bob@nih-csl.UUCP (04/22/87)
The original thin Ethernet specification calls for a host to
be connected directly to the segment trunk; i.e., no drop cable
allowed between the segment and host. Could anyone comment on the
use of a tee-connector on the thin segment with a length of RG-58
coax connected to the host? This makes a much neater, more desirable
connection for office space LAN design, but does it significantly
degrade network performance? Is this practice taboo? Will I grow
warts if I try it?
Bob Dew
Division of Computer Research and Technology
National Institutes of Health
Bethesda, Maryland 20894
(301) 496-5361foster@seismo.UUCP (04/23/87)
What you're proposing is officially a no-no, however I have heard (from a 3Com salesman) that people have tried it and gotten away with it but the practical limit seems to be about 3 feet. His position is "try it, if it doesn't work, go by the spec." There is also a "legal" way to do drop cables with thin ethernet but it requires the additional cost of xcvrs and xcvr cables. There is an adaptor (looks like a barrel connector that is threaded on one end) that allows a TCL "vampire" transceiver to be attached to a "T" connector in thin ethernet. I've tried these and they work fine. Can anyone tell me if the wires in a xcvr cable are "special" in any way? If I were to make up a cable from high quality "RS-232C" "type" cable (AWG 22 I think, maybe 24's) would I have a reasonable chance of success? The xcvr cable conductors appear to be AWG 20 to my untrained eye, perhaps if I doubled the 22/24's? My reason for wanting to do this? I have about 30 unused "serial" cables (12 conductor) running through a secure area and it would be a lot easier to rewire the ends than to run new cables through the area. Thanks in advance, Glen
preuss@mimsy.UUCP (04/23/87)
Along these same lines, could you use a DEC DESTA to hook onto a thick drop cable and hook a segment of thin to it? thanks donp -- Uucp: ...seismo!umcp-cs!preuss Arpa: preuss@Maryland 301-496-2265
wunder@hpcea.UUCP (04/24/87)
The Ethernet spec allows for 3cm stub to get from the center conductor to the guts of the transceiver. This includes wiring inside the transciever. The BNC tee eats up a lot of that 3cm. If you put a drop cable on the tee, a wave on the coax will see 25 Ohms at that point, and half the volts will be reflected. In addition, half of the wave goes down the drop cable, and half the wave goes down the coax. So, you lose 75% of your signal at each drop. In practice, it works for the first two or three stations, then everything falls apart. Don't do it. wunder
phil@amdcad.AMD.COM (Phil Ngai) (04/24/87)
In article <43715@beno.seismo.CSS.GOV> foster@beno.CSS.GOV.UUCP (Glen Foster) writes: >Can anyone tell me if the wires in a xcvr cable are "special" in any >way? If I were to make up a cable from high quality "RS-232C" "type" >cable (AWG 22 I think, maybe 24's) would I have a reasonable chance of >success? The xcvr cable conductors appear to be AWG 20 to my untrained >eye, perhaps if I doubled the 22/24's? The considerations in the xcvr cable are: 1) dc resistance. you have to power the transceiver 2) ac signal attenuation. can't let the signal get too small. note that you are driving lines terminated with 78 ohms 3) impedance. the characteristic impedance of the pairs (they are twisted, aren't they?) shouldn't be too much different than 78 ohms or reflections may be a problem 4) shielding. you don't want too much cross talk between the transmit pair and the receive pair. I wouldn't worry about esoteric stuff like velocity of propagation. The requirements are all documented in the Ethernet book. The big questions are 1) do you have shielded twisted pairs 2) how long is the run? Assuming you do have shielded twisted pairs, I would look up the spec and then see how the gauge of the existing wire compares with the spec. If it isn't too outrageously different, I'd make up a cable and try it. If it is too long but not more than twice as long, I'd try doubling up the wires. Since you have 4 signals and only 12 conductors, I'd probably leave Collision as a single. Best thing is to try one and see if it works. -- Phil Ngai, {ucbvax,decwrl,allegra}!amdcad!phil or amdcad!phil@decwrl.dec.com
tony@trlamct.UUCP (04/24/87)
Re: > Comment on the use of a tee-connector on the thin ethernet segment with > a length of coax connected to the host? Will I grow warts if I try it? YES. The bus cable is a valuable shared resource that should be kept reliable. The wiring might be changed, such as adding another node, or someone lengthening their cable. Then prior sins could manifest themselves and GROW WARTS. Adding open circuit stubs of coax that are significant fractions of a wavelength long is liable to create highly reactive (capacitive or inductive) line characteristics, especially at multiples of a quarter wavelength. To avoid this, one could keep the stub less than say, a tenth of a wavelength, and the line will appear slightly capacitive. A 100MHz component has a wave- length of 3m ie. a cable of less than one foot. We have two wall connectors per node, and thus each host has two cables and a T piece. That is, lots of connectors per node and these are all in series with the bus. The user can also inadvertently unplug a connector and break the bus. Apart from direct cables machine to machine, and securing the connectors, there is no really neat way to implement thin ethernet that is simple and safe. The brochures (wrongly) make it sound trivial. -- Tony Thomas, Applied Mathematics & Computer Techniques Telecom Australia Research Laboratories, P.O. Box 249 Clayton, 3168. ACSnet tony@trlamct D D
fey@scotty.dccs.upenn.edu.UUCP (04/24/87)
Yes, our standard wiring for thinwire hookup includes this method. A DESTA connection to RG-58 via BNC connectors works fine. -mlf-
parker@epiwrl.UUCP (Alan Parker) (04/25/87)
Is there a good source of information about the rules of Thin Ethernet? Things like length of runs, methods of taps, etc. Thanks. Alan Parker, Entropic Processing, Inc. ...!seimso!epiwrl!parker
henry@utzoo.UUCP (Henry Spencer) (04/28/87)
> Best thing is to try one and see if it works.
But if it does work and you put it into use, for heaven's sake document
it prominently so your successor will know that something nonstandard
has been done!
--
"If you want PL/I, you know Henry Spencer @ U of Toronto Zoology
where to find it." -- DMR {allegra,ihnp4,decvax,pyramid}!utzoo!henrystevel@dartvax.UUCP (04/28/87)
In article <470001@hpcea.HP.COM> wunder@hpcea.HP.COM (Walter Underwood) writes: >The Ethernet spec allows for 3cm stub to get from the center >conductor to the guts of the transceiver. This includes wiring >inside the transciever. The BNC tee eats up a lot of that 3cm. ----------- I wonder why Cheapernet boards always need tees. I'm building a Ethernet/Cheapnet board, and I'm putting TWO BNC connectors on it. Will this work? If it doesn't, my fallback is to just use one of the connectors, and add the tee... I'll be back with the results in a month or so.
jerry@oliveb.UUCP (Jerry F Aguirre) (05/01/87)
In article <6083@dartvax.UUCP> stevel@dartvax.UUCP (Steve Ligett) writes: >I wonder why Cheapernet boards always need tees. I'm building >a Ethernet/Cheapnet board, and I'm putting TWO BNC connectors >on it. Will this work? If it doesn't, my fallback is to just >use one of the connectors, and add the tee... I'll be back with >the results in a month or so. The problem with this method is that disconnecting the system will take down the entire segment because you must interrupt the cable connection. People are always disconnecting small systems to move or repair them. With a "T" connector you can unscrew the "T" from the system without interrupting service to the rest of that segment. Of course the average user is just as likely to unscrew both cables from the "T" as unscrew the "T" from the system.
bdale@winfree.UUCP (Bdale Garbee) (05/08/87)
In article <907@oliveb.UUCP> jerry@oliveb.UUCP (Jerry F Aguirre) writes: >With a "T" connector you can unscrew the "T" from the system without >interrupting service to the rest of that segment. Of course the average >user is just as likely to unscrew both cables from the "T" as unscrew >the "T" from the system. The fix for this is easy. Two pieces of heat-shrink tubing. It's not that hard to remove if you ever "need" to unhook the two cables from the T, and it leaves only one connection point for the user to choose from when unhooking his machine to take home for the weekend... -- Bdale Garbee, N3EUA phone: 303/593-9828 h, 303/590-2868 w uucp: {bellcore,crash,hp-lsd,hpcsma,ncc,pitt,usafa,vixie}!winfree!bdale fido: sysop of 128/19 packet: n3eua @ k0hoa, Colorado Springs