Bug 218609 - snd-hda-intel sofhdadsp inaccurate speaker sound missing 0.8sec WAKEUP TIME FROM POWERSAVE STATE on Tiger Lake notebook MSI GP76 11UG 11800h
Summary: snd-hda-intel sofhdadsp inaccurate speaker sound missing 0.8sec WAKEUP TIME ...
Status: RESOLVED DUPLICATE of bug 218709
Alias: None
Product: Drivers
Classification: Unclassified
Component: Sound(ALSA) (show other bugs)
Hardware: Intel Linux
: P3 high
Assignee: Jaroslav Kysela
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2024-03-17 06:11 UTC by FRANK S
Modified: 2024-04-11 05:38 UTC (History)
0 users

See Also:
Kernel Version: 6.8.1 upstream
Subsystem:
Regression: No
Bisected commit-id:


Attachments
lspci > lspci.txt (1.81 KB, text/plain)
2024-03-18 03:41 UTC, FRANK S
Details
Screenshot of Manajaros installed ALSA/sound related packages (116.41 KB, image/png)
2024-03-18 09:11 UTC, FRANK S
Details
short video of system settings audio test (3.58 MB, video/mp4)
2024-03-22 05:44 UTC, FRANK S
Details
output of alsa-info.sh (64.98 KB, text/plain)
2024-03-24 10:58 UTC, FRANK S
Details
sudo dmesg > dmesg.txt (87.88 KB, text/plain)
2024-03-25 10:32 UTC, FRANK S
Details
oem28.inf file from w10 driver (82.23 KB, text/plain)
2024-03-27 05:31 UTC, FRANK S
Details
oem29.inf file from w10 driver (16.59 KB, text/plain)
2024-03-27 05:32 UTC, FRANK S
Details
oem30.inf file from w10 driver (54.72 KB, text/plain)
2024-03-27 05:32 UTC, FRANK S
Details
record of the audio test with and without background audio workaround (901.42 KB, audio/mp4)
2024-03-28 11:05 UTC, FRANK S
Details

Description FRANK S 2024-03-17 06:11:31 UTC
Sound fragments do not get played correctly and are swallowed somewhere. Could this be a DMA problem ?
Affected is speaker playback for example the sound test "front left, front right" already.
I cannot test headphones, as I have USB one. USB headphone has no problem.
Manjaro, but also Ubuntu show the same problem. Thus I compiled upstream 6.8.1 kernel to test and report.
On recently tested W10 the problems do not occur. Also on Android podcasts and youtube media are error free which cause problems on the notebook.
Firefox and Brave both show the problem.
Comment 1 FRANK S 2024-03-17 06:14:14 UTC
Thank you for looking into this.
Comment 3 FRANK S 2024-03-18 03:41:18 UTC
Created attachment 306001 [details]
lspci > lspci.txt
Comment 4 FRANK S 2024-03-18 09:11:53 UTC
Created attachment 306006 [details]
Screenshot of Manajaros installed ALSA/sound related packages
Comment 5 FRANK S 2024-03-22 05:44:25 UTC
Created attachment 306023 [details]
short video of system settings audio test
Comment 6 FRANK S 2024-03-24 10:58:42 UTC
Created attachment 306031 [details]
output of alsa-info.sh

output of alsa-info.sh
Comment 7 FRANK S 2024-03-25 10:32:54 UTC
Created attachment 306037 [details]
sudo dmesg > dmesg.txt

sudo dmesg > dmesg.txt
Comment 8 FRANK S 2024-03-27 05:31:36 UTC
Created attachment 306046 [details]
oem28.inf file from w10 driver
Comment 9 FRANK S 2024-03-27 05:32:04 UTC
Created attachment 306047 [details]
oem29.inf file from w10 driver

oem29.inf file from w10 driver
Comment 10 FRANK S 2024-03-27 05:32:34 UTC
Created attachment 306048 [details]
oem30.inf file from w10 driver

oem30.inf file from w10 driver
Comment 11 FRANK S 2024-03-27 17:52:30 UTC
KDE System Settings Audio Test doesn't work correctly.
ALSA Sound tests also produce same kind of flaw. incomplete played sound fragments.
Some Videos on youtube also show the flaw, audio streams with interruptions.

On the other hand not all media playback is affected of it.

I tried position_fix 1,2 and 5 not effective.

As I see that too many Bugs get opened in much worse state of audio problems, I consider my case not worth mentioning.

so I'll close it.
Comment 12 FRANK S 2024-03-27 23:25:58 UTC
I reopen this report, because I seem to have understood what might be going on:

The sound takes approximately 0.8 seconds to wake up from a power save state. This can be prevented by keeping audio busy playing some long sound piece "in the background".

Please take a look at report 11839. It describes what I do see now - 16 years later. 11839 has been closed a unreproducible after hanging some years.

I tried the following to solve the problem myself:

1. options snd_hda_intel power_save=0 or 1000   >reboot,visible in /sys/module/snd_hda_intel/parameters, but no effect

2. options snd_hda_intel power_save_controller=N   >reboot, visible in /sys/module/snd_hda_intel/parameters, but no effect

I raised the importance to high to get some attention. If you could give me a hint what I'm doing wrong ?
Comment 13 FRANK S 2024-03-28 10:03:58 UTC
how can a power management blacklist entry for gp76's ALC298 on snd-hda-intel be made ?
how can power save state be prevented ?
Comment 14 FRANK S 2024-03-28 10:48:16 UTC
OK, I've set power_save=0 and power_save_controller=0 in intel_hda.c and built the kernel. restarting the system I verified the settings again in sys/module and they were correctly set.
But the problem wasn"t solved that way.

My amateur conclusion is, the ALC298 and sofhdadsp have their own power management settings somewhere. I don"t know.

what I'll do now is record the audio situation with my mobile and without the workaround of playing a continuous wav stream an how it changes the ability of the system to do the audio test front left/right. More I can't do now. I hope ALSA developer team will come up with a solution/hints.
Comment 15 FRANK S 2024-03-28 11:05:28 UTC
Created attachment 306053 [details]
record of the audio test with and without background audio workaround

record of the audio test with and without background audio workaround
Comment 16 FRANK S 2024-03-28 11:42:23 UTC
the workaround also corrects the problem with longer youtube videos. These videos are not only affected the first 0.8 seconds. The ALC298 or whatever component starts sleeping evr and ever again during the minutes long videos.
Butit would be nice if we had a solution for the phenomenon...
Comment 17 FRANK S 2024-04-11 05:38:10 UTC

*** This bug has been marked as a duplicate of bug 218709 ***

Note You need to log in before you can comment on or make changes to this bug.