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][CML] pause/release headset capture will cause pulse noise in recorded wave #2224
Describe the bug
pause/release headset capture with SPACE will cause pulse noise in the recorded wave.
To Reproduce
run arecord via headset pipeline: arecord -Dhw:0,0 -f dat -r 48000 -i -vvv -c2 ./headset.wav
press SPACE in pair(double press)(see picture above)
after several pairs of press, the noise should be triggered.
if you don't have correct UCM, the headset capture may be muted, please use the ucm file attached.
Reproduction Rate
within several pairs of SPACE press, you will see the pulse noise in the terminal. should be easy to reproduce on the platform specified below.
Environment
Branch name and commit hash of the 2 repositories: sof (firmware/topology) and linux (kernel driver).
once the pulse is seen, it will exist along the whole recording.
once the pulse is seen, unplug headset, it disappears, insert jack back, it appears again.
once the pulse is seen, continue double press SPACE, it may disappear and appear again.
Some of kcontrols which may change data route have influence on the pulse noise and the recording, such as ADCDAT Mux and Stereo1 ADC L Mux.
once the pulse is seen, change Stereo1 ADC L Mux from ADC1 L to ADC1 R, and change it back, the pulse disappears, but press SPACE will trigger the noise again.
sometimes, no need to press SPACE, the pulse noise appears at the beginning of recording
tried nocodec kernel and tplg, this issue not reproduced.
@bardliao any insights here, it look like a codec driver (pause/release) issue if the noise disappears when 2, 4, 5 and 7 is tried ? Should be add realtek to issue ?
@bardliao any insights here, it look like a codec driver (pause/release) issue if the noise disappears when 2, 4, 5 and 7 is tried ? Should be add realtek to issue ?
@lgirdwood Yes, it could be. @aiChaoSONG Could you check if MCLK is active when the recording is paused? If not, please check if MCLK starts before I2S BCLK. IIRC, MCLK starts before BCLK is a requirement of rt5682. @shumingfan can you take a look?
We had done some experiments like loopback the signal inside/outside of codec when pulse noise happened. We heard the sound from the headset mic to the headphone. The sound is OK.
Another test is that AP(Audio Precision) outputs the signal to SoC instead of rt5682i codec. This issue still happened.
This issue we are discussing on Google issue tracker and please check the website below. https://partnerissuetracker.corp.google.com/issues/145640941
@aiChaoSONG As I know, Realtek suspects that is a layout or HW design issue. They provided a solution to ODM and are waiting for new HW version to verify. @shumingfan Is there any update on the status of new HW version?
@bardliao
I think there is no new progress for now. Realtek HW AE left a comment on Google issue tracker last week.
I post his commit here.
[From Realtek Albert Chen]
The issue phenomenon is completely different between ALC1011 and ALC5682.
ALC1011 request bclk output earlier than data is to prevent some sounds disappearing while start playback.
And currently issue is random noise happened during recording in CML and measurement I2S signal waveform on PCB, it’s a signal reflection on the clock can’t solved by add delay between CLK start and data.
FYI, Only for reference following my thinking.
There are two methods maybe can process signal reflection phenomenon . One is to reduce I2S driving strength and the other is to modify the I2S traces on PCB layout.
If platform can’t adjust I2S driving strength, only waiting for ODMs to modify the PCB layout for verify this issue.
Activity
aiChaoSONG commentedon Dec 18, 2019
I also observed:
ADCDAT MuxandStereo1 ADC L Mux.Stereo1 ADC L MuxfromADC1 LtoADC1 R, and change it back, the pulse disappears, but press SPACE will trigger the noise again.lgirdwood commentedon Dec 18, 2019
@bardliao any insights here, it look like a codec driver (pause/release) issue if the noise disappears when 2, 4, 5 and 7 is tried ? Should be add realtek to issue ?
bardliao commentedon Dec 19, 2019
@lgirdwood Yes, it could be. @aiChaoSONG Could you check if MCLK is active when the recording is paused? If not, please check if MCLK starts before I2S BCLK. IIRC, MCLK starts before BCLK is a requirement of rt5682. @shumingfan can you take a look?
shumingfan commentedon Dec 19, 2019
@aiChaoSONG @bardliao I want to involve our hardware AE to debug this issue.
He wants to know which is customer and project name to this issue?
shumingfan commentedon Dec 25, 2019
We had done some experiments like loopback the signal inside/outside of codec when pulse noise happened. We heard the sound from the headset mic to the headphone. The sound is OK.
Another test is that AP(Audio Precision) outputs the signal to SoC instead of rt5682i codec. This issue still happened.
This issue we are discussing on Google issue tracker and please check the website below.
https://partnerissuetracker.corp.google.com/issues/145640941
aiChaoSONG commentedon Jan 13, 2020
this issue is not a sof issue nor kernel issue, close for now
aiChaoSONG commentedon Feb 25, 2020
Codecs seems fine, reopen
slawblauciak commentedon Feb 25, 2020
Is this reproducible on the high DSP clock?
bardliao commentedon Feb 26, 2020
@aiChaoSONG As I know, Realtek suspects that is a layout or HW design issue. They provided a solution to ODM and are waiting for new HW version to verify. @shumingfan Is there any update on the status of new HW version?
shumingfan commentedon Feb 26, 2020
@bardliao
I think there is no new progress for now. Realtek HW AE left a comment on Google issue tracker last week.
I post his commit here.
[From Realtek Albert Chen]
The issue phenomenon is completely different between ALC1011 and ALC5682.
ALC1011 request bclk output earlier than data is to prevent some sounds disappearing while start playback.
And currently issue is random noise happened during recording in CML and measurement I2S signal waveform on PCB, it’s a signal reflection on the clock can’t solved by add delay between CLK start and data.
FYI, Only for reference following my thinking.
There are two methods maybe can process signal reflection phenomenon . One is to reduce I2S driving strength and the other is to modify the I2S traces on PCB layout.
If platform can’t adjust I2S driving strength, only waiting for ODMs to modify the PCB layout for verify this issue.
bardliao commentedon Feb 26, 2020
Let's close it for now since Realtek is working with Google. We can reopen it if Realtek needs our help.
keyonjie commentedon Mar 11, 2020
Hi @aiChaoSONG can you try the method provided by Albert to check if the noise is from ADC via the 5682 A-D D-A loopback feature?
Please following sequence setup.
aiChaoSONG commentedon Mar 11, 2020
@keyonjie
so noise may not be produced by codec
same result on https://github.com/thesofproject/linux/issues/1604
keyonjie commentedon Mar 11, 2020
Thanks @aiChaoSONG
So next step let's try if we can reproduce similar issue with nocodec SSP LBM feature.
Using any other verified inputs(e.g. input from AP, from rt5651, DMIC injector, ...) to reproduce it can help for analyzing also.
aiChaoSONG commentedon Mar 12, 2020
@keyonjie I tried nocodec kernel and tplg:
none of the above test reproduced the issue. Very wired.
lancedkhou commentedon Mar 12, 2020
@aiChaoSONG
The noise symptom might be low reproduction rate in a certain sku of SoC.
Normally I will use test scripts to reproduce.
Let me mail you test scripts to see whether it helps.