materials.diffusers.measured_scattering

Scattering coefficients as the books print them, one row per surface.

A geometric room acoustics model asks for one number per surface per band, the scattering coefficient, and has no way of working it out. The rest of this subpackage computes it: .reverberation_room_scattering runs the ISO 17497-1 procedure on four reverberation times, and .design predicts a polar response from a well sequence. This module holds what other people measured, so that a modeller who has neither a turntable nor a boundary element solver has somewhere to start, and so that a measurement made here has published values to sit beside.

ISO 17497 is two documents and two quantities. Part 1 measures the scattering coefficient, the fraction of reflected energy that leaves the specular direction, from four reverberation times in a room with a turning table; it says nothing about where that energy goes. Part 2 measures the diffusion coefficient, how even the polar response is, and a surface can score high on one and low on the other. A catalogue that put them in one field would let a caller pass a diffusion coefficient to a model that wants a scattering coefficient, which is a silent error, so they are two classes and two catalogues: PUBLISHED_SCATTERING here, and the diffusion coefficients of ISO 17497-2 where they belong.

These tables are set in one-third octave bands, so each band is a field of its own, scattering_coefficient_1000 and so on, with the band’s centre frequency in hertz as the suffix and no unit because the quantity has none. Every hedge of CatalogueRow is keyed by field name, so a band the page leaves empty says so through why_missing rather than answering zero, and a zero here would read as a perfectly specular surface. ScatteringCoefficientSpectrum.bands and spectrum hand the row back as a spectrum for the caller who wants one.

A row is a surface of that description, measured once, by one team, in one room. Cox prints no uncertainty and no laboratory, and the spread between teams is not small: the same battens, 10 cm high and 10 cm wide on a 20 cm period, appear twice in Appendix D, credited to two different papers, and read 0.28 and 0.44 at 630 Hz. Nothing is wrong with either; that is how wide the method is. A coefficient above one is likewise not an error but what the formula gave, since ISO 17497-1 derives it from a ratio of reverberation times and puts no ceiling on the result.

The surfaces are described, not named. A row reads "h = w = 10 cm, L = 2h" and means nothing without the group heading above it, so group carries that heading and scattering_named matches on either.

Auto-generated from the source docstrings by scripts/generate_api_docs.py (make api-docs). Do not edit by hand.

Constant (mapping).

Constant (tuple).

SCATTERING_BANDS_HZ = (100, 125, 160, 200, 250, 315, 400, 500, 630, 800, 1000, 1250, 1600, 2000, 2500, 3150, 4000, 5000)

scattering_named(name: str) -> tuple[ScatteringCoefficientSpectrum, ...]

Every published row whose description or group contains name.

Parameters

Returns: The rows that match, in the order the tables are read, which is empty when no page has one.

ScatteringCoefficientSpectrum(
    *,
    name: str,
    source: str,
    table: str = '',
    variant: str = '',
    basis: Mapping[str, str] = ...,
    approximate: frozenset[str] = frozenset(),
    derived: Mapping[str, str] = ...,
    converted: Mapping[str, tuple[str, str]] = ...,
    carried: Mapping[str, str] = ...,
    ranges: Mapping[str, tuple[float | None, float | None]] = ...,
    bounded_above: frozenset[str] = frozenset(),
    bounded_below: frozenset[str] = frozenset(),
    reported: Mapping[str, tuple[float | tuple[float, float], ...]] = ...,
    unquantified: Mapping[str, str] = ...,
    uncertainty: Mapping[str, float] = ...,
    not_derivable: Mapping[str, str] = ...,
    misprinted: Mapping[str, str] = ...,
    attributed_to: Mapping[str, str] = ...,
    group: str = '',
    note: str = '',
    scattering_coefficient_100: float | None = None,
    scattering_coefficient_125: float | None = None,
    scattering_coefficient_160: float | None = None,
    scattering_coefficient_200: float | None = None,
    scattering_coefficient_250: float | None = None,
    scattering_coefficient_315: float | None = None,
    scattering_coefficient_400: float | None = None,
    scattering_coefficient_500: float | None = None,
    scattering_coefficient_630: float | None = None,
    scattering_coefficient_800: float | None = None,
    scattering_coefficient_1000: float | None = None,
    scattering_coefficient_1250: float | None = None,
    scattering_coefficient_1600: float | None = None,
    scattering_coefficient_2000: float | None = None,
    scattering_coefficient_2500: float | None = None,
    scattering_coefficient_3150: float | None = None,
    scattering_coefficient_4000: float | None = None,
    scattering_coefficient_5000: float | None = None,
)

One measured surface of a published table, band by band.

The geometry is part of name, as the page prints it, and the family it belongs to is group: pulling h and L into fields would mean deciding what the height of a randomly arranged array of blocks is, and the pages do not agree on which letters they use.

The band fields and the reading method come from ScatteringBands, which a computed row carries too. What this class says, and what its name is for, is that the numbers were measured. It does not inherit from the predicted class and the predicted class does not inherit from it, so a caller narrowing on the type gets a straight answer.

Attributes

ScatteringCoefficientSpectrum.bands() -> tuple[int, ...]

The bands this row prints a value for, in hertz.

ScatteringCoefficientSpectrum.basis_of(field_name: str) -> str

What the source says this field is, one of CATALOGUE_BASES.

The five are measured, declared, calculated, estimated and extended.

Parameters

Returns: The field’s own entry in basis, else the row’s, else the empty string, which means the source does not say.

classmethod

ScatteringCoefficientSpectrum.from_printed(**cells: Any) -> Self

A row built from the cells its page prints, completed and marked.

The one path that works anything out. The cells are what the page prints, under the field names of the class and with the hedges each cell carries, as a data file writes them. A figure written under a unit the class takes as an alias of its own (an AbsorptionAreaSpectrum takes absorption_area_125_ft2 for absorption_area_125_m2) is converted on its digits with an exact factor and rounded once, and converted records the figure and its unit. The row is then built and held to the contract the class docstring lists, so every cell is checked before any arithmetic reads it. Last, the class fills what follows from those cells (a modulus from a plate speed, a density and a Poisson ratio), never over a cell that holds a value or one the row says something else about, and derived says how each filled value was reached and, when the cells it rests on do not share one basis, the basis of each. Cells the arithmetic cannot take are refused rather than turned into a value that would be wrong: a modulus of 1 GPa and a shear modulus of 0.1 GPa give a Poisson ratio of 4, which no isotropic solid has, and a row whose cells are not meant to give a value says so in not_derivable, which keeps the arithmetic from running.

Cls(...) stays literal: it holds what it is given and works nothing out. To change a cell of a row and have what follows from it follow again, change it in printed_fields and build again here; dataclasses.replace would copy the derived values as they were:

cells = row.printed_fields()
cells["density_kg_m3"] = 2400.0
row = type(row).from_printed(**cells)

Parameters

Returns: The row, with what follows from its cells filled in.

Raises

ScatteringCoefficientSpectrum.is_approximate(field_name: str) -> bool

Whether the page prints this field with a ~.

Parameters

Returns: True when the page rounded the cell on purpose.

ScatteringCoefficientSpectrum.is_derived(field_name: str) -> bool

Whether this library computed this field instead of reading it.

Parameters

Returns: True when the page did not print it and the value follows from cells that it did. derived says how. A value the page prints in another unit, or gives by reference to another of its rows, answers False: the number is the page’s, and converted or carried says so.

ScatteringCoefficientSpectrum.printed(
    field_name: str,
    *,
    wanted_by: str = 'the caller',
) -> float

One quantity this page prints, or a refusal that says what it had.

Every quantity of a row is optional, because the pages print different columns, so a caller passing one into a function that requires a float has to narrow it. Doing it here beats an assertion at each call site: the refusal names the field, who wanted it and what the page had in that cell, which is the difference between a cell the book left empty and a cell holding the word “model”.

Parameters

Returns: The value, as a float.

Raises

ScatteringCoefficientSpectrum.printed_fields() -> dict[str, Any]

The cells the page prints, as from_printed takes them.

Every field, the name, the citation, the table and every hedge included, except the values this library derived, derived itself, and a field left at a default that holds nothing (a quantity the page leaves out, an empty text or hedge). A text field whose default says something is kept even when it holds no text: the per of an area per unit is a person unless the row says otherwise, and a row that leaves it empty has said otherwise. A value converted from the page’s unit and one the page gives by reference to another of its rows are the page’s, so they stay, with converted and carried beside them.

For every row from_printed builds, and so for every packaged one, type(row).from_printed(**row.printed_fields()) is the row again, and changing a cell before building it again is how a row is edited without carrying a derived value that no longer follows from it. A derived passed to the literal constructor is the caller’s own; it is left out here with its value, like every derived one, so building again gives back only what the class works out.

Returns: A new dictionary of constructor keywords. The values are the ones the row holds, frozen as the row holds them.

ScatteringCoefficientSpectrum.scattering_coefficient(band_hz: int) -> float

The coefficient in one band, or a refusal that says what the page had.

Parameters

Returns: The scattering coefficient, dimensionless.

Raises

ScatteringCoefficientSpectrum.spectrum() -> dict[int, float]

The row as {band_hz: value} over the bands it prints.

A band the page left empty, or printed as something other than a number, is left out rather than filled with a zero; why_missing on that band’s field says which it was.

ScatteringCoefficientSpectrum.values_at(
    frequencies_hz: ArrayLike,
) -> NDArray[np.float64]

The row’s value at each of frequencies_hz, or a refusal.

The adapter for a function that takes one value per band by position (the absorption of a surface in a reverberation formula, the transmission loss of a panel at the frequencies an enclosure model asks for), because an array handed over by position cannot say which band is missing and a row can. Each frequency is read as the band of these tables whose centre is nearest on a logarithmic scale, and matches it when it lies within a sixth of the spacing between bands: a sixth of an octave for octave bands, an eighteenth for one-third octave bands. That takes the nominal centre, the exact base-ten one and the exact base-two one alike (160 Hz, 158.49 Hz and 157.49 Hz are one band) and never reaches the next band. A band the row does not print is refused, as printed refuses it, with what the page had there; it is never read as zero and never filled from its neighbours.

Parameters

Returns: A new array of the same shape, one value per frequency.

Raises

ScatteringCoefficientSpectrum.why_missing(field_name: str) -> str

Why this field is None, in the page’s own terms.

A catalogue that answers None and stops is asking the caller to guess whether the material has no such property, whether the book measured it and printed a dash, or whether the cell holds something that is not a number. Each of those is a different answer.

Parameters

Returns: What the page had in that cell, or the empty string when the field is not missing at all. A field the page has no column for and this library cannot derive, because the cells it would need are themselves a range, answers that it does not follow.

Raises