11
\$\begingroup\$

I am working on a PCB that is very crowded, and has high gain amplifiers working between 300kHz and 500kHz

Typically I would use Mu metal or similar for shielding at this frequency, but obviously nobody makes Mu metal PCBs. So I have a choice of solid or hatched pours. External shields are not an option.

I don't have any controlled impedance tracks.

My only worry is the high frequency AC magnetic fields. We use copper mesh shielding in our RF cages, which works rather better than I expected. I suspect this is due to the shorted turns.

I asked a couple of shielding companies, but they don't characterize their meshes for this sort of application.

Can someone point me to data that would indicate whether solid or meshed copper pours would perform better in this situation?

\$\endgroup\$
  • \$\begingroup\$ Numerous planes (GND or VDD) build up internal reflections and boost the magnetic shielding, if the magnetic fields are steady-state. You can test this, with several double-sided raw-stock (unetched) PCBs, a transmitter coil (with 50 ohm to prevent shorting the generator) and a receiver coil to scope or spectrum analyzer. \$\endgroup\$ – analogsystemsrf Feb 17 '17 at 4:51
17
\$\begingroup\$

My only worry is the high frequency AC magnetic fields

It's really all about a thing called skin depth: -

enter image description here

Graph taken from this wiki page

So, for example, at 100 kHz, copper has a skin depth of about 0.2 mm and this means a 1mm thick screen forms a fairly effective shield against magnetic fields leaking out or leaking in.

I don't think that (even) 2 oz copper on a PCB is going to be that good whether solid or hatched. 2 oz copper is about 0.07mm thick so maybe you will get a little attenuation.

At 300 kHz it's in that borderline area where you might get a reduction of a couple of dB but if you are expecting a few tens of dB then it's very unlikely.

At 500 kHz (where the skin depth is about 0.09 dB) you might see a 5 dB reduction. Having said that, every dB counts so it might just be enough.

\$\endgroup\$
13
\$\begingroup\$

Solid would perform better, all other things being equal, but perhaps not significantly better.

Since the 'holes' in your mesh will be a tiny fraction of a wavelength, the mesh should behave similarly to a thinner (higher resistivity) solid copper layer when measured from a relatively large distance away compared to the 'holes'.

The 'shorted turns' you mention are just eddy currents which will occur in either case.

\$\endgroup\$
  • \$\begingroup\$ Hi Sphero! Indeed, we are at about 650 meters, so it's all just a speck. I have a front end amp which is two stages of G=40 feeding my receiver, and I am worried that I will radiate enough to get picked up by the antenna, a ferrite rod. I can shield that from E field but I dont think that would accomplish much. \$\endgroup\$ – user103218 Dec 1 '16 at 16:26
  • \$\begingroup\$ So am I hearing that there really isn't any effective shielding here without ferrous material? \$\endgroup\$ – user103218 Dec 4 '16 at 3:30
  • \$\begingroup\$ 1mm aluminum could be almost perfect at 300-500khz. \$\endgroup\$ – Spehro Pefhany Dec 4 '16 at 3:38
0
\$\begingroup\$

Depends on whether you have repetitive sinusoids, or repetitive pulses with fast edges. For sinusoids, we are trained in the limitations of SkinDepth. But fast-edges is the reality for embedded systems; lacking theory, I take measurements of square-waves coupling THROUGH foil, and find 50dB attenuation with 150nanosecond delay....through the foil.

Here are solutions for standard sinusoidal interferers.

With poor control over the magnetic fields, you can reduce the victim's loop areas. Thus opamps with the least possible height above the PCB are best choices. No DIPs allowed. And run GND under the packages, to be right under the piece of metal to which the silicon die is attached.

For those Resistors and Capacitors, surround them with GNDed chunks of copper, to have Eddy Currents develop (are your interferers repetitive or transients?) and thus partially cancel. And have GND pours right under the Rs and Cs, to minimize the loop area; you need to tie the pours very close by to the upper GND, again to minimize the loop areas.

With repetitive magnetic interferers, with partial transmission (Skin Depth not doing much good) you will also get partial REFLECTION. Multiple planes under critical opamps/Rs/Cs will implement multiple magnetic reflections, and provide better shielding of fields approaching from behind the opamps.

With your frequency-of-interest being nearly 1MHz, the Opamp PSRR will be poor. Thus large capacitors on the VDD+/VDD- pins, with 10_ohm resistors to the central bulk supply are useful. The central power will experience lots of magnetic-field noise, and you want to use LPFs to greatly reduce that repetive noise. 10uF and 10 ohms is 100uS tau, or 1.6KHz F3db, a 50dB reduce in 500KHz trash.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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