Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utcs!mnetor!seismo!rochester!ritcv!cci632!rb
From: rb@cci632.UUCP
Newsgroups: net.micro,net.micro.atari16,net.micro.ti
Subject: Re: Optimum transmission speed over telephone lines
Message-ID: <218@cci632.UUCP>
Date: Sun, 13-Jul-86 02:32:03 EDT
Article-I.D.: cci632.218
Posted: Sun Jul 13 02:32:03 1986
Date-Received: Wed, 16-Jul-86 03:29:39 EDT
References: <1207@mhuxo.UUCP>
Reply-To: rb@ccird1.UUCP (Rex Ballard)
Organization: CCI, Rochester Development, Rochester, NY
Lines: 68
Xref: utcs net.micro:14723 net.micro.atari16:1291 net.micro.ti:210
Summary: Would you believe almost 19200 baud analog.

In article <1207@mhuxo.UUCP> gjphw@mhuxo.UUCP (WYANT) writes:
>
>  While considering the purchase of a 1200 baud modem, I began wondering
>if there is some optimum data rate for modem operations over the U.S.
>telephone lines.  I have seen a few advertisements for 2400 baud modems,
>and read mention of 9600 bps modems becoming available.  Where does
>the retransmission rate due to lost bits negate the higher data
>transmission rate?
>What might serve to be the most efficient data rate
>using ordinary (unconditioned) telephone lines?

Well let's first consider the theoretical maximum.
The yellow book indicates that the analog value of the line will be
"polled" 9600 times/second, and a dynamic range of 64 db can be
represented in 1 db steps (7 binary bits, but only six are "reliable").
So in theory, you could jam a d/a converter and get nearly 23 Kbaud.
Of course this isn't quite reliable for real communications, and
the phone company might have fits.  This technique has been used
for several "illicit" perposes, most commonly fooling the phone
company into thinking you got a wrong number, when in fact, during
the few seconds you were connected, you had sent a 3 page report. :-).

Now, to be legal, the fastest legal standard is the V.33 or "Trellis
Coding" standard.  Most modem makers are combining this with mnp
protocol, and a relatively standard version of compression.
In addition, the mnp error checking can be shut off for situations
where external error checking is needed (long modem->computer lines,
multiplexors,...).

Vanilla V.33 give a generic full duplex 9600 baud rate, with reasonable
error rates for "terminal mode" types of transmission, where a lost
bit doesn't cost too much.  The compression nearly doubles the data
rates in many situations, but multiplies the effects of errors.  The
MNP protocol allows error detection/correction.
Most of these modems also "double buffer" the line, allowing synch or
asynch in, and synch or asynch out.  This is necessary because after
compression and error checking/correction the rate can vary from
19200 baud in "perfect" conditions to as low as 10Kb in "very noisy
conditions" (including data that won't compress nicely).


>Would the long distance
>digital telephone trunks be a more serious limiting factor than local or
>short haul analog lines?

Actually just the opposite.  The distance to the pcm converter provides
the greatest possiblity for LC filtering, standing waves, ringing, and
several other problems.  The digital lines are very easy to get good
repeaters for, and the loss of a millivolt can be recovered.  Not so
on the analog lines.  Even in trellis coding, a millivolt or an extra
pf on the line can mean a 300 baud drop in speed.

Let's hope that telephone companies can get ISDN (digital/analog)
services available at reasonable rates.  Then we can talk in terms
of 56 Kb/line, over multiple lines if desired.  It would be nice
if a parallel modem interface, rather than serial was a valid option as
well.

>
>  Any suggestions?

I'm writing this at home, so there may be some errors, so,

Any corrections?
>                                 Patrick Wyant
>                                 AT&T Bell Laboratories
                                  ^^^^
Why do I feel this is a sucker question? :-).