[alsa-devel] Grey noise with HDA Intel PCH

[Takashi Iwai]

At Fri, 22 Aug 2014 18:18:28 +0200,
Julian Wollrath wrote:
> 
> Am Fri, 15 Aug 2014 08:02:38 +0200
> schrieb Takashi Iwai <tiwai at suse.de>:
> 
> > At Thu, 14 Aug 2014 18:29:19 +0200,
> > Julian Wollrath wrote:
> > > 
> > > > > > > > Thanks.  The setup looks OK, the path is:
> > > > > > > >   DAC (0x02) -> mixer (0x0c) -> mixer (0x14)
> > > > > > > > and the mixer mutes the loopback path (index 1).
> > > > > > > > 
> > > > > > > > Is it the state where you still hear the static noise?
> > > > > > > Yes, the alsa-info.sh script was executed, while hearing the
> > > > > > > static noise.
> > > > > > 
> > > > > > OK.  And if you mute "Speaker" or "Headphone", the noise is
> > > > > > gone, too, right?
> > > > > No, if I mute "Speaker" (the internal laptop speaker) it is not
> > > > > gone. It is only gone, if I mute "Master" or have a headphone
> > > > > plugged in and mute "Headphone" or "Master".
> > > > 
> > > > Interesting.  The "Speaker" mute changes "Speaker Playback
> > > > Switch", and as you can see in alsa-info.sh output, this toggles
> > > > the mute of the speaker pin NID 0x14, i.e. the endpoint.  If
> > > > "Master" influences, it means to mute *both* Headphone and
> > > > Speaker mutes would work. Could you check this?
> > > Muting both Headphone and Speaker does not work for the static
> > > noise on the speaker.
> > 
> > Then check what's the difference in codec proc content between the two
> > cases: Master=mute and Headphone+Speaker=mute.
> > 
> > > > > More fascinating, it is not present, if the speaker is not muted
> > > > > and I disable powersaving via "echo '0' >
> > > > > '/sys/module/snd_hda_intel/parameters/power_save'" but even
> > > > > then, it is still present on the headphone.
> > > > 
> > > > Hm.  Is the noise audible when you playback a PCM stream, too?
> > > > For example, you can play a silent stream.
> > > Yes, then it is audible.
> > > 
> > > > > > > > Also, what if you use the module option for snd-hda-intel
> > > > > > > > model=nofixup or model=generic?
> > > > > > > That did not change anything either. Passing the "mixer_nid
> > > > > > > = 0" hint also did not get rid of the static noise, when
> > > > > > > loading snd-hda-intel with model=nofixup resp.
> > > > > > > model=generic.
> > > > > > 
> > > > > > Did you reboot with setting the option in /etc/modprobe.d/*?
> > > > > > Reloading the module might not work for such a problem.
> > > > > No, I just unloaded every sound-related module and than loaded
> > > > > the module via "modprobe snd-hda-intel model=..." again.
> > > > 
> > > > Then always test with reboot.  The problem is about the
> > > > vendor-specific setups, and it's often sticky unless the cold
> > > > boot. At best, do the cold boot.  The warm boot doesn't cure
> > > > always.
> > > Ok, I tested with a reboot but that did not help.
> > > 
> > > > > > In either way, there should be some difference in alsa-info.sh
> > > > > > output, e.g. more (or less) mixer items with the model option.
> > > > > The difference is their, all the "Dock ..." entries were
> > > > > missing, when loading with the model="..." option.
> > > > 
> > > > Yes, and you should also see the difference in the kernel
> > > > messages.
> > > > 
> > > > Another thing to check is to swap the DAC assignment.  There was a
> > > > similar problem on some Sony laptops, and we had to swap the DAC
> > > > assignment since the hardware seems to have some implicit
> > > > assumption of the DAC.  A test patch is below.
> > > Thanks, but sadly the patch did not help.
> > 
> > I have no much other clue, so far.  The rest you can test is to toggle
> > EAPD, toggle GPIO pin, and change the different pin setups, e.g. put
> > VREF or HP amp bit on any (even unused) pins.
> Toggling the EAPD got rid of the noise but I also did not get any sound
> output anymore. Playing with the other pins got me nowhere. Seems like
> I have to live with the noise.

Possibly a combination of something might work, but it needs really
many trial-and-error.  In anyway, if you hit a good state
coincidentally, don't forget to take alsa-info.sh snapshot.


Takashi