1514526 - DynamicsCompressorNode doesn't seem to have a tail time

Status: NEW

(Core :: Web Audio, defect, P3)

Description

[chrisguttandin]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_13_6) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/71.0.3578.98 Safari/537.36

Steps to reproduce:

I scheduled a very small (only a few samples) AudioBuffer with an AudioBufferSourceNode. I connected that ABSN to a DynamicsCompressorNode. I connected that DCN to the destination. I noticed that no signal ever reached the destination. That changed when I modified the AudioBuffer to have at least 257 samples. Given that observation I guess that a DynamicsCompressorNode is outputting silence once the input sends silence even though their is still some audio in the buffer.


Actual results:

The following code snippet does not log any values when executed in Firefox:

const oac = new OfflineAudioContext({ length: 500, sampleRate: 44100 });
const audioBuffer = new AudioBuffer({ length: 3, sampleRate: 44100 })
const audioBufferSourceNode = oac.createBufferSource();
const dynamicsCompressorNode = oac.createDynamicsCompressor();

audioBuffer.copyToChannel(new Float32Array([ 1, 1, 1 ]), 0)

audioBufferSourceNode.buffer = audioBuffer;

audioBufferSourceNode.start(0);

audioBufferSourceNode.connect(dynamicsCompressorNode).connect(oac.destination);

oac.startRendering()
    .then((renderedBuffer) => {
        const channelData = renderedBuffer.getChannelData(0);

        for (let i = 0; i < channelData.length; i += 1) {
            if (channelData[i] !== 0) {
                console.log(i, channelData[i]);
            }
        }
    });


Expected results:

It log three values in Chrome:

264 0.3056469261646271
265 0.30571678280830383
266 0.3057866394519806

Comment 1

[Karl Tomlinson (:karlt)]

https://searchfox.org/mozilla-central/rev/1a4a7745fd4a14402a0de0e9abc040724d4fcf8f/dom/media/webaudio/DynamicsCompressorNode.cpp#76

Comment 2

[Paul Adenot (:padenot)]

Additionally, we need a bit of code that feeds the underlying compressor code the right amount of channels, otherwise an assert blows up, and in release, silence is output.

Comment 3

[Paul Adenot (:padenot)]

I remember finding the solution for this, it was something simple.