Playing a 50 ms AudioFrame takes less time than expected

[microbit-carlos]

Using the latest version (at the time of writing) of the recording & playback branch: 0b06914
Hex: https://github.com/microbit-foundation/micropython-microbit-v2/actions/runs/8416237764?pr=163

Playing a 50 ms AudioFrame takes less time than expected:

>>> audioframe_50_ms = audio.AudioFrame(duration=50)
>>> audioframe_50_ms[-1] = 255
>>> ms_start = microbit.running_time(); audio.play(audioframe_50_ms); ms_end = microbit.running_time();
>>> ms_end - ms_start
42
>>> # Should be more than 50ms
>>> len(audioframe_50_ms) * 1000 / 7812
53.25141

Using time.ticks_ms() instead:

>>> from time import ticks_ms
>>> ms_start = ticks_ms(); audio.play(audioframe_50_ms); ms_end = ticks_ms()
>>> ms_end - ms_start
44

[dpgeorge]

The exact time taken by audio.play() is somewhat set by CODAL and the way it queues data for playback, via timing of the pullRequest()/pull() methods.

I suspect the following:

• The actual time taken to play the data on the speaker is correct.
• The time for the audio.play() call is always less than the above speaker output time, because the method returns once the last few frames are queued.
• The time taken for the method to return varies a bit in a non-deterministic way due to internal timing/synchronisation of CODAL's audio pipeline.

If you play increasingly large audio frames, then the time for audio.play() also increases in proportion to the length of the audio frame.

[microbit-carlos]

Based on this:

• Play 'until done' should actually wait until audio is complete lancaster-university/codal-microbit-v2#231 (comment)

The CODAL isPlaying() method should indicate real sound playback, but when calling audio.is_playing() right after it does indicate no more sound should be playing:

>>> t = running_time(); audio.play(audioframe_50_ms); audio.is_playing(); running_time() - t
False
44

Do you think this is an issue in the CODAL side?

[microbit-carlos]

I'll try replicate this on the CODAL side first as well.

[dpgeorge]

Note that MicroPython doesn't use uBit.audio.isPlaying(). Instead we track whether there's any more data to go out to CODAL. If not, then the audio is "finished".

[microbit-carlos]

Ah, thanks Damien. The reason we looked into implementing uBit.audio.isPlaying() is that we found situations were the audio playback ends up triggering the microphone, for example in programmes like this one:

from microbit import *

while True:
    audio.play(Sound.HELLO)
    # Wait until we hear something loud to do something else
    while microphone.sound_level() < 60:
        pass
    # At this point audio.play() has triggered the microphone sound level
    sleep(100)

[dpgeorge]

This should be now fixed on the audio-recording branch, waiting for audio to finish playing now uses uBit.audio.isPlaying().