0
\$\begingroup\$

I am laying out a pcb with a number of external connectors that will have RF common-mode filters at each connector, shunting any common-mode signals to the chassis. However, I'm not sure what is the best way to connect the filters to the chassis to minimize radiated noise within the enclosure.

  1. There are a number of standoffs that could be used for this, but should all of the CM currents from each filter be run through traces and to a single chassis connection? Or should I use small isolated ground islands and connect them to separate, local standoffs that way?
  2. If I should use a trace to a single point, should it be routed in a particular way (I was thinking along the edge of the board)? Any tips on how to do this?
\$\endgroup\$
5
  • \$\begingroup\$ Will this cable be exposed to RF or ESD or arc weld noise that you do not want that current to conduct on signal grounds or radiate near high impedance inputs? If so then shunt cap or feedthru /w cap needs to be nearest entrance to case or ground input. Or is it just line f ingress reduction? Otherwise the CM choke may just Raise CM higher than DM impedance up to some f where they are equal. Z. Need more details. \$\endgroup\$
    – D.A.S.
    Commented Nov 30, 2017 at 21:18
  • \$\begingroup\$ Yes, the first one, don't want cables acting as antennas and picking up a bunch of RF to get into signal/signal grounds. I'm using the CM filter shown in this post: electronics.stackexchange.com/questions/342564/… \$\endgroup\$
    – User7251
    Commented Nov 30, 2017 at 23:07
  • \$\begingroup\$ That post not this has any specs, yet a detailed design with improper source and load impedance. Thus erroneous attenuation and response with poor rejection in UHF band. \$\endgroup\$
    – D.A.S.
    Commented Nov 30, 2017 at 23:42
  • \$\begingroup\$ Ok, so what I take from that is the answer to one of my questions in the other post, that simulating a filter without source and load impedances is basically meaningless. \$\endgroup\$
    – User7251
    Commented Nov 30, 2017 at 23:45
  • \$\begingroup\$ Take a look at how Ethernet rejects stray RF with the CM choke and 1nF cap to ground to avoid AC ground loop hum. \$\endgroup\$
    – D.A.S.
    Commented Nov 30, 2017 at 23:53

1 Answer 1

2
\$\begingroup\$

(I'm assuming that your chassis is, or is close enough to, a full enclosure. If it isn't — if the incoming wires/cables do not pass through it — things may be more complicated.)

You want to minimize the inductance in the chassis connection and minimize the capacitance to the rest of your circuit. Thus, the connection from filter to chassis should be as short as possible. Do not route it any distance on the board unless you must.

The best form of this is where the filter is itself chassis mount, such as an IEC socket with integral line filter, a feedthrough capacitor, or filter components soldered directly to a panel-mount connector. Failing that, give each filter its own direct connection to the chassis; do not join it to your board's ground plane.

There are a number of standoffs that could be used for this,

Standoffs are not a great choice because they are usually perpendicular to the signal routing, resulting in unnecessarily large inductive loop area. Instead, use a separate wire connecting to the chassis close to where the wire or connector enters.

If I should use a trace to a single point, should it be routed in a particular way (I was thinking along the edge of the board)?

A trace around the edge could be a poor substitute for a metallic enclosure if you didn't have one. You're trying to keep this signal off of your board — don't bring it closer than you have to.

\$\endgroup\$
1
  • \$\begingroup\$ Yep, as short, fat and direct as possible, you want to minimise the loop area inside the box. \$\endgroup\$
    – Dan Mills
    Commented Nov 30, 2017 at 21:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.