psychoacoustics.quality.roughness_ecma

Psychoacoustic roughness per ECMA-418-2:2025 (4th ed., Sottek Hearing Model).

Clean-room implementation of the roughness signal chain of ECMA-418-2:2025 (Clause 7). The shared auditory front-end (Clause 5: outer/middle-ear filter, 53-band gammatone-like filter bank, half-wave rectification, compressive nonlinearity to the specific basis loudness N'_basis(l, z) of Formula 25) is reused from phonometry.psychoacoustics.loudness.ecma; this module adds the roughness-specific chain:

• roughness-specific zero-padding (Clause 5.1.2.2) and segmentation (Clause 5.1.5.2) with the fixed block/hop $s_{\mathrm{b}}=16384$ / $s_{\mathrm{h}}=4096$ (Clause 7.1.1);
• the Hilbert envelope of each critical-band block and a factor-32 downsampling to 1500 Hz (Clause 7.1.2, Formula 65);
• the scaled envelope power spectrum Phi_E,l,z(k) (Clause 7.1.3, Formulae 66-67);
• the two-step noise reduction of the envelope spectra (Clause 7.1.4, Formulae 68-71);
• the four-stage spectral weighting (Clause 7.1.5, Formulae 72-96): peak picking with a quadratic-fit modulation-rate refinement and a bias correction, the high-modulation-rate weighting, the fundamental modulation-rate estimation, and the low-modulation-rate weighting;
• the interpolation to 50 Hz, the distribution-dependent nonlinear transform with the calibration constant c_R, and the asymmetric time smoothing (Clause 7.1.7, Formulae 103-110); and
• the average specific roughness R'(z) (Clause 7.1.8), the time-dependent roughness R(l50) (Clause 7.1.9, Formula 111) and the representative 90th-percentile single value R (Clause 7.1.10).

The optional entropy weighting of Clause 7.1.6 requires an external rotational speed signal and is not implemented (see notes-ecma418-2-roughness.md). The API is monaural: the quadratic-mean binaural combination of Formula (112) (Clause 7.1.11) is not implemented — analyse each channel separately.

The calibration constant c_R of Formula (104) is the standard’s tabulated value (not reverse-fit), and the chain reproduces the Clause 7 reference point exactly: a 1 kHz carrier 100 %-amplitude-modulated at 70 Hz with an overall sound pressure level of 60 dB SPL computes to 0.9999 asper against the defined 1 asper. Note the level convention: Clause 7 states the sound pressure level of the signal (its overall RMS level), not the level of the unmodulated carrier — a fully modulated signal whose carrier alone sits at 60 dB is +1.76 dB hot overall and reads ~4 % high.

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

EcmaRoughness(
    roughness: float,
    specific_roughness: np.ndarray,
    bark: np.ndarray,
    centre_frequencies: np.ndarray,
    time: np.ndarray,
    roughness_vs_time: np.ndarray,
    specific_roughness_vs_time: np.ndarray,
    field: str,
)

Result of an ECMA-418-2:2025 (Sottek) roughness calculation.

roughness is the single representative roughness R in asper (the 90th percentile of R(l50), Clause 7.1.10). specific_roughness is the average specific roughness R’(z) in asper/Bark_HMS over the 53 auditory bands (Clause 7.1.8), with bark the critical-band-rate scale z (0.5..26.5 Bark_HMS) and centre_frequencies the band centre frequencies F(z). time and roughness_vs_time hold the time-dependent roughness R(l50) at 50 Hz (Formula 111); specific_roughness_vs_time is the time-dependent specific roughness R’(l50, z) (Formula 109) of shape (n_times, 53). field records the assumed sound field.

EcmaRoughness.plot(
    ax: Axes | None = None,
    *,
    language: str = 'en',
    **kwargs: Any,
) -> Axes | np.ndarray

Plot the roughness result (see phonometry._plot.psychoacoustics).

Draws the time-dependent roughness R(l50) and a specific-roughness heatmap. Requires matplotlib (pip install phonometry[plot]).

roughness_ecma(
    signal_in: Signal | np.ndarray,
    fs: float | None = None,
    field: Literal['free', 'diffuse'] = 'free',
) -> EcmaRoughness

Psychoacoustic roughness per ECMA-418-2:2025 (Sottek Hearing Model).

Parameters

Returns: An EcmaRoughness with the single value R (Clause 7.1.10), the average specific roughness R’(z) (Clause 7.1.8) and the time-dependent roughness R(l50) (Formula 111).

A 1 kHz carrier 100 %-amplitude-modulated at 70 Hz with an overall level of 60 dB SPL yields 1 asper (Clause 7 calibration; reproduced to 0.9999 asper with the tabulated c_R of Formula (104)).