A regular client of mine was a value-added-reseller of computing
and communications equipment. They installed bar code and computing
equipment into mature manufacturing plants, such as paper mills,
steel mills, etc. On several occasions, their customers experienced
problems with equipment that was connected to their central computer
over twisted pair cable.
The first case presents an example of an extreme failure. The
company, located in Chicago, had two buildings across the street
from each other. One of the buildings housed most of the
communications equipment, and there were no problems. The building
across the street housed one PC that needed to communicate with the
main computer on the other side of the street. A twisted pair cable
was used to form an RS-422 link between the two machines. Every time
the customer tried to connect this link, the RS-422 printed circuit
board in the remote computer's machine was damaged. Clearly, there
was a major electrical problem. I suspected that the two buildings
did not receive their electrical power from the same source. In
fact, we learned that the two buildings were powered by separate
sub-stations. This explained the problem. The potential difference
between the neutral lines powering the two buildings was great
enough that the common-mode voltage between the two communicating
computers was greater than the damage threshold for the RS-422
chips. Therefore, when the twisted-pair cable connected the two
machines, the RS-422 chips were destroyed on the remote computer.
The fix was easy. I recommended they place transformer-coupled
modems between the two computers. These devices were readily
available as plug-in attachments to the PC. The problem was gone.
I was called for another situation involving a remote printer in
a manufacturing plant. The system did not experience communication
problems anywhere in the plant except for one printer in one room. I
suspected a power problem. I spoke with an electrician there, and he
told me the plant was powered through a bank of several
transformers. I asked him to check the wiring of the transformers to
ensure that all of the neutral lines were connected to each other.
After he checked, he called me back, and told me that one of the
neutral lines had not been connected. Then he informed me that after
he connected it, the printer problem vanished.
The printer did not work properly before because there was a
potential difference between its neutral power line and the neutral
lines of the other equipment in the system. In effect, the potential
difference caused errors in the data signal sent to the printer by
the controlling computer.
