Controlling uncertainty in test data
is very important if the data is to be useful. The case of
insertion of an RJ-45 plug on an Ethernet cable into a jack is used to illustrate how a
hidden uncontrolled variable can affect tests results, in this case
making the results meaningless.
Discussion: Uncontrolled hidden
variables can cause severe problems in testing of electronic circuits.
The case of cable discharge and RJ-45 connectors makes a case in point.
Electrostatic discharge from Ethernet cables carrying a static charge
many Ethernet ports in recent years as a charged cable is connected to
equipment. When electronic circuits are
located outside of the protecting Ethernet transformers, such as when
power is supplied over an Ethernet connection through the transformer
centertaps, protecting the exposed circuitry is very important as is
characterizing the event so adequate protection can be designed into
Figure 1 shows an RJ-45 Ethernet plug being inserted into a jack. Note
that it is not being held straight, but at an angle. It turns out there
is a little play in the jack such that when the plug is inserted at an
angle, one pair of the Ethernet connection makes contact first. If the
plug is angled in the other direction, the other Ethernet pair will
make contact first. If there are electronic circuits connected to one
pair and just ground on the other pair through the transformer
centertaps, the insertion angle may cause a charged cable to connect to
the electronics first or ground first.
During a charged cable test I was conducting, one of the Ethernet
pairs was connected to ground and the other to electronics through the
centertaps of the Ethernet transformers, If the plug was angled one
way, the circuit would survive the highest charge voltage level that
could be applied to the cable. When the plug was angled the other way,
the circuit would fail at a definite charge voltage on the cable, a
much smaller value. Until this was realized, the data was a scatter
plot with some ports failing at a much lower voltage than others.
Similar effects can occur on all types of testing where an unknown,
that is uncontrolled, variable has significant effects on the outcome
of the test. The example used here of an RJ-45 plug and a static charge
on the cable is just one possible example.
If the effect of the uncontrolled variable is more subtle than my
example here, real trouble can result. The bad data may not be realized
as such and result in incorrect decisions on designs or errors in
standards. I have seen both.
Uncontrolled variables in a test can lead to unacceptable uncertainty
in the outcome of the test. A sign that this may be happening is data
that seems to make no sense. Subtle errors may be even more dangerous.
Other articles on this website and other sources related to this topic are:
If you like the information in this article and others on this website, much more information is available in my courses.
to see a listing of upcoming courses on design, measurement, and
troubleshooting of chips, circuits, and systems.