On Thu,4/16/2015 8:48 AM, Don Jones wrote:
Sorry for taking to the field or sharing here what works for me in my
EMC/Lightning Lab on a daily basis. We make tens of thousands of
dollars every year off those who fail to ground properly. That figure
does not include the cost for damage to test samples and careers. Over
the years I've had several know it all's take up software engineering
instead.
It is my experience that the remote ground rods help in several ways,
including CM suppression.
The ONLY benefit of a connection to earth is to provide an alternate
path for current on the chassis that chassis that should not be there in
the first place. The earth is NOT a sump into which noise is dumped.
An example of when a ground rod can help is a generator, like a Honda
2000i, that puts out common mode noise at HF. Bonding the generator
chassis to a ground rod gives the current somewhere else to go other
than the power cable. BUT -- that function is also easily served by a
common mode choke formed by winding the power cable multiple turns
around a suitable ferrite core. Indeed, the combination of rod and choke
would work better than either alone.
And here's the other factor -- soil conductivity varies widely from
pretty good to really bad. Here in the Santa Cruz Mountains, I've
measured hundreds of Ohms between two 8 ft rods driven their full length
about 8 ft apart. The top soil here is pretty loamy -- tall redwoods
have been shedding lots of duff for centuries -- but underneath is lots
of granite. The rod-to-rod resistance is far lower than the rod to earth
resistance. The resistance between the two ends of my 550 ft Beverage is
almost 50K Ohms. In sandy soil, it would be even higher.
A major cause of this fuzzy thinking is that we use the word "ground" in
so many different ways, including circuit common.
When we connect the shield of a cable to the shielding enclosure of a
piece of gear, we are NOT "GROUNDING" it, we are BONDING it to the
shielding enclosure, making the cable an effective shield. It does not
matter whether the equipment has a connection to earth. Without that
shield bond, the cable is not shielded. If the cable shield goes inside
the equipment and then to the chassis, we've got a classic "Pin One
Problem," which conducts signals into and out of the box. When we bond
the shield to the shielding enclosure, shield current goes to the power
system green wire -- it stays outside the box.
Neil Muncy, ex-W3WJE (SK) taught EMC classes for many years, and
published landmark papers on EMC. He discovered the mechanism we call
Shield Current Induced Noise and also what we call The Pin One Problem.
In his classes, he emphasized that the only reason for an earth
connection is lightning and electrical safety. And he told his classes,
"get yourself a six-pack, a comfortable chair, and a pair of really good
binoculars and set yourself down at the end of a runway at your major
airport, and CALL ME COLLECT when you see one of those jets taking off
trailing a ground wire."
I suspect that this "failure to ground properly" you're talking about is
really "The Pin One Problem" -- connectors mounted to a PC board but not
to the shielding enclosure. To get to the shielding enclosure, shield
current must travel along the equipment common bus (or "ground plane),
where it is then injected into gain stages at the whim of the circuit
layout specialist.
Remember that on that "ground" layer, at all frequencies above audio,
the current of a circuit trace returns in a narrow region directly under
the trace, forming a transmission line. AND -- if that "ground" layer
happens to be broken under a trace (to add a trace that wouldn't fit on
another layer, for example), that return path is broken, so the return
current flows all over the box, creating a large magnetic loop and
antenna that couples into circuitry at random.
73, Jim K9YC
Vice-Chair Working Group on EMC,
Standards Committee of the Audio Engineering Society
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|