Struct AecConfig 

pub struct AecConfig {
    pub tail_ms: u32,
    pub delay_ms: u32,
    pub mu: f32,
}

Tunables for Aec::new. All three fields have the module’s tested defaults via Default.

Fields

tail_ms: u32

Length of the modeled echo tail, in milliseconds. Rounded up to a whole number of 10 ms partitions. Must cover the room’s actual reverberant echo plus any residual misalignment left after delay_ms — see the module docs on bulk delay.

delay_ms: u32

Bulk delay compensation applied to the far-end reference before it enters the filter, in milliseconds — models playback pipeline latency between “handed to the speaker” and “captured by the mic”.

mu: f32

NLMS step size. The theoretical bound for a power-normalized update is 0 < mu < 2; the practical bound on narrowband bursty far ends (speech through a speaker) measured far below it: 0.25+ diverges within seconds, 0.1 reaches a bad marginal equilibrium (gradient-noise misadjustment above the echo level), 0.05 holds through every measured stress with ERLE +8..11 dB on the worst-case harmonic proxy and 12+ dB on broadband far ends. Default 0.05 — raise it only with evals/dspbench watching.

Trait Implementations

impl Default for AecConfig

fn default() -> Self

Returns the “default value” for a type.