# In-situ Road-Surface Absorption

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

Canonical: https://docs.jmrp.io/phonometry/materials/surfaces/road-absorption/
Language: en
Alternate: https://docs.jmrp.io/phonometry/es/materials/surfaces/road-absorption/index.md
Version: phonometry 4.0.0
Install: pip install "phonometry[full]"
Author: José Manuel Requena Plens (https://orcid.org/0000-0003-1250-6212)
Published: 2026-07-28
Updated: 2026-11-16T21:36:13Z
Standards: ISO 13472-1:2002; ISO 13472-2:2010
Verified: 3 conformance checks (ISO 13472-1:2002; ISO 13472-2:2010) https://docs.jmrp.io/phonometry/reference/conformance/#in-situ-road-absorption-iso-13472
License: MIT (https://github.com/jmrplens/phonometry/blob/main/LICENSE)
Summary: Road-surface absorption measured in situ: ISO 13472-1 subtraction with the Adrienne window, the ISO 13472-2 spot tube, and how to choose between them.

---

How much sound a road surface absorbs decides how loud its traffic is, and a
core cannot be cut out of a motorway and carried to a laboratory. ISO 13472
measures the absorption *in situ*, on the pavement as built, twice over. The
**subtraction technique** of Part 1 fires an impulse from a loudspeaker down
onto an extended surface and separates the incident and reflected components
of one microphone's impulse response in time; the **spot method** of Part 2
seals a short tube onto a small patch and hands the mathematics to the
impedance-tube transfer-function method. This guide covers both, from the
mandatory geometry and the Adrienne window to the sampled area and the
plane-wave limits of the tube, and closes with the choice between them.

## 1. Subtraction technique (ISO 13472-1)

Out in the field there is no reverberation room. ISO 13472-1 measures the sound
absorption of a road surface (or any extended flat surface) *in situ* by firing
an impulse from a loudspeaker at height $d_\mathrm{s}$ down onto the surface and recording
the impulse response at a microphone at height $d_\mathrm{m}$. The **incident** and
**reflected** components are separated in time with an Adrienne window; their
transfer function gives the reflection factor and hence the absorption.

![ISO 13472-1 in-situ road absorption by the subtraction technique: a loudspeaker at 1.25 m and a microphone at 0.25 m above the road surface, with the direct and road-reflected ray paths and a free-field reference measurement, the reflected component isolated by an Adrienne time window](https://docs.jmrp.io/phonometry/media/diagram_insitu_subtraction.svg)

**Geometrical spreading (Clause 4.1).** The reflected wave travels farther than
the direct wave, so it is attenuated by the geometrical-spreading factor

$$
K_\mathrm{r} = \frac{d_\mathrm{s} - d_\mathrm{m}}{d_\mathrm{s} + d_\mathrm{m}},
$$

which equals $2/3$ for the mandatory geometry $d_\mathrm{s} = 1.25$ m, $d_\mathrm{m} = 0.25$ m.
Both heights are held to within 0.005 m through the averaging, and the
acoustic centres of source and microphone lie on one line normal to the
surface with the microphone axis parallel to it. The absorption follows from
the windowed incident and reflected spectra $H_\mathrm{i}$, $H_\mathrm{r}$:

$$
\alpha(f) = 1 - \frac{1}{K_\mathrm{r}^2}\left|\frac{H_\mathrm{r}(f)}{H_\mathrm{i}(f)}\right|^2.
$$

### 1.1 What is subtracted

The microphone hears the direct wave and the road reflection in the same
impulse response, and the mandatory geometry puts them only
$2 d_\mathrm{m} / c \approx 1.5$ ms apart, far too little to gate apart at the bottom
of the reported band. ISO 13472-1 therefore requires the **subtraction
technique** (Clause 4.2) rather than plain time windowing, and that is the
step the two arrays this page's functions consume come out of:

1. Rig the source and the microphone so that their separation **cannot**
   change: the standard asks for a fixed, stable mechanical connection
   between them, because the subtraction is a cancellation of two nominally
   identical waveforms.
2. Measure over the road and average until the impulse response is stable:
   **at least 50** responses (Clause 4.3), which is what beats the background
   noise and the meteorological variation.
3. Repeat the identical measurement with the rig taken away from any
   reflecting surface. That free-field response **is** $h_\mathrm{i}$. Correct any
   residual time shift between the two takes, and make them within a short
   time of each other (the standard recommends under 10 minutes) because
   the delay is temperature dependent.
4. Subtract: $h_\mathrm{r} = h_\text{road} - h_\mathrm{i}$. Everything the two takes share
   (the direct wave, the loudspeaker, the microphone, the whole chain)
   cancels, and what survives is the surface reflection alone.
5. Window $h_\mathrm{i}$ and $h_\mathrm{r}$ with the **same** window applied at their own
   arrival times ($T_{\mathrm{w},\text{direct}} = T_{\mathrm{w},\text{reflected}}$), transform,
   and form $\alpha$ with $K_\mathrm{r}$.

Two acceptance conditions ride on that chain. The signal-to-noise ratio must
exceed **10 dB in every one-third-octave band from 250 Hz to 4 kHz**
(Clause 7.4), and no calibration of level is needed at all: the method is a
ratio of two transfer functions from the same chain (Clause 6.1). But the
Annex B reference measurement of section 1.2 is not optional for the same
reason. The snippets below start from an `hi` and an `hr` that stand in for
that pair; in the field they are the free-field reference and the windowed
road reflection, in that order.

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

# A band-limited incident impulse response and a synthetic road reflection
# hr = Kr * r0 * delayed(hi): a reflection of magnitude r0 = 0.4, delayed by the
# extra path, and scaled by the geometrical-spreading factor Kr.
fs, n = 48000.0, 4096
t = np.arange(n) / fs
hi = np.zeros(n)
hi[:64] = np.hanning(64) * np.cos(2.0 * np.pi * 1500.0 * t[:64])

kr = materials.geometric_spreading_factor()          # (ds - dm)/(ds + dm) = 2/3
hr = kr * 0.4 * np.roll(hi, 96)

# Narrow-band absorption, then reduced to one-third octaves over 250-4000 Hz.
alpha = materials.in_situ_absorption_coefficient(hi, hr)   # 1 - (1/Kr^2)|Hr/Hi|^2
freq = np.fft.rfftfreq(n, 1.0 / fs)
bands = materials.one_third_octave_absorption(alpha, frequencies=freq)
centres, band = bands.frequencies, bands.absorption
print(round(kr, 4))                # 0.6667
print(round(float(band[2]), 3))    # 0.84  (alpha = 1 - 0.4^2 = 0.84)
```

### 1.2 The reference surface (Annex B)

The raw ratio $|H_\mathrm{r}/H_\mathrm{i}|^2/K_\mathrm{r}^2$ assumes that the electro-acoustic chain and
the geometry factor cancel exactly. They do not: the loudspeaker and the
microphone have their own responses along paths of different length, and the
mandatory $K_\mathrm{r} = 2/3$ is a plane-wave idealisation of a finite geometry. At
the bottom of the band, where a dense road absorbs a few percent, that
residual error is comparable with the quantity being measured, which is why
Annex B makes a **reference measurement mandatory**, not advisory.

The fix is one extra measurement per set-up. Lay a highly reflecting plate
(plane, smooth, dense and jointless, and large enough to cover the whole
maximum sampled area) on the same spot, and repeat the entire procedure over
it with the rig geometry unchanged. The measured pressure reflection factor of
the reference is then the chain's own error function $e(f)$, because the plate
itself reflects everything; dividing the road's measured factor by it removes
$e(f)$ and, since the geometry is identical, the $K_\mathrm{r}$ factor with it:

$$
\alpha_\text{road}(f) = 1 -
  \left|\frac{Q_{p,\text{road,meas}}(f)}{Q_{p,\text{ref,meas}}(f)}\right|^2 .
$$

```python
import numpy as np

# `materials` is imported by the block above. Two complex pressure reflection
# factors from the identical set-up: one over the road, one over the reference
# plate. Their ratio is Annex B.
freq = np.array([250.0, 500.0, 1000.0, 2000.0, 4000.0])
q_road = np.array([0.97, 0.94, 0.88, 0.72, 0.55]) * np.exp(-0.05j * freq / 1e3)
q_ref = np.array([1.02, 1.01, 1.00, 0.99, 0.98]) * np.exp(-0.05j * freq / 1e3)
print(np.round(materials.absorption_reference_corrected(q_road, q_ref), 3))
# [0.096 0.134 0.226 0.471 0.685]

# Uncorrected, the same road data would read:
print(np.round(1.0 - np.abs(q_road) ** 2, 3))
# [0.059 0.116 0.226 0.482 0.698]
```

The plate must itself be verified: a sample of its material is measured in an
impedance tube per ISO 10534-1 or ISO 10534-2 to confirm an absorption
coefficient below **0.05** over the reported range, which is what licenses the
assumption $|Q_{p,\text{ref}}| = 1$. A portable reference surface must also be
rigid enough that it cannot resonate. In practice the reference is the first
thing to reach for when a measured spectrum shows small negative absorptions
in the lowest bands: as the printed values above show, the correction moves
the 250 Hz band by more than the value itself.

**Adrienne window (Clause 6.4).** The time window that isolates the reflection
mandates only a sharp leading edge, a 5 ms flat portion and a cosine-squared or
Blackman-Harris trailing edge; the exact durations are reported per measurement,
not fixed, so they are configurable here.

```python
from phonometry import materials

# Default: 0.5 ms leading edge, 5 ms flat top, 5 ms Blackman-Harris trailing.
w = materials.adrienne_window(48000.0)
print(w.shape[0])          # 504 samples at 48 kHz
print(round(float(w.max()), 3))   # 1.0  (flat top and edges meet at unity)
```

That length is the method's single free parameter, and it is a trade with two
ends. The **total** window duration sets the lowest frequency the transform
can resolve: 10.5 ms is about 95 Hz, and a 220 Hz lower limit needs roughly
5 ms of flat portion, which is why the reported band starts at 250 Hz and not
lower. Lengthening it buys low-frequency validity, and every millisecond added
also admits reflections from the mast, the vehicle, the operator and anything
else within the corresponding radius, and enlarges the patch of road that
contributes to the answer. The practical rule: choose the longest window that
stays clear of the first parasitic arrival visible in the raw impulse
response, apply the same window to the free-field reference and to the
reflection, and **report the shape and the durations with the result**, since
the standard fixes neither and two laboratories using different windows are
not measuring the same thing. A spectrum that rises implausibly in the lowest
reported bands is the usual sign of a window that is too short, or of a
parasitic arrival inside it.

![Three stacked panels on a common time axis: the free-field reference impulse response with the Adrienne window placed on the direct sound; the response measured over the road where the direct and reflected arrivals overlap, their 1.47 millisecond separation dimensioned, the subtracted surface reflection and the 10.5 millisecond Adrienne window drawn over it with a parasitic arrival just outside the gate; and the resulting windowed incident and reflected spectra with the 95 hertz reciprocal-window limit marked and the region below the reported 250 hertz band shaded](https://docs.jmrp.io/phonometry/media/adrienne_window.svg)

*The whole method in three traces. The two arrivals are 1.47 ms apart on the
mandatory geometry, so no gate separates them; the middle panel's red trace is
what subtraction leaves once the free-field reference has been taken away.
The green gate is the Adrienne window at its default 0.5 + 5 + 5 ms: its
total duration sets the 95 Hz floor marked in the bottom panel, and its 5 ms
flat portion is the window width Annex A turns into the 1.34 m radius of road
that has to be clear. The parasitic arrival at 15.8 ms is outside the gate; it
would not be if the window were lengthened to reach lower in frequency.*

**Show the code for this figure**

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

# `materials` is imported above. The window itself, placed at the reflected
# arrival of the mandatory geometry (2 dm / c after the direct sound).
fs = 48000.0
w = materials.adrienne_window(fs)
t = np.arange(w.size) / fs + (1.5 / 340.0) - 0.5e-3
fig, ax = plt.subplots()
ax.plot(t * 1e3, w)
ax.axvline(1.0 / 340.0 * 1e3, linestyle=":", label="direct arrival")
ax.axvline(1.5 / 340.0 * 1e3, linestyle="--", label="reflected arrival")
ax.set_xlabel("Time [ms]")
ax.set_ylabel("Window")
ax.legend()
plt.show()
```

**End-to-end spectrum.** `in_situ_absorption_spectrum` runs the whole chain (the
windowed incident and reflected impulse responses to the narrow-band absorption
and on to one-third-octave bands) and returns a plottable
`InSituAbsorptionResult`:

```python
import numpy as np
from phonometry import materials
from scipy.signal import firwin, lfilter

# A synthetic-but-realistic measurement. hi is a unit incident impulse; the road
# reflection hr = Kr * r0 * roll(hi, shift) uses the geometrical-spreading
# factor Kr, a mildly frequency-dependent r0 (a gentle low-pass, so a porous
# surface reflects less as frequency rises) and the reflected-path delay
# shift = round(2 dm / c * fs).
fs, n = 48000.0, 8192
kr = materials.geometric_spreading_factor()           # (ds - dm)/(ds + dm) = 2/3
hi = np.zeros(n)
hi[0] = 1.0
taps = firwin(41, 1200.0, fs=fs)
taps = taps / taps.sum()
shift = int(round(2.0 * 0.25 / 340.0 * fs))     # reflected-path delay 2 dm / c
hr = kr * 0.85 * np.roll(lfilter(taps, 1.0, hi), shift)

result = materials.in_situ_absorption_spectrum(hi, hr, fs)
print(result.frequencies[[0, -1]].astype(int))     # [ 250 4000]
print(np.round(result.absorption[[0, 6, 12]], 2))  # [0.31 0.65 1.  ]
result.plot()   # alpha(f) bar chart over 250-4000 Hz (needs matplotlib)
```

![An in-situ one-third-octave road-surface absorption spectrum computed by the reflection-factor route from a synthetic road reflection, rising from about 0.3 at 250 Hz to near 1.0 above 2 kHz](https://docs.jmrp.io/phonometry/media/insitu_absorption.svg)

*The absorption rises with frequency because the surface reflects less of the
high-frequency energy, exactly as the low-pass reflection factor $r_0(f)$
dictates through $\alpha = 1 - (1/K_\mathrm{r}^2)\,|H_\mathrm{r}/H_\mathrm{i}|^2$.*

**Show the code for this figure**

```python
import matplotlib.pyplot as plt
import numpy as np
from scipy.signal import firwin, lfilter
from phonometry import materials

# A synthetic-but-realistic measurement. hi is a unit incident impulse; the road
# reflection hr = Kr * r0 * roll(hi, shift) uses the geometrical-spreading
# factor Kr, a mildly frequency-dependent r0 (a gentle low-pass, so a porous
# surface reflects less as frequency rises) and the reflected-path delay
# shift = round(2 dm / c * fs).
fs, n = 48000.0, 8192
kr = materials.geometric_spreading_factor()           # (ds - dm)/(ds + dm) = 2/3
hi = np.zeros(n)
hi[0] = 1.0
taps = firwin(41, 1200.0, fs=fs)
taps = taps / taps.sum()
shift = int(round(2.0 * 0.25 / 340.0 * fs))     # reflected-path delay 2 dm / c
hr = kr * 0.85 * np.roll(lfilter(taps, 1.0, hi), shift)
result = materials.in_situ_absorption_spectrum(hi, hr, fs)

# result is the InSituAbsorptionResult computed above. One line:
result.plot()
plt.show()

# By hand: a bar chart of alpha over the one-third-octave bands.
freqs = result.frequencies
positions = np.arange(freqs.size)
fig, ax = plt.subplots()
ax.bar(positions, np.nan_to_num(result.absorption), width=0.7, color="#1f77b4")
ax.set_xticks(positions)
ax.set_xticklabels([f"{f:g}" for f in freqs], rotation=45, ha="right")
ax.set_xlabel("Frequency [Hz]")
ax.set_ylabel("Absorption coefficient alpha")
ax.set_ylim(0.0, 1.0)
ax.set_title("In-situ road-surface absorption (ISO 13472-1)")
plt.show()
```

**Maximum sampled area (Annex A).** The finite time window limits how much of the
surface contributes to the reflection. The maximum sampled area is a circle whose
radius the library computes from the geometry and window width; the Annex A worked
example ($d_\mathrm{s} = 1.25$ m, $d_\mathrm{m} = 0.25$ m, $c = 340$ m/s, 5 ms flat window) gives
about 1.34 m.

```python
from phonometry import materials
print(round(materials.max_sampled_area_radius(5.0e-3), 3))   # 1.343  (metres)
```

The whole arrangement fits in one to-scale drawing. `plot_in_situ_geometry`
draws the standard set-up with that sampled radius on the surface, and a
measured `InSituAbsorptionResult` that retained its heights redraws its own
with `result.plot_geometry()`.

![To-scale side view of the in-situ absorption set-up: the loudspeaker on its mast 1.25 m above the hatched road surface, the microphone on the same vertical at 0.25 m, the dashed surface-reflected path below it and the 1.34 m radius of the sampled area dimensioned on the surface](https://docs.jmrp.io/phonometry/media/insitu_setup_geometry.svg)

*The standard geometry to scale: source at 1.25 m, microphone at 0.25 m on the
same vertical, and the 5 ms window turned into the 1.34 m radius of road that
actually contributes to the reflection.*

**Show the code for this figure**

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

# The standard geometry: source at 1.25 m, microphone at 0.25 m, and the
# 1.34 m sampled-area radius of the 5 ms window.
materials.plot_in_situ_geometry()
plt.show()

# A measured spectrum retains its heights and redraws its own set-up:
#   result = materials.in_situ_absorption_spectrum(hi, hr, fs)
#   result.plot_geometry()
```

### 1.3 Before you press record

That radius is not a number to admire; it is a clearance you walk. Compute it,
pace it out, and check what is inside. ISO 13472-1 Clauses 7 and 8 give the
gates that decide whether a measurement may be made at all, and they, not the
signal processing, are what usually invalidates a roadside campaign.

- **The maximum sampled area must be clear.** Any object other than the
  pavement (a fence, a rock, a barrier, a parked car) is a reflecting object.
  None may be inside the sampled area, and anything outside it must be farther
  from the microphone than $d_\mathrm{s}$, that is farther than 1.25 m. The microphone
  stand itself must not influence the measurement.
- **The surface must be visually homogeneous** and free of changes in the
  material, and it must be **dry**. On a surface with a significant void
  content, verify that the pores are dry too, not just the surface. Wet
  measurements are legitimate for research, and their results cannot be used
  to classify or qualify a surface.
- **Wind** at microphone height must not exceed 5 m/s.
- **Temperature**: ambient air between 0 °C and 35 °C, and the road surface
  between 0 °C and 50 °C.
- **The reference surface of section 1.2 must cover at least the same
  sampled area**, so the radius sizes the plate you have to carry as well as
  the clearance you have to find.

**The measuring system (Clauses 5 and 6).** The whole system, end to end,
must meet at least the requirements of a type 2 instrument of IEC 60651
(Clause 5.1). It is an electroacoustic chain: a signal generator, a power
amplifier and a loudspeaker, a microphone with its amplifier, and an analyser
that can cross-correlate and pass between the time and the frequency domains.
The loudspeaker is a single or a coaxial cone in a closed cabinet, with a
frequency response smooth enough that its impulse response in a free field is
no longer than 2 ms (Clause 5.2). The test signal is short and repeatable,
with a low peak-to-RMS ratio (the crest factor), typically below 2, and enough
energy in every one-third-octave band from 250 Hz to 4 kHz: a maximum-length
sequence or a short sweep (Clause 5.3). The overall impulse response is recovered by
cross-correlating the electrical source signal with the microphone signal
(Clause 6.3), and the sampling frequency is above 40 kHz (Clause 6.2), because
the subtraction needs the exact waveform to detect a change of delay.

**The run, in order (Clause 9).** Clause 9 puts the measurement in fifteen items,
a) to o), and the order is part of the method. The numbers below follow its
letters.

1. **Check the surface and the weather** against Clauses 8.1 to 8.3, the gates
   above. If they are not met, the measurement cannot be carried out.
2. **Place the rig** as Clause 7.2 asks: the source at 1.25 m and the
   microphone at 0.25 m above the surface, on one line normal to it. The
   safety note of Clause 7.5 applies here: on a road with traffic the method
   can be hazardous, and the standard leaves the safety practice to its user.
3. **Compute the radius of the maximum sampled area** (Annex A) and check that
   no reflecting object is inside it. If one is, the measurement cannot be
   performed.
4. **Select the sound source and the test signal** (Clauses 5.2 and 5.3).
5. **Generate the test signal.**
6. **Sample the total signal** the microphone receives, at the sampling
   frequency of Clause 6.2.
7. **Average** the microphone response until the impulse response is stable:
   at least 50 averages (Clause 4.3).
8. **Record the free-field impulse response** with the rig taken away from any
   reflecting surface that could influence it, and the geometry kept as it was
   (Clause 4.2).
9. **Isolate the reflected path** by the subtraction technique. A suitable
   time window cancels the parasitic reflections (Clause 6.4).
10. **Window and transform.** Multiply the direct and the reflected impulse
    responses by the time window and compute the power spectra of the two by
    Fourier transform. Keep the complex spectra if the complex reflection
    factor is wanted.
11. **Compute the sound power reflection factor**, with the geometrical
    spreading factor $K_\mathrm{r}$ (Clause 4.1 and Annex C).
12. **Repeat items 1 to 11 on the reference surface** and apply Annex B, the
    correction of section 1.2.
13. **Average the narrow-band absorption linearly into one-third-octave
    bands** (Clause 4.1 and Annex B). That average is the absorption
    coefficient of the road surface.
14. **Repeat at other points** of the road surface if necessary.
15. **Write the report** (Clause 11; Annex E prints an example).

**What the report carries (Clauses 10 and 11).** Clause 11 lists fifteen items
as well. Besides the reference to the standard, the testing organization, the
date and place and the signature of the person responsible, they are:

- the test site, as a drawing or pictures that show the road surface, the
  set-up and any reflecting object near the maximum sampled area;
- the road surface: its age, the measurement conditions and its composition
  (number of layers, thicknesses, material, porosity);
- its dryness and its temperature;
- the weather during the test: wind speed and direction, air and road surface
  temperatures;
- the test arrangement, on a scale drawing or a dimensioned sketch with the
  positions of the source and the microphone;
- the equipment used to measure and to analyse (name, type, serial number and
  manufacturer) and a description of the sound source;
- the type and characteristics of the anti-aliasing filter, and the sample
  rate;
- the shape and the lengths of the time windows;
- the results, as a graph and as a table of the absorption coefficient in
  one-third-octave bands from 250 Hz to 4 kHz, rounded to two decimal places
  (narrow bands may be added);
- the uncertainty of the results.

Clause 10 expects a standard uncertainty of 0.05 over the whole frequency
range under normal conditions, and more over the full range of temperature and
wind the standard allows or on thick absorptive layers. The report states it
as an expanded uncertainty for a coverage probability of 95 %; the standard's
example is $U = 0.1$ with $k = 2$. A result reported without the window shape
and durations, and without the reference measurement, is not an ISO 13472-1
result.

### 1.4 When the site forces an angle

Everything above is the **mandatory normal-incidence geometry**: source and
microphone on the same vertical over the surface. Annex F provides for an
oblique arrangement, and the library follows it. Two things change and both
are exposed:

- The geometrical-spreading factor acquires an angular form,
  $K_{\mathrm{r},\theta}^2 = 1 - \cos^2\theta\,(1 - K_\mathrm{r}^2)$, computed by
  `geometric_spreading_factor_angle` and used automatically when
  `in_situ_absorption_spectrum` is given an `incidence_angle_rad`. Both take the
  angle in **radians**, unlike the degrees the rest of this page quotes.
- The sampled area stops being the circle in the figure above. It becomes the
  intersection of the surface with an ellipsoid of revolution whose foci are
  the source and the microphone, so it is an **ellipse stretched along the
  direction of incidence**, whose major axis `msa_major_axis` returns from the
  window width and the projected source-to-microphone distance. At 30° the
  1.34 m circle becomes an ellipse 3.43 m along its major axis, so the clear
  zone to be checked grows with the angle and the measurement averages over
  more pavement, and less symmetrically.

```python
import numpy as np

# `materials` is imported above. Thirty degrees off normal, with the same two
# heights: the spreading factor rises towards one and the sampled area stops
# being a 1.34 m circle. Both helpers take the angle in RADIANS.
print(round(float(materials.geometric_spreading_factor()), 4))           # 0.6667
print(round(float(materials.geometric_spreading_factor_angle(
    np.radians(30.0))), 4))                                              # 0.7638

# The projected source-to-microphone distance on the surface, (ds + dm) tan(theta),
# is what stretches the ellipse.
dp = 1.5 * np.tan(np.radians(30.0))
print(round(dp, 3))                                                      # 0.866
print(round(float(materials.max_sampled_area_radius(5.0e-3)), 3))        # 1.343
print(round(float(materials.msa_major_axis(5.0e-3, dp)), 3))             # 3.432
```

Keep the normal-incidence geometry wherever the site allows it: it is the
arrangement the comparison with the spot method assumes, and the one published
road data are quoted for.

## 2. Spot method (ISO 13472-2)

For smaller patches, ISO 13472-2 seals a short circular tube onto the surface and
measures the absorption with the two-microphone transfer-function method of
ISO 10534-2. The library provides the spot-method geometry and validity helpers;
the transfer-function DSP itself is the impedance-tube routine
`two_microphone_impedance` (see the [Impedance Tube guide](https://docs.jmrp.io/phonometry/materials/absorbers/impedance-tube/)).

![ISO 13472-2 spot method: a short circular tube sealed onto the road surface with a loudspeaker at the top and two microphones flush in the tube wall at spacing s, measuring absorption over 250 to 1600 Hz via the ISO 10534-2 two-microphone transfer-function method](https://docs.jmrp.io/phonometry/media/diagram_spot_tube.svg)

The apparatus is a portable impedance tube stood on end: a loudspeaker at the
top, two flush-mounted microphones at spacing $s$ in the tube wall, and the
road surface itself as the sample. A test fixture with a ring of deformable
material (a rubber O-ring or similar) makes an airtight seal between the tube
mouth and the pavement, so the patch under the mouth, not the leak around it,
terminates the tube; the underside of that sealing device defines the plane
of reference at which the reflection factor is evaluated. The method is
written for **reflective** surfaces, dense asphalts and test tracks per
ISO 10844 rather than open porous layers: the standard declares it unreliable
once the measured absorption coefficient exceeds 0.15, and no longer calls a
surface reflective above 0.10, because on an absorbing surface the reflected
wave the two microphones must resolve all but vanishes. The tube samples only
the circle under its mouth, so one reading is a spot in the statistical sense
too: on regular roads the standard spaces measurement positions over the
area of interest with an emphasis on the wheel tracks (as a guideline, one
position every 10 m), on a surface that must be smooth, flat, free of debris
and dry.

**Plane-wave limits (Clause 5.4).** The tube supports only plane waves below

$$
f_\mathrm{u} = 0.58\,\frac{c_0}{d},
$$

with $d$ the tube diameter, and the microphone spacing $s$ must sit between
$0.05\,c_0/f_\mathrm{min}$ and $0.45\,c_0/f_\mathrm{max}$. The reported range is the
one-third-octave bands 250–1600 Hz, and the standard's own definition of a
frequency range is by band *edges*: those centre frequencies mean 220 Hz to
1800 Hz in narrow bands, which is where the two arguments below come from.

```python
from phonometry import materials

# Upper usable frequency of a 100 mm tube and the valid spacing window.
print(round(materials.spot_tube_upper_frequency(0.100, speed_of_sound=343.0), 1))      # 1989.4 Hz
# 220 Hz and 1800 Hz are the outer edges of the 250 Hz and 1600 Hz bands.
s_min, s_max = materials.spot_microphone_spacing_bounds(
    speed_of_sound=343.0, f_min=220.0, f_max=1800.0)
print(round(s_min, 3), round(s_max, 3))    # 0.078 0.086  (metres)
```

Three numbers, three different limits, and they do reconcile. The 1989 Hz is
the plane-wave ceiling of the 100 mm bore; the reported range stops one band
below it because a band is only usable when its *whole* width is inside the
valid range, and the 1600 Hz band reaches 1800 Hz. The 250 Hz floor is a
**spacing** limit, not a length limit: below $f_\mathrm{min}$ the two microphones are
closer than 5 % of a wavelength apart and there is no phase difference left to
resolve. Tube length answers a different question: it must be long enough for
a plane wave to form, which Clause 5.4.2 meets by putting the microphones at
least $3d$ from the source (a minimum length of 480 mm for the 100 mm tube).
The resulting spacing window, 78 mm to 86 mm here, is why the standard
specifies $(81 \pm 4)$ mm for a single microphone pair. It is a narrower
window than the
250–4000 Hz of the subtraction technique. Within it, the reduction is the
ISO 10534-2 chain unchanged: the transfer function between the two
microphones separates incident and reflected waves, and the standard's
Annex A additionally subtracts the internal losses of the system (measured
once against a rigid plate) so the tube's own dissipation is not billed to
the road; that correction is `spot_internal_loss_correction`.

### 2.1 A position, step by step

The tube is portable and the procedure is short, which is exactly why it is
worth writing down: almost everything that goes wrong with a spot measurement
goes wrong in the two minutes before the source is switched on. Clauses 6 to 8
give the order.

1. **Check the gates first.** The surface must be smooth, flat, free of debris
   and dry; the air between 5 °C and 30 °C and the road surface between 5 °C
   and 45 °C, out of direct sunlight and away from strong heat sources; test
   at night where that cannot be met. If any of these fails, the measurement
   cannot be carried out.
2. **Warm up.** Switch the system on and leave it running for **at least
   15 minutes**. The method lives on phase accuracy, and the electronics, the
   loudspeaker coil and the tube itself all have to be thermally stable.
3. **Calibrate the microphones** by the ISO 10534-2 Clause 7.5 procedure,
   before each measurement series and after each hour of measurement, for
   every microphone position used. That calibration is what corrects the phase
   and amplitude mismatch between the two capsules.
4. **Take the reference measurement** on a totally reflective specimen:
   a 10 mm steel plate is the standard's own example. Its one-third-octave
   absorption must come out below **0.03**; the result feeds the Annex A
   internal-loss correction, and it is taken at the same location immediately
   before or after the series.
5. **Measure the background** with the source off, and require at least
   **10 dB** signal-to-noise in every band against it. Stay at least **25 m**
   from passing heavy vehicles and motorcycles, and discard any run a passing
   vehicle disturbed.
6. **Seal.** Lay a bead of sealant in the fixture's groove, sized to the
   surface texture: a few millimetres on a smooth surface, thicker on a
   coarse one. Press the fixture onto the road: the right amount is when a
   little material squeezes out. If almost none does, lift the fixture, check
   for a **continuous circular impression**, add sealant and repeat. Then
   mount the tube without moving the fixture and check for leaks around the
   perimeter, for instance by listening through an open tube.
7. **Run and check.** Average until the result is stable, or over 50 sweeps or
   bursts if there is no live read-out. If the signal-to-noise ratio is too
   low, raise the source level, remove the noise source or wait.
8. **Verify the seal again.** Lift the fixture and confirm the sealing
   material was in tight contact in a closed circle. If in doubt, re-seal and
   repeat the position: an air leak reads as absorption and there is nothing
   in the data to distinguish the two.

Two more things about the reduction. Systems commonly carry **two microphone
pairs** and splice their narrow-band results, each pair over the part of the
band where its spacing is optimal (`spot_microphone_spacing_bounds` gives that
window for any pair). And the Annex A internal-loss correction is applied
**after** the narrow-band results have been averaged into one-third octaves
and **before** negative bands are clipped to zero: the standard deliberately
lets the FFT produce negative narrow-band values and clips only at the end.
Finish by repeating at **at least four positions** and reporting the mean and
the standard deviation per band.

## Subtraction or spot?

The two parts of ISO 13472 are complements, not competitors, and Part 2's own
introduction says both should give similar results between 315 Hz and
1600 Hz while their fields of application differ:

- **Surface type.** The subtraction technique handles the full range from
  reflective to highly absorbing (porous asphalt included); the spot method
  is scoped to reflective surfaces and declared unreliable above a measured
  absorption of 0.15, where its confined reflected wave is no longer strong
  enough to resolve.
- **Sampled area.** The subtraction technique averages over a patch whose
  Annex A clear zone extends to the 1.34 m maximum sampled-area radius of
  section 1 (about 5.6 m² of pavement, with the area actually contributing
  at each frequency set by the time window), a fair sample of texture and
  joints; the spot method sees only the patch under
  the tube mouth, so it trades spatial representativeness for the ability to
  sit on a narrow strip, a wheel track, or a laboratory core.
- **Bandwidth.** 250–4000 Hz against 250–1600 Hz, which matters because the
  tyre-road noise the measurement usually serves peaks around 1 kHz but has
  content beyond the spot method's ceiling.
- **Logistics.** The subtraction rig needs a mast, a free-field reference
  measurement and a pause in traffic; the spot tube needs a flat, sealable
  patch and minutes per point. Surface types and acceptance criteria aside,
  that practical difference is usually what decides.

![Two panels: on top, the reported one-third-octave ranges of the two methods drawn as bars on a common log-frequency axis, 250 to 4000 hertz for the subtraction technique and 250 to 1600 hertz for the spot tube, with the 315 to 1600 hertz interval where Part 2 expects them to agree shaded; below, the plane-wave ceiling of the spot tube falling as the bore grows, the 100 millimetre worked example marked at 1989 hertz and the region below the 1800 hertz top band edge shaded to show that a bore above 111 millimetres no longer covers the reported range](https://docs.jmrp.io/phonometry/media/insitu_method_windows.svg)

*Why the choice is made on bandwidth as much as on logistics. The spot tube's
ceiling is set by its own bore, and the bore has two jobs that pull against
each other: it must be wide enough to seal on a textured pavement and narrow
enough to keep plane waves up to the top of the reported range. At 111 mm the
two requirements meet; beyond it the 1600 Hz band is no longer covered. The
shaded interval on top is where Part 2's introduction expects the two methods
to agree, and it is also the interval a comparison between them should be
judged on.*

**Show the code for this figure**

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

# `materials` is imported above. The plane-wave ceiling against the bore.
bores = np.linspace(0.055, 0.155, 300)
ceiling = [materials.spot_tube_upper_frequency(d, speed_of_sound=343.0) for d in bores]
fig, ax = plt.subplots()
ax.plot(bores * 1e3, ceiling)
ax.axhline(1800.0, linestyle="--")     # top edge of the 1600 Hz band
ax.set_xlabel("Tube diameter d [mm]")
ax.set_ylabel("Plane-wave ceiling [Hz]")
plt.show()
```

Both methods report the same quantity, the normal-incidence absorption
coefficient in one-third-octave bands, so a low-absorption lane can be
surveyed with the spot tube and anchored with a subtraction measurement at a
few positions.

## What this guide covers

- **Covered.** ISO 13472-1:2002's subtraction technique: the Clause 4.1 geometrical spreading
  (`materials.geometric_spreading_factor`), the Clause 6.4 Adrienne time window
  (`materials.adrienne_window`), the narrow-band and one-third-octave reduction
  (`materials.in_situ_absorption_coefficient`,
  `materials.one_third_octave_absorption`,
  `materials.in_situ_absorption_spectrum`) and the Annex A sampled-area radius
  (`materials.max_sampled_area_radius`), with the Annex B reference-surface
  correction (`materials.absorption_reference_corrected`) and the Annex F
  oblique-incidence case (`materials.geometric_spreading_factor_angle`,
  `materials.msa_major_axis`, and the `incidence_angle_rad` argument of
  `materials.in_situ_absorption_spectrum`); and ISO 13472-2:2010's spot-method
  plane-wave frequency limit, microphone-spacing geometry and Annex A
  internal-loss correction (`materials.spot_tube_upper_frequency`,
  `materials.spot_microphone_spacing_bounds`,
  `materials.spot_internal_loss_correction`).

- **Not covered.** Both in-situ road standards have since been revised (ISO 13472-1:2022, ISO
  13472-2:2025); only the earlier 2002 and 2010 editions are implemented here,
  as the reference notes above flag. ISO 13472-2's own transfer-function
  processing is not implemented in this module either: only the geometry,
  validity and correction helpers are provided, and the actual two-microphone
  DSP is the ISO 10534-2 routine of the [Impedance Tube
  guide](https://docs.jmrp.io/phonometry/materials/absorbers/impedance-tube/). Neither the
  acquisition itself nor the field gates are enforced anywhere in code: the
  averaging, the free-field reference, the signal-to-noise checks, the seal
  verification and the weather limits of sections 1.1 to 1.4 and 2.1 are the
  operator's, and the library only sees the two arrays that come out of them.

## See also

- [Impedance Tube](https://docs.jmrp.io/phonometry/materials/absorbers/impedance-tube/): the ISO 10534-2
  two-microphone method the spot tube defers its mathematics to, and the
  laboratory instrument the spot method miniaturises.
- [Diffusers and Their Coefficients](https://docs.jmrp.io/phonometry/materials/diffusers/diffusers/): the
  other family of surface measurements, grading where a surface sends its
  reflection rather than how much it absorbs.
- [Sound Absorption Measurement and Rating](https://docs.jmrp.io/phonometry/materials/absorbers/absorption-measurement/):
  the laboratory (reverberation-room) absorption measurement and its rating,
  for materials that can be brought indoors.
- [Outdoor Sound Propagation](https://docs.jmrp.io/phonometry/environment/propagation/outdoor-propagation/): where
  the absorption of the ground surface enters the propagation models.
- API reference: [`materials.surfaces.road_absorption`](https://docs.jmrp.io/phonometry/reference/api/materials/road-absorption/).
- Theory: [In-situ road surface absorption](https://docs.jmrp.io/phonometry/reference/theory/materials-surfaces/#in-situ-road-surface-absorption-iso-13472-1-iso-13472-2): the ISO 13472 subtraction and the geometry that makes an in-situ measurement possible at all.
- [Published catalogues](https://docs.jmrp.io/phonometry/reference/catalogues/ground-surfaces/): the
  flow resistivity and the model parameters the books print for grass, snow,
  asphalt and the rest of the ground surfaces.

## References

- **International Organization for Standardization. (1998).** [*Acoustics — Determination of sound absorption coefficient and impedance in impedance tubes — Part 2: Transfer-function method*](https://www.iso.org/standard/22851.html) (ISO 10534-2:1998).
  The two-microphone transfer-function reduction the spot method reuses, implemented in the Impedance Tube guide.
- **International Organization for Standardization. (2002).** [*Acoustics — Measurement of sound absorption properties of road surfaces in situ — Part 1: Extended surface method*](https://www.iso.org/standard/35387.html) (ISO 13472-1:2002).
  The subtraction technique of section 1: the mandatory geometry, the Adrienne window and the Annex A sampled-area radius. Since revised as ISO 13472-1:2022 (https://www.iso.org/standard/77032.html); the 2002 edition is the implemented one.
- **International Organization for Standardization. (2010).** [*Acoustics — Measurement of sound absorption properties of road surfaces in situ — Part 2: Spot method for reflective surfaces*](https://www.iso.org/standard/32304.html) (ISO 13472-2:2010).
  The spot method of section 2: the plane-wave limits, the microphone spacing window and the deferral to ISO 10534-2. Since revised as ISO 13472-2:2025 (https://www.iso.org/standard/84925.html); the 2010 edition is the implemented one.
