Maven 2.0a18 16kbps audio over a 28.8kbps link

Geoff Rehn (EST)
Sat, 26 Nov 1994 05:26:09 -0500


>Changes in version 2.0a18:

>* Tired of fiddling with frustratingly CPU-intensive DSP algorithms like
>GSM and CELP and trying to cram them onto a 68040, I opened up a digital
>communication textbook and read up on something called Delta Modulation. I've
>implemented a codec for delta-mod, which uses only 16 kilobits per second of
>network bandwidth and doesn't require any fancy floating-point either. This
>may be in reach of SLIP connections, which is what a lot of people have been
>asking for. Quality is somewhat poor; there's a great deal of quantization
>noise in the received audio which sounds like an AM radio station tuned
>slightly off, but the speech is quite intelligible. Comments about this coding
>are welcome.

and also...

>-mod is a somewhat crude delta-modulation encoding. It is quite
>bandwidth-efficient, using only half the network bandwidth of DVI and one
>quarter that of law. It does not however go to any great pains to preserve
>audio quality, so using this encoding will result in somewhat noisy audio, not
>unlike that of an AM radio tuned slightly off-station. The quality is however
>plenty good for speech. You should use this when you have very low bandwidth
>available to you and making efficient use of this bandwidth is more important
>than the quality of the audio. Since this encoding algorithm was dreamed up by
>the author from some information in Cooper and McGillem, Modern Communications
>and Spread Spectrum, it is not likely to be understood by other conferencing
>tools. It was also not present in versions of Maven older than 2.0a18.
>Therefore, use this encoding only when you're sure all the other participants
>in the conference are using Maven 2.0a18 or newer.

----

I have tested Maven 2.0a18 and the delta-mod 16kbps encoding over a
28.8kbps SLIP link and I can report that I am very satisfied with a Maven
only connection. The voice audio is readily intelligible (bounced off a
reflector, which some report gives poorer quality audio). There is, as
Charlie says, some background noise but ... what the heck! audio is
possible over dialup!

I must report though that I am disappointed in the CU-SeeMe / maven
combination: I had predicted and hoped that Maven 16kbps audio would work
if the CU-SeeMe video were _paused_ while Maven audio is being sent. The
quality of the audio in this case is not anywhere near as good as Maven
alone.

Even if the video is _stopped_ the Maven audio still is good. I don't
understand this but it seems that even if video is stopped but a connection
to a reflector (or whatever) is still in place, some bandwidth is still
used by CU-SeeMe. Comments?

However, of course if one _disconnects_ and then uses Maven, the results
are fine.

I played around with maven 2.0a18 and a 14k4 only link and some of you
might be pleasantly surprised with the results - it's almost there,
possibly even acceptable.

So, all you dialup Mac CU-SeeMe-ers, get a copy of the latest Maven 2.0a18
and check out the delta-mod 16kbps audio. Over a 28.8kbps modem link, it is
more than fine.

(Please don't ask me where to get a copy: do an archie search for your
nearest site. I'm in Oz!)

CU-SeeMe developers: any chance of including Charlie's new encoding in the
CU-SeeMe Mac audio? It will cut down on bandwidth usage substantially.

So thanks, Charlie, for this big step forward.

Anyone who wishes to test a Maven 16kbps link up, please email me
(rehn@cleo.murdoch.edu.au). We can connect to our Murdoch Maven enabled
reflector here in Western Australia.

Anyone who wants to do a CU-SeeMe, I have a camera here at home over this
weekend and that 28.8kbps modem for a loan.

best wishes all.

Geoff R.

PS. How will this delta symbol transmit across email?

---
Geoff Rehn
Academic Services Unit  Murdoch University
voice: 09 360 6308  fax: 09 310 4929
Email: rehn@cleo.murdoch.edu.au