Shielding ground - which side of my CM choke?

[jrs45]

I have a project I'm working on that is powered by 24VDC via an external adapter.  There is a CM choke isolating local board GND (and power) from the DC input wires.

I assume I should be connecting my shield to the local PCB ground, not on the other side of the CM choke (on the DC negative input)?  Seems to be the way things are done but I wasn't sure, since on-board noise on the GND planes and whatnot might cause problems?

[T3sl4co1l]

What shield?

Tim

[coppercone2]

if you should float the shield by the impedance of the choke or ground the shield at a elevated impedance?

[jrs45]

What shield?

Tim

The electrical shield - the enclosure box itself. I could connect it to the local PCB ground, or to the DC input's negative terminal (on the other side of the CM choke).  Pretty sure the first is "what's done" but I wanted to check.

[jrs45]

if you should float the shield by the impedance of the choke or ground the shield at a elevated impedance?

Something like that.  The PCB has its own ground, and this is isolated from the DC input by a CM choke for conducted EMI purposes.  For reducing radiated EMI, does the physical chassis shield go to the local PCB ground, or the DC input negative terminal (which is probably floating anyway, being fed from a switchmode power supply). 

Perhaps since the negative DC input terminal is floating anyway, it doesn't matter?  There is no earth present (or needed, no mains involved).  Seems people connect it to the local PCB gnd but I wasn't sure if that was always the best case, or why.

The PCB is fairly noisy, lots of current being switched quickly, so I want to get it right.

[T3sl4co1l]

What shield?

Tim

The electrical shield - the enclosure box itself. I could connect it to the local PCB ground, or to the DC input's negative terminal (on the other side of the CM choke).  Pretty sure the first is "what's done" but I wanted to check.

Yes, definitely PCB ground.  I wanted to make sure whether it was a cable shield or something else.

The PCB is fairly noisy, lots of current being switched quickly, so I want to get it right.

Oh...  So it will also depend on where PCB ground is.

Tim

[jrs45]

[/quote]

Oh...  So it will also depend on where PCB ground is.

[/quote]

Well, it is where it is.  It's a Class D audio amp, fwiw.  But there aren't too many realistic choices.  Either PCB gnd or the DC negative input, which is on the other side of a CM choke.  Thanks.

[penfold]

Make sure you watch out for the safety aspect also, if the DC supply is a battery, then less problem, if it is derived from a mains supply somehow, then connecting the case to the DC supply isn't a wise idea. "Grounding" the chassis via a low value safety rated capacitor will mitigate the "connected to supply" issue somewhat.

If you are worried about EMC emissions then planning to be able to chose whether you connect via inlet side ground, filtered ground or not at all would be a very good plan. If you are more concerned about incoming CM noise being radiated by your enclosure then filtered side ground is better, if you are more worried out your circuit causing problems then inlet side ground would be better. It would be quite easy to measure with an oscilloscope probe used as a near field probe when you're testing it. You may be surprised at which one will work better for you.

[T3sl4co1l]

Point being, even if the PCB layout has a ground plane, the shield should be connected away from noisy sources, and close to whatever all signals are referenced to.

Since you have three connections (power, signal and output), you can have common mode noise between any combination of them.  The best way to eliminate that, is to bring all three to a common location and apply filtering there.  So, the outputs have LC filters, the final Cs of which should be in this location; the input could have a CMC, and/or use a diff amp for some CMRR and range (and should be filtered to modest bandwidths anyway: only use what bandwidth you need!); and the power source could have a CMC, with filter caps in this location.  Finally, shield also connects in this location, preferably with wide connections: copper pours, EMI spring stock, multiple screws, etc.

When grouping all the signals together isn't feasible (often, input/power and output are on opposite ends of the board), the shield should connect in both respective locations, so that signals can be filtered to it as much as possible.  The ultimate goal is to have everything that penetrates the shield, filtered to that very same point.  This is easier to ensure when everything is grouped together.

Tim

[penfold]

Given that the choice is between the DC neg side and PCB gnd (DC neg after the CM choke), the noisy currents generated by the circuit will produce a voltage drop through the CM choke (even if it due to its diff mode impedance). My point is that without exact details on the circuit setup and where exactly the CM return path is, one cannot directly determine whether or not from a radiated emissions perspective if the pre or post CM choke DC neg will be noisier.

Even if the signal input and output is completely isolated and the best grounding in the world between them within the circuit, there will still be a greater noise voltage on "system" side of the CM choke simply by virtue of having some additional impedance from the choke and some noisy currents flowing. The situation would be reversed if the intention was to prevent radiation of common mode noise present on the DC input, but still, without a common mode return path, the CM choke won't be having its desired effect.