What's the most effective way to prevent EM noise from coupling onto a cable and radiating out?

Question

I've a PCB radiating noise inside a metal enclosure (this is Product 1 from this question here). Unfortunately, I can't change the PCB itself, so I'm having to tackle the problem by other means. I can shield the metal enclosure to contain the noise, but I also need to deal with the cables exiting the enclosure. I've got a few questions regarding the best way to prevent the noise from coupling to the cables:

1. Regarding the cable shields, is it best to connect the shield to ground at the point at which they connect to the PCB, or at the point at which they exit the enclosure?
2. I'm guessing ferrites are best placed right at the point at which the cable exits the enclosure. Is this correct?
3. Does it make much difference placing the ferrite on a shielded cable compared to an unshielded one?
   3. a. Is it best to place the ferrite before the cable shield starts, or after?

Answer 1

1. Shields must generally be connected, across 360 degrees of the opening, directly to the enclosure. Any connectors and connector shells used must maintain such multipoint, all-around connection.

   Shield stubs, drain wires, and single-point connections have parasitic inductances that convert the shield into a tuned antenna circuit - you really don't want that!

   Be especially wary of poor contact along the circumference of the connector. When using D-Sub connectors, ensure that the male connector has multiple indentations around the shell, so that there is a multi-point contact around the circumference of the female connector.

2. Ferrites placed on the entire cable are common-mode filters. You must have some reason to believe that the radiation is due to common mode noise. Given that the frequencies involved have meter-long wavelengths, it'll work just as well to run the conductors exiting the connector, inside the enclosure, through a ferrite - ideally try to squeeze a couple of turns. The little stub between the ferrite and the connector will not be a good enough antenna to pick up the noise radiated within the enclosure.

3. The shield must be continuous with the enclosure, so the (a) subquestion doesn't apply. The filtering efficiency depends on how the currents split between the shield and the internal wires. If they split perfectly equally, then the ferrite will have no effect, since the net common mode current is zero.

Also note that double braided shields have more than twice the performance of similar single braided ones. This of course makes difference when you maintain proper shielding continuity.

All of the above is very general - as others have said, it's very hard to talk about the details without seeing some pictures and board layout.

Answer 2

Nothing concrete without seeing circuits and PCB layouts: -

1) If the metal box is the most direct connection to earth then the cable shield is best connected directly to the metal box because the PCB "earth" may not be exactly at earth potential due to feed wires for power.

2) I would consider placing ferrites as close to the source as possible, that being the PCB.

3) I don't think it makes much difference but if the interfering signal is predominantly either common-mode or differential then it might.