[linux-audio-user] A most bizarre problem.

Malcolm Baldridge linux-audio at paypc.com
Thu May 13 08:03:18 EDT 2004


A few weeks into my admittedly bizarre Jack-based telephone logging +
music-on-hold system, I ran into an interesting "bug".

But first, the hardware in play:

Motherboard: ECS P4IBMS (Pentium 4, i845 chipset)

00:1f.5 Multimedia audio controller: Intel Corp. 82801BA/BAM AC'97 Audio
Card: Intel 82801BA-ICH2                                                   
                                 
Chip: Realtek ALC200/200P rev 0    

ALSA-1.0.3b w/1.0.4 drivers + libraries, jackd is from late April 2004 CVS.

My jackd invocation line is as follows:
   jackd -R -d alsa -S -C -p 4096 -M -n 2 -r 11025

I have an AT&T MERLIN phone system in my closet along with my main home
server (see above hardware description).  I have my PCM (Line out) plugged
into the MERLIN system's music-on-hold port, that works well.

I have two phone lines hooked up to JD Audio Phone Patch boxes to provide
local + remote side level correction (trans-hybrid technology), which feeds
into my Line In jack [each patch box provides a mixed in/out audio source on
a mono output, so I have Line 1 feeding Left, and Line 2 feeding Right]. 
This too works beautifully with my vox-activated audio logging program which
I'd adapted from one of the jackd sample clients. [It's ended up quite
something else, but not too bad if I say so myself.]

PROBLEM: When I have someone on hold on Line 2, say, and someone else is
talking on Line 1, the person on Line 2 hears the music-on-hold *AND* the
conversation of Line 1!!

It gets better.  If I dial out on Line 3 [lines 3+4 are outbound only], put
that call on hold, and there are people talking on Lines 1+2, the caller at
the other side of Line 3 hears not only the music-on-hold, but BOTH Line 1 +
2's conversations!!!

YIKES!!!  This was a big shocker.  Naturally, I never did this sort of
testing, being more concerned with DC bias, filtering curves and bandpass
settings, and which compressors to use for the online-browsable phonelogs as
well as the losslessly-compressed (FLAC of course) data archival logs.

Obviously the PCM output is being mixed into whatever feeds into the Line
input of the AC97 codec.  This is annoying, and I can see no way of
switching it off.  I've explored every Mute/on/off option in the alsamixer,
alas the dark fire did not avail me.  

I see nothing about "hardware monitoring" as an option for this chipset,
however Hammerfall and others are mentioned as supporting this feature;
nothing in the intel-8x0 driver suggest this is a supported feature.

I'm relatively new to PC audio hardware, but has the LINE-IN been fed into
the PCM output mixer from the beginning?  Is this a funky chipset bug or
weird issue or interaction?  I know that the output mixer is fed audio from
the CD and the WAVE device, but it seems a bit strange that the LINE-IN port
is also routed through in this way.  I can't try this under Windows, as this
machine has been kept clean from the Dark Software of Udûn since its deployment.

I obviously can't mute the PCM output, since that's the music-on-hold.

After much experimentation, I figured I'd try the Microphone input instead.
 It seems noisier, even with the +20dB turned off.  The impedance is quite a
bit different, though fortunately I can get by with the gain set to the bare
minimum [3 in the ALSA mixer, which is the smallest non-zero value I'm
allowed].  But the cross-surveillance bonus to the music-on-hold problem is
gone.

Maybe I can do something tricksy like emit the music-on-hold on the right
channel only, and hook up Line 2's audio feed into the left channel of the
LINE-IN.  This is a cruel hack which depends on the mic+mix functions in the
codec doing their job properly, which is a huge ask.  Ah well.

I'm back in action, only now I have only one channel of recording
capability. :( :(  

Is there a mixer option/setting to disable this "pass-through" routing of
LINE-IN to the LINE-OUT sound port?

In Mordor where the PC platform lies,

=MB=

-- 
A focus on Quality.




More information about the Linux-audio-user mailing list