The Adventure of the Lossy NetworksThe Adventure of the Lossy Networks

A well-known electrical phenomenon explains why school networks were on the fritz

By Radcliffe Cutshaw, Contributing Writer

Soon after they introduced the PC, IBM developed a machine called the PCJr (”PC Junior”), which was intended for kids and less costly applications. Many grade schools and high schools purchased the popular machine for computer labs.

The PCJr used an analog TV for a monitor with 40 columns of display. New functions could be added by means of a function adaptor that mounted on the side of the machine. One such function that IBM engineers added was a networking adaptor that used 75-ohm coaxial cable.

I was working for the Broadband Networking Division of IBM in Boca Raton when I was assigned to find out why some of the installed networks in schools were working only sporadically or not working at all. Yet, the PCJr and the network adaptor worked fine and, in fact, worked on other networks.

The networks and the PCJrs used only coaxial cable. I requested samples of the five cable types used in the networks and set up a network of three computers in the lab.

First, I tested all the cables and discovered that the network failed when one particular brand of cable was used. Next, I looked at the signals using a spectrum analyzer. It turned out that the PCJr’s network adaptor used baseband communication between the machines. In fact the designers put raw RZ (Return to Zero) digital data, including dc, on the cabling. With a data rate of about 300 Kbaud, signal energy was concentrated below 500 KHz. The offending cable showed a significant drop in amplitude at about the data rate — 300 KHz. Further tests showed that all brands of cable met the industry specifications, which are specified from 1 MHz to 1 GHz.

I suspected that the problem was due to skin effect, which is a phenomenon through which the higher the frequency, the more energy tends to occupy the surface of the conductor. The skin effect becomes significant at higher currents, even at 60 Hz. Skin depth in cm=6.62/(f^1/2) for copper at 20C, with the ac resistance varying directly as the square of frequency and inversely as the conductor size. The high frequency energy drops rapidly to zero over several skin depths.

75 ohm coaxial cable is made in the most inexpensive way possible.  It consists of a shield of aluminum foil wrapped around a dielectric layer surrounding a copper-plated steel central wire.

I needed more data to support my conjecture, so I sent samples of each of the coaxial cables to a mechanical measurement unit that IBM had on the Boca Raton site. The results came back: The copper plating on the offending cable was less than half the thickness of the plating on the other cable brands.

A calculation showed that the skin effect allowed energy at the peak frequency to flow in  the higher resistance steel core (due to the thinner copper plating), causing the losses. I submitted my final report detailing the cause and stating that this brand of cable should not be used in the networks and should be removed and replaced in the networks in which it was installed.

I heard nothing for two years and had just accepted a buyout from the company when I was asked to travel nationwide to supervise the replacement of the non-functioning cable. I declined

Mr. Cutshaw, a serial entrepreneur, is currently a private consultant specializing in RF and analog design and development. He has been involved in many areas of engineering throughout his career.