COBBLED CABLING

by Wayne M. Krakau - Chicago Computer Guide, October 1995

It’s time again for what is becoming a regular event - my annual rave-out on LAN cabling! Why, you may ask, am I making this an annual event? That’s an easy one. It’s because my rate of discovering bad cable plants versus good ones has risen from two out of three to about five out of every six new networks that I encounter -- and the only reason the ratio isn’t worse is because I am including cases where I was called in by a cabling company for non-cabling troubleshooting immediately after a complete rewiring job!

Since my last series on cabling, I have received a continuous stream of inquiries on violations of LAN cabling standards. Most calls involve requests for, in essence, a special dispensation from the standards (and by implication, the laws of physics), along with a demand for an exact enumeration of just how much degradation will occur for a given set of standards violations.

As to the first issue, I usually refer the questioner to his or her favorite deity. (With the stipulation that I get rights to the concession stand if there is a major transgression of the laws of physics. If needed, I can refer them to a team of lawyers in Los Angeles who have a proven record of a willingness to promote absolutely anything without fear of divine - or even judicial - retribution.)

As to the second issue, I refer the questioner to The Psychic Hotline. The people there are better equipped for that type of prognostication. They may also give advice as to an alternate career path for those who are in charge when the aforementioned standards violations cause a network to collapse. (Learn how to say "Do you want fries with that?")

These are case studies of cable plant problems, accompanied by my semi-coherent ramblings. What’s scary about these cases is that they are all from one three-month period!

The first case started with a complaint about incomplete overnight tape backup jobs. After inspecting the workstation holding the tape drive, I found an invalid configuration. The CONFIG.SYS and AUTOEXEC.BAT, the tape drive setup, and even the network card setup were a mess. The settings had been changed repeatedly during multiple conversations between the administrator and different technical support representatives of the tape drive manufacturer, causing a gradual drift into this unstable configuration.

After initial testing, I spent some time completely reconfiguring the workstation. Then I tested the tape drive by backing up the C drive. That test was successful. I started another test, this time by backing up a portion of the F drive, the DOS representation of the volume SYS on the file server. This test worked, but it was alarmingly slow. It was so sluggish that I figured that the workstation could easily lose its network connection in midstream when backing up all of the network drive. My theory matched the description of the initial problem that was reported before all of the configuration fiddling.

This led me to examine the file server. It was running quite efficiently (luckily, since I’m the one who last reconfigured it). Finally, I checked the cable plant. I examined the individual cable segments of this Thin Ethernet network and found several that were plain RG-58 coaxial cable instead of the specified RG-58 A/U. Those extra couple of letters make a big difference in electrical characteristics, and the lack of them can cripple a network.

I also found several bad terminations. Some were poorly applied screw-on connectors (which seldom make a very high quality connection). Others were improperly installed crimped connectors. One termination even had a two-inch strip of completely exposed wire braid! The outer sheathing was completely cut away.

A quick check with my LANCat (Datacom Technologies, Inc., 800-468-5557) showed the biggest problem. The LANCat is a LAN cable testing device that includes a Time Domain Reflectometry feature (TDR is a kind of radar). It showed a total length of nearly 1,000 feet! The Thin Ethernet standard requires a maximum trunk segment length of 185 meters - about 607 feet. The fact that the signal got through at all is a tribute to the resilience of the underlying specification.

We managed to get the network down to less than 800 feet after painstaking segment-by-segment testing (since nobody new where the cable was routed) by removing a completely redundant segment. We also discovered that most of the segments had a large coils of excess cable stashed in the ceiling. This means that the client company can clean up its cable plant by replacing nonstandard cable, cutting off the surplus cable, and reterminating with professionally crimped connectors. This will fix both the length and faulty termination problems. The notes from the tests will be used to create cable plant documentation. Interestingly, the cable was strung by the client’s computer maintenance company.

The second case also involves Thin Ethernet. I was inspecting a LANtastic network for a possible upgrade and expansion to a NetWare 4.1 system. The client mentioned that the current system was not reliable. Workstations locked up in ways that I recognized as indicative of losing their connection to the file server, possibly due to cabling problems. The client also told me that the cable was strung by their previous computer vendor - a dead giveaway for cabling problems. (Just for good measure, the computers weren’t even set up competently!)

I immediately noticed that the coaxial cable went up into the ceiling and was just randomly dragged across the suspended ceiling as needed, even though it wasn’t the mandated plenum cable. Plenum cable is sheathed in Teflon, which doesn’t burn as easily as the PVC (polyvinyl chloride) that sheaths regular cable, and doesn’t give off poisonous gas - mostly cyanide - when it burns, like standard cable does. Using loose, non-plenum cable in suspended ceilings used for ventilation is a violation of building and fire codes.

I also noticed that the cable was a mix of a couple of segments of the specified RG-58 A/U coaxial cable along with multiple segments of RG-62 cable - the kind used for ARCnet LANs and IBM 3270 series mainframe terminals! To add spice to the mix, there were multiple "opens". This is a term used to describe a point where a cable has an open, exposed socket, not filled by another cable or a terminator. A terminator is a resistor of the appropriate resistance - 50 Ohms for Thin Ethernet - encased in a sort of cap that fits on the open end of a cable segment.

The most common open is usually a T-connector (a sort of three-way intersection for cable) left dangling after a workstation has been moved. The side of the T connected to the workstation’s network card is left open. The correct procedure is to replace the T-connector with a barrel (inline) connector that has only two sides, thereby removing the offending open socket.

This site had multiple opens created as the company redecorated and renovated its office. They did, however, have nice, attractive crimped connectors on all of their cables. (That’s one in a row!) Considering the mix of 50-Ohm (impedance) RG-58 A/U and 92-Ohm RG-62 cable, and adding in the numerous opens, I was quite surprised that they could get any work done between the workstation disconnections! This client has decided to scrap the existing cable plant and has drawn up plans to switch completely to 10Base-T (Ethernet over Unshielded Twisted Pair - UTP - cable).

Next month I will continue this theme with examples that show that even professional phone cablers - as opposed to the (alleged) computer people in this article - can’t be expected to dabble successfully in high-speed (LAN) data cabling.

1995, Wayne M. Krakau