[net.lan] Ethernet Transceiver Cable question...

dan@rna.UUCP (Dan Ts'o) (11/30/84)

xyzzy

	Is there a minimum length for an Ethernet transceiver cable ?

	The scenerio is as follows. We have a VAX 780 and a VAX 750, each
with an Interlan Ethernet board connected to a DEC DELNI. The 750 has
the Interlan 10 foot flat transceiver cable connected to a 50 foot
round transceiver cable which connects to the DELNI. The 780, however,
only has the 10 foot flat transceiver cable connected to the DELNI, i.e.
no 50 foot transceiver cable. Operationally, the 750 looks fine, the
Carrier Sense and Collision Detect LEDs blink on and about once every
couple of seconds. (Probably rwho packets, etc. no users are using the
Ethernet). The 780, however, though mostly functional from the user
standpoint, has its CD and CS LEDs flashing wildly and occassionally
there are delays and even dropped connections at the user level.
	If I interpose a 50 foot transceiver cable in the 780's hookup,
every thing looks identical to the 750. Now, I imagine that the 780 with
the short cable is colliding with itself and retransmitting alot.
However a brief glance at the Ethernet specs shows no mention of a
minimum transceiver cable length. Is there one ? Or maybe just a property
of Interlan boards...
	Thanks.

					Cheers,
					Dan Ts'o
					Dept. Neurobiology
					Rockefeller Univ.
					1230 York Ave.
					NY, NY 10021
					212-570-7671
					...cmcl2!rna!dan

rpw3@redwood.UUCP (Rob Warnock) (12/01/84)

There is no official minimum transceiver cable length. However, what
you are seeing may be some interaction between the Interlan controllers
and the DELNI. In particular, if one side is configured for Ethernet
Version 1.0 and the other for 2.0, there is a spec on the "idle state"
of a transceiver signal pair that may be causing trouble. The 1.0 units
expect the other side to maintain a solid differential "1" when idle;
2.0 (and IEEE 802.3) specifies a "decay to idle" which is more-or-less
an ECL "tri-state" (no differential voltage).

Now if the "squelch" on the 1.0 side is set too low (or is missing),
such a "tri-state" condition is a candidate for picking up noise,
causing your erratic performance. Adding 50 feet of cable may put
enough capacitive load on the lines that the noise problem goes away.

Alternatively, there is a spec on the decay-to-idle time, which one
side or the other may be violating in the absence of enough capacitive
load.


Rob Warnock

UUCP:	{ihnp4,ucbvax!amd}!fortune!redwood!rpw3
DDD:	(415)572-2607
Envoy:	rob.warnock/kingfisher
USPS:	510 Trinidad Ln, Foster City, CA  94404