You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[BUG][TGL][DSM] one additional sample will be captured after playback ends #2880
Describe the bug
This issue happens when playback ends first. From the captured wav, there's one more cycle of single is captured. After more tests, we found that this issue only happens with s32_le, no reproductions when testing with S16_LE.
changed the title [-][BUG][TGL][DSM] one additional cycle of single will be captured after playback ends[/-][+][BUG][TGL][DSM] one additional sample will be captured after playback ends[/+]on Apr 27, 2020
On TGL-Delbin-I2S-02 , using https://sof-ci-storage.sh.intel.com/release/sof/releases/tgl/v11.0-rc3/c6249fff6493_chrome_release/ ,issue can be reproduced in below steps:
1.enable DSM loopback
Run:
amp_enable.sh.txt
2.run playback(only for 3 seconds ) and capture at same time.
aplay -Dhw:0,0 -f s32_le -c 2 -r 48000 sine_48k_32bit_10s_stereo_l5_r2.wav
-d 3 & arecord -Dhw:0,0 -f s32_le -c 2 -r 48000 capture.wav -d 5
3.check the wav file, form the snapshot you can see, ~16ms after playback, there's one more cycle of single is captured.
Activity
[-][BUG][TGL][DSM] one additional cycle of single will be captured after playback ends[/-][+][BUG][TGL][DSM] one additional sample will be captured after playback ends[/+]keqiaozhang commentedon Apr 27, 2020
Duplicated as #2701
YvonneYang2 commentedon Jun 9, 2020
Issue can still be reproduced with test recipe as below with command:
Kernel:https://github.com/thesofproject/linux/commits/topic/sof-dev commit:5cb79ee6f
Revert 6a7c533
Revert cd0a323+ thesofproject/linux#2062 + hack commit thesofproject/linux@6769df9
FW: https://github.com/thesofproject/sof/tree/tgl-006-drop-stable + #3006
lgirdwood commentedon Jun 10, 2020
Looks like 4ms of data coming 16ms after end of playback. Does this happens in SSP loopback mode ? i.e. is it a codec issue or a FW issue.
sinahuang commentedon Jun 24, 2020
Error also can be reproduced with https://github.com/thesofproject/sof/tree/tgl-007-drop-stable on chrome book with codec 98373 in I2S mode.
lgirdwood commentedon Jun 24, 2020
@tlauda is it possible that DMA continues to drain pipeline (or parts thereof) ?
Liviali155 commentedon Jul 8, 2020
On TGL chromebook in SDW mode with codec 98373 FW:https://github.com/thesofproject/sof/commits/releases/tgl/v7.0-rc7 ,commit: 33cfa90
aplay -Dhw:0,2 -f s16_le -c 2 -r 48000 sine_48k_16bit_10s_stereo_l5_r2.wav -d 3 &sleep 0.5; arecord -Dhw:0,3 -f s16_le -c 4 -r 48000 record.wav -d 3
The wave of capture file as below:

sof-logger0708.log
dmesg0708.log
record.zip
Liviali155 commentedon Jul 9, 2020
Error still can be reproduced with https://github.com/thesofproject/sof/tree/releases/tgl/v7.0-rc8 ,commit: 2057409 on TGL chrome laptop I2S with codec 98357,the same phenomenon with FW:https://github.com/thesofproject/sof/commits/releases/tgl/v7.0-rc7 ,commit: 33cfa90
ClarexZhou commentedon Jul 27, 2020
Error still seen on https://github.com/thesofproject/sof/tree/releases/tgl/v7.0-rc10
yongzhi1 commentedon Oct 19, 2020
Hi, @Liviali155 , can you verify the same with https://github.com/thesofproject/sof/releases/tag/releases%2Ftgl%2Fv10.0-rc1 and cherry-pick 20fbbe77e47421dae? In my test, the last channel is always filled with 0, I am using SDW based sof-tgl-sdw-max98373-rt5682.mp4. Thanks!!
lgirdwood commentedon Oct 20, 2020
@slawblauciak @bkokoszx request for TGL release v10
WYL-12 commentedon Dec 1, 2020
On TGL-Delbin-I2S-02 , using https://sof-ci-storage.sh.intel.com/release/sof/releases/tgl/v11.0-rc3/c6249fff6493_chrome_release/ ,issue can be reproduced in below steps:

1.enable DSM loopback
Run:
amp_enable.sh.txt
2.run playback(only for 3 seconds ) and capture at same time.
aplay -Dhw:0,0 -f s32_le -c 2 -r 48000 sine_48k_32bit_10s_stereo_l5_r2.wav
-d 3 & arecord -Dhw:0,0 -f s32_le -c 2 -r 48000 capture.wav -d 5
3.check the wav file, form the snapshot you can see, ~16ms after playback, there's one more cycle of single is captured.