Post by AlanPost by TimPost by Gene E. BlochPost by Tim10base2 was for all intents and purposes one long piece of coax
strung between all of the systems, with a terminator at each end.
Wherever one wanted to connect to the cable, a simple T connector
was inserted, and a drop of the same cable was run to the system
being connected. As the previous poster found out, one MUST have
that drop cable connected between the T and the network card, or
dire things happened to the electrical balance of the coax and very
little if any data would get through. If the cable was disconnected
from the network card, it had to be terminated, or the cable
removed from the T connector and a terminator place there instead.
The whole system was very sensitive to impedance mismatch, which
would cause either loss of signal or 'ghost packets', as the
electrical signal would hit the end of the unterminated cable and,
in effect, reflect of the end of the wire and cause interference
with the valid packet being transmitted. The resulting 'collision'
would cause all of the stations wishing to transmit to back off for
their random time and them retransmit. The same thing would happen
to those packets, with the end result that no data would get
through.
Thanks, that was clear.
Although I learned quickly that what I had done was not a good idea,
I don't recall hearing or reading such a clear explanation of it
before.
Obviously I had a vague idea, but I never really looked into it.
Your case was a little bit different. The NIC (Network Interface Card)
was designed so that it would, in effect, 'suck in' any electrical
signal that appeared at it's input so that there would be no
reflection due to impedance mismatch. As long as the coaxial drop
cable from the T connector was in place, this worked great. If the T
connector was connected directly to the NIC, the NIC would 'suck in'
enough of the signal from the main cable that there would be little if
any left to propagate past the connector. In effect, breaking the
single LAN into two at that point. If the sending and receiving
stations were both on the same side of the station in question, they
could communicate. But no traffic would get past the connection in
question. Made for some good troubleshooting problems once in a while.
Not quite.
The Ethernet cable was indeed a long coaxial cable, terminated at each
end in its characteristic impedance (50 ohms). However, a connection
to that cable consisted of a T connector with the sides connected to
the cable in each direction, and the center connected directly to the
transceiver. There was only the shortest possible length of connection
between the transceiver and the electronics of the transceiver - the T
connected directly to the BNC connector on the transceiver without an
intervening coaxial cable..
Since the transceiver was tapped across the conductors of the coaxial
cable, it had to be a very high impedance when receiving, so it would
"suck" only the smallest amount of energy from the cable.
Fortunately, since it was a powered device, this was not generally a
huge problem.
If there had been a coax tail from the T, the 50 ohm cable on any input
would have seen two 50 ohm loads in parallel on the other inputs, and
would have seen 25 ohms at that point, causing a mismatch that would
have reflected about 11 percent of the signal. That would have been
bad.
Given that, many of the transceivers of the time were built with the T
built right into them. Look up the asante friendlynet adapter and you
may find pictures of both forms, one with a BNC connector on the end
opposite the cable, or the later version with two BNC connectors on
opposite sides near the far end from the cable (where the T function
was built in).
There was a cable involved from the transceiver to the computer, but it
was a multi-conductor cable that carried the power and signals between
the computer and the transceiver.
Even a small amount of reflection at the connection point to the coaxial
cable would be a cause for problems, as there were typically several
of these on a single cable segment, and the cumulative effect of their
reflections at various points on the cable could cause the entire
segment to fail.
Alan
You are right, I stand corrected. I actually did put a drop cable in the
'how it's done'. Now that you correct me, I remember having to find a
would reach to the NIC. At least it wasn't the old 10Base2 cable. There
transceiver drop cable to the workstation. I forget how far apart the
marks were - I want to say 15 feet. Anyone else remember? Network guys
have it easy these days. Just run some CAT6 or fiber from the closest
having to crimp my own CAT3 crossover cables to connect hubs together.
feet, and you could only have no more than y hubs in any cable path.