Howard Johnson ([email protected])
Thu, 31 Oct 1996 08:56:53 -0800 (PST)

I received this question following one of my
recent seminars, and thought it was of general
interest. Would anyone else care to take a
crack at generating a response to Mr. Etamad?

>My question is regarding shielded cables, when only one side of the
>shield is grounded as oppose to both side. How much difereces in shield
>we may have? Would like to know your thoughts.
>Thank you,
>Hamzeh Etemad ([email protected])

My answer:
- - - - -
I assume you have two boxes, connected with a shielded

When the shield is connected to ground at only one end,
you have made an *electrostatic* shield but not a *magnetic*

If your cable includes any high-impedance wires
carrying a large dv/dt, the electrostatice shielding
effect helps to terminate their electric field lines
emanating from these wires, thus attenuating emissions.
In audio products, operating at impedance levels of 1K
or 10K ohms, this type of shielding is effective. In
digital products, you rarely have such signals.

If your cable includes any low-impedance wires
carrying a large di/dt, the magnetic shielding
effect helps to provide counter-circulating current paths
which cancel out the magnetic fields produced by your circuits,
thus attenuating emissions. In digital products,
most near-field radiation is magnetic in nature.
Digital products most often require a magnetic shield (that is,
a path for counter-rotating electric currents) to combat emissions.

There is a relevant picture on page 315, Figure 9.14 of the book
"High-Speed Digital Design" (ISBN 0-13-395724-1).
It shows current going out on one set of wires, and returning
along the shield. In your design, you probably have some additional
explicit ground return wires in the cable
bundle (or maybe you are using balanced signaling). In either
case, the function of the shield is the same.
The shield serves to provide an alternate,
low impedance path for stray return current to get back to its source,
staying tightly bunched near the outgoing signals.
You see, some of the current you send to the other box will
return to its source along the wiring you have provided inside the cable
bundle, and some won't. Anything that doesn't return in the cable bundle still
has to get back to its source somehow. If you don't provide a path for
it to flow along the shield, it will take some other path (probably
a big loop going from the destination box, capacitively coupling
through the earth, and back to the source box). That big
loop kills your emissions performance.

The purpose of grounding the shield at both ends is to provide
a better path for this stray returning signal current.
When you disconnect the shield at one end it's like cutting the
drain wire in Figure 9.14. There's no way for currents
to pass through the shield. It won't work.

Conclusion: your shield, if terminated only at one
end, will provide little or no shielding effect.

Unfortunately, grounding two boxes together with a metallic
shield creates some U.L. safety problems. You must ensure
that both chassis connected to the shield are served by a
Green wire ground connected to the same Earth potential.
Green wire ground stakes located in various parts of a building
can have different 60-Hz potentials (as much as a few
volts). Interconnecting a metallic shield between two
chassis connected to different Green wire ground systems
can result in noticeable sparks, and can force
current to flow in the Green wire. That's a violation
of safety principles. In the absence of safety faults,
no current is ever supposed to flow in the Green wire.
Tracing Green wire current is a primary tool used to
debug electrical safety faults.

If both boxes are part of the same rack-mounted chassis, you're
OK. If both boxes are in the same room, provide a way for one
to plug into the other to solve the problem.

- Bye
Dr. Howard Johnson, Signal Consulting, Inc.
16541 Redmond Way, Suite 264, Redmond, WA 98052
U.S. tel (206) 556 0800 // fax 206 881 6149 // email [email protected]

***Please note our new company name and domain name***
(formerly Olympic Technology Group, Inc.)