Newsgroups: comp.dcom.modems
Path: utzoo!utgpu!watserv1!watdragon!rose!ccplumb
From: ccplumb@rose.uwaterloo.ca (Colin Plumb)
Subject: Re: PEP vs. v.32
Message-ID: <1991Feb28.050506.7297@watdragon.waterloo.edu>
Sender: daemon@watdragon.waterloo.edu (Owner of Many System Processes)
Organization: University of Waterloo
References: <1991Feb27.185340.3897@shaman.com>
Date: Thu, 28 Feb 1991 05:05:06 GMT
Lines: 43

jiro@shaman.com (Jiro Nakamura) wrote:
>
>  I have a question for the peanut gallery. I know that PEP uses a number
>of different frequencies for DAMQAM, achieving an average throughput of
>10,000 to 18,000 bps in one direction, with a smaller opposite channel
>in the opposite direction (I think it was about 30bps?)

Well, on local connections I usually get 18,500 bps or so raw, but
there is some protocol overhead for error correction.  I think the
reverse channel is more than 30 bps, though.

>  Now, this is great raw throughput. But a v.32 achieves 9600 bps in
>each direction, so theoretically it *could* do 19200 in one direction.
>No? If v.32 was set up as a half-duplex system with all of its resources
>going one way, it could do 19200 easily. 

No!  V.32 uses the same frequencies to send in both directions, and a complex
echo cancelling scheme whereby you listen carefully, and subtract off your
own shouting to hear the comparative whisper of the other end.  Putting it
all in one direction would require complete re-engineering, and while it
is probably possible to achieve 19,200 bps unidirectional, the existence of
V.32 doesn't particularly help.  You'd have to eliminate the echo-cancellation
and use the improved noise margins to use a denser constellation (512
points) if you wanted it to resemble V.32.  But decoding a 512-point
constellation is hugely computationally expen$ive.

>  I don't know if v.32bis is full duplex or not, but it does 14,400 bps.
>  I would think PEP, using more of the available frequency bandwidth
>should be able to get consistently 18,000 bps or higher. Why not? 

It is full duplex, using the same echo-cancelling scheme.  Again, it uses
the entire frequency spectrum in both directions.  BTW, this makes V.32
and V.32bis connections very hard to wiretap.

Your assertion that PEP uses more of the available frequency bandwidth
may be true, but I for one don't know it.

Toby Nixon posted a while ago that Telebit was going to present a paper
to the V.fast commitee about a PEP-like 28,000 baud full duplex modem.
There were other proposals in the same wonderful performance range for
the meeting... I wonder what happened?
-- 
	-Colin
