NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/sys/ibm/pc/hardware/34419

Internet Message Format | 1992-12-29 | 3.4 KB

Path: sparky!uunet!dtix!darwin.sura.net!zaphod.mps.ohio-state.edu!cs.utexas.edu!rutgers!att-out!cbfsb!cbnewsg.cb.att.com!rnichols
From: [e-mail] (robert.k.nichols)
Newsgroups: comp.sys.ibm.pc.hardware
Subject: Internal modems & UARTS (Was: Re: Question about serial port speed)
Message-ID: <[e-mail]>
Date: 30 Dec 92 02:17:06 GMT
References: <[e-mail]> <[e-mail]>
Sender: [e-mail]
Organization: AT&T
Lines: 58

In article <[e-mail]> [e-mail] writes:
...
>If your I/O board has a real 16450, you can replace it with with a
>16550 - assuming it is socketed, or if you know which end of a
>unsoldering tool to hold.  If not, you can either add another I/O board
>(you might have to figure out how to disable a port on the existing
>one), or you can simply buy an internal modem, which has its own UART
>and gives you an additional serial port independent of the two you
>already have.  The internal modem is by far the easiest and cheapest
>option, if you can do without the flashing lights of an external.  Some
>internals only have a 16450 - ask before buying.
...

The more I think about this, I am driven to the conclusion that it is
meaningless to talk about the type of UART emulated in a high-speed
modem.  Here's my line of thought -- I'd be interested to hear other
opinions, or perhaps even facts (but only if opinions will not
suffice ;-).

(Note: I'm discussing only INTERNAL modems here.  Also, I am
concentrating on the incoming data and ignoring the transmit path.)

Any modem which implements compression or error correction uses a local
microprocessor to run the link protocol.  Thus, there are two
interfaces:  (1) The serial line interfaces to the local microprocessor
via a "UART" of some sort.  The type of UART used here is of concern
only to the modem designer, as the PC never uses it directly.  (2) The
local microprocessor has an interface to the PC bus.  It is unlikely
that a UART would be employed here, since the data is in parallel form
on both sides of the interface.

The second interface would most logically take the form of a FIFO of
some sort, which could conceivably be implemented in the firmware in
the modem, rather than in hardware.  In fact, the only way I can
envision an interface that did NOT effectively implement a FIFO would
be if (a) a one-byte register were used as the data interface, and
(b) the microprocessor in the modem deposited data in this register
in a manner based solely on the timing of the (emulated) DTE speed
setting.

It is condition (b) which makes little sense to me.  Why would timing
be used as a stimulus when it would be no more difficult to use a
trigger generated when the PC actually read the previous data from the
register?  Using this latter trigger would result in a
firmware-implemented FIFO.

So, I'm confused.  It would seem that a modem designer would have to go
out of his/her way to design an internal modem that did NOT present a
FIFO interface to the PC.  In any event, it is meaningless to talk
about the type of UART used in an internal modem, because there simply
is no UART involved in the critical interface.

I've never designed a modem in my life, so I'm sure someone can make
substantial corrections to the above logic.

Bob Nichols
AT&T Bell Laboratories
[e-mail]