Open Source Toolkit Documentation:

Communicator Audio/Telephony Server (CATS)

License / Documentation home / Help and feedback

CATS is a cross-platform Communicator-compliant audio server. CATS implements basic multi-platform desktop audio capability, and is based on an object-oriented, extensible software framework that allows alternate implementations to be developed as plugins.

Once you've compiled CATS, if you've downloaded CATS as part of the Open Source Toolkit, you can run CATS in a simple full-duplex echo example.

Status

History and plans

This implementation was inspired by the JDAS Java desktop audio server, which never worked completely due to shortcomings with JavaSound. It is a fairly faithful reimplementation of the basic ideas in C, using a better cross-platform audio solution (PortAudio).

The CATS server currently incorporates preliminary support for the ComputerFone telephony device, and an initial cut at voice activity detection. We have not yet incorporated echo cancellation.

Version history

Installation

Make sure you've followed the instructions below about setting up PortAudio.

The CATS server requires threads for PortAudio, but is not known to be thread-safe enough that the individual timed tasks can be run as threads themselves. The core Communicator Makefile support for this configuration (thread-aware, but not tasks as threads) should be transparent, but isn't. Just in case you ever wonder why the Unix Makefile is more complicated than normal.

Unix

If you have downloaded CATS as part of the Open Source Toolkit, then CATS should build automatically after you configure the toplevel Makefile.settings file. If you have downloaded CATS as a standalone package, you should modify the file cats_settings.make as instructed, then type make.

Windows

We've provided a MSVC workspace file for the CATS server which makes four assumptions:
You can change these assumptions, at least in MSVC 6.0, in the following locations:
You should change the settings for both the debug and release versions.

Windows runtime notes

The MDTM library compiles to a DLL on Windows, and it must be in a location where it can be found at runtime. Your options are:
If you choose the second option, the command will look something  like this:
C:\> set PATH=C:\communicator\mdtm\lib\win32\Release;%PATH%

Telephony support

We've attempted to design CATS with telephony in mind. In particular, the audio GUI server is set up to mimic telephone control, for maximum simplicity of testing on the desktop. As for actual hardware, we have been working for a short while on incorporating a robust port of the ComputerFone, a low-cost single-line telephony device. MIT used the ComputerFone successfully for a number of years on the Solaris platform, with somewhat less success on Linux. It is possible to enable our ComputerFone support in Makefile.settings.<archos> by uncommenting the ENABLE_CFONE line. However, the current implementation is quite preliminary; it recognizes phone events fairly reliably, and we've achieved audio connectivity, and it appears to shut down cleanly, but testing has been extremely minimal. We'd like to thank MIT for allowing us to use their code as a guide, but we, of course, are responsible for any shortcomings. We hope to return to this work in the future; we include it in case anyone's feeling brave and wants to try it, or use it as a guide for other possible telephony integration.

Apropos of nothing, for telephony testing, we've greatly valued a bit of hardware that MIT pointed us to: a telephone line simulator manufactured by Skutch Electronics.

Usage, message set, messages issued

See the core CATS documentation for a description of the use and behavior of the CATS server.

Compiling and using PortAudio

The current version of PortAudio is version 18. Version 19 will feature a much-modified API, which we are certain will not work. Version 18 is known to enable CATS on Sparc Solaris, Windows and Intel Linux.  No other platforms have been tested.

Sparc Solaris

The Solaris port of PortAudio, which we helped to contribute, is not available in the original version 18 release. You must grab the CVS snapshot of the version 18 tree, which you can find at http://www.portaudio.com/usingcvs.html. The Solaris port works with the default Solaris audio drivers and libraries.

Building PortAudio on Sparc Solaris

Do not type make; first, there's no default Makefile, and second, the default compilation attempts to install the library in /usr/local/lib, which you may not have access to. You really only need the PortAudio library. Do this:
% cd <root of PortAudio distribution>
% make -f Makefile.solaris clean # If you've previously compiled from the same source on a different platform
% make -f Makefile.solaris sharedlib
Note that the compilation currently only supports shared libraries. You shouldn't have to set LD_LIBRARY_PATH when you run CATS, because the Makefile templates should install the run-time dependency in the executable.

If you intend to compile from the same source on multiple platforms, make sure you move the shared library, which can be found in pa_unix_oss/libportaudio.so, to a safe place. You can try setting LIBINST in Makefile.solaris and doing make libinstall, but we haven't had the need to do that. Wherever you end up putting libportaudio.so should be the value ofPA_LIB for CATS in cats_settings.make (if this is a standalone installation) or in the toplevel Makefile.settings.<ARCHOS> (if CATS is part of the Open Source Toolkit).

Intel Linux

PortAudio on Linux relies on the low-cost commercial Open Sound System(TM) drivers available at http://www.opensound.com. On Intel Linux, you must also use the CVS snapshot of the version 18 tree, but for different reasons: the released code assumes that usleep() returns an integer, but on Linux it returns void. You can find this snapshot at http://www.portaudio.com/usingcvs.html.

Building PortAudio on Intel Linux

Follow the instructions for Sparc Solaris, except the relevant Makefile is called Makefile.linux. All the same comments apply.

We have discovered that with at least one Intel Linux audio card, all frames-per-buffer values except one yield audio choppiness with recording. We encountered this problem on one of our laptops.

Windows

On Windows, you can use the stock PortAudio v18 distribution.

Building PortAudio on Windows

Download portaudio_v18.zip, and also download the zip file for Windows for MSVC. Follow the instructions on the README. Change the name of the toplevel directory to portaudio, if you intend to follow the default assumptions outlined here.  If you follow the directions, you should end up with a file WinVC\WinVC.dsw. You can build the required libraries using this workspace.

Known bugs

The ComputerFone implementation is still pre-alpha and essentially unusable in its current form.

License / Documentation home / Help and feedback

Last updated October 4, 2002