2
\$\begingroup\$

I want to get some knowledge in RF PCB design. Therefore I have some questions about this, in order to make some clarification.

What is the purpose of the golden ring around the PCB, often seen on RF designs, like this?

bare PCB Image source

I suppose it is used as some kind of shield. But, what is it protecting - the board from the external EM sources or the environment from the board's EM radiation? Why is it golden and not masked (painted)? Are there some rules for designing such rings (e.g. ring width, shape, via mechanics, ...)? And, finally, in some cases, there are rings that are not completely golden, like around the CC3200 in the following image below. What is the difference?

TI CC3200 Launchpad PCB Image source

\$\endgroup\$
  • 2
    \$\begingroup\$ You can insert pictures into your question. The google searches you linked aren't really clear - no one here knows which "golden ring" in which picture you are talking about. \$\endgroup\$ – JRE Dec 26 '17 at 14:34
  • \$\begingroup\$ You should consider if the usual FR4 PCB material may be used for your circuit or a special HF PCB material with lower losses should be prefered. Those special HF materials are more expensive than FR4. \$\endgroup\$ – Uwe Dec 26 '17 at 20:40
  • \$\begingroup\$ @JRE - Agreed. I've now added the images "inline". \$\endgroup\$ – SamGibson Dec 28 '17 at 14:31
  • \$\begingroup\$ The rings in the first picture may be intended as landing pads for RF gaskets on a shielding enclosure. The fastener holes seem to indicate that the enclosure is of a robust construction and not a simple soldered tin-plated can. \$\endgroup\$ – KalleMP Dec 30 '17 at 12:26
7
\$\begingroup\$

I believe the first image you posted is meant to accommodate a machined EMI shield that is screwed down. You can see that it would provide EMI shielding as well as isolating the circuits from each other. The metal is plated so it makes contact with the shield (maybe just ENIG).

enter image description here

The second one you posted is for a cheaper TI part. It looks like this design could accommodate a soldered down sheet metal shield for EMI. These are much cheaper and if this small can be placed during automated assembly.

enter image description here

It's easier to take off the more expensive screwdown shield, but you can get some sheet metal ones with pull off tops. They also make board clips now that you can put on your board and snap the shield into.

\$\endgroup\$
5
\$\begingroup\$

The conductive ring goes to a shield. There should be lots of vias from the ring to an inner-layer ground plane. If the vias are not visible, they might be implemented as filled vias, and you can see them in an X-ray view of the board. The number of vias required is not easy to generalize, but I use a minimum of 10 per inch. These vias should have no thermal relief. In general, any two conductors on an inner layer that traverse from outside the ring to inside the ring should have two vias separating them.

In my circuits, the purpose of the rings is for mounting shields, as shown in the answer by Some Hardware Guy. Most often, the purpose of the shields is to provide RF isolation between the circuits. For example, shields keep the output of an RF filter or amplifier isolated from the input. Also, the shields help prevent external interference from switching power supplies, digital circuits, etc. Less often, shields prevent leakage of RF signals to the outside environment, for example to meet FCC interference requirements.

These shields don't do much unless there is some sort of gasket or other way to get a very low resistance contact all along the shield and the board. Just screwing the shield down really tight is insufficient. Seams are leaky. Since solder-mask is an insulator, it is kept off of the ring to allow for a good electrical contact.

The ring needs to be wide enough to allow room for the footprint of the shield and gasket.

The TI layout linked in your question looks like an RF circuit block that was copied into the design. It doesn't look like it is set up to mate with a shield. Here is a sketch from a brief article about designing radios into products: enter image description here

\$\endgroup\$
  • \$\begingroup\$ On the TI board it looks like the shield should be soldered in place. \$\endgroup\$ – Mike Dec 29 '17 at 8:19
3
\$\begingroup\$

As an aside, launching off TomAnderson's link, slowing the digital EDGES by 10:1, or 100:1 if you can, greatly reduces the interference with the radios.

You may be able to avoid shielding, if you slow down the edges.

If you need to survive external trash (FCC susceptabiiity testing) or merely local interferers (switch regs), you may need shielding.

Here is illustration of slowing the edges, showing the internal SPECTRUM of the interference. Using 10MHz clock with 1 nanosecond edges, the harmonic at 90MHz is 16 milliVolts amplitude. After running that clock thru a 10MHz RC LPF, which has 16 nanosecond time constant, the 90MHz harmonic is attenuated to 2 milliVolts, or 8:1, or 18dB weaker interference. At 900MHz, another 20dB reduction occurs. In the WiFi band, another 8dB occurs, for total 46dB, because you slowed both edges of the digital signals.

enter image description here

\$\endgroup\$
0
\$\begingroup\$

Lets examine RF interference. Suppose you have a 10MHz MCU and a 433Mhz radio link. Will the MCU clock interfere with the radio?

Yes. The 43rd harmonic seems to be the culprit. But the real culprit will be the Q of the radio's frontend LC circuits.

Slow edges on that 10MHz MCU clock will poorly correlate (or convolve) with the 433MHz circuits, even if high Q at 433.

Fast edges on that 10Mhz MCU clock will superbly correlate, and transfer energy, into the 433MHz circuits, whether lowQ or highQ. And the high Q circuits will ring for longer, overloading the front end low-noise-amplifier, causing blocking.

Thus we have identified 2 degrees of freedom for operating a radio near powerful digital signals: (1) ensure the digital edges are slow ramps; (2) ensure the radio's input LC circuits (matching, etc) are low Q, so the energy injected by correlation/convolution will also be low, or will dampen out in just a small portion of a bittime; (3) reduce the bitrate of the link, so blocking by induced LC ringing will not trash an entire bit's energy.

\$\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.