Play computed audio

CircuitPython’s audio outputs play samples, and a sample is filled from C, so CircuitPython has no built-in way to play a stream your Python code or a graph of nodes renders as it goes. A Ring is a sample you write into. The output pulls it in C, block by block, while your loop keeps it topped up:

import audiobusio, board
from array import array
from audiopump import Ring

ring = Ring(sample_rate=22050, channel_count=1, frames=256, capacity=16)
block = array("h", bytes(512))           # 256 mono frames
while ring.space() >= len(block) * 2:    # fill it before play()
    ring.write(render(block))
out = audiobusio.I2SOut(board.I2S_BCLK, board.I2S_WCLK, board.I2S_DOUT)
out.play(ring)
while True:
    while ring.space() >= len(block) * 2:
        ring.write(render(block))

It plays through audiobusio.I2SOut, audioio.AudioOut, audiopwmio.PWMAudioOut and an audiomixer voice like any other sample. When your code falls behind, the output plays a block of silence rather than stopping, and ring.stats()[3] counts those underruns; a write that doesn’t fit is short and returns the bytes it took. capacity is how many blocks the ring holds, which is also its latency.

Size the ring for what your output takes at once, and fill it before play(). An output that takes a large first gulp underruns at the start however fast you write, so leave room beyond one block. Blocks of 256 frames with capacity=16 hold 186 ms at 22,050 Hz.

It is the same audiopump.Ring MicroPython’s pump pulls, with the same methods (write, space, level, stats, clear, deinit), so one producer runs on both. On CircuitPython the module carries only Ring: the board’s output is the pump.