Fernando Ariel Gont wrote in a message to Mike Bilow:
 MB> Let's say that each card puts a 1 volt DC bias on its signal.
 MB> Then, by measuring the average DC bias, you can deduce how many
 MB> cards are transmitting simultaneously.  If a transmitting card
 FAG> But ... how many cards can transmit simultaneously? Only
 FAG> one, can't it? 
More than one card can transmit, but then the data will be garbled.
 FAG> So if each card put 1 volt DC bias, and not taking into
 FAG> account the colision moment, there's no more voltage in the
 FAG> line than 1 volt DC, is there?
Once the collision has occurred, the frame sent by both will be garbled, and 
each will have to be resent separately.
 FAG>> Two machines were waiting for the carrior to get free. The line
 FAG>> becomes free, and both machines notice it at the same time. They
 FAG>> both try to send their packets, but at the same time!
 MB> Right.  They use the DC bias measurement to detect this.  The
 MB> frame is lost anyway, but the time the cable is unavailable is
 MB> minimized  -- Mike
 FAG> Aha... But why do the cards keep sensing the carrior?
 FAG> Wouldn't the cards that try to use the line after the first
 FAG> one has started, detect it instead of having to have the
 FAG> first card that used the line keeping sensing the carrior?
Assuming they are working correctly, yes, but...
Remember that electricity takes a finite amount of time to travel on the 
wire. The phase velocity in coax could be as low as 66% of the speed of light 
in vacuum, which works out to about 200 million meters per second.  A 10 Mbps 
Ethernet signal, obviously, sends one bit in 0.1 millionths of a second.  In 
that time, the electrical signal can only go about 20 meters!  If the nodes 
on the Ethernet segment are more than 20 meters away from each other -- and 
the maximum length of 10Base-5 coax is 50 times that -- then it is physically 
impossible for the nodes to be aware that they have begun transmitting 
simultaneously until the electricity from the other one reaches them.  Add 
repeaters into the mix, and collisions become a common occurrence.
 FAG> Once again, thank you very much, Mike!! :)
You're very welcome.
 
-- Mike
--- 
---------------