This article describes a problem with a school lab network and how the
problem (after two attempts) was solved. The network has a PhoneNET
backbone of 2100 feet with a repeater in the middle. Workstations include
25 Apple IIe systems, 3 Macintosh systems, 1 Apple IIGS, a few AppleTalk
ImageWriter II printers, and one AppleTalk ImageWriter LQ printer. Devices
are connected off of the backbone with RJ-11 jacks/wires as small trunks.
In the diagrams below, nodes (jack boxes on backbone) are the numbers 1
through 22. R denotes a Farallon repeater. "f.s." denotes a Macintosh SE
with HD20 SC File Server.
1-2-3-4-5-6-7-8a=R=8b-9-10-11-12-13-14-15-16-17-18-19-20-21-22
|
f.s.
The problem was that prior to installation of the repeater, various Apple
IIe systems would only intermittently be able to start up off of the server
("f.s." located at node 7); particularly, those toward the ends of the
backbone.
Solution #1: Adding a Repeater
------------------------------
A repeater was installed by separating node 8 into 2 nodes (8a and 8b) and
placing the repeater between the two nodes. While installing node 8, the
engineer
noticed a loose, but electrically continuous connection, within the RJ-11
box. Thinking nothing of it, he installed the repeater and tightened the
screws. All worked great--everyone gave all of the credit to the repeater
(Farallon had said, all along, that we shouldn't need one with only 2100
feet of backbone. We had to try something, so we bought one.)
4-5 Weeks Later
---------------
A teacher's class was disrupted by a loud "popping" noise from an Apple IIe
in her classroom. The screen inverted, became garbled, and returned to
normal. Coincidentally, the network quit working, and machines could no
longer start up on the network--everything went back to the way it was four
or five weeks earlier.
This was too much of a coincidence and pointed to the repeater. (All the
board, computer, and ohmmeter checks of the backbone tested fine.)
Furthermore, after taking out the repeater, the engineer got a few more of
the machines to start. All the machines within 700 feet of the repeater
and the file server started up regularly.
After many hours of lost labor time and much frustration, the engineer took
the only approach that he could think of to narrow down the cause of the
problem; he asked for a replacement repeater. But once installed, there
was no change. Additional tests, inspections, ohm checks, continuity
checks, and termination checks failed to give a clue.
Solution #2: Damaged, Loose Wiring
----------------------------------
Although the engineer had been warned repeatedly to make sure that there
were no extra resistors on the network, after seeing the ohmmeter readings
of the backbone, he defied logic and put another 100-ohm resistor where the
highest resistance measurement was taken (node 15: 80.1 ohms with the
repeater installed). The result: five more machines came up (nodes 16
through 20).
He decided to put in another resistor (maximum of four, now), and there was
no further change. This came as a surprise, because node 20 now started
with lightning speed (it was sluggish before), while node 21 just sat there
looking "dumb." That's when it was clear that it was time to crawl along
the wires again to see what was going on.
The loose wire idea came up, so an inspection was made of node 21. There
it was. The pair of wires coming into the box (#21) had been damaged to
the point that they were "broken" but not fully severed. This pair had
been all but fully separated from the wiring posts. This fit the confusing
symptoms of passing the ohmmeter checks, but possibly causing data
reflections in an AC mode setting. After reconnecting the wires, all was
fine. (Reflection checks still passed the requirements.)
The installation at the box did not have a complete plastic molding
installed to protect the wires from the young students' feet. The wire had
been kicked and stretched to the breaking point. A plan for molding
installation is in the works.
The extra resistors were removed; everything worked fine, but a little
slower on the ends. One resistor was placed at the high-resistance point
again (node 18) with the result that the end units start up very fast (as
if they were next to the server).
In retrospect, it appears that the repeater was not needed in the first
place. Its installation could have been avoided if the wires had been
tightened. (Reminder: ohmmeter checks can still test okay on bad or loose
connections. It's the AC component, the +5V and -5V, that counts).
However, the school is planning to add a new lab to the network and will
eventually need the repeater anyway.
It appears that all of the other coincidences were merely coincidences.
Most likely, a power surge occurred that disrupted the computers. The
cable could have been kicked even before the surge, because the low traffic
on the network at the time may have caused it to go unnoticed).