Random inverted audio channels

Random inverted audio channels

[Alexandre BOUIN]

Hi alsa-devel list,

I'm currently working on an internal project. 
Here is some details :
This embedded project includes a AT91SAM9260 CPU and a WM8900 codec.
Audio driver is using ASoC API. 
Audio data could be either switch from I2S or DSP A, but I prefer the second
one.
Audio Codec is master.

We have an issue : when we play a right channel stereo track, sound is
played on right channel, but sometimes on left.
This issue could happen right after starting kernel, or later after 10-20
sounds played.
With Frank Mandarino, we've checked any issue from SSC(audio bus) or PDC
(dma) controlers, but no much success.

Any idea to help us debugging this issue would be welcome !

Regards,
Alex.

Re: Random inverted audio channels

[Mark Brown]

On Fri, Feb 15, 2008 at 05:39:49PM +0100, Alexandre BOUIN wrote:

> This embedded project includes a AT91SAM9260 CPU and a WM8900 codec.

JOOI, is this the WM8900 driver in our dev branch or another one?  It
shouldn't make any difference either way what driver is used for the
codec, though, so long as it is generating the relevant clock signals.

> We have an issue : when we play a right channel stereo track, sound is
> played on right channel, but sometimes on left.
> This issue could happen right after starting kernel, or later after 10-20
> sounds played.
> With Frank Mandarino, we've checked any issue from SSC(audio bus) or PDC
> (dma) controlers, but no much success.

> Any idea to help us debugging this issue would be welcome !

The most likely issue here is a DMA underrun but from what you say you
may already have ruled that out?  It would be helpful if you could run
through the debugging you've already done here.

One thing to try if your design allows it would be to try making the
CPU the master.  That should reduce the number of variables, at least.

Re: Random inverted audio channels

[Timur Tabi]

Alexandre BOUIN wrote:

> We have an issue : when we play a right channel stereo track, sound is
> played on right channel, but sometimes on left.
> This issue could happen right after starting kernel, or later after 10-20
> sounds played.

I've had a similar problem.  In my case, the clock from the codec to the chip 
was unstable.  When I attached a 5pF capacitor to it, the clock stabilized and 
the problem went away.

Try using a probe to validate the clock signal.

-- 
Timur Tabi
Linux kernel developer at Freescale

Re: Random inverted audio channels

[Alexandre BOUIN]

> > This embedded project includes a AT91SAM9260 CPU and a WM8900 codec.
> 
> JOOI, is this the WM8900 driver in our dev branch or another one?  It
> shouldn't make any difference either way what driver is used for the
> codec, though, so long as it is generating the relevant clock signals.

We are porting a WM8900 driver from a 8731 one. It fit quite well our
requirements and should not be the same as your dev branch.
Sincerely, It didn't know you're working on it.

> > We have an issue : when we play a right channel stereo track, sound
> is
> > played on right channel, but sometimes on left.
> > This issue could happen right after starting kernel, or later after
> 10-20
> > sounds played.
> > With Frank Mandarino, we've checked any issue from SSC(audio bus) or
> PDC
> > (dma) controlers, but no much success.
> 
> > Any idea to help us debugging this issue would be welcome !
> 
> The most likely issue here is a DMA underrun but from what you say you
> may already have ruled that out?  It would be helpful if you could run
> through the debugging you've already done here.

I had traced the addresses sent into PDC (soc/at91-pcm.c), and are always
the same.
What other traces are you looking for ?

> One thing to try if your design allows it would be to try making the
> CPU the master.  That should reduce the number of variables, at least.

You're right I should try this way, so it would confirm if I have or not an
issue with my LRC clock.

Re: Random inverted audio channels

[Mark Brown]

On Mon, Feb 18, 2008 at 09:38:12AM +0100, Alexandre BOUIN wrote:

> We are porting a WM8900 driver from a 8731 one. It fit quite well our
> requirements and should not be the same as your dev branch.
> Sincerely, It didn't know you're working on it.

No problem at all - I was just asking from curiosity.

> > The most likely issue here is a DMA underrun but from what you say you
> > may already have ruled that out?  It would be helpful if you could run
> > through the debugging you've already done here.

> I had traced the addresses sent into PDC (soc/at91-pcm.c), and are always
> the same.
> What other traces are you looking for ?

Have you checked to make sure that the PDC is not reporting any errors?

Re: Random inverted audio channels

[Alexandre BOUIN]

> > We are porting a WM8900 driver from a 8731 one. It fit quite well our
> > requirements and should not be the same as your dev branch.
> > Sincerely, It didn't know you're working on it.
> 
> No problem at all - I was just asking from curiosity.
> 
> > > The most likely issue here is a DMA underrun but from what you say
> you
> > > may already have ruled that out?  It would be helpful if you could
> run
> > > through the debugging you've already done here.
> 
> > I had traced the addresses sent into PDC (soc/at91-pcm.c), and are
> always
> > the same.
> > What other traces are you looking for ?
> 
> Have you checked to make sure that the PDC is not reporting any errors?

at91_pcm_dma_irq (soc/at91/at91-pcm.c) doesn't reports that there is
underrun or overruns.

I've seen the same swap behavior with codec configured as slave.

Re: Random inverted audio channels

[Alexandre BOUIN]

> I'm currently working on an internal project.
> Here is some details :
> This embedded project includes a AT91SAM9260 CPU and a WM8900 codec.
> Audio driver is using ASoC API.
> Audio data could be either switch from I2S or DSP A, but I prefer the
> second one.
> Audio Codec is master.
> 
> We have an issue : when we play a right channel stereo track, sound is
> played on right channel, but sometimes on left.
> This issue could happen right after starting kernel, or later after 10-
> 20 sounds played.

Hi list,

I'm back on this issue. I have few questions about it and I'd hope someone
could answer : 
- does someone have already used AT91 SSC controller in slave mode ? Can he
send me his register configuration and tell if he already heard and L/R
channels swaps ?
- is there a way to play a sound at the lowest software level, maybe without
using alsa or dsp devices ?
- where is the software audio buffer ?

As you can see I really want to trace any difficulties to find where comes
this problem.
As always, if you have any other ideas to test, I'll be glad to hear them.

Thanks and regards,
Alex.