Hello Bob.

11 Jul 04 10:37, you wrote to me:

 JB>> maybe the impedance mismatch is a problem, the ethernet cards
 JB>> are designed to drive 25 ohm load when they're sending. (50 on
 JB>> each end of the run)

 BL>  That's not how it works.

 BL>  If you hook a string of computers to a 50-ohm cable, each able to
 BL> send and recieve, then you have to send using a constant courrent
 BL> source (>> 50 ohms) and receive with a high impedance
(>> 50 ohms).
 BL> That's a recipe for disaster if you don't terminate, becasue the
 BL> *voltage* can rise to high levels (above the actual sending voltage).

yeah, that over voltage will happen for any transmitter with a source
resistance above a few ohms.

 BL>  The *ideal* situation, is to terminate both ends of the cable with
 BL> 50 ohms. All the computers see a 25-ohm load and source, and
 BL> reflections along the cable are absorbed by the matched load at
 BL> either end. Of course, what happens *along* the cable at each site
 BL> is another matter. You then design the transmitter to provide enough
 BL> *current* into the 25-ohms (as you said) to generate enough *voltage*
 BL> to drive the receiver. The sytem is a compromise that just doesn't
 BL> work, which is why they have changed to the hub method and individual
 BL> cables.

yeah,

 BL>  The point I am making is that if you only terminate on one end, the
 BL> losses along the cable effectively terminate the other end.

that didn't happen here... cable too short?

 BL> The voltage doubles (at worst), but even a Chinese designer would
 BL> allow that much tolerance in the system, and in any case, losses in
 BL> those *really* skinny cables will make sure it never happens. With
 BL> only *two* computers at distances over 3m, it is actually better to
 BL> only terminate one end, and it doesn't matter which end.

I got that working with a 1m cable, but I had to put 25 ohms of termination
on it (two 50 ohm terminators at one end). 50 wasn't enough.

I don't have a 3m cable - only 1m and 10m.

 JB>> And while they're sending they also need to be able to detect
 JB>> if another card is sending... I'd bet that removing the
 JB>> terminators would mess with that.

 BL>  How do they do that? They *have* to pause, because two
 BL> transmitting simultaneously would mean at best a 2:1 shift in level,
 BL> and 2:1 is not nearly enough. Do they use carriers, or something? With
 BL> a carrier, levels wouldn't matter a rat's arse.

they use NRZ coding a 1 is a positive followd by a negative,  and a 0 is is
a negatie volatge followd by a positive.
(for 10Mbps each pulse is 50ns, so each bit is 100 ns)

a string of 1s or 0s is results in a 10Mhz square wave, and a string of
alternating 1s ans 0s (10101010) resuilts in a 5MHz square wave

as you see it's a lot like phase modulation applied to a square wave.

So is there a carrier ?  yeah, sort of.

As for how colliosions are detected (two terminals sending at the same time)
i can't be sure, but it seems to involve the impedance of the network.

Ii seems to me that a current sensing circuit in the output circuit could
detect is there was another card sending at the same time by the output
current being not characteristic of a 25 ohm load...

Jasen

--- GoldED+/LNX 1.1.4.7