Repository navigation
[BUG][CML] Pulse Noise on speaker pipeline in pause/release test #2203
Description
Describe the bug
run aplay via speaker pipeline, and press Space to pause/release playback, you can here pulse noise(sounds like "pa"). check dmesg and sof-logger, no clear errors found.
To Reproduce
- run
aplay -Dhw:0,5 -f dat -i -vvv -c 2 ./longwave.wav - Press SPACE to pause/release playback
MAKE SURE WAVE FILE CAN OUTPUT SOUND, DO NOT PLAY /dev/zero
if not reproduced with above procedures, please try the script attached. Because the sound of press space may cover the pulse noise.
Reproduced Rate
should be very easy to reproduce on the platform described below
Environment
- Branch name and commit hash of the 2 repositories: sof (firmware/topology) and linux (kernel driver).
- Name of the topology file
- Topology: {sof-cml-rt1011-rt5682.tplg}
- Name of the platform(s) on which the bug is observed.
- Platform: {CML with rt5682 and rt1011 codecs}
Error Logs
No clear error seen in logs
sof-test-pause-release.zip
spk_pulse_noise_pause_release_soflogger.log
spk_pulse_noise_pause_release_dmesg.log
Activity
lgirdwood commentedon Dec 16, 2019
@aiChaoSONG can you paste in audacity screenshots showing the pop (palying sine wave). Please also attach small wavs of pop during sine wave.
aiChaoSONG commentedon Dec 18, 2019
@lrgirdwo @mmaka1 I run some test with nocodec tplg, and the pause/release interval is 0.2s, and there is surely some pulse here, please see the picture and attached wave. seems we lost some samples. but if we change pause/release interval to 0.5s, things are much better, see attached wave.
headset_interval_0.2s.zip
headset_interval_0.5s.zip
lgirdwood commentedon Dec 18, 2019
@aiChaoSONG potentailly looks like stale data in the DAI FIFO (that's replayed at unpuase). What interface HDA, SSP are you playing from ?
@tlauda it could be we are pausing the DAI prior to FIFO being empty ?
aiChaoSONG commentedon Dec 19, 2019
@lgirdwood , it's SSP
lgirdwood commentedon Dec 19, 2019
@tlauda SSP trailing bytes ?
tlauda commentedon Mar 13, 2020
@ktrzcinx @aiChaoSONG I see some fixes have been merged. Do you still see the issue?
aiChaoSONG commentedon Mar 13, 2020
@tlauda I can not hear the noise, but we still have very tiny amount of data lost, if this doesn't matter, we can close this now.
tlauda commentedon Mar 13, 2020
@aiChaoSONG This is expected on SSP. Transfer can end in the middle of the link list item and next one starts from the next item, so some small amount of data is lost.
aiChaoSONG commentedon Mar 13, 2020
@tlauda Fine, let's close this.
close for fixed by #2364