[BUG]: [Multiple Capture - No-codec mode]: Observed dragging and metallic noise from data captured from cascaded (second level) copier

With the PR #6783 verified the cascaded mode with no-code mode and the data captured from the PCM's of copier.host.18.1 and copier.host.20.1 has dragging and mettalic noise.

@sathya-nujella

@sathya-nujella @RanderWang @RDharageswari

@lyakh

@lyakh

@jairaj-arava

@jairaj-arava Can you please share the FW version in your test? And are you using Chrome kernel?
We don't observe this issue with sof-dev and SOF main in daily testing on MTL. So hope to have a cross check with you.

@mwasko

@mwasko fyi.

@jairaj-arava

@jairaj-arava Can you please share the FW version in your test? And are you using Chrome kernel? We don't observe this issue with sof-dev and SOF main in daily testing on MTL. So hope to have a cross check with you.

@mengdonglin I tried with both origin/mtl-003-drop-stable (top commit smart_amp_test: fix data size check)
and sof/main (top commit smart_amp: eliminat pops when switch device between speaker and headphone).

Regarding kernel, we see reboot issue with sof-dev kernel on our boards, hence @bardliao has posted changes on top of our internal kernel 6.1 https://github.com/bardliao/os.chrome.internal.kernel-bleeding-edge/commits/mtl_6_1_upstream along with two patches from https://patchwork.kernel.org/project/alsa-devel/patch/20230209142123.17193-3-peter.ujfalusi@linux.intel.com/ that supports the mutliple catpure feature.

But still we see the issue.

@lyakh

@lyakh fyi

I can reproduce it on MTL RVP

@jairaj-arava

@jairaj-arava can you try this one #7145 ? It fixes this bug

@jairaj-arava

@jairaj-arava can you try this one #7145 ? It fixes this bug

@RanderWang , I tried the #7145 fix but still I see the same issue with both no-codec and codec mode.

@jairaj-arava

@jairaj-arava is this happening only with the DMIC input?
Wondering if the cascaded copiers on SSP capture also show the same problem (should be easy to test in nocodec mode with SSP lbm bit set).

@jairaj-arava

@jairaj-arava @btian1 ok - that looks like channel swapping in the copier, there are some copier PRs in flow for speed up and simplification, can we retest after merging these PRs as the issue may be easier to resolve (with the simplifications merged). @mwasko fyi.

@lgirdwood sure, pls help mention them here if possible to keep track of them in case if they are few.

@jairaj-arava

@jairaj-arava , sorry, we don't have a mtlp nocodec with sound card and dmic injector device, HW rework is also not easy due to there is an conflicts between mtlp DMIC and ssp, on tglu, the reproduce rate is not 100%, only sometimes on first boot.
if possible, please share your remote and then I can have a check on your device.

@lgirdwood , @mwasko I don't have rootcause for this issue yet, if you have draft PRs, could you paste link here?

@jairaj-arava

@jairaj-arava , sorry, we don't have a mtlp nocodec with sound card and dmic injector device, HW rework is also not easy due to there is an conflicts between mtlp DMIC and ssp, on tglu, the reproduce rate is not 100%, only sometimes on first boot. if possible, please share your remote and then I can have a check on your device.

@lgirdwood , @mwasko I don't have rootcause for this issue yet, if you have draft PRs, could you paste link here?

@btian1 The draft PRs I mean are your and @ranj063 copier PRs that simplify/optimize.
From the channel swapping screenshot it does look like we have a miconfigured format and in/out mappings between our copiers as the channel swap is constant size and repeating synchronously.

@jairaj-arava

@jairaj-arava , sorry, we don't have a mtlp nocodec with sound card and dmic injector device, HW rework is also not easy due to there is an conflicts between mtlp DMIC and ssp, on tglu, the reproduce rate is not 100%, only sometimes on first boot. if possible, please share your remote and then I can have a check on your device.

@lgirdwood , @mwasko I don't have rootcause for this issue yet, if you have draft PRs, could you paste link here?

Sure, I will make the full setup ready for you today to access remotely.

@jairaj-arava

I tried the setup shared by Jairaj, however, there is no input, only background noise, @jairaj-arava , possible to use an dmic injector to a sound card, then the dmic input will be the playback sinetone, it will be much better for judgement.

@lgirdwood , issue reported in Feb, nothing merged at that time, should not related with the optimization patches.

@jairaj-arava

I tried the setup shared by Jairaj, however, there is no input, only background noise, @jairaj-arava , possible to use an dmic injector to a sound card, then the dmic input will be the playback sinetone, it will be much better for judgement.

@lgirdwood , issue reported in Feb, nothing merged at that time, should not related with the optimization patches.

Yesterday tried with @btian1 on the call, could repro it on the setup I shared.

@jairaj-arava

@jairaj-arava ， today, both Bard and I tried on our setup, issue can't be reproduced with daily build(mine is Aug 2th).

I would prefer below steps for chrome reported issue(based on different bios/core boot, and kernel driver):

CCG chrome validation team found audio issue on chrome book.
CCG validation engineer book an SSE linux CI machine to reproduce.
if reproducible on CI machine, transfer issue to SOF dev team.
if not reproducible, then it maybe caused by board/core boot/chrome audio kernel driver, need CCG related team to fix.

Regarding this issue, our MTLP need disable SSP1 in bios before recording, I will sent you offline.

@bardliao

Hi @bardliao and @btian1,
Today I tried on other board and could easily repro the issue on this board too. I am using the Chrome kernel (different from SOF-DEV) and also we didn't disable any SSP1 from the coreboot (I don't see any reason to disable it). Let us sync up again on this today.

@btian1

this issue should only be reproducible on nocodec with dmic board.

today, I tried with tgl nocodec and cmd shared from jairaj:

play a sine tone:
speaker-test -c 2 -t sine -Dhw:1,0

capture 10, 27, 28 simultaneously in another window.
arecord -d 60 -c 4 -f S32_LE -r 48000 -Dhw:0,27 32bit_27.wav & arecord -d 40 -c 4 -f S32_LE -r 48000 -Dhw:0,28 32bit_28.wav & arecord -d 20 -c 4 -f S32_LE -r 48000 -Dhw:0,28 32bit_10.wav

check the captured file

TGL board is fine with latest config, due to our current mtlp board are all configed with HDA, change to nocodec may cause CI issue.
This means, at least, there is no issue in copier for module copy part, since both of them share same code, next suspect area is tplg, will compare the tplg between Jairaj and mine later.
Next Monday， I will move one mtlp board to nocodec mode and have a check also.

@btian1 did you get a chance check on this? Btw you see the issue only when you start the recording first with DMIC 10 PCM (keep it running) and immediately with 27/28 PCM. If you test it in the reverse i.e. recording with 27/28 PCM first and later with PCM 10, you will not see the issue.

@bardliao
As discussed, pls start with PCM10 first and then PCM 27/28 then you will see the issue. If you start 27/28 and then PCM10, you will not see the issue.

@jairaj-arava

@jairaj-arava , issue is rootcaused and double confirmed by you and me, and it was fixed by: 6bcaf89

However, this PR#7890 is used to convert copier module interface, merged in July 12th, #7890 depend on #7740.

There are two solutions:

close this issue and mark it as mtl_006 will fix this issue.
if you want to port it, please port Copier: preparation work for convert copier to module interface #7740 and Convert copier to use the module interface #7890 together to mtl_005 branch.

Thanks Bard for the help on this.

@btian1

Hi @btian1,
Seems porting to MTL-005 there is other patches dependency apart from the #7740. We are aiming for MTL-006 and will discuss with the fw team on this.

Since, the PR #7890 fixes the issue, let's close this and will confirm it once we have the MTL-006 release.

yes, before #7740, there are copier optimization patches, maybe > 50, so it is big task to port it from main to mtl-005 branch, we can just wait for mtl-006 release and test again, if still have issues in mtl-006, we will check again.

Navigation Menu

Repository navigation

[BUG]: [Multiple Capture - No-codec mode]: Observed dragging and metallic noise from data captured from cascaded (second level) copier #7074

38 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[BUG]: [Multiple Capture - No-codec mode]: Observed dragging and metallic noise from data captured from cascaded (second level) copier #7074

Description

Activity

jairaj-arava commented on Feb 12, 2023

bardliao commented on Feb 13, 2023

mengdonglin commented on Feb 14, 2023

lgirdwood commented on Feb 14, 2023

jairaj-arava commented on Feb 15, 2023

lgirdwood commented on Feb 15, 2023

RanderWang commented on Feb 22, 2023

RanderWang commented on Feb 22, 2023

jairaj-arava commented on Feb 23, 2023

plbossart commented on Feb 23, 2023

38 remaining items

jairaj-arava commented on Jul 31, 2023

btian1 commented on Aug 1, 2023

lgirdwood commented on Aug 1, 2023

jairaj-arava commented on Aug 1, 2023

btian1 commented on Aug 2, 2023

jairaj-arava commented on Aug 3, 2023

btian1 commented on Aug 8, 2023

jairaj-arava commented on Aug 8, 2023

jairaj-arava commented on Aug 9, 2023

btian1 commented on Aug 11, 2023

jairaj-arava commented on Aug 14, 2023

btian1 commented on Aug 15, 2023

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions