# Microphone Characterisation (IEC 60268-4)

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

Canonical: https://docs.jmrp.io/phonometry/devices/electroacoustics/microphones/
Language: en
Alternate: https://docs.jmrp.io/phonometry/es/devices/electroacoustics/microphones/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: 2026-11-12T12:09:00Z
Standards: IEC 60268-4:2014; IEC 60263:1982; Recommendation ITU-R BS.468-4
Verified: 17 conformance checks (IEC 60268-4:2014; Recommendation ITU-R BS.468-4) https://docs.jmrp.io/phonometry/reference/conformance/
License: MIT (https://github.com/jmrplens/phonometry/blob/main/LICENSE)
Summary: Microphone characterisation by IEC 60268-4: sensitivity in mV/Pa, directional patterns, inherent noise in dB(A) and dB(468), and the data sheet.

---

Every acoustic measurement in this documentation starts at a microphone, and
IEC 60268-4 is the standard that fixes how the device itself is described:
what its sensitivity refers to, how its directional pattern is stated, and
how its self-noise and overload point bound the levels it can read. This
guide covers the conventions where datasheet comparisons go wrong (the
sensitivity references and the noise weightings), the directional patterns
with their directivity index, and the **rated-characteristics report** that
renders the standard's data sheet from a measured free-field response, with
every panel drawn to the IEC 60263 scale conventions. The distortion and
frequency-response measurements themselves live in
[Electroacoustics](https://docs.jmrp.io/phonometry/devices/electroacoustics/electroacoustics/); the loudspeaker
counterpart has [its own guide](https://docs.jmrp.io/phonometry/devices/electroacoustics/loudspeakers/).

## 1. Sensitivity and its references (IEC 60268-4 clause 11)

The sensitivity $M$ is the output
voltage per unit sound pressure, quoted in mV/Pa; the sensitivity *level* is
$L_M = 20\log_{10}(M / 1\,\text{V/Pa})$ dB, which is negative for every real
microphone (a
50 mV/Pa studio condenser sits at −26 dB re 1 V/Pa). The classic pitfall is
the reference: −26 dB re 1 V/Pa and +34 dB re 1 mV/Pa are the *same*
microphone, 60 dB apart on paper, so a sensitivity level without its reference
is meaningless. The standard also distinguishes free-field, diffuse-field and
pressure sensitivity; for the same capsule they diverge at high frequency,
so the stated type matters as much as the number.

Behind the number sits the transducer. A **condenser** capsule (externally
polarised or electret) reads the sound pressure as a capacitance change and
needs powering, but converts with high sensitivity, typically some tens of
mV/Pa for a studio diaphragm and a few mV/Pa for a small measurement
capsule; virtually every measurement microphone is a condenser because the
stretched-membrane capsule is stable enough to calibrate. A **dynamic**
(moving-coil) microphone generates its voltage by induction, needs no
power and survives abuse, but converts far less efficiently, around
1-3 mV/Pa, which is why a dynamic on a quiet source runs out of
preamplifier gain before a condenser does. The rated sensitivity of the
standard is quoted at the 1 kHz reference frequency (clause 11.3),
into the rated load, and the stated type (free-field, diffuse-field or
pressure) is part of the rating, not a detail.

### 1.1 The three sensitivity types, and how to choose one

The three types differ only in the **field the capsule is calibrated in**, and
they exist because a microphone is not a point. At frequencies where the
capsule is no longer small against the wavelength it perturbs the field it is
measuring, and the pressure on the diaphragm rises above the undisturbed
pressure for sound arriving on the reference axis, by several decibels for a
half-inch capsule above about 5 kHz, and higher still for a quarter-inch one.
The three definitions are three ways of dealing with that:

- **Pressure sensitivity** (11.2.4) is the ratio of output voltage to the sound
  pressure actually present at the acoustic entry. It is what a coupler, a
  pistonphone or an electrostatic actuator realises, and what a capsule flush
  in a boundary reads.
- **Free-field sensitivity** (11.2.1) is referred to the *undisturbed* pressure
  of a plane progressive wave arriving on the reference axis, so the capsule's
  own pressure build-up is corrected out.
- **Diffuse-field sensitivity** (11.2.2) is referred to sound arriving from all
  directions with random phase; the standard defines it as the r.m.s. of the
  free-field sensitivities over all directions, so
  $L_{M,\text{diff}} = L_{M,\text{ff}} - D$ with $D$ the directivity index of
  section 2.

![Three panels on one capsule glyph: free field, with a plane wave arriving on the reference axis and the two admissibility rules r at least d and r at least d squared over lambda; diffuse field, with rays arriving from every direction and the definition of the diffuse-field sensitivity as the r.m.s. of the pattern; and pressure, with the capsule closed into a coupler whose cavity is small against the wavelength. Below them the bench: the anechoic room and the half-wavelength plane-wave rule, the substitution calibration against a reference microphone at the same point, the simultaneous-comparison variant with its plus or minus 1 dB qualification, and the overall accuracy of plus or minus 2 dB or better](https://docs.jmrp.io/phonometry/media/diagram_microphone_references.svg)

**The selection rule.** A free-field capsule pointed at a known source in a
free or near-free field; a random-incidence capsule in a reverberant room, and
for an IEC 61672 class 1 sound level meter used indoors; a pressure capsule in
a coupler or flush in a boundary. Using a free-field capsule at grazing
incidence, or a pressure capsule on axis, costs several decibels in the top two
octaves and nothing at all below 1 kHz, which is why the error survives an
A-weighted level check and appears only in a spectrum. The size trade-off runs
the other way: a smaller capsule keeps its field type honest to a higher
frequency and pays for it in sensitivity and self-noise.

**The bench behind the number.** IEC 60268-4 clause 4.2.2 defines rated
conditions for a microphone: the specified resistive load, the rated power
supply, a free field (5.5.2) at zero degrees to the reference direction, and an
undisturbed sound pressure of 1 Pa (94 dB) at the reference point, at the
1 kHz measurement frequency. Clause 5.5.2 fixes the geometry: free-field
conditions realised in the open air, in an anechoic room or in a duct, with the
spherical-wave referral valid only where $r \ge d$ and $r \ge d^2/\lambda$ for
an effective source diameter $d$ (preferably beyond three times the largest
dimension of the radiating surface) and a spherical wave counted as plane only
at least half a wavelength from its centre of curvature at the lowest measured
frequency. Clause 5.6.2 fixes the calibration: the **substitution** method, in
which the microphone under test and a calibrated reference occupy the same
point one after the other, gives the highest accuracy; the simultaneous
comparison at two nearby points may be used only after showing it agrees with
substitution within ±1 dB. Clause 5.7 caps the whole thing: an overall accuracy
of ±2 dB or better shall be obtained. Every number in the fiche of section 4
inherits that bound. Turning a rated sensitivity into calibrated levels on a
working chain is [Calibration and dBFS](https://docs.jmrp.io/phonometry/signals/metrology/calibration/).

## 2. Directional pattern and the directivity index (clause 13)

The directional response is stated as the pattern of sensitivity against
the angle of incidence, normalized to the reference axis. The classic
first-order family mixes an omnidirectional (pressure) term and a cosine
(pressure-gradient) term, $M(\theta) \propto (1 - b) + b\cos\theta$: $b = 0$
is the omnidirectional capsule, $b = 1/2$ the cardioid with its null at 180°,
and $b = 1$ the figure-of-eight with nulls at ±90°. One number condenses the
pattern: the **directivity index** $D = 20\log_{10}(M_0/M_\mathrm{diff})$ compares the
reference-axis sensitivity with the diffuse-field sensitivity obtained from
the clause 11.2.2 a) integral over a rotationally symmetric pattern
(clause 13.2.2; the index is anchored on two clauses because the
$M_\mathrm{diff}$ it divides by is the one clause 11.2.2 a) defines, as the
r.m.s. of the pattern over a rotationally symmetric solid angle). An
omni scores 0 dB, the ideal cardioid and figure-of-eight both
$10\log_{10} 3 = 4.8$ dB: in a diffuse room field, a cardioid picks up three
times less reverberant power than an omni of equal axial sensitivity, which
is exactly the ratio a talker-to-microphone distance calculation wants.
Real patterns hold their textbook shape only over the middle of the band;
at high frequency the capsule's own size makes any microphone directive,
which is also why the free-field and diffuse-field sensitivities of one
capsule diverge there.

The whole family is six values of one coefficient. The directivity indices
below are what `microphone_characteristics` returns for each pattern through
the clause 11.2.2 a) integral (the section-4 snippet prints three of them), and
the random-energy efficiency is the fraction of a diffuse field the capsule
still picks up:

| Pattern | $b$ | $M(180°)$ | Null | Directivity index | Random-energy efficiency | Distance factor |
| :--- | ---: | ---: | ---: | ---: | ---: | ---: |
| Omnidirectional | 0 | +1.00 | — | 0.0 dB | 1.000 | 1.00 |
| Subcardioid | 1/3 | +0.33 | — | 3.2 dB | 0.481 | 1.44 |
| Cardioid | 1/2 | 0.00 | 180° | 4.8 dB | 0.333 | 1.73 |
| Supercardioid | 0.63 | −0.26 | 126° | 5.7 dB | 0.269 | 1.93 |
| Hypercardioid | 3/4 | −0.50 | 110° | 6.0 dB | 0.250 | 2.00 |
| Figure-of-eight | 1 | −1.00 | ±90° | 4.8 dB | 0.333 | 1.73 |

![The six members of the first-order family on one polar axis over a 30 dB range: omnidirectional, subcardioid, cardioid, supercardioid, hypercardioid and figure-of-eight, each labelled with the directivity index the library computes for it, 0.0, 3.2, 4.8, 5.7, 6.0 and 4.8 dB respectively](https://docs.jmrp.io/phonometry/media/microphone_patterns.svg)

*The cardioid and the figure-of-eight return the same 4.8 dB from opposite
geometries: the index says how much reverberant power a capsule rejects, not
from where. That is why a pattern and an index are always quoted together, and
why the supercardioid and hypercardioid (which trade a rear lobe for a deeper
null off-axis) beat both on the number while sounding quite different behind.
The distance factor is the practical consequence: a cardioid may sit 1.73 times
further from a talker than an omnidirectional capsule for the same
direct-to-reverberant ratio.*

**Show the code for this figure**

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

# The pattern shapes. The directivity index annotated on each curve is the
# one `microphone_characteristics(...).directivity_index` returns for the
# same polar data, as in the snippet of section 4.
theta = np.radians(np.linspace(0.0, 360.0, 721))
ax = plt.subplot(projection="polar")
for b in (0.0, 1 / 3, 0.5, 0.63, 0.75, 1.0):
    mag = np.maximum(np.abs((1.0 - b) + b * np.cos(theta)), 10 ** (-1.5))
    ax.plot(theta, np.maximum(20 * np.log10(mag), -30.0))
ax.set_ylim(-30.0, 0.0)
plt.show()
```

The cosine term is a **pressure-gradient** term, and that has a consequence the
formula hides: close to a source the gradient carries a component that grows as
$1/r^2$ while the pressure term falls as $1/r$, so any capsule with $b > 0$
boosts low frequencies as it approaches the source. The **proximity effect** is
typically 6 to 10 dB at 100 Hz at 5 cm for a cardioid and larger for a
figure-of-eight. It is exploited deliberately in speech work, and it is the
reason a directional microphone cannot be used as a measurement device near a
source: the response depends on where it is held. Together with the frequency
dependence of a real pattern, that is why every measurement microphone in this
documentation is an omnidirectional pressure capsule.

## 3. Inherent noise: dB(A) and dB(468) (clause 17)

A microphone's electronics and the air load on its diaphragm set a noise
floor, expressed as the **equivalent noise level**: the sound pressure
level whose output would equal the weighted inherent-noise voltage,
$L_\mathrm{N} = 20\log_{10}\bigl((U_\mathrm{N}/M)/20\,\mu\text{Pa}\bigr)$. The standard states it
with two weightings,
and the numbers are far apart on paper. The **A-weighted** figure (RMS
detector) is the one most datasheets quote, e.g. 14 dB(A) for a good studio
condenser. The **ITU-R BS.468-4** figure weights the same noise with the
curve that peaks +12.2 dB near 6.3 kHz and reads it with a quasi-peak
detector, so it penalises the hiss and crackle the ear actually notices;
for typical microphone noise it comes out roughly 10 dB above the
A-weighted number for the same capsule. Neither is wrong: they are
different weightings of the same voltage, and comparing a dB(A) figure from
one datasheet with a dB(468) figure from another silently flatters the
first by that margin. IEC 60268-1 still cites the curve by its earlier name,
CCIR Recommendation 468, which is why older datasheets print the same figure
as dB(CCIR); the library calls it by the current document and spells the
weighting `"468"` wherever it takes one. The signal-to-noise ratio re 1 Pa
(94 dB SPL) is
derived from the same equivalent noise level, and the overload sound
pressure level (clause 15.2) bounds the usable range from above. The
ITU-R BS.468-4 weighting curve itself is exposed as `weighting_bs468` in the
[electroacoustics distortion set](https://docs.jmrp.io/phonometry/devices/electroacoustics/electroacoustics/),
where it also weights THD.

![Upper panel: one inherent-noise band spectrum falling with frequency, drawn raw, A-weighted and ITU-R BS.468-4 weighted, so the 468 curve is seen lifting the 2 to 10 kHz bands the A curve has begun to suppress. Lower panel: the band-summed levels of the two weightings for that capsule and for a hissier one whose noise rises towards 20 kHz, with the network-only difference annotated for each, plus 2.4 dB and plus 8.3 dB respectively](https://docs.jmrp.io/phonometry/media/microphone_noise_weightings.svg)

*How much of the gap the network alone accounts for depends on the shape of
that capsule's noise: 2.4 dB for the falling 1/f spectrum of the example, 8.3 dB
for a hissier capsule. The rest of the customary 10 dB comes from the ITU-R BS.468-4
**quasi-peak detector**, which reads a noise higher than an r.m.s. detector
does; that detector is
[`broadcast.quasi_peak_meter`](https://docs.jmrp.io/phonometry/devices/broadcast/quasi-peak/) and it
reads a recording, while `weighting_bs468` is the network only, so a level
computed with the network and an r.m.s. sum is still not a dBqps figure and must
not be compared with one from a datasheet.*

The band spectrum drawn there is the one the section-4 snippet passes as
`MicrophoneNoise(spectrum=...)`, weighted with `weighting_bs468` from the
[electroacoustics module](https://docs.jmrp.io/phonometry/devices/electroacoustics/electroacoustics/) and
with A-weighting, then summed band by band.

**How $U_\mathrm{N}$ is obtained** (clause 17.2). The microphone is isolated from sound,
wind, shock, vibration and electric or magnetic fields, but it must stay in
**acoustical operating mode**: the standard warns explicitly that replacing the
transducer element with an equivalent circuit does not measure the microphone,
because the element contributes noise of its own. A 40 to 60 dB preamplifier
keeps the microphone's own noise dominant, so no correction for the
instrumentation is needed. Then comes the check that decides whether the number
is usable: replace the microphone with a resistor at room temperature equal to
its rated impedance and require the measured output to be **less than one
third** of the value measured with the microphone, so that the instrumentation
and any residual external sound inflate the result by less than 10 %. Report
which weighting and detector produced the figure: A-weighted r.m.s.
(IEC 60268-1 6.2.1), ITU-R BS.468-4 quasi-peak (6.2.2), or unweighted third octaves
(6.2.3), because the three differ by about 10 dB on the same capsule and the
clause asks for the quasi-peak measurement to be among them.

### 3.1 What else bounds the usable range

The **overload sound pressure level** (15.2) is the pressure of a plane wave at
which the microphone's amplitude non-linearity reaches a declared limit, for
any frequency in the effective range and any direction of incidence; the
standard notes that no common limit exists and that data sheets typically
declare 0.5 % or 1 %. Producing the field is the hard part of the measurement:
a 140 dB level has to come from something whose own distortion is well below
the capsule's: in practice a small pressure chamber or a compression driver on
a coupler, with the source's own distortion characterised first against a
reference microphone. Sweep the level with an ordinary loudspeaker and the
curve plotted is the loudspeaker's. The clause's own way round the problem is
to use the difference-frequency measurement of 14.4.2 instead, which at least
minimises the influence of the source's non-linearity.

Three further limits are rated characteristics in their own right, and all
three are quoted as **equivalent sound pressure levels**, so they can be
compared band for band with the inherent noise above:

- **Mechanical vibration** (19.2): the equivalent sound pressure produced by a
  stated r.m.s. acceleration in the direction of maximum influence, with the
  direction of minimum influence also stated, measured with a gliding frequency
  up to 250 Hz. This is stand and handling noise, and it is why a shock mount
  is specified rather than assumed.
- **Wind** (19.3): the equivalent sound pressure for a stated wind velocity and
  direction, the reference value being 10 m/s. It is the quantity a
  windscreen's effectiveness is stated against.
- **Transient "pop" effect** (19.4): the plosive burst of a close talker, which
  a pop filter or a change of working distance addresses.

## 4. Microphone characteristics report (IEC 60268-4)

The microphone companion of the
[loudspeaker report](https://docs.jmrp.io/phonometry/devices/electroacoustics/loudspeakers/): the rated
characteristics IEC 60268-4
defines around a measured free-field frequency response are gathered into a single
**microphone characteristics** result that renders the standard's
rated-characteristics data sheet. Four of the numbers are computed from the
standard's own definitions rather than merely repeated:

- **Sensitivity level** (11.1). The rated free-field sensitivity $M$ (mV/Pa,
  at the 1 kHz reference frequency of 11.3) as a level,
  $L_M = 20\log_{10}(M / 1\,\text{V/Pa})$: 12.5 mV/Pa is $-38.1$ dB re 1 V/Pa.
- **Effective frequency range** (12.2). The band over which the response,
  normalized to 0 dB at the reference frequency, stays within the stated
  $\pm$ tolerance; the edges are the interpolated tolerance crossings.
- **Directivity index** (13.2.2). $D = 20\log_{10}(M_0/M_\mathrm{diff})$ with the
  diffuse-field sensitivity from the 11.2.2 a) integral over a rotationally
  symmetric pattern; the ideal cardioid returns $10\log_{10} 3 = 4.8$ dB.
- **Equivalent noise level** (17.2). The weighted inherent-noise voltage over
  the rated sensitivity as a sound pressure level,
  $L_\mathrm{N} = 20\log_{10}\bigl((U_\mathrm{N}/M)/20\,\mu\text{Pa}\bigr)$, with the
  signal-to-noise ratio re 1 Pa (94 dB SPL) derived from it. The overload
  sound pressure level (15.2) is read from a distortion-against-level curve
  at the stated THD limit.

The four input curves in the snippet below are synthetic, but each stands for a
named measurement and it is worth knowing which before collecting data. The
**free-field response** comes from a substitution calibration against a
reference microphone in an anechoic room (5.6.2 a), or is synthesised from an
electrostatic-actuator pressure response plus the manufacturer's free-field
correction, and is normalised to 0 dB at the 1 kHz reference frequency the
report marks. The **polar cuts** come from rotating the microphone on a
turntable in the same room, one cut per stated frequency, in a plane containing
the reference axis, with the distance, the sound pressure and the frequency
held constant while $\theta$ is stepped (13.1.2 a). The **inherent-noise
spectrum** is the third-octave breakdown of the clause 17.2 measurement, with
the analyser bandwidth stated, and the single `voltage_v` is the weighted r.m.s.
of the same measurement. The **distortion-against-level curve** comes from
stepping the sound pressure of a high-level source while the distortion at the
output is measured (15.2.2), and the overload level is read off it where the
curve crosses the declared limit.

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

freqs = np.geomspace(20, 20000, 400)
response = -10 * np.log10(1 + (30.0 / freqs) ** 4)      # low-frequency roll-off
response -= 10 * np.log10(1 + (freqs / 19000.0) ** 8)   # high-frequency roll-off
response += 2.0 * np.exp(-(np.log2(freqs / 9000.0) ** 2) / 0.3)  # presence region

angles_rad = np.linspace(0, 179, 359)
cardioid = 20 * np.log10((1 + np.cos(np.radians(angles_rad))) / 2)
noise_f = np.geomspace(20, 20000, 31)

result = electroacoustics.microphone_characteristics(
    response, 12.5, frequencies=freqs, tolerance_db=3.0,          # 12.5 mV/Pa at 1 kHz
    directivity=electroacoustics.MicrophoneDirectivity(polar=(angles_rad, cardioid), frequency=1000.0),
    noise=electroacoustics.MicrophoneNoise(                             # A-weighted, V
        voltage_v=1.25e-6,
        spectrum=(noise_f, 6.0 + 12.0 * np.log10(1000.0 / noise_f)),
    ),
    overload=electroacoustics.MicrophoneOverload(
        distortion=(np.linspace(100, 140, 81),
                    0.5 * 10 ** ((np.linspace(100, 140, 81) - 130.0) * 0.08)),
        thd_percent=0.5,
    ),
    electrical=electroacoustics.MicrophoneElectrical(
        rated_impedance_ohm=150.0, min_load_impedance_ohm=1000.0,
        powering="Phantom P48 (IEC 61938)", supply_current_ma=3.1,
    ),
)
print(round(result.sensitivity_level, 1))            # -38.1 dB re 1 V/Pa
print(tuple(round(x) for x in result.effective_range))  # Hz
print(round(result.directivity_index, 1))            # 4.8 dB (cardioid)
print(round(result.equivalent_noise_level, 1))       # 14.0 dB(A)
print(round(result.signal_to_noise_ratio_db, 1))        # 80.0 dB re 1 Pa

# The three canonical patterns of section 2, through the same 11.2.2 a)
# integral: b = 1/2 and b = 1 return the same index from opposite shapes.
for b in (0.0, 0.5, 1.0):
    polar = 20 * np.log10(
        np.maximum(np.abs((1 - b) + b * np.cos(np.radians(angles_rad))), 1e-3))
    family = electroacoustics.microphone_characteristics(
        response, 12.5, frequencies=freqs, tolerance_db=3.0,
        directivity=electroacoustics.MicrophoneDirectivity(polar=(angles_rad, polar),
                                                           frequency=1000.0))
    print(b, round(family.directivity_index, 1))
# 0.0 0.0
# 0.5 4.8   b = 1/2 and b = 1: one index from opposite shapes
# 1.0 4.8

result.report("microphone.pdf", metadata=ReportMetadata(measurement_standard="IEC 60268-4"))
```

The fourth print closes a chain the page has left open twice. An inherent-noise
voltage of 1.25 µV at the output of a 12.5 mV/Pa capsule is an acoustic
pressure of 100 µPa at its diaphragm, which is $20\log_{10}(100/20) = 14.0$ dB
above the 20 µPa reference: this microphone would read 14 dB(A) in a perfectly
silent room, and cannot measure anything quieter than about 24 dB(A) without
contributing to the answer. The signal-to-noise ratio the result also carries
is simply 94 dB minus that figure (80.0 dB) because the standard's reference
is 1 Pa. The same voltage read through ITU-R BS.468-4 with a quasi-peak
detector would be reported about 10 dB higher for the same capsule, so this
number is 14 dB(A) and must never be set beside a dB(468) figure from another
sheet.

One trap the API reproduces and the page has to close: `MicrophoneNoise.voltage_v`
is the inherent-noise voltage **as measured through the weighting network**, not
a raw wideband r.m.s., and `weighting=` records which network that was. It
defaults silently to `'A'` (IEC 60268-1 6.2.1). The field is a *label, not a
filter*: the function re-weights nothing, so declaring `weighting="468"` while
passing an A-weighted voltage produces a correctly computed level with the
wrong name on it, exactly the error section 3 warns about, now printed on a
report. That stays true now that the quasi-peak detector exists: clause 17
takes a stated voltage, not a recording, so there is nothing here for
[`quasi_peak_meter`](https://docs.jmrp.io/phonometry/devices/broadcast/quasi-peak/) to run on and
the field remains a record of how the supplier measured. Quoting both figures
for the same capsule is the honest form, since they are different questions and
both are legitimate.

`microphone_characteristics` returns a `MicrophoneCharacteristics` with the
computed `sensitivity_level`, `effective_range`, `directivity_index`,
`equivalent_noise_level`, `max_spl`, `signal_to_noise_ratio_db` and
`diffuse_field_sensitivity_level`, and a `.report()` that writes the fiche.
The free-field response is drawn with its tolerance band, the
reference-frequency marker and the effective-range markers to the IEC 60263
proportion (one frequency decade equal to 25 dB), and the directional pattern
on the IEC 60263 25 dB reference circle; the inherent-noise spectrum and the
distortion-against-level curve (with the THD limit and the overload level
marked) feed the secondary panels. A `requirement` in the metadata is checked
as a maximum permitted equivalent noise level.

The rated characteristics are also available interactively through `.plot()`,
which draws **one concept per figure** with the same panel code the report
composes, selected by `quantity`. Passing an axes draws on it:

```python
result.plot()                        # free-field response (default)
result.plot(quantity="directivity")  # polar pattern on the 25 dB circle
result.plot(quantity="noise")        # inherent-noise band spectrum
result.plot(quantity="distortion")   # THD vs sound pressure level
```

![Microphone free-field relative response with its shaded tolerance band, the reference-frequency marker and the effective-range markers on a nominal-frequency axis](https://docs.jmrp.io/phonometry/media/microphone_response.svg)

*The shaded band is the tolerance the reader declares (`tolerance_db`, 3 dB
here) and the effective-range edges of clause 12.2 are its interpolated
crossings, so a tighter tolerance shortens the quoted range: two microphones
cannot be compared on range alone without it. The curve is normalised to 0 dB
at the 1 kHz reference frequency of clause 11.3, which is where the rated
sensitivity is quoted, and the lift near 9 kHz here is a deliberate presence
peak rather than a defect.*

**Show the code for this figure**

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

freqs = np.geomspace(20, 20000, 400)
response = -10 * np.log10(1 + (30.0 / freqs) ** 4)      # low-frequency roll-off
response -= 10 * np.log10(1 + (freqs / 19000.0) ** 8)   # high-frequency roll-off
response += 2.0 * np.exp(-(np.log2(freqs / 9000.0) ** 2) / 0.3)  # presence region

result = electroacoustics.microphone_characteristics(response, 12.5, frequencies=freqs, tolerance_db=3.0)
result.plot()   # quantity="response" (the default)
plt.show()
```

![Microphone cardioid directional pattern at 1000 Hz on the IEC 60263 25 dB reference circle](https://docs.jmrp.io/phonometry/media/microphone_directivity.svg)

*Drawn on the IEC 60263 25 dB reference circle of clause 3, so the full radius
is 25 dB and each ring is 5 dB; the cardioid's rear rejection is the depth of
its 180° null. This is one cut at one frequency: a real capsule holds the
textbook shape only through the middle of its band, which is why clause
13.1.2 a) asks for the octave centres from 125 Hz to 16 kHz. It is the panel
the 4.8 dB directivity index of the fiche is computed from.*

**Show the code for this figure**

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

freqs = np.geomspace(20, 20000, 400)
response = -10 * np.log10(1 + (30.0 / freqs) ** 4)
angles_rad = np.linspace(0, 179, 359)
cardioid = 20 * np.log10((1 + np.cos(np.radians(angles_rad))) / 2)

result = electroacoustics.microphone_characteristics(
    response, 12.5, frequencies=freqs, tolerance_db=3.0,
    directivity=electroacoustics.MicrophoneDirectivity(polar=(angles_rad, cardioid), frequency=1000.0),
)
result.plot(quantity="directivity")
plt.show()
```

![Microphone inherent-noise equivalent band-level spectrum against frequency](https://docs.jmrp.io/phonometry/media/microphone_noise.svg)

*The band breakdown behind the single equivalent-noise figure, and note the
ordinate: these are equivalent **sound pressure** band levels, not voltages.
The rising low-frequency end is the amplifier's 1/f noise, which A-weighting
discounts heavily, which is why the A-weighted total is dominated by the mid
bands and why the same capsule reads so differently through the 468 network.*

**Show the code for this figure**

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

freqs = np.geomspace(20, 20000, 400)
response = -10 * np.log10(1 + (30.0 / freqs) ** 4)
noise_f = np.geomspace(20, 20000, 31)

result = electroacoustics.microphone_characteristics(
    response, 12.5, frequencies=freqs, tolerance_db=3.0,
    noise=electroacoustics.MicrophoneNoise(
        voltage_v=1.25e-6,
        spectrum=(noise_f, 6.0 + 12.0 * np.log10(1000.0 / noise_f)),
    ),
)
result.plot(quantity="noise")
plt.show()
```

![Microphone total harmonic distortion in percent against sound pressure level, with the THD limit and the overload sound pressure level marked](https://docs.jmrp.io/phonometry/media/microphone_distortion.svg)

*The knee is where the capsule's motion stops being linear, and the overload
sound pressure level of clause 15.2 is simply the reading of this curve at the
declared limit, so a "max SPL" quoted at 0.5 % and one quoted at 3 % differ by
several decibels for the same microphone, and neither means anything without
its percentage.*

**Show the code for this figure**

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

freqs = np.geomspace(20, 20000, 400)
response = -10 * np.log10(1 + (30.0 / freqs) ** 4)
spl_axis = np.linspace(100, 140, 81)

result = electroacoustics.microphone_characteristics(
    response, 12.5, frequencies=freqs, tolerance_db=3.0,
    overload=electroacoustics.MicrophoneOverload(
        distortion=(spl_axis, 0.5 * 10 ** ((spl_axis - 130.0) * 0.08)),
        thd_percent=0.5,
    ),
)
result.plot(quantity="distortion")
plt.show()
```

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

**Report:** [IEC 60268-4 microphone characteristics example report (PDF)](https://docs.jmrp.io/phonometry/media/iec60268_4_microphone_example.pdf): One-page IEC 60268-4 microphone rated-characteristics fiche: a header with the manufacturer and model, the rated-characteristics table (free-field sensitivity in mV/Pa and its level re 1 V/Pa, effective frequency range, rated and minimum load impedances, equivalent noise level, signal-to-noise ratio, maximum SPL at the stated THD limit, directivity index and phantom powering) beside the free-field frequency response with its tolerance band and effective-range markers, and the directional-pattern, inherent-noise-spectrum and distortion panels, all drawn to the IEC 60263 scale conventions.

Microphone characteristics fiche (MicrophoneCharacteristics.report), the IEC 60268-4 rated-characteristics table beside the free-field response, with the cardioid directional pattern, inherent-noise and THD-against-level panels drawn to the IEC 60263 25 dB-per-decade and 25 dB reference-circle conventions.

## What this guide covers

- **Covered.** The three sensitivity types and the rated conditions, geometry, substitution
  calibration and ±2 dB accuracy bound behind them (clauses 4.2.2, 5.5.2, 5.6.2,
  5.7, 11.2), the clause 17.2 inherent-noise procedure with its one-third
  residual check, and the clause 15.2/19.2-19.4 limits. The IEC 60268-4:2014
  rated microphone characteristics via `microphone_characteristics` and its
  `.report()`: the sensitivity level re 1 V/Pa (11.1/11.3), the effective
  frequency range against a stated tolerance (12.2), the directivity index
  through the 11.2.2 a) diffuse-field integral (13.2.2), the equivalent noise
  level and signal-to-noise ratio (17.2), the overload sound pressure level read
  off a distortion curve (15.2) and the rated impedances and powering (9/10),
  with the response, polar, noise and distortion panels drawn to the IEC 60263
  scale conventions.

- **Not covered.** The laboratory's own instrumentation: the anechoic room and its qualification,
  the reference microphone and its calibration certificate, the high-level
  source that produces a 140 dB field, and the turntable. The *conditions* those
  instruments have to satisfy are in sections 1.1, 3 and 3.1, because they are
  what the reported numbers mean; the acquisition itself is not implemented
  here. The ITU-R BS.468-4 quasi-peak detector now is, in [Quasi-peak programme
  meter](https://docs.jmrp.io/phonometry/devices/broadcast/quasi-peak/), but it reads a recording and
  clause 17 takes a stated voltage or level, so `weighting="468"` on this page
  stays a label on a supplier's figure, and an inherent-noise level computed from
  `weighting_bs468` and an r.m.s. sum is still not a dBqps figure. The implemented edition is frozen at IEC 60268-4:2014; the superseding
  2018 revision is not the one checked. Acoustic calibration of a working chain,
  sensitivity into pascals and dBFS, lives in [Calibration and
  dBFS](https://docs.jmrp.io/phonometry/signals/metrology/calibration/).

## See also

- [Electroacoustics](https://docs.jmrp.io/phonometry/devices/electroacoustics/electroacoustics/): the IEC 60268-3
  distortion set, THD+N and SINAD, the ITU-R BS.468-4 weighted THD and the H1/H2
  frequency-response estimators that produce the measured curves.
- [Loudspeaker Characterisation (IEC 60268-5)](https://docs.jmrp.io/phonometry/devices/electroacoustics/loudspeakers/):
  the companion rated-characteristics report for the reproducing side of
  the chain.
- [Calibration and dBFS](https://docs.jmrp.io/phonometry/signals/metrology/calibration/): turning a
  microphone's rated sensitivity into calibrated sound pressure levels.
- [Build a sound level meter](https://docs.jmrp.io/phonometry/signals/sound-level-meter/): the
  measurement chain a characterised microphone feeds.
- API reference: [`electroacoustics.microphone`](https://docs.jmrp.io/phonometry/reference/api/electroacoustics/microphone/).

## References

- **Beranek, L. L., & Mellow, T. J. (2012).** *Acoustics: Sound fields and transducers*. Academic Press. https://doi.org/10.1016/C2011-0-05897-0
  ISBN 978-0-12-391421-7. The transducer physics behind sections 1 and 2: condenser and moving-coil transduction and first-order directional patterns.
- **International Electrotechnical Commission. (1982).** [*Scales and sizes for plotting frequency characteristics and polar diagrams*](https://webstore.iec.ch/en/publication/1176) (IEC 60263:1982).
  The scale proportions of the characteristic graphs: one frequency decade equal to 25 dB on the ordinate (clause 2), and the polar diagram plotted on a 25 dB reference-circle radius (clause 3).
- **International Electrotechnical Commission. (2014).** [*Sound system equipment – Part 4: Microphones*](https://webstore.iec.ch/en/publication/1221) (IEC 60268-4:2014).
  The rated microphone characteristics of this guide: the free-field sensitivity and its level re 1 V/Pa (11.1/11.3), the frequency response and effective frequency range (12.1/12.2), the directional pattern (13.1) and the directivity index (13.2.2) through the 11.2.2 a) diffuse-field integral, the overload sound pressure level (15.2), the equivalent sound pressure level due to inherent noise (17) and the rated impedances and power supply (9/10). Since revised as IEC 60268-4:2018 (https://webstore.iec.ch/en/publication/32039); the 2014 edition is the implemented one.
- **International Telecommunication Union. (1986).** [*Measurement of audio-frequency noise voltage level in sound broadcasting*](https://www.itu.int/rec/R-REC-BS.468/en) (Recommendation ITU-R BS.468-4).
  The weighting network and quasi-peak detection behind the dB(468) inherent-noise figure of section 3.
