1
\$\begingroup\$

When using a PCB mount EMI shield, can you put decoupling capacitors on the back side of the board?

Well of course you can put them there, but will it make the shield essentially useless? Or will the effect of placing the capacitors on the back be unnoticeable?

An answer that could quote dB/uV results from an actual measurement on any design where decoupling capacitors were placed opposite the shield (for a power supply, CPU, or anything else) would be great.

For context, I am putting the shield around four switching power supplies.

1) +15V to +5V/250mA which is LT8606IMSE running at roughly 500KHz.
2) +15V to +3.3V/250mA which is LT8606IMSE running at roughly 500KHz.
3) +15V to -15V/50mA which is ADP2360ACPZ-R7 running at variable frequency up to 450KHz.
4) +15V to 7.5V out 500mA which is LT8611EUDD running at about 900kHz.

There are several capacitors 0603, 0805, 1206 size that I would like to place on the bottom to make the design more compact.

Since my EMI shield is surface mount I could try and put a footprint for one on both sides as a fall back plan (although I am not sure how well that would work out in the reflow ovens).

\$\endgroup\$
3
  • \$\begingroup\$ Insufficient info. What is your present emission level dBuV? what is the loop circumference of exposed bottom current loop and rise time, Ipk? Generally it is long tracks and gaps and not just PTH pads to caps and ground plane that cause issues. Do you have any measurement tools? \$\endgroup\$
    – D.A.S.
    Commented Dec 30, 2019 at 2:31
  • \$\begingroup\$ @TonyStewartSunnyskyguyEE75 Thanks for the comment. Because the board has not been designed yet I can't make measurements, and also I don't have tools at the moment. The idea is to place the SMD capacitors directly under the chips. In this case the loop runs through the board rather than on its surface. The loop area would be the trace length (including SMD capacitor body length) x the board thickness. This is about 150 mil x 63 mil. From the graphs in the datasheets the rise rates look to be about 1 to 2V/ns on the switching nodes. \$\endgroup\$
    – user4574
    Commented Dec 30, 2019 at 3:55
  • \$\begingroup\$ Thats a pretty small area of emissions up to 350MHz. No Sweat. \$\endgroup\$
    – D.A.S.
    Commented Dec 30, 2019 at 4:17

1 Answer 1

Reset to default
1
\$\begingroup\$

It seems to me that you are kind of asking how you should make the topping of a cake so that the cake will be excellent.

You might not need an EMI shield at all if you take care on component placement and layout. In general, proper signal routing has much greater influence on EMI reduction than shielding. So the shield is the last thing you should care about.

So my answer is: Place the caps where they are best for the circuit, then care about the shield.

\$\endgroup\$
2
  • \$\begingroup\$ I sort of agree with this. My goal is to have low noise, so its better to minimize how much I generate to begin with. I don't know if I will need a shield or not. In past layouts it has been a lot of work to rip up stuff after the fact to try and make space for a shield on a second version of the PCB. I think that ultimately I will just make the layout as tight as possible, and if that results in double sided layout then I will put a shield footprint on both sides so that its there if I need it. \$\endgroup\$
    – user4574
    Commented Dec 30, 2019 at 18:32
  • \$\begingroup\$ Most important for buck type switching regulator is the layout of the input capacitor C_in, the switch and the diode and a proper star-GND concept. If you do a very good job here, then a shield might not be necessary. If you do a bad job here, the shield will probably be useless because of the traces leaving the shield radiating EMI. \$\endgroup\$
    – Stefan Wyss
    Commented Dec 30, 2019 at 20:53

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.