@audio/k
Status calibrated
npm1.2.1licenseMIT
ITU-R BS.1770-4 K-weighting: high-shelf + RLB highpass
ITU-R BS.1770-4 K-weighting — the loudness pre-filter behind LUFS
import kWeighting from '@audio/weighting-k'
Time-domain K-weighting: two cascaded stages redesigned per sample rate from the
canonical ITU-R BS.1770-4 Annex 1 analog prototype constants (De Man 2014, "Evaluation
of Implementations of the ITU-R BS.1770 Loudness Algorithm" — the same design pyloudnorm
uses). Stage 1 is a spherical-head high shelf (G=3.999843853973347 dB,
Q=0.7071752369554196, fc=1681.974450955533 Hz); stage 2 is the RLB weighting highpass
(Q=0.5003270373238773, fc=38.13547087602444 Hz, unnormalized {1,-2,1} numerator at any
fs). At fs=48000 this reproduces the spec's published coefficient table to ~1e-11.
kWeighting(data) // in place, fs=48000
kWeighting(data, { fs: 44100 })
let sos = kWeighting.coefs(44100) // 2 SOS sections (shelf, then RLB highpass)
kWeighting.response(2000, 48000) // |H(2000)| — shelf boost
| Param | Default | |
|---|---|---|
fs |
48000 |
sample rate — SOS recomputed when it changes mid-stream |
| Static | Returns | |
|---|---|---|
.coefs(fs) |
SOS |
2-section cascade (high shelf + RLB highpass) at rate fs, for analysis/plotting |
.response(f, fs) |
number |
|H(f)| — exact magnitude of this atom's own .coefs(fs) cascade, not an independent approximation |
.ms(channels, params) |
Float64Array |
mean square of each channel's K-weighted block (input untouched; state per channel on params, same params every block): the BS.1770 measurement loop, fused, float64 |
Use when: implementing integrated/momentary/short-term loudness (LUFS) per ITU-R BS.1770 / EBU R128 — this is the mandated pre-filter, applied before mean-square gating and channel summing.
Part of @audio/weighting — the weighting family umbrella.
MIT © audiojs