Hi I sorted of had an epiphany when thinking about EMI shielding a PCB...

If the frequency of the system on a PCB was constant ( lets 20MHz digital system or something) could I take advantage of that fact and design a trace around the perimeter that will prevent that frequency from being transmitted in or out?

For example a circular trace with a diameter that is half the wavelength, an EM wave would induce a current in the loop equal and opposite to itself causing the net energy seen inside the loop to be 0. Right?

Of course this only applies if the wave is polarized in the correct direction but since all the PCB traces are also only polarized that direction it should not matter.

(I know for MHz a half wave is fairly large and impractical but just ignore that part of it)

  • 1
    \$\begingroup\$ Besides that you never have only one frequency in a system, doing that in the GHz range seems not feasible, as well as what about radiating perpendicular to the pcb? \$\endgroup\$ – PlasmaHH Nov 27 '17 at 12:16
  • 2
    \$\begingroup\$ If the emitted EM waves are 100% predictable and if you could generate the exact opposite and if the waves emit the same in all directions then yes, this could theoretically work. But in the real world none of this is true so it is not going to work. Metal shields do work so that's why they're used. \$\endgroup\$ – Bimpelrekkie Nov 27 '17 at 12:28
  • \$\begingroup\$ Good luck in predicting the antenna pattern for the RF, most guys I know that do antenna's get excited when their model matches the antenna better than 50% \$\endgroup\$ – Voltage Spike Nov 27 '17 at 22:43

I worked for a company that put ground layers at top and bottom and stitched them together all round the board perimeter, trying to make a Faraday cage. The product still radiated a lot. Turned out to be the 68000 chip die was radiating through the plastic package which was picked up by the adjacent power cord and emitted from there.
I think EMC follows Murphy law: for some reason it always radiates where you do not expect it.
You should have been suspicious: if it was that easy to solve EMC everybody would have done it. :-)

  • \$\begingroup\$ Sad truth. No matter what You do on PCB if You forget about wires they will be great antennas. \$\endgroup\$ – Jakub Rakus Nov 27 '17 at 12:46
  • 2
    \$\begingroup\$ Obvious lies. The 68000 is the greatest chip ever made, it would never cause any problems. \$\endgroup\$ – pipe Nov 27 '17 at 12:50

You are making the assumption that the generated EMI from the loop is linear across the loop, it isn't. It is stronger closer to the loop and weaker in the centre of the board.

Add to that the fact that whatever noise the rest of the board is making will be at any number of base and harmonic frequencies distributed and concentrated at various points across the board, it is impossible to noise cancel everything.

Further, your assumption that everything is on plane is also incorrect. The parts are on a different plane and pins and vias are all on a different axis. Also, unless this thing works in isolation, you will have cables attached to it going somewhere.

As such, although you may be able to attenuate it a bit, cancelling what is inside the loop is impossible. Worse, you may actually set up a beat frequency that radiates worse than having nothing.

You also have to remember the field generated from the loop is not just on the inside of the loop, but all around it. That is, on the outside too. It may be real quiet on the board, but outside the box it will be humming.


In very round numbers, pc boards are planar structures, while radiated emissions are spherical.


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