Are AudioBufferSourceNode.loopStart and .loopEnd subsample-accurate?

[joeberkovitz]

The spec should say whether the loop points specified for an AudioBuffer are subsample-accurate, or whether they are rounded to the nearest sample.

Most commercial sampler applications do not support subsample-accurate loop points and require loop points to be at exact sample frames, for efficiency's sake.

[cwilso]

"No." :)

[mdjp]

TPAC RESOLUTION: No.

[rtoy]

Curiously, the spec currently says loopStart and loopEnd are rounded to the nearest frame. I didn't look to see when this change was made.

However, I think Chrome does do subsample accurate loopStart and loopEnd. Here's an example

var context = new OfflineAudioContext(1, 100, 12800);
var src = context.createBufferSource();
var b = context.createBuffer(1, 3, context.sampleRate);
var d = b.getChannelData(0);
d[0] = 0;
d[1] = 1;
d[2] = 1;
src.buffer = b;
src.loop = true;
src.loopStart = 0.5 / context.sampleRate;
src.loopEnd = 2 / context.sampleRate;

src.connect(context.destination);
src.start();
context.startRendering().then (function (buffer) {
    console.log(buffer.getChannelData(0));
});

If there were no subsampling going on, the output should always be either 0 or 1. In Chrome, some samples are 0.5, indicating that subsampling is being done.

Presumably this is also true with Safari (webkit), Edge, and Opera too. I didn't test this though.

[padenot]

Gecko snaps to frames.

[joeberkovitz]

This is a duplicate of #332, which itself was closed via #400. So, closing.

[hoch]

Note: the resolution is reverted by #915.