Repository navigation
[BUG] distortion during playback on MTL-NOCODEC #6837
Description
Describe the bug
we observed the waveform distortion on MTL-NOCODEC, we have a similar issue SOF#6727 on mtl-stable-002, but it has been fixed on 002 branch, so let's use this bug to track the distortion issue on main branch.
To Reproduce
~/sof-test/test-case/check-alsabat.sh -p hw:sofnocodec,0 -c hw:sofnocodec,0 -C 2
Reproduction Rate
100%
Environment
Kernel Branch: topic/sof-dev
Kernel Commit:
c41b1735d3da
SOF Branch: main
SOF Commit:
e6f72dee8289
Platform: MTL-NOCODEC
Activity
lgirdwood commentedon Dec 19, 2022
Looks like we are missing a fix from mtl-002 @mwasko ?
juimonen commentedon Jan 31, 2023
This is because of missing link clock setting in mtl/ace platforms in ssp driver.
mtl 003 branch has a hard coded quick fix for "link clock", but that is not in sof main.
"non hard coded" fix is coming from:
zephyrproject-rtos/zephyr#53297
after this, there also needs to be a topology change for mtl to configure/set the link clock.
wszypelt commentedon Feb 6, 2023
Reproduction test under implementation
keqiaozhang commentedon Feb 22, 2023
This issue still exists in CI.
lgirdwood commentedon Feb 22, 2023
ok, looks like we could still be missing a MTL branch patch in main @mwasko fyi.
juimonen commentedon Mar 7, 2023
So we seem to have 2 overlapping bugs here:
fredoh9 commentedon Mar 10, 2023
I tested with mtl-004(2eec99b). alsabat test failed there also but waveforms are totally different. @abonislawski Do we miss any patch on main branch? Upper one is mtl-004, lower one is main branch(85367b7).
alsabat_wav_capture.zip
abonislawski commentedon Mar 11, 2023
@fredoh9 I assume you didnt use new topology patch from Jaska in main-branch test?
On mtl-004 it will still work with old topology but this additional commit is not intended for main branch. On main-branch new topology (Jaska PR) is required, it should work the same but not sure if this solution is already fully verified
fredoh9 commentedon Mar 11, 2023
You mean this one, #7060 ? I will take a look and check with it also.
kv2019i commentedon Mar 13, 2023
MTL only, moving to v2.6 milestone.
keqiaozhang commentedon Mar 20, 2023
I tried PR#7060 on MTL-NOCODEC, but it doesn't help.
@fredoh9 , please help to double confirm it.
kv2019i commentedon Mar 21, 2023
#7060 is expected to help with the issue. @juimonen tested this with ok results, but the resulting audio file still has gaps related to smart-amp pipeline (potentially the remaining issue seen is #7285). Test with #7060 and without smart-amp, has not been done.
plbossart commentedon Mar 21, 2023
We really need the ability to take the SmartAmp out of the nocodec topology.
Adding more features needs to come with the ability to come back to a less complex part.
juimonen commentedon Mar 21, 2023
yeah, tested today with #7060 in "jf mtl nocodec 3" device - and ssp0 (smart amp pipe) has these very rare gaps in the right channel (loopback feedback) and ssp1 is clean. so this needs to be divided into 2 issues 1. link clock setting, that #7060 fixes 2. smart amp gaps
keqiaozhang commentedon Mar 23, 2023
Confirmed with #7060.
Waveform is distorted at the beginning of both SSP0 channels.
There're also a lot of glitches in right channel of SSP0.
I will file a new bug after #7060 merged.
These 2 problems are not reproduced on SSP2.
lrudyX commentedon Mar 24, 2023
Close as original issue has been resolved. @keqiaozhang please create new ticket and provide IPC logs with payloads for FW reproduction.
keqiaozhang commentedon Apr 3, 2023
#7060 is still under review, reopening this bug.
keqiaozhang commentedon Apr 3, 2023
Verified this issue again with PR #7060, the result is same as before.
At beginning:
Some glitches in the right channel:
ranj063 commentedon Apr 4, 2023
These are the replacements for #7060