On May 16, 2017, at 11:47 PM, Shoppa, Tim via cctalk
<cctalk at classiccmp.org> wrote:
Al just recently put this up on Bitsavers, November 1974 drawings for the first 2.94MHz
Ethernet transceiver:
http://bitsavers.trailing-edge.com/pdf/xerox/alto/ethernet/Ethernet_Transce…
Neat to see the 15 pin AUI to Thicknet transceiver (well, a lower bandwidth version of
the 10MHz ones I grew up with) drawn out so clearly. Also shows something I've never
seen in real life, an "Ethernet Dummy Transceiver" which is, I guess, something
like a two-port DELNI ? (obviously showing my DEC introduction to AUI Ethernet there.)
It seems that way. One interesting aspect in this design, if I read it correctly, is that
there isn't a "collision detect" function in the transceiver the way there
is in the 10 Mb D/I/X Ethernet. Instead, the transceiver delivers "TRDATA"
(received data) and "TROTHER" which is the XOR of received and transmit data.
So it looks like TROTHER != 0 means collision. In the later Ethernet, collision is
detected in the transceiver and signaled to the NIC with a signal that's stretched out
as needed. For example, in the DELNI schematics (also on Bitsavers) you can see a state
machine to do that -- unlike the combinatorial logic in this "dummy
transceiver". I wonder how it was done in the 10Base5 transceivers; I have a vague
memory that it's partly an analog process but that may be confusion. (I may be mixed
up with the "packet voltmeter" -- an A/D project at DEC to build a device that
would make a map of a coax segment by measuring the voltages of each station's packets
using sensors at both cable ends.)
paul