Module companding 

Audio companding: the non-linear µ-law and “A”-law compression curves used by the ITU-T G.711 telephony standard to compress a wide-dynamic-range linear audio sample down to a lower bit depth with minimal perceptual loss (Steven W. Smith, “The Scientist and Engineer’s Guide to DSP”, Ch. 22, Eq. 22-1 / 22-2).

All functions operate on samples normalized to -1.0..=1.0.

Functions

a_law_compress_f32

Compresses a normalized linear sample x (-1.0..=1.0) using “A”-law companding (Eq. 22-2): a linear segment near zero, transitioning to a logarithmic curve.

a_law_expand_f32

Expands an “A”-law-compressed sample y (-1.0..=1.0) back to a normalized linear sample, inverting a_law_compress_f32.

mu_law_compress_f32

Compresses a normalized linear sample x (-1.0..=1.0) using µ255-law companding (Eq. 22-1), expanding resolution for small amplitudes at the expense of large ones.

mu_law_expand_f32

Expands a µ-law-compressed sample y (-1.0..=1.0) back to a normalized linear sample, inverting mu_law_compress_f32.