Re: [alsa-devel] ca0106 SPDIF Output Problems

3 Aug 2008


      On Thu, 2008-07-03 at 22:09 +1000, Ben Stanley wrote:
...
Hi,
I am having some problems getting sound out of a ca0106 card using
SPDIF. I have enabled SPDIF output using alsamixer. Some programs
produce sound, and others stubbornly don't.
After much messing around, this problem is now understood. I have a
workaround. I want to start a discussion about the best way of solving
it permanently.
It turned out to be that the 'audio' bit of the SPDIF AES setting was
set to 'non-audio'. This has the unfortunate effect of preventing any
PCM stream from working, while Dolby Digital and DTS work just fine.
This explains the reported symptoms perfectly.
(I eventually tracked down the problem by examining the state and
meaning of all of the registers used by the ca0106 driver, for working
and non-working streams. I narrowed my search by diffing the register
dumps and focusing on the differing register values.)
The non-audio bit is bit 2 in AES0. Setting AES0=0x2 asserts the
'non-audio' bit (DD or DTS), while setting AES0=0x0 indicates audio
(PCM) data.
It seems that MythTV understands how to set the AES settings, and
correctly sets the non-audio bit when playing DD and DTS streams. When
it has finished playing such a stream, it restores the previous AES
settings. Unfortunately, if MythTV crashes before restoring the
settings, the non-audio setting remains. Other programs which don't pay
attention to the AES settings then fail to output any sound (because
their PCM data is incorrectly marked as 'non-audio').
Once the 'non-audio' bit is set in such a manner, not even a re-boot
fixes the issue. This is because the state of the driver is saved by
alsactl to the file /var/lib/alsa/asound.state on shutdown, and restores
it on boot. My temporary fix was to edit (This information is probably
specific to ca0106)
control.18 {
                comment.access 'read write'
                comment.type IEC958
                comment.count 1
                iface PCM
                name 'IEC958 Playback Default'
                index 1
                value
'0282000200000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000'
        }
The fix is to clear the 2 in the first byte of the hexadecimal value. In
my case, the value is now
'0082000200000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000'
After editing and saving the file, re-load the configuration using the
command
# alsactl restore
If you have more than one card, check the man page for the appropriate
incantation.
If you don't want to modify your settings permanently, you can use the
following commands to test if this problem affects you by manually
setting the AES values on the command line, as follows (this is probably
generic to all cards with SPDIF):
For example, this works for me:
speaker-test --device=iec958:AES0=0x0,AES1=0x92,AES2=0x10,AES3=0x2
--channels=2
and this produces no audio output for me (by setting the non-audio
flag):
speaker-test --device=iec958:AES0=0x2,AES1=0x92,AES2=0x10,AES3=0x2
--channels=2
So in summary, the problem occurred because most applications don't seem
to understand/assert AES settings for SPDIF. In the case where it is set
to 'audio', this is fine, but where the registers are set to
'non-audio', this causes *no audio output* for all PCM streams.
Now for some hard questions. I want to start some discussion around
these points before I try to solve this problem permanently.
* Whos responsibility is it to set the state of the audio/non-audio bit?
ALSA-driver? alsa-lib? application? I believe it should be *set*
appropriately for each opened stream.
* Is there a good reason why iecset does not work with device names like
hw:0,0 , hw:0,1 , hw:0,2 etc? (I could probably fix this...)
On Thu, 2008-07-03 at 22:09 +1000, Ben Stanley wrote:
...
Hi,
I am having some problems getting sound out of a ca0106 card using
SPDIF. I have enabled SPDIF output using alsamixer. Some programs
produce sound, and others stubbornly don't.
I need some hints about what to try next, what information I could look
at, to be able to diagnose some factor common to the failures, and
hopefully the cause.
I am familiar with part of the driver, I have already written a patch
which used to enable 44.1kHz output on an earlier Ubuntu.
Anyway, the details are all recorded below. I would appreciate it if a
patient person could have a look through and throw me some ideas.
Thanks,
Ben Stanley.
I did a git checkout yesterday and built alsa 1.0.17rc3 on Ubuntu 8.04
kernel 2.6.24-19-generic.
I used the following configure options:
./configure --with-debug=full --with-cards=ca0106
--with-moddir=/lib/modules/`uname -r`/alsa-driver-orig
(I am using the --with-moddir to avoid clobbering the distro alsa
modules. I am selecting which modules to use by editing /etc/depmod.conf
and changing the priority order, and then re-running depmod. I can
verify the result by checking /var/lib/`uname -r`/modules.dep .)
Everything builds and installs without any warnings or errors.
So, after all that, I tested three applications:
MythTV
xine
speaker-test
My sound output is via SPDIF to a Yamaha RX-V557 digital amplifier.
I only get sound out of MythTV. I get nothing out of xine (playing
flacs) or out of speaker-test. However, I do get signal out of xine
playing DVDs.
Details:
speaker-test --device-hw:0,0 --channels=2 --rate=48000
This should work, but instead it produces no SPDIF output. The amplifier
shows no digital signals.
root@mythtv:/proc/asound/card0/pcm0p/sub0# cat hw_params 
access: RW_INTERLEAVED
format: S16_LE
subformat: STD
channels: 2
rate: 48000 (48000/1)
period_size: 4096
buffer_size: 16384
================================================================
Testing with xine playing a 44100Hz stereo S16_LE flac:
root@mythtv:/proc/asound/card0/pcm0p/sub0# cat hw_params 
access: MMAP_INTERLEAVED
format: S16_LE
subformat: STD
channels: 2
rate: 48000 (48000/1)
period_size: 1024
buffer_size: 8192
The amplifier says 'Unknown Digital', NO SOUND PRODUCED.
================================================================Testing
with xine playing 'Riverdance' DVD in DTS surround with AC3 Passthrough:
Sound output WORKS
Amplifier says 48kHz DTS
root@mythtv:/proc/asound/card0/pcm0p/sub0# cat hw_params 
access: RW_INTERLEAVED
format: S16_LE
subformat: STD
channels: 2
rate: 48000 (48000/1)
period_size: 2048
buffer_size: 16384
================================================================
Running MythTV: Configured to open ALSA:hw:0,0 directly.
root@mythtv:/proc/asound/card0/pcm0p/sub0# cat hw_params 
access: MMAP_INTERLEAVED
format: S16_LE
subformat: STD
channels: 2
rate: 48000 (48000/1)
period_size: 8192
buffer_size: 16384
The amplifier identifies this as a 48kHz PCM signal.
Sound output WORKS.
================================================================
As a general observation, this problem was present in Ubuntu 7.10, but
was 'intermittent' (I used to be able to get speaker-test and xine
playing flacs to work most of the time.) The failures now seem to be
permanent after upgrading to Ubuntu 8.04, whether latest alsa-git is
installed or not.
Now, I need some ideas for what to look for here. I can't see
differentiating factor between things that work and things that do not.
Perhaps I need to look at some registers.
I don't know what relationship the files at 
/proc/asound/CA0106/ca0106_reg*
have to the registers declared in
alsa-kmirror/pcm/ca0106/ca0106.h
but I think I understand the registers from the header file.
Ideas please?
Once I get this sorted out I can test my patch allowing 44100Hz sampling
rates to be output to SPDIF. At the moment this bug is preventing me
from testing...
Thanks,
Ben Stanley.

Alsa-devel mailing list
Alsa-devel@alsa-project.org
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel