Path: utzoo!attcan!uunet!husc6!rutgers!tut.cis.ohio-state.edu!cwjcc!neoucom!wtm
From: wtm@neoucom.UUCP (Bill Mayhew)
Newsgroups: comp.dcom.modems
Subject: Re: Courier HST/IX
Summary: They said to expect a TB killer in 3rd qtr of 1988
Keywords: 30,000 bps, USR
Message-ID: <1414@neoucom.UUCP>
Date: 16 Nov 88 16:21:16 GMT
References: <1414@nmtsun.nmt.edu>
Organization: Northeastern Ohio Universities College of Medicine
Lines: 60


The HST/IX mentioned sounds like something that some people from
USR that I talked to last June referred to as a "TB killer".  They
claimed taht it would have 2x the thoughput for the same price as a
TB modem.  They also said that it would have uucp spoofing.  It
sounds like they may have delivered.  By the way, I heard (but
can't verify) that some of the people working on the TB killer used
to work for Telebit.

I'm not sure that I'm ready to change yet.  We don't have any
machine here that can honestly feed a continuous stream of data
into the modem at 30 Kbaud with the uucp g protocol, even with
spoofing.  We use a separate 3b1 to handle all our traffic that
goes out through the TB modems, as the tty drivers of BSD 4.2 are
pretty lousy.  It was easier to go to the 3b1 for us than to try to
rehack the drivers.  We're switching over to Ultrix, so it will be
interesting to see if we can get better performance for our Vaxen's
serial ports.

Interestingly, the HST/IX is starting to get pretty close to what
Shannon's theorm gives as the ultimate amount of data that can be
sent though a dial-up phone circuit.

The Shannon-Hartley law states:

H = B log_2( 1 + ( P / No B ) )  in bits per second

H : channel capacity in bps
B : bandwidth of channel in Hz
P : received power in watts
No : single sided noise power spectral density in watts / Hz

This, of course, assumes that the noise received has a gaussian
distribution, which may or may not be true in a phone circuit.  I
guestimated that an average phone cirucit has a usable bandwidth of
about 2600 Hz.  Tyical S/N ratio for phone lines these days is
about 45 dB.  Putting that into the above gives me ~38,800 bps.  I
work in medical electronics as opposed to telephony, so apologies
in advance if I've made stupid assumptions here.

Reference: _Satelite Communications_.  Pratt, Timothy and Charles
Bostian, Wiley Books, New York. 1986.  p 284.

I don't have Shannon's paper, but you might want to look it up:

"A Mathematic Theory of Communications", _Bell System Technical
Journal_, Part I.  1948.  pp 623-656

Another possibility might be that the HST/IX has some nifty
data compression sceme, that gives a high thoughput if the data
resembles english (or some other language) text.  A Trailblazer can
give astronomically high throughputs if you feed it 100K of the
same character with compression enabled.  One must look into what
sort of data it is upon which the transfer statistics are based.
Please note!!  I'm not saying at all that USR is fibbing; just be
sure to check it out; Shannon certainly says it is possible.

--Bill

return mail path:  ...!lll-winken!scooter!neoucom!impulse!wtm