Repository navigation
[BUG][BYT] alsabat frequency analysis fails due to waveform distortion #1988
Description
Describe the bug
alsabat frequency analysis fails occasionally due to waveform distortion in BYT_MB_NOCODEC platform. This is not 100% but easy to reproduce, 1-2 failures out of 10 test. NOCODEC mode use SOF firmware SSP loopback.
To Reproduce
Run alsabat test and check test result
-
simple test
alsabat -Dhw:0,0 -c 2 -r 48000 -f s16_le -
save to file and standalone analysis
alsabat -Phw:0,0 -c 2 -r 48000 -f s16_le --standalone -n 240000 &
sleep 0.5
arecord -Dhw:0,0 -c 2 -r 48000 -f s16_le rec.wav;sync
alsabat --local --file rec.wav
Reproduction Rate
1-2/10
Expected behavior
Alsabat test should return success, 0.
Impact
DEVICE_TEST will fail for PR test and daily test
Environment
- Branch name and commit hash of the 2 repositories: sof (firmware/topology) and linux (kernel driver).
There is no particular build, this issue exist more than a month. But I tested today's build. - Name of the topology file
- Topology: sof-byt-nocodec.tplg
- Name of the platform(s) on which the bug is observed.
- Platform: baytrail, BYT_MB_NOCODEC
Screenshots or console output
Attached multiple pictures
sof debugging files are attached as a zip
test_outputs.zip
Activity
[-][BUG][BYT] alsabat test fails due to waveform distortion[/-][+][BUG][BYT] alsabat frequency analysis fails due to waveform distortion[/+]tlauda commentedon Oct 22, 2019
Could you check whether #1984 improves it?
fredoh9 commentedon Oct 22, 2019
@tlauda sure, i can pick up #1984 and update the results here
fredoh9 commentedon Oct 22, 2019
@tlauda I updated only sof build. booting fine loading firmware and tplg is fine. But any aplay have hw_param failed with IPC error
1022_new_pr1984_ipc_error.log
tlauda commentedon Oct 22, 2019
Weird. It passes CI flawlessly and I've also run it today on BYT nocodec.
tlauda commentedon Oct 23, 2019
@fredoh9 Could you retry? Nobody else sees this problem with #1984.
fredoh9 commentedon Oct 23, 2019
Sure, i will plan to build your PR and re-test again. I will update the result here
fredoh9 commentedon Oct 23, 2019
I'm able to test with #1984, tlauda:topic/dma-separate-buffers-new.
Got better and less occurrence than before, but still I'm seeing alsabat frequency analysis failure.
tlauda commentedon Oct 23, 2019
@fredoh9 Could you attach recording screenshot? Does it look the same as previous ones?
fredoh9 commentedon Oct 23, 2019
@tlauda yes and no. I found a missing samples problem. The others wasn't noticeable waveform like that. I will attach wavefiles and logs.
fredoh9 commentedon Oct 23, 2019
I ran 100 loop on two BYT_MB_NOCODEC devices. failures are 6/100 and 8/100. It was much worse before.
fred-mb (6/100, attached 3 files)
fred-mb.zip
Some of them after zoom in, i could see missing samples. alsarec_140341.wav doesn't have visible defect.
jf-byt-mb3(8/100, attached 2 files)
jf_mb3.zip
one or two samples seems missing.
Hope this help @tlauda . you can reproduce this easily with the command in the description. Alsabat generate 997Hz sinewave by default.
tlauda commentedon Apr 29, 2020
@fredoh9 @wenqingfu Does this bug still exist? I haven't seen any failure in CI for a long time.
lgirdwood commentedon May 7, 2020
Ack, lets close.