[vox-tech] Modem question . . . I think it is possible

[Shawn Neugebauer]

Your idea is reliant on each device producing a signal that is
synchronous (received at the same time) with the signals produced
by the other two devices.  It is one thing to do this with many
transmitters and a single receiver; it is quite another matter to 
do this with multiple transmitters and two receivers.

Without synchronous signaling, each device, while able to cancel
its own transmission through the use of echo cancellation, 
effectively receives two signals---co-channel interference.  
Although it is technically feasible to separate two signals of the 
type transmitted by a V.34 modem, it is very, very difficult in 
practice.

I think there would be significant problems with other aspects
of V.34 communication (many of which are not linear and are not
separable in the manner you describe), as well.

The idea was worth considering, but I don't think there's much
merit in tweaking the V.34 spec to accommodate multi-user
communication---there are other physical-layer communication
schemes better suited for this.  If one wanted to do this, one
would be better served directly designing a system for it.

shawn.

> I know the post is a few days old but I was intrigued with the idea.
> modems capable of the v.34 standard have something called an auxiliary
> channel by which it can exchange information on such things as line
> quality.  It can code using nonlinear, precoding, and 16-trellis along
> with most modern being able to do 32 and 64 state trellis depending of
> course on what type of bit (dibit etc) you want to transmit at.  Here is
> the trick: Because of the ability to use such things as assymetrical
> transmission and the use of trellis coding, and echo cancellation, each
> modem will have an idea of sender and reciever.  For the first modem, the
> other modem will be a sender and the other end of the ppp will be a
> sender.  This essentially sets what could be a master slave model in order
> to do correct splitting, with hopefully little interference (via the
> trellis encoding).  Here is how it would work:
> 
> Computer 1 is master
> Computer 2 is slave
> 
> ppp ISP is ISP
> 
> Initiate the connection with the master to ISP and send a request for
> 16-state trellis encoding at 14400 bps.  Ask for 16-state send and
> 16-state receive.  Put the master on 32-state trellis recieve
> without confirmation via the auxiliary making each other
> baud recieved as null or noise.
> 
> Place slave in 16-state trellis before connecting.  Because of the
> trellis, it will interpret the data being sent as noise and adjust phase
> correctly to the nulls.
> 
> When it sends out a request, the ISP with 16 state trellis which is synced
> with the master will interpret it as noise and ignore it.  However, the
> master will take the signals in as extra data.  As long as the master can
> split every other bit and maintain parity and whatever, it has now set up
> two channels, one 16 bit with the other laptop and one 16 bit with the
> ISP.  Each other end will take the 'excess' data as noise and ignore it
> because it is not in sync.  You now have two linked connections over one
> line.
> 
> Truly, the slave still goes to the master which in turns goes to the ISP,
> however, this would be a sane operation of 14400 bps and a technical modem
> split.  Implementing it however, is another thing . . . shouldn't be that
> difficult but also shouldn't be that worthwhile.  Anyway, that is all for
> now.
> 
> 
> Sincerely,
> 	Christopher J. McKenzie
> 
> 	cjm@ucdavis.edu
> 	mckenzie@cs.ucdavis.edu
> 	H: +1 818.9917724
> 	C: +1 818.4293772
> 	1815 Mesa Ridge Ave
> 	Westlake Village, CA 91362
> 
> _______________________________________________
> vox-tech mailing list
> vox-tech@lists.lugod.org
> http://lists.lugod.org/mailman/listinfo/vox-tech
>