[AG-TECH] Alternative mechanism for telco bridging

[Robert Olson]

See http://www-unix.mcs.anl.gov/~olson/AG/TelcoBridging.html for this 
document. I set this up in the ANL Workshop node last week for a meeting, 
and it worked quite well.
--bob

--------------------------------------------------

Alternative Telephone Bridging Solution

The standard mechanism used in the Access Grid for telephone bridging – 
using the Gentner matrix to configure a three-way link between room audio, 
network audio capture PC, and telephone interface – has some serious 
usability problems. Most serious is that obtaining an ideal balance of 
levels between the three audio realms can be quite difficult, and simple 
adjustments (say to the local microphones) can have undesired consequences 
on telephone audio levels.

An alternative mechanism dedicates an audio card and instance of the rat 
audio capture tool to the telephone interface. In this configuration, a 
second audio capture PC (this can be the Display Machine in a standard AG 
node, if it has an audio card) is connected to the Gentner in the same way 
that the standard audio capture PC is. It can use the second audio channel 
on the Matchmaker MM100 if the node is not configured for stereo 
connections; otherwise, it is recommended that a second MM100 be installed.

The Gentner matrix is then set up as follows. I assume below that the 
display computer has been attached to the Gentner for telco bridging purposes.

<Table deleted for email purposes,see the webpage>

In other words, the telco and display computer are interconnected, and the 
audio capture and the room audio (speakers and microphones) are 
interconnected. The telco is completely independent of room audio and the 
audio capture computer. If you have the telco output routed to the echo 
cancellation reference output, remove that routing (as that audio will not 
directly appear in the room audio and this setting will likely confuse the 
echo cancellation).

We install rat on the display computer (I did this by copying from a 
separate computer which had the PIG software installed. It is recommeded to 
use the rat that comes with the PIG because the stock Windows rat will 
introduce a roughly 1-second latency). In my testing I invoked the Windows 
rat by hand with the proper address; however, it should be possible to 
install the arm-eventlistener from the PIG software suite on the display 
machine as well. I have not tested this, however. The command looks like this:

rat -t 127 <ipaddress>/<port>

You can find <ipaddress> and <port> from the arm-eventlistener window on 
the audio capture machine if you are running rat by hand.

Once this rat is up and running, you should be able to connect to a 
telephone call as usual, and see incoming level on the display machine rat 
when there is audio coming from the telephone. When network audio comes in, 
you should hear this audio on the telephone line. Note that in this 
configuration, the audio from the local room is network audio from the 
point of view of the telephone interface. This makes setting telephone 
levels much easier – the telephone audio you hear in your node is exactly 
the telephone audio that others will hear. I advise turning off silence 
suppression on the telephone audio rat. Adjust the levels on that rat for 
comfortable audio levels to and from the telephone (you will likely have to 
get feedback from the people on the telephone as to what an appropriate 
level is for them to listen to).