MLT 4.81 MAXIMIZER - REFERENCES AND SCOPE
=======================================

Original stereo-linked limiter implementation. No source code, coefficient table,
recording or impulse response has been copied from another product.

ITU-R BS.1770, Algorithms to measure audio programme loudness and true-peak
audio level: https://www.itu.int/rec/R-REC-BS.1770
Used for the distinction between sample peaks, reconstructed peaks and average
loudness. MLT's windowed-sinc detector is its own implementation, not the reference
coefficient table, and no certified compliance with this recommendation is claimed.

MathWorks, limiter / Dynamic range limiter:
https://www.mathworks.com/manual/audio/ref/limiter-system-object.html
Conceptual reference for limiting, makeup gain and attack/release envelopes.
No MATLAB code or proprietary implementation was ported. MLT uses linked stereo
and an original held-minimum / finite-average envelope rather than that product.

W3C Web Audio API:
https://www.w3.org/TR/webaudio-1.0/
API reference for AudioWorklet, AudioParam scheduling, OfflineAudioContext and
worklet registration. No ScriptProcessor or per-sample main-thread processing.

The FIR coefficients are generated from sinc and Blackman window equations in
the processor constructor. Audio samples are not passed through this FIR; it is
only a peak detector. The output is delayed audio multiplied by a common gain.

Tested on Windows Edge/Chromium. Actual DAC reconstruction, lossy codecs, later
processing, device output gain and other browsers can produce different peaks.
Use conservative headroom, place the device last and check the delivered file.