bbh@mtek.uucp (Bud Hovell) (07/16/90)
This request for help was originally posted to the unix-pc and comp.sys.
att groups. Several persons helpfully replied (as shown below), but we
still haven't a clue about how to further proceed.
We're getting a bit desperate out here, folks.
*Any* suggestions would be *most* appreciated! :-)
############################ original posting ############################
Ok - this one for the *real* wizards! Except I just *know* I'm gonna
hate myself when someone points out the (probably foolishly simple)
explanation of this frustrating little mystery. :-)
We have just occupied new offices, which were professionally wired for
both communications (Merlin) and data, linked by 4-pair connections to
standard RJ-45 jacks/plugs.
We set up STARLAN between two 3B1 computers thus:
3B1/STARLAN<cable>RJ45<-------->punchblock<-------->RJ45<cable>3B1/STARLAN
..........in other words, an electrically straight-
line connection interrupted only by the punchblock and RJ45 connections.
Simple, no? Pretty damn near the most fundamental circuit one can imagine,
short of a straight copper wire.
Piece a' cake. However, referring again to the diagram......
3B1/STARLAN<cable>RJ45<-------->punchblock<-------->RJ45<cable>3B1/STARLAN
^ ^ ^ ^
| | | |
A B C D
The funny thing is that it doesn't work. Total run is *well* under the max.
400 feet allowed. The installers have spent *****HOURS****** trying to figure
out how cum:
* Line A-B tests good.
* Ditto Line C-D.
* Ditto Line A-D.
* Ditto STARLAN-to-STARLAN, at the RJ45 jacks.
* ...but it fails when connected thru as diagrammed. >-|
We even had them bring in a really nifty Tek recording O-scope to look for
environmental funnies. Did a full-spectrum sweep....
Nuthin.
Re-punch the wires. Punch the wires directly on top of each other on
the punchboard connector. Add another punchboard and move the wiring
over to it. Run new cable C-D. Verify *all* connections at the jacks
(twice). Double-check polarity (!?!). (Someone suggests garlic, but we
finally agree that is for something else). Stamp feet. Curse vehemently.
Nuthin.
Some kind of impedence problem? Doesn't seem likely. Mismanufactured
AT&T cable? Faulty RJ45 stuff? Umm, well....wouldn't it show up? This
stuff all looks clean as a hound's tooth.
Weird. So....
We ran about a 60' cable out on the floor, put on RJ45s, plugged it in
to the two STARLAN boards, and it cranked right up and ran FINE. Ditto
with shorter cable of, say, 6 feet.
**************************************************
Now get ready, folks, cause HERE COME D'GOOD PART:
**************************************************
We now take the 60' cable and place it IN SERIES (i.e., replacing one of
the standard cables between the wall and the STARLAN board on one end)...
........AND STARLAN WORKS!!!!! Flawlessly.
Total cable length is now 60' *longer* than before, and relies upon all the
same installed wiring and connectors that failed before.
Three of us guys standin' there looking *really* tired and amazed. :-)
CURRENT STATUS: Right now, STARLAN works great, so long as I have an extra 60'
of cable stuffed down beside my desk. Just *why* I need have an extra 60' of
cable in order to enable STARLAN is quite an aggravating puzzle. Then, too,
the cable keeps getting wrapped around my ankle while I'm working.
IN SUMMARY: Oliver Wendell Holmes once commented that there are some situ-
ations to which the only reasonable response is: "Well I'll be God-damned!"
I must most-fully concur.
Any help out there, guys?
################################# replies #################################
There have been three replies of interest, quoted below:
From one person:
> I have wired a starlan before. Check to see if your terminating resistors
> are there...."
Terminating resistors???? Are reflected signals a potential problem on a
starlan system? Even if the wiring is healthy? I understand that Ethernet
gets such resistors, but I've read no mention anywhere in the documentation
of such a need on this (or any other) starlan installation.
Are reflected signals a possibile explanation if there is a minor defect in
the wiring - a defect that does not show up on ordinary continuity and
other typical circuit-testing?
From another:
> No answer!
> But, while I've seen poor in-wall wiring and other problems, the
> inexpensive wires from the NAU to the wall jack have often been
> culprits. I'n not looking at your mail... SURELY you switched
> those wires out? And if you replace your 60' wire with a NEW
> short one you still get the trouble?
Yup - wires have been repeatedly switched, and all other cables fail.
The in-wall jacks were verified by the installer (and his supervisor)
to be correctly pinned-out.
> Sounds like the problems I used to have with Laser Writer II's:
> We use the same wires used with StarLAN with varying D25 shells
> to run much of our standard RS-232 traffic in an office. These
> wires seem to have 4 different shades of plastic cover reflecting
> 4 manufacturers or their most recent spools of wire. Anyway,
> ffrom some systems I could only drive the LW's reliably with
> two of the colors of wire!
[ ... ]
Yeah - I keep wondering if there is something weird about the wire itself.
STARLAN isn't an experimental product, and it seems odd that ours should
behave in a way that no one else seems to have experienced in similar
installations.....
And from yet another:
> Here's a silly thing I didn't really think of ... and it kind of makes
> sense here. Are your cables MIRROR image or STRAIGHT THROUGH?
> (ie. What is the connection from A-D? Does pin-1 go to pin-1 of each
> RJ45 on each end?)
I should have mentioned that we verified this - yeah, I'm afraid it *is*
straight through. :-(
We are absolutely stumped about what to do next. I also just got handed a bill
for $650 for the installers to trouble-shoot the installation (made by them
two months ago for about the same price, and not tested until now). They seem
to reason that if they cannot prove there is a problem on their installed
wiring, then the problem must be elsewhere - though they cannot suggest where
else or how this might be - and that it is *our* problem to pay *them* to
solve. Empirically, however, the fact that the line *does* work when the
longer cable is installed leaves some serious doubt about the installation
itself, since the starlan cards obviously are working.
################################# HELP!! ###################################
Bud
____________________________________________________________________________
UUCP: ...{tektronix|sun}!nosun!mtek!bbh ...bbh%mtek@nosun.West.Sun.Com
MOTD: "...never forget that most soldiers hate warriers." - W. E. B. Griffinalana@hprnd.HP.COM (Alan Albrecht) (07/19/90)
/ hprnd:comp.dcom.lans / bbh@mtek.uucp (Bud Hovell) / 11:34 am Jul 15, 1990 / >We're getting a bit desperate out here, folks. > >*Any* suggestions would be *most* appreciated! :-) > > Bud >____________________________________________________________________________ >UUCP: ...{tektronix|sun}!nosun!mtek!bbh ...bbh%mtek@nosun.West.Sun.Com >---------- Maybe you are having a problem with crosstalk. I expect you have mainly been checking continuity. There are several other important parameters to check. You could try using an HP Wire Test instrument to verify your cable. This was developed specifically to test wire for compatibility to 10 Base T. Your cable installer may have one, otherwise check with HP. I believe we offer a Wire Testing service or if you have a big installation you may want to consider buying one. Alan Albrecht
pat@hprnd.HP.COM (Pat Thaler) (07/19/90)
A couple of questions: ATT has marketed Ethernet over twisted-pair products as StarLAN 10. Since you say only StarLAN, I assume you are talking about 1 Mbit StarLAN products. Is that correct? One reason I am confused is because you mention 400 feet as the distance limit when 1 Mbit can support 250 m. Is the wire you were used from the jack to the 3B1 truely twisted pair or is it flat? In what way does it not work? Do you get collisions, CRC errors, or does the data just not get throught? Do both directions of communication fail to work or only one. You may be to the point where it will be hard to pin down the cause of the problem without testing the wire or by testing only DC parameters of the wire. On terminating resistors: AT&T and some other manufacturers produced a bus version of StarLAN (the standard doesn't cover a bus, only a point-to-point link). When using a bus version, terminating resistors are needed. I believe that in the AT&T products those resistors are actually built into the interface cards and are automatically connected when needed (based on which jacks have connectors plugged in) so as a user you would not necessarily be aware of them. (Bus products generally have two RJ-45 jacks, so the bus goes through the card.) Some other bus type products require a terminatting resistor module to be plugged into the unused RJ-45 jack at each end of the bus. Since you don't appear to be using a hub, I assume you are using the AT&T cards in bus fashion. Also because you mention a distance limit of 400 feet. 1BASE5 (the name of the standard covering StarLAN) allows a distance of 250 m (about 800 feet) from the card to the hub. A bus version used without a hub would cut that roughly in half. In non-bus versions, the transmitter and receiver terminate the line. Pat Thaler