@audio/defeedback-notchbank

import notchbank from '@audio/defeedback-notchbank'

A pool of narrow peaking-EQ cuts sitting directly in the signal path — pure IIR, zero added latency. deploy(freq) activates a free slot at that frequency (or, if the pool is full, steals the shallowest cut); a repeated deploy at a frequency already covered deepens that notch instead of adding a new one (residual howl still growing → cut harder, down to maxDepth). Coefficients morph linearly over ramp samples on every deploy, so cuts fade in/deepen without clicks.

let notches = notchbank({ fs: 48000, size: 12, Q: 30, depth: -9 })

notches.deploy(2340)              // cut a howl at 2340 Hz
let out = notches.process(chunk)  // apply the active cuts, in place
notches.notches()                 // → [{ freq, gain }, ...] currently active

Returns { process(chunk): chunk, deploy(freq): void, notches(): Array<{freq, gain}> }. process mutates and returns chunk in place.

Use when: the direct audio path of an AFS chain — pair with @audio/defeedback-analyzer + @audio/defeedback-tracker to decide when to deploy.
Not for: detection — this module only applies cuts, it has no howl criteria of its own.

Part of @audio/defeedback — the defeedback family umbrella.

MIT © audiojs