> On Feb 10, 2025, at 11:46 AM, Jim Brain <[email protected]> wrote:
>
> On 2/10/2025 10:08 AM, Paul Koning wrote:
>> ...
>> The difference is that the 202 standard was designed to run half duplex over
>> a standard phone line, or full duplex if you had a 4 wire (leased line)
>> circuit. It's a very simple device that actually works at any speed up to
>> 1200 bps (or a hair more, as PLATO did). The 212 modem using QPSK is a
>> clocked system, but it can carry 1200 bps full duplex over a single phone
>> line, with half the channel bandwidth used for one direction and half for
>> the other.
> I realize it's extremely late and probably of no value except for historical
> purposes, but having a way to visualize the various standards in this space
> with respect to duplex, baud/bps rate, etc. would be of so much value. Like
> the poster I replied to, how a modem worked always seemed so oblique,
> especially as the speeds increased beyond 9600, even without the added
> complexity of things like MNP and the negotiation "dance" later modems held
> on the line. It was fascinating to hear about and use, but I always felt I
> should know more about it. Yet, most material in the day either waved a hand
> over the whole topic, or tried to regurgitate the CCITT documentation.
> Specifically, in your above statement, I'm still struggling to understand the
> duplex aspect of the various standards. As a ham operator and having went
> through my EE degree, I understand duplex, but since I always thought of the
> phone line as a full duplex medium, how it would be used as a half duplex
> channel eludes me. I'm OK with some terminology simplification, as shown
> above, if it could help show how the bandwidth of the telephone line was
> divided up in the various standards and how a 202 standard managed to emulate
> a full duplex conversation (if it actually did this) over the half duplex 2
> wire telephone circuit. (And I use emulate in a loose sense, I suppose.
> Back in the day, when IBM and LU.2 was a thing, I worked at a company that
> created a general comms package that could pass data over various protocols,
> including TCP/IP, LU6.2, NetBIOS, IPX, and LU.2, which I believe was half
> duplex. But, the generic package promised full duplex comms, so we (not me,
> but the team) had to build a way to emulate a full duplex connection over
> that half duplex technology. It worked, at least well enough to support the
> apps used with it, but even it was "magic" to me, and I read all the source
> code)
Yes, a phone line is full duplex, to humans. What actually happens is that the
line carries both directions, superimposed -- you hear both sides in the
headphone. That works well for us because we know how to sort the two signals.
A telephone line modem can be done three ways, roughly:
1. Half duplex: the channel is only active in one direction at a time. (202
modem)
2. Full duplex band split: one direction uses the lower half of the bandwidth,
the other the upper half. (212 modem)
3. Full duplex full bandwidth: each direction uses the full bandwidth but
carefully subtracts the local signal from what is seen on the line to arrive at
the signal from the other side. (V.32 etc., I believe)
The "negotiation" you're referring to may include things like equalization and
calibrating the local side suppression for case (3), since those things vary
from one connection to the next (and, possibly, over time even for one
continuing connection).
A 202 modem, on a POTS line, did not emulate full duplex. It could only run
half duplex, with the usual RS-232 modem control signal handshakes to do line
turnaround. Some (perhaps most or all, I don't know) of 202 modems can be
connected to a 4-wire channel, essentially two phone lines in parallel where
each pair carries one direction, the two pairs in opposite directions. You
could get that sort of line from Ma Bell but it would be a permanent circuit,
not a dialed phone line.
As for datacomm protocols like TCP/IP and such: IP sits above the data link
layer so it's not really in the picture. Looking at DECnet, the original data
link (DDCMP) has support for both half and full duplex. Interestingly enough,
multipoint involves one station (the "master") polling the others
("tributaries") for traffic, so it seems an obvious half duplex setup, but
DDCMP 4.0 supports multipoint both on half and full duplex lines.
Above the data link you'd typically see what looks like a full duplex system,
but if the data link is half duplex then at that layer the "both directions at
the same time" property disappears. Since it's packet switching with
indeterminate timing that is fine. If you want hard-synchronous transmisson it
isn't, but that isn't in the domain of IP.
The distinction doesn't just appear in modems: DDCMP over "integral modem" coax
links can be half duplex (one coax) or full duplex (two coax cables). Original
coaxial cable Ethernet is half duplex, while the introduction of twisted pair
enabled full duplex but at least in the lower speeds doesn't require it. The
various token schemes (802.5, 802.4, FDDI) are all inherently half duplex.
paul
