NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/dcom/modems/11450

Internet Message Format | 1992-07-30 | 4.6 KB

Path: sparky!uunet!mcsun!uknet!icdoc!cc.ic.ac.uk!carrion.cc.ic.ac.uk!vulture
From: [e-mail] (Thomas Sippel - Dau)
Newsgroups: comp.dcom.modems
Subject: Re:    Serial and Parallel interface ??????
Message-ID: <[e-mail]>
Date: 29 Jul 92 11:05:55 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]>
Reply-To: [e-mail]
Organization: Imperial College of Science, Technology and Medicine
Lines: 69
Nntp-Posting-Host: cscgc

In article <[e-mail]>, [e-mail] (Floyd Davidson) writes:
 --
 -- The original question was how to speed up modems and why not use
 -- a parallel interface.  Simple, it won't make anything any faster.
 --
Well, no, the original question as I can recall was if it was suitable to use
THE parallel port rather than THE serial port. I.E. the ports as found in real
existing PCs, with 8250/16450/16550 uarts etc.

Then there was some noise about parallel interfaces being inherently faster.
This is intuitively true (2 wires in parallel transport twice the number of
bits that a single wire of the same ilk does).

But it is not true that serial comms is always slower than parallel, and
examples were given for that. Parallel data transfer has the inherent problem
of skew, i.e. one bit travelling faster than another one meant to be "parallel"
with it. And it needs careful design to overcome this problem, and it might
be more useful to spend that hardware and ingenuity to get the bits faster
through a single wire in the first place.

When talking about real existing parallel ports, there are mostly the PC and
some workstations, while other workstations or the MAC have not sported parallel
ports. Since the PC/AT (1984) those parallel ports were bi-directional, and
are therefore suitable as general purpose data transport mechanisms. They have
a handshake protocol that prevents data loss - each byte is declared ready
by the transmitter and acknowledged by the receiver.

I went into our museum to get a 1983 manual for an Epson FX80, which specifies
0.5 microseconds for valid data before and after a 0.5 microsecond strobe
pulse, and banging in another half of a microsecond for good measure
suggests that a byte transfer cycle should take no more than two microseconds.
giving half a megabyte per second (peak) transfer capbility.

At those rates data skew could be managed in a ribbon cable up to 6 meter
long with 1983 technology. Now there are millions of PCs out there who have
parallel ports designed to work with that. I will not claim that each and every
one of them can push half a megabyte per second through the parallel port,
but it should be possible to get more then the 2-3 kilobytes that really
existing PCs can get through their current serial ports. And that is without
major surgery like a hacking away some part of an integrated PC chip set and
putting a 16550A in. (Try finding an 8250 or 16450 in a notebook PC :-)

That half a megabyte transfer capacity is twice as much as the current Premium
Rate ISDN is designed for, so it will for some time not be the major bottleneck
for wide area data communications. I would therefore like to see modems and
other peripherals produced for it, and produced with PUBLISHED interface
protocols so that others can design new ports on computers to make even better
use of those peripherals than currently existing PCs with currently existing
parallel ports can.

I know that there are high performance Multiport-serial, Ethernet, SCSI, FDDI
X25, 3270, you-name-it comms cards for PCs. I also know that some of them cost
more than a PC cou could fit them in - not difficult at a time when a company
over here has just -given away- 10000 20 MHz 386 PCs to people who bought three
software packages (Lotus 123 and friends) at a price still below list.

The advantage of the parallel port is not that its parallel, but that it has a
clear and easy to understand hand-off for data, and that it can be used at
very low speeds as well as at moderately high ones. But its biggest advantage
is that it is there, that it is cheap, and that it is dedicated to whatever
it is doing.

                                    Thomas
--
*** This is the operative statement, all previous statements are inoperative.
*   email: cmaae47 @ cc.ic.ac.uk (Thomas Sippel - Dau) (uk.ac.ic.cc on Janet)
*   voice: [phone] x4937 or 4934 (day), or [phone] (fax)
*   snail: Imperial College of Science, Technology and Medicine
*   The Center for Computing Services, Kensington SW7 2BX, Great Britain