```

**One 10 s block** (13.6.2.3). The 100 values are de-trended with a cubic fit
and transformed with a rectangular DFT of 0.1 Hz resolution into the power
spectrum of Equation (12), $S_{xx} = |F\{x\}|^2 / m^2$. The highest local
maximum inside the range where the blade passage frequency is expected is the
fundamental. Its prominence, Equation (13), is the line over the mean of the two
lines beyond each neighbour; a block without a maximum in the range, or with a
prominence under 4, is "prominence-failed". The second and third harmonics join
if a maximum sits at, or within one or two lines of, twice and three times the
fundamental and both the fundamental alone and the harmonic alone swing more
than 1.5 dB. Three lines about each kept component, with their negative
frequency mirrors, are transformed back, and the modulation depth is the $L_5 -
L_{95}$ of that series.

```python
t = np.arange(100) * 0.1
rng = np.random.default_rng(1340)
levels = np.round(
    38.0
    + 1.2 * np.sin(2 * np.pi * t / 47.0)                # the wind drifting
    + 2.0 * np.sin(2 * np.pi * 0.8 * t)                 # the swish, 0.8 Hz
    + (4.0 / 6.0) * np.sin(4 * np.pi * 0.8 * t + 0.9)   # and its harmonic
    + rng.normal(0.0, 1.0, t.size),
    1,
)
block = environment.amplitude_modulation_block(
    levels, modulation_frequency_range_hz=(0.5, 1.1)
)
print(block.status.value, block.fundamental_frequency_hz)   # valid 0.8
print(round(block.prominence, 1), block.harmonic_frequencies_hz)  # 79.9 (1.6, 2.5)
print(round(block.modulation_depth_db, 2))                  # 4.95
```

![Two panels of one 10 s block. Left: the power spectrum from 0.1 Hz to 4.9 Hz as bars, the range of the fundamental from 0.5 Hz to 1.1 Hz between dotted lines, the fundamental at 0.8 Hz standing far above everything else, the estimated second and third harmonics at 1.6 Hz and 2.4 Hz as dashed lines, and the three lines kept about the fundamental and about each harmonic filled; the title gives the fundamental of 0.8 Hz and a prominence of 79.87. Right: the de-trended 100 ms series over 10 s swinging about 3 dB either way with noise, the reconstructed series drawn over it as a smooth periodic curve, and its L5 and L95 as dashed lines; the title gives L5 minus L95 of 4.95 dB](https://docs.jmrp.io/phonometry/media/wind_turbine_modulation_block.svg)

*Left: the third harmonic is found at 2.5 Hz, one line above its estimate of
2.4 Hz, which 13.6.2.3 f) allows within two lines. Right: the reconstruction
keeps the blade passage and drops the noise.*

**Show the code for this figure**

```python
fig, (left, right) = plt.subplots(1, 2, figsize=(13.0, 5.4))
block.plot(left)
block.plot(right, kind="series")
plt.show()
```

**The 10 min rating** (13.6.3). Sixty blocks make a period; with at least 30
valid ones, the rating is the 90th percentile of their depths, with the mean and
the mode of their fundamentals. A period with fewer is rated 0 dB and gives no
modulation frequency. The rating feeds the adjustment of Figure A.1.

```python
t = np.arange(6000) * 0.1
rng = np.random.default_rng(1363)
depth = 3.0 + 2.5 * np.sin(2.0 * np.pi * t / 600.0) ** 2
gust = (t >= 240.0) & (t < 330.0)                  # 90 s of gusts, no swish
depth[gust] = 0.0
series = (
    38.0
    + 1.2 * np.sin(2 * np.pi * t / 47.0)
    + 0.5 * depth * np.sin(2 * np.pi * 0.8 * t)
    + depth / 6.0 * np.sin(4 * np.pi * 0.8 * t + 0.9)
    + rng.normal(0.0, 1.0, t.size)
)
series[gust] += rng.normal(0.0, 2.5, int(gust.sum()))
period = environment.amplitude_modulation_period(
    np.round(series, 1), modulation_frequency_range_hz=(0.5, 1.1)
)
print(period.n_valid_blocks, round(period.rating_db, 2))    # 55 5.84
print(round(period.mean_modulation_frequency_hz, 2), period.mode_modulation_frequency_hz)  # 0.8 0.8
```

**The bins** (13.6.4). Each period is a data point of its rating in each band,
binned by 1 m/s and 30° like everything else. The mean of each bin counts the
0 dB of the unrated periods and leaves out the excluded ones, and each bin
reports the band with the highest mean, with its count, the number of non-zero
ratings and the share above 3 dB, 6 dB and 9 dB (13.7).

```python
ratings = np.array(
    [[2.1, 4.6, 3.0], [0.0, 0.0, 0.0], [1.8, 5.2, 3.9],
     [2.5, 3.9, 4.4], [0.0, 0.0, 0.0], [3.1, 2.2, 5.0]]
)   # bands 1, 2 and 3 of six periods
binned = environment.bin_amplitude_modulation(ratings, [6.2, 6.4, 5.9, 7.1, 7.3, 6.8])
print(binned.wind_speeds_m_s, binned.counts)           # [6. 7.] [3 3]
print(binned.selected_bands, np.round(binned.selected_ratings_db, 2))  # [2 3] [3.27 3.13]
```

![Two panels. Left: the modulation depth of each of the sixty 10 s blocks of a 10 min period against time, rising and falling in two humps between about 2.5 dB and 6 dB, with five blocks around 280 s to 330 s marked invalid at 0 dB during a gusty minute, and the 10 min rating of 5.84 dB drawn as a line; the title reads n = 55/60. Right: the mean rating of the worst band in each 1 m/s bin of a 600-period survey against wind speed from 4 m/s to 10 m/s, rising from about 2.4 dB to 5.4 dB at 7 m/s and falling again, each point drawn as the number of the band it selected on a chip: 1 at 4 m/s and 5 m/s, 2 at 6 m/s and 7 m/s, 3 from 8 m/s to 10 m/s, with the legend naming band 1 as 50 Hz to 200 Hz, band 2 as 100 Hz to 400 Hz and band 3 as 200 Hz to 800 Hz](https://docs.jmrp.io/phonometry/media/wind_turbine_modulation_period.svg)

*Right: which band carries the swish changes with the wind, and 13.6.4 picks it
per bin; the chip over each point names the band, 1 for 50 Hz to 200 Hz, 2 for
100 Hz to 400 Hz and 3 for 200 Hz to 800 Hz. The unrated periods pull the means
down where the swish is rare.*

**Show the code for this figure**

```python
rng = np.random.default_rng(1364)
survey_speeds = rng.uniform(3.5, 10.45, 600)
survey = np.column_stack([
    np.clip(1.5 + height * np.exp(-(((survey_speeds - peak) / 2.0) ** 2))
            + rng.normal(0.0, 0.8, 600), 0.0, None)
    for peak, height in ((5.0, 2.5), (7.0, 4.0), (9.0, 4.5))
])
audible = np.exp(-(((survey_speeds - 7.0) / 2.5) ** 2))
survey[rng.uniform(0.0, 1.0, 600) < 0.35 * (1.0 - audible)] = 0.0
fig, (left, right) = plt.subplots(1, 2, figsize=(13.0, 5.4))
period.plot(left)
environment.bin_amplitude_modulation(np.round(survey, 2), survey_speeds).plot(right)
plt.show()
```

### The TS against the working group's report and code

Clause 13 says it implements the AMWG Final Report with more specific outputs
and the worst case band in each bin, and the 10 s analysis is that report's
step for step. Two differences follow from the TS's wording, and the TS is what
the library implements:

- **A period with fewer than 30 valid blocks** is rated 0 dB and kept in the
  bin mean (13.6.3, 13.6.4); the report discards it (its Figure 4.2.2). 13.6.4
  also leaves out "any excluded periods due to pf or other exclusions". Read
  literally, that would drop the very periods 13.6.3 rates 0 dB, whose blocks
  failed the prominence test, and the "including 0 values" of the same sentence
  and the count of non-zero values it asks for would mean nothing. The library
  reads it as the periods the practitioner excludes, the `excluded=` flags of
  `bin_amplitude_modulation`.
- **The band** is chosen per bin (13.6.4); the report chooses one band for a
  whole survey by regressing the bands' ratings against each other.

A period of exactly 30 valid blocks is rated, as 13.6.2.2 ("at least 50 %") and
the report ("at least 50 % (i.e. 30)") say; the bullets of 13.6.3 say "greater
than" and "less than" and leave it out, which is in the errata register.

The working group published its code (v1.4) with three sample series. It was
used here as an oracle and never as a source: the library reproduces the
printed sample results, prominence 44.64, fundamental 0.70 Hz and a 10 s rating
of 8.38 dB for the 100 Hz to 400 Hz series in the 0.4 Hz to 0.9 Hz range, and
agrees with the unmodified code on synthetic blocks and periods, as the
[conformance report](https://docs.jmrp.io/phonometry/reference/conformance/) shows. Where the two
diverge the TS decides, and the rows pin it: a period of 29 valid blocks is
rated 0 dB. One detail belongs to the code alone: it compares the frequencies
of its spectrum lines, computed as multiples of 0.1 Hz, with the range limits
as floating-point numbers, so an upper limit of 0.6 Hz, 0.7 Hz, 1.2 Hz or
1.4 Hz drops its own line (six, seven, twelve and fourteen times 0.1 Hz come
out a little above the limit; sixteen times 0.1 Hz is exactly 1.6 Hz, and that
line is kept). The library works in line indices and keeps every limit's line,
as the TS's "between" reads.

## The rating level (Annex A)

$L_\mathrm{r} = L_\mathrm{eq} + K$, with $K$ the most severe of the tonal, the
amplitude modulation and the impulsive adjustment, only one at a time (A.1).
The TS's examples are Table A.1 for tones (ISO 1996-2 Table J.1,
[`tonal_adjustment_from_mean_audibility`](https://docs.jmrp.io/phonometry/environment/assessment/environmental-levels/)),
ISO/PAS 1996-3 for impulses
([`impulse_adjustment`](https://docs.jmrp.io/phonometry/environment/assessment/impulsive-sound/)) and
Figure A.1 for amplitude modulation: nothing below a depth of 3 dB, 3 dB at
3 dB, rising to 5 dB at 10 dB and flat above.

```python
k_am = environment.amplitude_modulation_adjustment(period.rating_db)
rating = environment.wind_turbine_rating_level(
    41.0, tonal_adjustment=2.0, amplitude_modulation_adjustment=k_am
)
print(round(k_am, 2), rating.governing, round(rating.rating_level, 1))  # 3.81 amplitude_modulation 44.8
```

![Two panels. Left: the amplitude modulation adjustment of Figure A.1 against modulation depth from 0 dB to 12 dB, zero up to 3 dB, stepping to 3 dB there, rising linearly to 5 dB at 10 dB and flat beyond, with the 10 min rating of 5.84 dB marked on it at an adjustment of 3.8 dB. Right: the three candidate adjustments as bars, tonal 2 dB, amplitude modulation 3.8 dB outlined and hatched as the governing one, which the legend names as the one applied, the most severe of A.1, impulsive zero, under the title L_r = 41.0 + 3.8 = 44.8 dB](https://docs.jmrp.io/phonometry/media/wind_turbine_receptor_rating.svg)

**Show the code for this figure**

```python
depths = np.linspace(0.0, 12.0, 481)
fig, (left, right) = plt.subplots(1, 2, figsize=(13.0, 5.4))
left.plot(depths, environment.amplitude_modulation_adjustment(depths))
left.plot([period.rating_db], [k_am], "o")
left.set(xlabel="Modulation depth [dB]", ylabel=r"Adjustment $K_\mathrm{am}$ [dB]")
rating.plot(right)
plt.show()
```

## Where to stop looking for tones (12.5.2.4)

Tones in a band that ISO 9613-1 attenuates by 20 dB or more over the distance
to the nearest turbine are typically of no concern, and every higher band is
attenuated more, so the lowest such band sets the top of the search. Table 7
lists it for five atmospheres and four distances; the function computes it for
any.

```python
search = environment.upper_tone_search_frequency(
    600.0, temperature_c=10.0, relative_humidity_percent=50.0
)
print(search.upper_frequency_hz)   # 4000.0, as Table 7 prints
```

The row of Table 7 printed for 10 °C and 80 % holds the frequencies the rule
gives at −10 °C and 80 % (not those of the table's own −10 °C row, which is at
100 %); at 10 °C and 80 % the function gives 10, 6.3, 5 and 4 kHz where the table prints
10, 5, 2.5 and 2 kHz. That row, with eight other printed defects of the TS, is
in the [errata register](https://docs.jmrp.io/phonometry/reference/errata/).

![The ISO 9613-1 attenuation over 600 m at 10 degrees C and 50 % relative humidity for each one-third octave band from 50 Hz to 10 kHz on logarithmic axes, rising from 0.06 dB at 50 Hz to about 130 dB at 10 kHz, the 20 dB line dashed, and the upper search frequency of 4 kHz marked where the attenuation first reaches it](https://docs.jmrp.io/phonometry/media/wind_turbine_tone_search.svg)

**Show the code for this figure**

```python
search.plot()
plt.show()
```

## What this guide covers

- **Covered.** The calculations IEC TS 61400-11-2:2024 writes down: the wind speed between
  heights of Annex K (Equations (K.1) and (K.2), Tables K.1 to K.3) and the
  binning wind speed of 9.3.2.1 and 9.3.2.3; the bin averages and uncertainties of
  Equations (1) to (9) with Table 3, the background correction of Equations (6)
  and (7) under the rule of 11.7, and the prediction uncertainty of Equations (10)
  and (11); the low frequency level of Equation (C.1) with Tables C.1 to C.3 and
  C.5; the emergence of Equation (J.1); the amplitude modulation of clause 13, the
  10 s block, the 10 min rating and the bins, with the higher bands of Annex G.3;
  the rating level of A.1 with the adjustment of Figure A.1; and the upper tone
  search frequency of 12.5.2.4. Every result plots.

- **Not covered.** The measurement campaign itself: the selection of intervals, the filtering of
  extraneous sound and the audio auditioning of Annex F are the practitioner's.
  The tonal audibility procedure of 12.5 (3 s spectra, the frequency grouping of
  12.5.2.6 and the 1 min mean of 12.5.2.7) is not implemented; the library's
  tonal audibility is that of ISO/PAS 20065:2016 on the
  [tone audibility](https://docs.jmrp.io/phonometry/perception/psychoacoustics/tone-audibility/) page.
  Nor are the impulsivity of clause 14 beyond the ISO/PAS 1996-3 adjustment, the
  infrasound of Annex D, the Swedish and Danish criteria as verdicts, or the
  Nord2000 propagation Annex C builds on. The faster rotors of Annex G.2, whose
  fundamental modulation frequency lies above 1.6 Hz, are outside it too: the
  functions refuse a range above the 1.6 Hz of 13.2.

## See also

- [Wind-turbine noise: sound power and tonal audibility](https://docs.jmrp.io/phonometry/environment/sources/wind-turbine-noise/): the IEC 61400-11 emission whose sound power levels feed Annex C and the prediction of 10.3.4.
- [Outdoor Sound Propagation](https://docs.jmrp.io/phonometry/environment/propagation/outdoor-propagation/): ISO 9613-1 air absorption and the ISO 9613-2 chain.
- [Impulsive-sound prominence](https://docs.jmrp.io/phonometry/environment/assessment/impulsive-sound/): the ISO/PAS 1996-3 adjustment of A.4.
- [Fluctuation strength and roughness](https://docs.jmrp.io/phonometry/perception/psychoacoustics/sound-quality/): the psychoacoustic view of modulation.
- API reference: [`environment.assessment.wind_turbine_receptor`](https://docs.jmrp.io/phonometry/reference/api/environment/wind-turbine-receptor/) and [`environment.assessment.wind_turbine_modulation`](https://docs.jmrp.io/phonometry/reference/api/environment/wind-turbine-modulation/).

## References

- **Institute of Acoustics, Amplitude Modulation Working Group. (2016).** *A method for rating amplitude modulation in wind turbine noise*.
  Final Report, Version 1, of 9 August 2016: the reference method clause 13 implements, with its 10 s and 10 min methodology. Where the TS departs from it, the TS is followed.
- **Institute of Acoustics, Amplitude Modulation Working Group. (2016).** [*IOA AM code*](https://sourceforge.net/projects/ioa-am-code/). SourceForge.
  The working group's reference code (v1.4) and its three sample series, the address the TS prints in 13.6.1. Used as an oracle only: the conformance rows reproduce its sample results and its output on synthetic blocks and periods.
- **International Electrotechnical Commission. (2024).** *Wind energy generation systems – Part 11-2: Acoustic noise measurement techniques – Measurement of wind turbine sound characteristics in receptor position* (IEC TS 61400-11-2:2024).
  The implemented edition (Edition 1.0, 2024-03): Annex K and 9.3.2 for the wind speeds, 10.3 and 11.7 for the bins and the background, Annex C for the low frequency level, Annex J, Annex A and 12.5.2.4, and the amplitude modulation of clause 13. Nine of its printed defects are in the errata register.
- **International Organization for Standardization. (1993).** [*Acoustics — Attenuation of sound during propagation outdoors — Part 1: Calculation of the absorption of sound by the atmosphere*](https://www.iso.org/standard/17426.html) (ISO 9613-1:1993).
  The air absorption of 12.5.2.4 and of the caption of Table C.2.
- **Miljøministeriet. (2019).** *Bekendtgørelse om støj fra vindmøller* (BEK nr. 135 af 07/02/2019).
  The Danish statutory order whose Table 1.4 is the source of the ground correction, the lightweight facade row and, from 10 Hz to 100 Hz, the air attenuation of Table C.2, set for 10 °C and 80 % relative humidity.
