geoffb@nrcvax.UUCP (G. Geoffrey Baehr) (10/24/85)
Keywords:repeater ethernet
In response to the question raised here about Ethernet repeaters, specifically
fiber-optic repeaters, I have some information to give.
I. Fiber Optic Repeaters:
Made by several manufacturers, I have seen Excelan's EXOS 1150 and Codenoll's
units. Both have a link range of ~ 2 km max, you must use a standard transceiveras the input/output tap to the repeater unit.
Siecor Inc also makes a fiber-optic repeater in the same class as the above.
II. IR Laser Repeaters:
Codenoll produces an IR Laser link ( 1 laser, modulator, demodulator)/per end
with a useful range of about 10km. It offers full speed ( 10mbit/20mhz) operation,however it is affected by fog. The price as I recall is about $15k.
It is intended primarily as a building to building link.
III. Microwave.
In short, no hope here. No manufacturer I have yet run across produces a
microwave unit with sufficient bandwidth to support Ethernet full speed
operations. We have tried General Electric GEMLINK's personally, good unit
but too difficult to modify IF amplifiers to give sufficient bandwidth. Many
unit available if you gateway down to < 1mbit .
IV. Gateways.
For building to building links, gateways are the answer. Telephone lines
conditioned for 56kbit usage are cheap (?), and gateway units have reached a
reliable state several years ago. HOWEVER a combined XNS (DECNET) and
TCP gateway is still a nice dream.So.....If you want to gateway down to a
slow speed link, gateway back out to another net , using XNS, you're in
good shape. Any other combination is 'interesting'.
V. Some notes:
Reading my Ethernet Specs from Xerox and 802.3 specs, the maximum co-axial
cable length permitted is ~ 500 meters, ie. loss cannot exceed 8.5db at 10Mhz.
No variation in impedance > 3 ohms permitted . This is for 1 segment. Repeaters
may be installed to increase the number of segments, however maximum round
trip delay for any bit may not exceed 51.2 microsec. You would really need to
know the velocity factor for the coax batch in use to get a max length for
your entire net, also repeater induced delays. Fiber optic links offer very
low propagation delays, somewhat secure physical nets and immunity from
impulse problems, ie. if lightning hits, adios to the repeater but not the
rest of the net. Note that the specs specifically recommend that the net not
be at ground potential (conected to the building ground). It should be isolated
from ground if at all possible. If necessary, one end ONLY may be grounded if
building codes require it. Any other ground connections will virtually assure
ground loops, unequal ground potentials and lots of problems.
Hope this helps.
"Any errors included are entirely the result of inadequate knowledge combined
with excessive verbosity"
--
Geoffrey Baehr - N6LXA Network Research Corp
USENET - inhp4!nrcvax!geoffb
ucbvax!calma!nrcvax!geoffb
{sdcsvax,hplabs}!sdcrdcf!psivax!nrcvax!geoffb
BELL - 805-485-2700
POST - 2380 N Rose Ave. Oxnard C.A. 93030
-------------------------------------------------------------------------------medin@lll-crg.ARpA (Milo Medin) (11/04/85)
Since they were mentioned, I thought I'd throw in some experience I have had with the codenoll 3030 fiber modems in use with repeaters. In short, they dont work. The 3030's by themselves work OK, as long as you plug them into an ethernet controller of the appropriate version number, things work out. To go long distances though, they deactivate CD. Depending on loading, this may not be a problem, but its ugly in any case. BGut when you combine them with a repeater such as a DEC DEREP-AA (local repeater), they don't work. Note we got our DEREP by buying it from codenoll as part of their repeater package. You pay $2000 extra for having a codenoll sticker on it, there are no internal differences. Well, the thing simply wouldnt work at all with the 3030's. After a long debate with the manufacturer, they came out and tried to make it work themselves. After cutting various things in the 3030's, it came up. BUT, we were showing anywhere from 16-50% collisions on the cable, where it used to be 1%! ANd, the really weird thing was that initailly deffered packets were close to zero, and multiple collisions were very high. We took the thing off the link and sent it back. In short, don't buy Codenoll for fiber-coax technology! Milo Medin NASA Ames Research Center