# Periodic Tests of a Sound Level Meter (IEC 61672-3)

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

Canonical: https://docs.jmrp.io/phonometry/signals/metrology/sound-level-meter-periodic-tests/
Language: en
Alternate: https://docs.jmrp.io/phonometry/es/signals/metrology/sound-level-meter-periodic-tests/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: 2026-11-16T21:36:13Z
Standards: IEC 61672-3:2013; IEC 61672-1:2013
Verified: 55 conformance checks (IEC 61672-3:2013; IEC 61672-1:2013) https://docs.jmrp.io/phonometry/reference/conformance/
License: MIT (https://github.com/jmrplens/phonometry/blob/main/LICENSE)
Summary: IEC 61672-3:2013 verdict on a working sound level meter from a laboratory's periodic tests, clause by clause against IEC 61672-1.

---

A sound level meter in service goes back to a laboratory every year or two
for its **periodic tests**, the short list of key checks IEC 61672-3:2013
keeps "to the minimum considered necessary" (Clause 1): a calibrator on the
microphone, the frequency weightings acoustically at three frequencies and
electrically at the octaves, the weightings and time weightings at 1 kHz, the
stability over half an hour and at the top of the range, the level linearity,
the tonebursts, the C-weighted peak and the overload indicator. The laboratory
measures; `metrology.verify_sound_level_meter_periodic` grades what it
measured, clause by clause, and writes the statement Clause 22 prescribes for
the outcome. It runs none of the tests.

This page grades the record of a synthetic class 1 meter whose one fault is a
0.25 ms toneburst, then shows what the verdict needs from a record to call
it complete.

![The periodic tests of a synthetic class 1 sound level meter, one slot per clause from 12 to 21 on a symmetric-logarithmic axis with a dashed line at zero: the distance from each result's deviation to its nearer acceptance limit is a short blue dash, and over the dashes of each clause the result that decides it is marked with its actual expanded uncertainty as error bar. Clause 13 spreads its 27 dashes from 0.7 dB to 3.7 dB and Clause 16 its 25 from 0.2 dB to 0.8 dB. Every clause is marked by a green diamond above the line, those of 14.2, 14.3, 15 and 21 within 0.1 dB of it, except clause 18, whose red cross at -0.4 dB is the maximum F-weighted response to the 0.25 ms toneburst; the title reads IEC 61672-3 periodic tests, class 1: not passed](https://docs.jmrp.io/phonometry/media/slm_periodic_verdict.svg)

*Every graded clause of the record in one slot, from the result's own
`.plot()`: each result's margin to its nearer acceptance limit is a dash, and
the result that decides the clause is marked over them. One toneburst below
its limit holds the verdict back.*

## 1. The rule every clause is graded by

IEC 61672-3 grades a result the way every IEC TC 29 instrument standard
written since 2013 does (4.1): the measured deviation from the design goal
within the acceptance limits of IEC 61672-1 **and** the laboratory's actual
expanded uncertainty, for a coverage probability of 95 %, within the maximum
permitted by Table B.1 of IEC 61672-1, both limits inclusive. The
uncertainty is not added to the deviation; the tolerance the meter has to
meet lies beyond the acceptance limit by that maximum. Each result is one
`metrology.verify_conformance`, and [Compliance and
verification](https://docs.jmrp.io/phonometry/signals/metrology/compliance-verification/) shows
the rule on the printed examples of IEC 61672-1 Table C.1.

A result measured with more uncertainty than its maximum "shall not be used
to evaluate conformance" (4.3). It is not a failure of the meter, and the
verdict does not pass while it is there: the verdict lists it under
`unusable`, apart from the results under `failed`.

The one exception is 4.4. The acoustical test of the frequency weighting
(Clause 12) and the electrical one (Clause 13) correct the indications with
the manufacturer's free-field or random-incidence correction data, and the
uncertainty of that data is part of the laboratory's budget. Without it, the
laboratory's uncertainty "shall not exceed" the maximum; when the total
exceeds the maximum only because of it, "testing may proceed", and the
result is one that did not conform, with that reason in the statement. The
record takes the uncertainty without the correction data beside the total
(section 6 shows one); without it, the total is all the verdict knows and 4.3
applies.

## 2. What each clause is judged against

| Clause | Test | Acceptance limits (IEC 61672-1) | Maximum uncertainty (Table B.1) |
| :--- | :--- | :--- | :--- |
| 12 | Frequency weighting with acoustical signals, at 125 Hz and 8 kHz relative to 1 kHz | Table 3 | 0.60 dB up to 4 kHz, 0.70 dB above |
| 13 | Every frequency weighting with electrical signals, 63 Hz to 16 kHz (class 1) or to 8 kHz (class 2) | Table 3 | 0.60 dB, 0.70 dB above 4 kHz, 1.00 dB above 10 kHz |
| 14.2 | C and Z against A at 1 kHz | ±0.2 dB (5.5.9) | 0.20 dB |
| 14.3 | S and time-averaged against F at 1 kHz | ±0.1 dB (5.8.3) | 0.20 dB |
| 15 | Long-term stability, 25 min to 35 min | ±0.1 dB, class 2 ±0.3 dB (5.14.2) | 0.10 dB |
| 16 | Level linearity at 8 kHz on the reference level range | ±0.8 dB, class 2 ±1.1 dB (5.6.5) | 0.30 dB |
| 17 | Level linearity including the level range control, at 1 kHz | ±0.8 dB, class 2 ±1.1 dB (5.6.5) | 0.30 dB |
| 18 | 4 kHz tonebursts of 200 ms, 2 ms and 0.25 ms | Table 4 | 0.30 dB |
| 19 | C-weighted peak: one cycle at 8 kHz, half cycles at 500 Hz | Table 5 | 0.35 dB |
| 20 | Overload indication, positive against negative half cycle | 1.5 dB (5.11.3) | 0.25 dB |
| 21 | High-level stability, 5 min near the top of the least-sensitive range | ±0.1 dB, class 2 ±0.3 dB (5.15.2) | 0.10 dB |

Two things in the table are as the pages print them and not what one might
expect. The maximum uncertainty of 14.3, 0.20 dB, is twice its acceptance
limit, 0.1 dB. And Part 3 grades the level linearity of Clauses 16 and 17
against 5.6.5 alone: the limit of 5.6.6 on a 1 dB to 10 dB change in level is
a pattern-evaluation requirement (IEC 61672-2, 9.8.1.7) that Part 3 never
mentions.

Clauses 10 and 11 are records, not tests: the indications at the calibration
check frequency before and after the adjustment, and the self-generated
noise, which "is reported for information only and is not used to assess
conformance" and without an uncertainty (11.1.2, NOTE 2). Clauses 16, 19 and
20 each hold a yes/no requirement besides the numbers: no overload or
under-range indication inside the linear operating range the manual states
(16.4), the C-weighted peak signals shall not cause an overload indication
(19.3, 19.5), and the overload indicator shall latch on (20.5).

## 3. Grading a record

`metrology.SoundLevelMeterPeriodicMeasurements` holds what the laboratory
measured, each graded result beside its expanded uncertainty and every field
named by its unit. The toneburst responses and the C-weighted peak
differences go in as measured, relative to the steady level; the verdict
takes the deviation from Tables 4 and 5 itself. Everything else goes in as
the deviation the clause defines. Clause 16 also carries what shows that its
steps cover the procedure of 16.3: over a linear operating range of 80 dB,
the 25 steps a laboratory takes, which the verdict's figure draws in one
slot. The meter's two level ranges, which no record can count, are declared
beside it; section 5 says why.

```python
from phonometry import metrology

record = metrology.SoundLevelMeterPeriodicMeasurements(
    # Clause 7: the conditions at the start and the end of the tests.
    static_pressures_kpa=[100.9, 100.7],
    air_temperatures_c=[22.8, 23.3],
    relative_humidities_percent=[48.0, 46.0],
    # Clauses 10 and 11: recorded, not graded.
    calibration_check_initial_level=93.8,
    calibration_check_adjusted_level=94.0,
    microphone_self_noise_level=16.9,
    electrical_self_noise_levels={"A": 11.8, "C": 13.9, "Z": 19.2},
    # Clause 12: 125 Hz and 8 kHz, relative to 1 kHz, after the corrections.
    acoustic_weighting_deviations_db=[0.3, -0.9],
    acoustic_weighting_uncertainties_db=[0.28, 0.45],
    # Clause 13: every weighting, at the nine octaves from 63 Hz to 16 kHz.
    electrical_weighting_deviations_db={
        "A": [0.1, 0.1, 0.0, 0.0, 0.0, 0.0, -0.1, -0.3, -0.9],
        "C": [0.2, 0.1, 0.0, 0.0, 0.0, 0.0, -0.1, -0.4, -1.2],
        "Z": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.2, -0.8],
    },
    electrical_weighting_uncertainties_db={
        w: [0.14] * 7 + [0.18, 0.25] for w in ("A", "C", "Z")
    },
    # Clause 14: C and Z against A, S and time-averaged against F, at 1 kHz.
    weighting_at_1khz_deviations_db={"C": 0.0, "Z": 0.1},
    weighting_at_1khz_uncertainties_db={"C": 0.12, "Z": 0.12},
    time_weighting_at_1khz_deviations_db={"S": 0.0, "eq": 0.05},
    time_weighting_at_1khz_uncertainties_db={"S": 0.12, "eq": 0.12},
    # Clause 15: the long-term stability.
    long_term_stability_db=0.04,
    long_term_stability_uncertainty_db=0.07,
    # Clause 16 at 8 kHz: the linear operating range and the starting point
    # the manual states, each step by its anticipated level, up from the
    # starting point in 5 dB steps and in 1 dB steps from within 5 dB of the
    # upper boundary, then down the same way, and the levels that first
    # indicated overload and under-range.
    linear_operating_range_db=[40.0, 120.0],
    linearity_starting_level=94.0,
    linearity_levels=[
        94.0, 99.0, 104.0, 109.0, 114.0, 119.0, 120.0, 121.0, 122.0,
        89.0, 84.0, 79.0, 74.0, 69.0, 64.0, 59.0, 54.0, 49.0, 44.0,
        43.0, 42.0, 41.0, 40.0, 39.0, 38.0,
    ],
    linearity_deviations_db=[
        0.0, 0.0, 0.1, 0.1, 0.1, 0.1, 0.0, -0.1, -0.2,
        0.0, 0.0, 0.0, -0.1, 0.0, 0.0, 0.1, 0.1, 0.1, 0.2,
        0.2, 0.3, 0.3, 0.4, 0.5, 0.6,
    ],
    linearity_uncertainties_db=[0.18] * 25,
    linearity_overload_level=123.0,
    linearity_under_range_level=37.0,
    # Clause 17 at 1 kHz: 5 dB above the first under-range indication on
    # both level ranges, the reference one first (17.4), and the reference
    # level held on the other one (17.3).
    range_linearity_deviations_db={"17.3": [0.1], "17.4": [0.1, -0.2]},
    range_linearity_uncertainties_db={"17.3": [0.18], "17.4": [0.18, 0.18]},
    # Clause 18: the responses at 200 ms, 2 ms and 0.25 ms (S: 200 ms, 2 ms).
    toneburst_responses_db={
        "F": [-1.1, -18.2, -30.4],
        "S": [-7.5, -27.4],
        "E": [-7.0, -27.1, -36.6],
    },
    toneburst_uncertainties_db={
        "F": [0.2, 0.2, 0.25],
        "S": [0.2, 0.2],
        "E": [0.2, 0.2, 0.25],
    },
    # Clause 19: one cycle at 8 kHz, then the two half cycles at 500 Hz.
    c_peak_differences=[3.9, 2.5, 2.2],
    c_peak_uncertainties_db=[0.3, 0.3, 0.3],
    c_peak_overload_indicated=False,
    # Clauses 20 and 21.
    overload_difference=0.4,
    overload_uncertainty_db=0.2,
    overload_latched=True,
    high_level_stability_db=0.03,
    high_level_stability_uncertainty_db=0.07,
)
meter = metrology.SoundLevelMeterFeatures(n_level_ranges=2)
result = metrology.verify_sound_level_meter_periodic(1, record, features=meter)
print(result.passes)  # False
print(result.failed)  # (('18', 'F, 0.25 ms'),)
result.plot()
```

The verdict is an object, not a boolean: `result.passes` is the answer, and
`bool(result)` raises rather than let `if verify_sound_level_meter_periodic(...):`
pass every meter. Each requirement is one `result.requirement(name)`, with
its results under `labels` and one `ConformanceVerification` each under
`verifications`, and `result.requirements` holds them in the order of the
standard.

## 4. Tonebursts and the C-weighted peak

The reference responses the toneburst test is judged from are those of
IEC 61672-1 Table 4, which its NOTES derive from Equation (7) for a maximum
time-weighted level,

$$
\delta_\mathrm{ref} = 10\,\lg\!\left(1 - \mathrm{e}^{-T_\mathrm{b}/\tau}\right)\ \mathrm{dB},
$$

with $\tau$ = 0.125 s for F and 1 s for S, and from Equation (8) for a sound
exposure level, $\delta_\mathrm{ref} = 10\,\lg(T_\mathrm{b}/T_0)$ dB with
$T_0$ = 1 s, both rounded to a tenth of a decibel. Their limits widen as the
toneburst shortens, and the S rows have limits of their own. The module
publishes the table as `IEC61672_TABLE_4`, keyed `"F"`, `"S"` and `"E"`:

```python
row = metrology.IEC61672_TABLE_4["F"][-1]
print(row.duration_ms, row.reference_response_db, row.limits_db(1))
# 0.25 -27.0 (-3.0, 1.0)
worst = result.requirement("toneburst").verifications[2]
print(round(worst.deviation, 2), worst.lower_limit)  # -3.4 -3.0
```

The meter read -30.4 dB for the 0.25 ms toneburst, 3.4 dB below the
reference, where class 1 allows 3.0 dB. Its uncertainty, 0.25 dB, is within
the 0.30 dB of Table B.1, so the result is usable and the meter fails the
clause.

![The eight toneburst results of clause 18 of the synthetic class 1 meter, as IEC 61672-1 Figure C.1 draws its examples: for each of F at 200 ms, 2 ms and 0.25 ms, S at 200 ms and 2 ms and E at 200 ms, 2 ms and 0.25 ms, the deviation from the reference response of Table 4 as a marker, its actual uncertainty as an error bar, the maximum-permitted uncertainty of 0.30 dB as a grey band behind it, and the acceptance limits of Table 4 as short orange bars, plus and minus 0.5 dB at 200 ms, +1.0 and -1.5 dB at 2 ms for F and E and -3.0 dB for S, and +1.0 and -3.0 dB at 0.25 ms. Every marker is a green diamond near zero except F at 0.25 ms, a red cross at -3.4 dB below its -3.0 dB limit; the title reads IEC 61672-3 §18: does not conform, over Toneburst response](https://docs.jmrp.io/phonometry/media/slm_periodic_toneburst.svg)

*The toneburst requirement, drawn by its own `.plot()`.*

The C-weighted peak test works the same way from Table 5: the difference
$L_\mathrm{Cpeak} - L_\mathrm{C}$ measured for one cycle at 8 kHz and for the
positive and the negative half cycle at 500 Hz, less the reference
difference of 3.4 dB, 2.4 dB and 2.4 dB. The record gives them in that
order, and the test's other requirement, that none of the three signals
causes an overload indication, as `c_peak_overload_indicated`.

```python
print(metrology.IEC61672_TABLE_5[2])
# PeakReference(signal='one cycle', nominal_frequency_hz=8000.0,
#   reference_difference=3.4, class_1_limits_db=(-2.0, 2.0),
#   class_2_limits_db=(-3.0, 3.0))
```

Table B.1, the maximum-permitted uncertainty of every requirement of
IEC 61672-1 and not only the periodic ones, is `IEC61672_TABLE_B1`, its
frequency bands with their printed ends; the frequency weightings are read
through `filters.weighting_class_limits`, which publishes Table 3.

## 5. What a complete test holds

A pass is a verdict on a complete record. 8.1 asks for every test of a
design feature IEC 61672-1 requires and the meter provides. A record shows
the features whose results it holds, and the verdict holds it to them:

- Clauses 7, 10, 11, 12, 13, 15, 16, 18 and 21 for every meter;
- frequency weighting A in Clause 13 always, and C as well for class 1,
  which has to provide it (5.1.9, 5.1.10 of IEC 61672-1), and for a meter
  with a C-weighted peak, which 5.1.10 has measure C-weighted time-averaged
  sound level too;
- every weighting any clause names, whether Clause 13, the self-generated
  noise of 11.2 or 14.2, tested in all three (C and Z compared with A in
  14.2);
- Clause 16 over the steps of 16.3: up from the starting point in 5 dB steps
  until within 5 dB of the upper boundary of the linear operating range, in
  1 dB steps from there up to, but not including, the first indication of
  overload, and down the same way to the first indication of under-range. A
  step wider than that, or a record without the levels to tell, leaves the
  clause incomplete; smaller steps cover more and are accepted. An overload
  or under-range indication inside the linear operating range, where the
  manual says the meter shows neither (IEC 61672-1 5.6.10), fails the clause
  (16.4);
- Clause 17 at 5 dB above the first under-range indication on every level
  range, the reference one included and first (17.4 tests "each level
  range"), and with the reference level held on every other one (17.3), NaN
  where a range does not display it. On the reference level range 17.3 has
  nothing to read: the reference level 17.2 set is the anticipated level
  there, a deviation of zero by construction (IEC 61672-1 5.6.3);
- the F, S and time-averaged displays in the toneburst tests of Clause 18,
  the sound exposure level calculated from the time-averaged one when the
  meter does not measure it (18.2), and the other way round: 14.3, with S,
  for a meter whose S-time-weighted toneburst Clause 18 holds, and with the
  time-averaged display for a meter whose overload indication Clause 20
  holds;
- Clause 20 for a meter that displays time-averaged sound level (20.1);
- the environmental conditions within 80 kPa to 105 kPa, 20 °C to 26 °C and
  25 % to 70 % (7.1), recorded at the start and the end at least (7.2).

The two level linearity items, drawn for the record of section 3: above, the
25 levels of `linearity_levels` in the order 16.3 takes them, and below, where
the readings of `range_linearity_deviations_db` fall on the two level ranges.

![Two panels. The upper one plots the anticipated level of the Clause 16 test at 8 kHz on a vertical axis ticked from 40 dB to 120 dB, with the linear operating range of the manual, 40 dB to 120 dB, shaded. From the starting point at 94 dB a staircase rises in 5 dB steps through 99, 104, 109, 114 and 119 dB and then in 1 dB steps through 120, 121 and 122 dB, and ends at a cross at 123 dB, the first indication of overload; a second staircase falls from the same starting point in 5 dB steps from 89 dB to 44 dB and in 1 dB steps from 43 dB to 38 dB, and ends at a cross at 37 dB, the first indication of under-range. Brackets mark the 5 dB below the upper boundary and the 5 dB above the lower one, where the steps become 1 dB. An inset draws the rise from 119 dB to 121 dB without the 120 dB step, 2 dB where 16.3 allows 1 dB, which the verdict lists as incomplete. The lower panel draws the two level ranges of the Clause 17 test at 1 kHz as two bars without a scale: on each, a cross at the first indication of under-range and a dot 5 dB above it for 17.4; a dashed line at the reference sound level crosses both bars, with an open circle where 17.2 sets it on the reference level range and a dot where 17.3 reads it on the other](https://docs.jmrp.io/phonometry/media/slm_periodic_linearity.svg)

*The steps of Clauses 16 and 17, read off the record of section 3 by the
library. The staircase is `linearity_levels` from `linearity_starting_level`,
the two crosses `linearity_overload_level` and `linearity_under_range_level`,
and the inset is the verdict on the same record without its 120 dB step; the
lower panel is the one reading of 17.3 and the two of 17.4 that
`range_linearity_deviations_db` holds.*

What a record cannot show is a feature the meter lacks: a meter with one
level range and a record that left Clause 17 out read the same, and so do a
meter without C-weighted peak and a record without Clause 19. The instruction
manual says which optional features the meter has, and
`metrology.SoundLevelMeterFeatures` carries that answer to the verdict,
through `features=`: `True` for a feature the meter has, `False` for one it
lacks, `None` (the default) for one not declared. A feature declared present
owes its tests as a shown one does, and one declared absent takes its
clauses out of the test. A feature neither declared nor shown keeps the
verdict from passing, as the record could be hiding a test left out.

The level ranges are a number, `n_level_ranges`, because a record cannot count
them either: a Clause 17 with one reading reads the same for a meter with two
level ranges and for a record that left the others out. `n_level_ranges=1`
takes Clause 17 out; a larger number is how many ranges 17.4 covers, the
reference one included, and 17.3 covers besides it, and until it is
declared, a meter whose Clause 17 shows several keeps the question open. A feature declared absent also takes
away the one IEC 61672-1 makes depend on it: a meter without frequency
weighting C measures no C-weighted peak sound level (5.1.10), so Clause 19 is
not applicable rather than undeclared, and a meter without F has no S
(5.1.9). A declaration without F, a time-averaged display and sound exposure
level is refused, as 5.1.9 asks one of them of every sound level meter.

`missing` names the clauses not measured, `incomplete` what a measured clause
is short of, `undeclared` the features the verdict cannot place,
`not_applicable` the clauses the declared features take out, and
`conditions_outside` the readings outside 7.1. The record of section 3 shows
every feature, and `meter` gives the number of level ranges, so nothing is
open. Without Clauses 17 and 19 it is open on both, until the meter is
declared with one level range and no C-weighted peak:

```python
import dataclasses

basic = dataclasses.replace(
    record,
    range_linearity_deviations_db=None,
    range_linearity_uncertainties_db=None,
    c_peak_differences=None,
    c_peak_uncertainties_db=None,
    c_peak_overload_indicated=None,
)
unsaid = metrology.verify_sound_level_meter_periodic(1, basic)
print([name for name, _ in unsaid.undeclared])
# ['n_level_ranges', 'c_weighted_peak']
declared = metrology.verify_sound_level_meter_periodic(
    1,
    basic,
    features=metrology.SoundLevelMeterFeatures(
        n_level_ranges=1, c_weighted_peak=False
    ),
)
print(declared.undeclared, [clause for clause, _ in declared.not_applicable])
# () ['17', '19']
```

The verdict's figure gives every graded clause one slot, however many
results it holds, and keeps one for every graded clause the record does not
hold, named under the slot: not measured, not declared or not applicable.

The same record without the S toneburst results, whose meter 14.3 shows to
have S, is incomplete:

```python
no_s = dataclasses.replace(
    record,
    toneburst_responses_db={"F": [-1.1, -18.2, -30.4], "E": [-7.0, -27.1, -36.6]},
    toneburst_uncertainties_db={"F": [0.2, 0.2, 0.25], "E": [0.2, 0.2, 0.25]},
)
short = metrology.verify_sound_level_meter_periodic(1, no_s, features=meter)
print(short.passes, short.incomplete[0][0])  # False 18
```

So is a Clause 16 of one step at the starting point, however well it
conforms: it covers neither the way up to the overload nor the way down to
the under-range.

```python
one_step = dataclasses.replace(
    record,
    linearity_levels=[94.0],
    linearity_deviations_db=[0.0],
    linearity_uncertainties_db=[0.18],
)
thin = metrology.verify_sound_level_meter_periodic(1, one_step, features=meter)
print(thin.passes, [clause for clause, _ in thin.incomplete])  # False ['16', '16']
print(thin.incomplete[0][1][:60])
# the steps go from 94 dB to the first indication of overload
```

And so is a Clause 17 whose 17.4 leaves out the reference level range, NaN
in its first place: 17.4 reads "for each level range", and a level 5 dB
above the first under-range indication is not the reference level 17.2 set
there.

```python
nan = float("nan")
no_reference = dataclasses.replace(
    record,
    range_linearity_deviations_db={"17.3": [0.1], "17.4": [nan, -0.2]},
    range_linearity_uncertainties_db={"17.3": [0.18], "17.4": [nan, 0.18]},
)
open_17 = metrology.verify_sound_level_meter_periodic(1, no_reference, features=meter)
print(open_17.incomplete)
# (('17', '17.4 was not measured on the reference level range, and it tests
#   every level range, the reference one included'),)
```

The conditions matter to the uncertainty too: below 97 kPa, and unless the
model's data say otherwise, 7.3 asks the laboratory to add a standard
uncertainty of 0.09 dB up to 3 kHz and 0.14 dB above it to its tests of the
frequency weightings. That component belongs to the laboratory's budget, and
the verdict reads the expanded uncertainty it is given.

## 6. The statement of Clause 22

`result.statement` is the text the documentation of the test carries,
chosen as Clause 22 chooses it and with the class and the year filled in.
The first of these that applies decides it:

- **a notice** when the tests were not performed within the conditions of
  7.1: tests outside them are not periodic tests to IEC 61672-3, whatever
  their results, and none of 22 r), s) and t) is theirs;
- **22 t)** when a result lies outside its limits, a check fails or a
  result exceeds its maximum only by the manufacturer's correction data
  (4.4): the meter
  "did not successfully complete the class 1 tests of IEC 61672-3:2013" and
  "did not conform to the class 1 specifications of IEC 61672-1:2013",
  followed by the tests not completed and why, in the words its NOTE
  suggests;
- **a notice** when a result is unusable (4.3) or the record is
  incomplete, as 22 r) and s) both need the results of all the periodic
  tests;
- **22 r)** when every result passes, evidence of the model's pattern
  approval under IEC 61672-2 is publicly available
  (`pattern_approval_public=True`) and the correction data of the acoustical
  test came from the instruction manual (`corrections_in_manual=True`): the
  meter "conforms to the class 1 specifications of IEC 61672-1:2013";
- **22 s)** when every result passes but either is missing: no general
  conclusion can be drawn about IEC 61672-1, because the periodic tests
  "cover only a limited subset of the specifications".

Both declarations default to `False`: 22 c) and 12.5 have the laboratory
state them, and the verdict assumes neither. With the 0.25 ms toneburst
brought within its limit, a public pattern approval and the correction data
of Clause 12 taken from the instruction manual, as 12.3 asks first, the same
meter earns 22 r). `corrections_in_manual=True` is what says where the data
came from: data from the manufacturer of the microphone or of the calibrator
(12.4), or a source left undeclared, would give 22 s) instead.

```python
fixed = dataclasses.replace(
    record,
    toneburst_responses_db={
        "F": [-1.1, -18.2, -29.6],
        "S": [-7.5, -27.4],
        "E": [-7.0, -27.1, -36.6],
    },
)
again = metrology.verify_sound_level_meter_periodic(
    1,
    fixed,
    pattern_approval_public=True,
    corrections_in_manual=True,
    features=meter,
)
print(again.passes)                                  # True
print(again.statement.endswith("of IEC 61672-1:2013."))  # True
```

A laboratory whose uncertainty at 8 kHz in Clause 12 is 0.85 dB, past the
0.70 dB of Table B.1 only because of the manufacturer's correction data,
gives the 0.66 dB it has without them beside the total. The result is then
one that did not conform, by 4.4, and the statement is 22 t) with the reason
the NOTE to 22 t) gives:

```python
over = dataclasses.replace(
    fixed,
    acoustic_weighting_uncertainties_db=[0.28, 0.85],
    acoustic_weighting_uncertainties_without_correction_data_db=[0.28, 0.66],
)
check = metrology.verify_sound_level_meter_periodic(
    1,
    over,
    pattern_approval_public=True,
    corrections_in_manual=True,
    features=meter,
)
print(check.unusable, check.over_max_by_correction_data)
# () (('12', '8 kHz'),)
reason = check.requirement("acoustic_weighting").failure_reason("8 kHz")
print(reason.endswith("uncertainty budget (4.4)"))  # True
```

**Show the code for these figures**

```python
import matplotlib.pyplot as plt

fig, ax = plt.subplots(figsize=(11, 6.2), layout="constrained")
result.plot(ax=ax)

fig, ax = plt.subplots(figsize=(10, 5.6), layout="constrained")
result.requirement("toneburst").plot(ax=ax)
plt.show()
```

## What this guide covers

- **Covered.** **IEC 61672-3:2013, the grading of a laboratory's results:** the conformance
  rule of 4.1 on every graded result of Clauses 12 to 21, with the acceptance
  limits of IEC 61672-1:2013 (Table 3, 5.5.9, 5.6.5, 5.8.3, Table 4, 5.11.3,
  Table 5, 5.14.2 and 5.15.2) and the maxima of its Table B.1; the unusable
  results of 4.3 and the exception of 4.4 for the uncertainty of the
  manufacturer's correction data; the yes/no requirements of 16.4, 19.3, 19.5
  and 20.5; the
  environmental conditions of 7.1 and 7.2; the completeness 8.1 asks for, from
  the features the record shows and those `SoundLevelMeterFeatures` declares
  present or absent, with what IEC 61672-1 5.1.9 and 5.1.10 make of a feature
  declared absent; the extent of Clause 16 over the steps of 16.3 and of
  Clause 17 over every level range (17.3, 17.4); and the statements of 22 r),
  s) and t). Tables 4, 5 and B.1 of IEC 61672-1 are published read-only.

- **Not covered.** The tests themselves: the library drives no calibrator, coupler, actuator or
  electrical input. The level linearity deviations of 16.2 (by the procedure of
  IEC 61672-2), the relative frequency weightings of 12.15 and 13.9 with their
  corrections, and the laboratory's uncertainty budget are inputs. The checks
  that are not a measured deviation are the laboratory's record: the instruction
  manual and markings of Clause 3, the preliminary inspection of Clause 5, the
  power supply of Clause 6, the calibrator's own conformance of Clause 9 (which
  `metrology.verify_sound_calibrator` grades against IEC 60942), the general
  test requirements of 8.2 to 8.5, among them the confirmation of 8.4 that an
  electrical output used for the tests reads as the display does (within the
  ±0.1 dB of IEC 61672-1 5.19.4), and the source of the correction data of
  12.2 to 12.6. The 22 l) statement, which the documentation carries when the
  free-field correction data come without an uncertainty, is not written.

## See also

- [Compliance and verification](https://docs.jmrp.io/phonometry/signals/metrology/compliance-verification/):
  the conformance rule of IEC TC 29, and what pattern evaluation and periodic
  tests each attest.
- [Free-field corrections of a sound level meter](https://docs.jmrp.io/phonometry/signals/metrology/free-field-corrections/):
  IEC 62585, the corrections the acoustical test of Clause 12 applies and
  their uncertainty.
- [Calibration and dBFS](https://docs.jmrp.io/phonometry/signals/metrology/calibration/): the
  sound calibrator of Clause 9 and its own verdict under IEC 60942.
- [Sound level meter](https://docs.jmrp.io/phonometry/signals/sound-level-meter/): the frequency
  and time weightings, and the design verdict the library gives its own
  chain.
- API reference: [`metrology.sound_level_meter`](https://docs.jmrp.io/phonometry/reference/api/metrology/sound-level-meter/).

## References

- **International Electrotechnical Commission. (2012).** *Electroacoustics — Methods to determine corrections to obtain the free-field response of a sound level meter* (IEC 62585:2012).
  Where the correction data of the acoustical test of Clause 12 come from, and their maximum uncertainties when the manufacturer states none (12.5).
- **International Electrotechnical Commission. (2013).** [*Electroacoustics — Sound level meters — Part 1: Specifications*](https://webstore.iec.ch/en/publication/5708) (IEC 61672-1:2013).
  The acceptance limits each clause of Part 3 points to: Table 3 (frequency weightings), 5.5.9, 5.6.5, 5.8.3, Table 4 (tonebursts), 5.11.3, Table 5 (C-weighted peak), 5.14.2 and 5.15.2, and the maximum-permitted uncertainties of Table B.1. Tables 4, 5 and B.1 are published by the module. Read from BS EN 61672-1:2013.
- **International Electrotechnical Commission. (2013).** [*Electroacoustics — Sound level meters — Part 2: Pattern evaluation tests*](https://webstore.iec.ch/en/publication/5709) (IEC 61672-2:2013).
  The pattern evaluation whose public approval turns a passed periodic test into a statement of conformance (22 r), and whose calculation procedure the level linearity deviations of 16.2 follow.
- **International Electrotechnical Commission. (2013).** [*Electroacoustics — Sound level meters — Part 3: Periodic tests*](https://webstore.iec.ch/en/publication/5710) (IEC 61672-3:2013).
  The implemented grading: the conformance rule of 4.1, the unusable results of 4.3 and the exception of 4.4 for the manufacturer's correction data, the environmental conditions of 7.1, what 8.1 makes a complete test, the tests of Clauses 10 to 21 and the statements of 22 r), s) and t). Read from BS EN 61672-3:2013, the English text of EN 61672-3:2013, which is IEC 61672-3:2013 unchanged.
- **International Electrotechnical Commission. (2017).** [*Electroacoustics — Sound calibrators*](https://webstore.iec.ch/en/publication/30045) (IEC 60942:2017).
  The calibrator of Clause 9, whose own periodic test the laboratory needs before the meter's.
