Tone generation length inaccurate on CPX, and varies between 5.3.1 and 6.0.0-rc.0

[mytechnotalent]

The new lightsleep feature might be causing issues with delays of a variety of types. Today was helping to debug a user's issue when he presented the following.

USER at 3:59 AM
I've noticed that using play_tone() with the Circuit Playground Express (see example below), CircuitPython 6.0.0_rc.0 adds a significant pause after each note compared to 5.3.1. Has anyone else experienced this?

from adafruit_circuitplayground.express import cpx
scale = [587, 659, 740, 880, 987, 1174]
while True:
  for note in scale:
    cpx.play_tone(note, 0.25)

It is possible time.sleep() which shuts down some things might be taking longer to come back and causing an issue with user experience.

[robertgallup]

The following code has similar results even not using the internal timing for notes:

import time
from adafruit_circuitplayground.express import cpx
scale = [587, 659, 740, 880, 987, 1174]

while True:
    for note in scale:
        cpx.start_tone(note)
        start_time = time.monotonic()
        while time.monotonic() - start_time < .2:
            pass
        cpx.stop_tone()

[dhalbert]

This seems to have to do with audio playing. The basic sleep delays are still accurate in 5.3.1 or 6.0.0-rc.2. But the audio play intervals on the SAMD21 are wildly off, and need further investigation.

Test program:

from adafruit_circuitplayground import cp
import time

for test in ("sleep", "tone"):
    for length in (0.05, 0.1, 0.25, 0.5, 1):
        start = time.monotonic()
        if test == "sleep":
            time.sleep(length)
        else:
            cp.play_tone(440, length)
        print(test, length, time.monotonic() - start)
    print()

Circuit Playground Express 5.3.1 output. Note the too-long tone lengths. The sleep lengths are fine. The output is pretty consistent over multiple runs.

sleep 0.05 0.0500488
sleep 0.1 0.100098
sleep 0.25 0.25
sleep 0.5 0.5
sleep 1 1.0

tone 0.05 0.198975
tone 0.1 0.261963
tone 0.25 0.437012
tone 0.5 0.637939
tone 1 1.13599

Circuit Playground Express 6.0.0-rc.0. Short tones are even longer.

sleep 0.05 0.0500031
sleep 0.1 0.0999985
sleep 0.25 0.25
sleep 0.5 0.500999
sleep 1 1.0

tone 0.05 0.362999
tone 0.1 0.444
tone 0.25 0.639999
tone 0.5 0.780998
tone 1 1.288

Circuit Playground Bluefruit 5.3.1. Tone lengths are much more accurate. (Note that audio is generated in a completely different way on this board.):

sleep 0.05 0.0499954
sleep 0.1 0.0999985
sleep 0.25 0.25
sleep 0.5 0.5
sleep 1 1.0

tone 0.05 0.0600052
tone 0.1 0.110001
tone 0.25 0.259995
tone 0.5 0.510002
tone 1 1.01

Circuit Playground Bluefruit 6.0.0-rc.0. Still fine:

sleep 0.05 0.0500031
sleep 0.1 0.100006
sleep 0.25 0.25
sleep 0.5 0.5
sleep 1 1.0

tone 0.05 0.0599976
tone 0.1 0.112
tone 0.25 0.261002
tone 0.5 0.512009
tone 1 1.01199

[robertgallup]

Great to have this confirmed. FWIW, it may not be obvious from the numbers, but this issue has significant audio impact on sound effects using play_tone(), start_tone() and stop_tone() on the CPX.