Metal enclosure grounding

[Mr Harry]

You don't say how the power supply circuits are arranged. If your circuit
ground is connected to one side of the mains, you absolutely must not
connect it to the box.

Mains comes into the grounded enclosure directly into the smps. the
output of the smps goes directly into the circuit board which does
indeed have digital connections to the outside world. the circuit
board does not have any connection to the enclosure.

i have spoken to the smps manufacturer who states it's ok to ground
the 0v output. however, they also state that the overall product could
possibly fail emc testing if i do this due to increased noise on the
earth wire.

i have taken the smps apart and there are indeed capacitors between 0v
and ground.

would a resistor (instead of a direct link) between 0v and ground
reduce any noise on the earth wire?

harry.

 

[Ken Smith]

Mains comes into the grounded enclosure directly into the smps. the
output of the smps goes directly into the circuit board which does
indeed have digital connections to the outside world. the circuit
board does not have any connection to the enclosure.

Chances are, you don't want any connection at the PCB to the enclosure.
If the chassis ground is connected to the power cord's earth connection by
a short wire near where the power cord comes into the box, this usually
gives the best EMI results.

The thing you are trying to avoid is causing RF currents to flow through
the housing.

i have spoken to the smps manufacturer who states it's ok to ground
the 0v output. however, they also state that the overall product could
possibly fail emc testing if i do this due to increased noise on the
earth wire.

This implies that the power supply puts a common mode noise out its output
wires. This is a common (pun intended) problem with switchers.

i have taken the smps apart and there are indeed capacitors between 0v
and ground.

Remember this RF we are working with here. Two points on the sheet metal
can be at very different voltages at high frequencies. Chances are, the
maker has put the capacitor at the point they discovered worked the best.

would a resistor (instead of a direct link) between 0v and ground
reduce any noise on the earth wire?

Putting resistors between grounds is an excelent way of reducing the
ringing etc that can cause tall peaks in the radiation. You can also use
high mu cores are RF-beads.

If you twist the output wires of the power supply together and run them
through a core, the impedance for common mode signals will be increased.
You can also to this with the digital lines. Since the AC signal and its
ground go through the same hole in the same core, there is no field in the
core and hence no differental impedance increase.

 

[Mac]

Chances are, you don't want any connection at the PCB to the enclosure.
If the chassis ground is connected to the power cord's earth connection by
a short wire near where the power cord comes into the box, this usually
gives the best EMI results.

The thing you are trying to avoid is causing RF currents to flow through
the housing.

Right, so you don't want the housing to be part of the circuit. So I would
try to provide a short, low impedance connection from the power supply
"ground" to the earth ground so the current doesn't have to go through the
box.

This implies that the power supply puts a common mode noise out its output
wires. This is a common (pun intended) problem with switchers.


Remember this RF we are working with here. Two points on the sheet metal
can be at very different voltages at high frequencies. Chances are, the
maker has put the capacitor at the point they discovered worked the best.


Putting resistors between grounds is an excelent way of reducing the
ringing etc that can cause tall peaks in the radiation. You can also use
high mu cores are RF-beads.

If you twist the output wires of the power supply together and run them
through a core, the impedance for common mode signals will be increased.
You can also to this with the digital lines. Since the AC signal and its
ground go through the same hole in the same core, there is no field in the
core and hence no differental impedance increase.

FWIW, I believe it is still standard practice in most places to connect
chassis ground to circuit ground anywhere that connectors take signals off
board. At the place I used to work, they created a chassis ground net that
went all around the circuit board on the edge where the off-board
connections were, and they treated all signals before they passed over
this chassis ground fringe. "Treating" means that any signals which could
be subjected to low-pass filters were low-pass filtered, and any
differential signals went through a common-mode choke and/or other
magnetics. Right at the crossover point, they usually put a capacitor
bridging from chassis ground to circuit ground. There was also a ferrite.
This seemed to work pretty well. I was only designing circuit boards, so I
don't know how they would have connected a power supply to earth ground,
but I bet they would have done it. And our equipment usually didn't have
too much problem passing FCC limits.

Also, I recently attended a one-day seminar on EMI/EMC issues sponsored by
Underwriter's Laboratories, and they seemed to advocate good grounding.
Good meaning low-impedance connections. They made the excellent point that
one net can't have noise on it. There is always a reference. And If you
can keep circuit ground at earth ground, and bypass all other voltage rails
to circuit ground, then you shouldn't have to worry about power supply
noise.

Note that I definitely agree with Ken that you don't want to allow stray
currents to traverse your box. And I can see how that could happen if you
connect the power supply output ground to circuit ground with an inductive
lead. So I am not proposing that you do that. I think you should connect
the power supply to the earth ground with the shortest possible
connection. Then connect circuit ground to chassis ground as I outlined
above.

By the way, not everyone at the seminar was buying what they guy was
saying. For example, he also said that cable shields should be connected
to the grounded chassis on both sides, largely using the same logic of
having only one ground.

Not everyone was convinced, because too many people there had seen bad
scans which, when they disconnected a shield from a cable, got better. The
guy leading the seminar said that that is indicitive of a more serious
problem somewhere else.

I think I have rambled a bit, but I hope I made at least one or two
coherent points.

regards,
Mac

 

[John Woodgate]

By the way, not everyone at the seminar was buying what they guy was
saying. For example, he also said that cable shields should be connected
to the grounded chassis on both sides, largely using the same logic of
having only one ground.

The idea is that the cable shield acts like a 'pipe' connecting two
'tanks' - no 'water', in the form of EMI, can leak out.

Not everyone was convinced, because too many people there had seen bad
scans which, when they disconnected a shield from a cable, got better.
The guy leading the seminar said that that is indicitive of a more
serious problem somewhere else.

Well, a problem, anyway. There are a number of such problems, that tend
to affect existing installations that are being modified rather than
well-designed new installations.

The first thing to realise is that this problem is, at first sight, an
*installation* problem, because more than one piece of equipment is
involved. However, the root cause MAY be a problem in one or more pieces
of equipment. Here are three examples:

* Input cable shield has a lower impedance connection to circuit ground
than to chassis ground, due to poor circuit design and/or construction
around the input connector. This injects hash from the shield into the
circuit, whereas it should go directly to ground via the chassis.

* Currents (may be quite large) circulate in the shield because the two
pieces of equipment are safety-grounded at different places. There may
be a ground potential difference, especially at power frequency, and the
loop formed by the safety grounding and the cable shield acts as a loop
antenna to pick up EMI.

In this case, remedial action on the installation itself may be too
costly to contemplate, so any solution that works may be acceptable. For
example, connecting the shield to chassis ground via a very low
inductance capacitor, and to circuit ground via a ferrite bead, may
work.

* emissions from one shield that is not grounded at one end can
partially cancel emissions from other cables or enclosures.