Design considerations for PCB to place RF shield

[madhu.wesly01]

Hello,

I want to place an RF shield/can for a PCB to reduce EMI, I am new to this topic.

Can anyone help me with design and layout considerations that I have to follow for placing the RF shield on the board?
For example, the distance between traces and solder mask opening of RF shield etc.

Thanks in advance

[CopperCone]

I made a thread about this recently but it turned into a derailing portapotty, you might find something useful there though

[madhu.wesly01]

I made a thread about this recently but it turned into a derailing portapotty, you might find something useful there though

I couldn't find any relevant information regarding my question

[mc172]

I don't really understand the question, especially the bit about solder mask opening in the context of traces? Surely your shielding can is going to touch the ground plane and stay well away from the traces... Can you post any pictures?

[madhu.wesly01]

I don't really understand the question, especially the bit about solder mask opening in the context of traces? Surely your shielding can is going to touch the ground plane and stay well away from the traces... Can you post any pictures?

Hello,

I want if there are any layout considerations I have to follow for placing RF shield on PCB

And what is the optimum clearance between components and shield ?

Thanks in advance

[CopperCone]

The tighter the shield is, the higher the high pass frequency on the waveguide will be. Each part is like a little launcher section I think.

I thought about it this way:

How much of an effect does putting a ground plane on the bottom and the top of the circuit have? I usually at least have a ground plane.

If you have two ground planes, does a extra box do anything really, between the planes? The capacitance of an enclosure even if you don't consider waveguide stuff is small considering the distance comparison between even a SMD part compared to a normal PCB thickness, even if you don't have a ground plane sandwich.

The dielectric constant of a PCB is much higher then of air, so I imagine even things like stripline would not be effected too much.

I figure the worst case of extra ground capacitance (im not sure you can even think of it this way too well) would be if you use tiny SMD parts and literary sandwich it down. But, the effect is uniform on the entire PCB, I think.

So, the effect would be that there might be a bit more coupling to ground. Everything should be in the near field (if you can think about it this way and not wave guide, I don't really know whats correct), so I don't think you would really have issues with things like reflections. I think the most imporant thing would be to keep shield impedance low/well matched to ground plane.

I do wonder if there is a case to analyze where the ground plane is really wimpy and the shield has a lower impedance, so you get some kind of significant internal shield current.. what comes to mind is datasheets that talk about RF isolators, where they recommend a floating capacitance between both sides to I guess reroute/concentrate emissions. But on a properly designed board the current should want to go into the ground plane more then anything I think. If the shield has crappy impedance then something might use it to capacitvely jump to another area of the PCB like in the recommended application of floating capacitance layers for use with RF isolators.

It gets more complicated with magnetics I think, since you start to introduce mutual inductance. In this case I have read on the forum *such as in my LISN threads* that you should really have good spacing between the magnetic elements and the enclosure, I think teslacoil said some kind of rule of thumb of distance related to magnetic shielding. It was basically some kind of factor related to the dimension of the inductor to reduce the mutual inductance to low levels.. I believe that following this rule basically doubled my enclosure size.

I bet there are other considerations but I think this is a somewhat meaningful basic analysis.

If you look at a RF can ground, you will see a higher end ones or ones for higher frequency use basically make a VIA sandwich. I suspect that unless you have some kind of shitty ground then basically your performance will get better the lower the impedance is between the shield and your PCB reference. I would suspect that if you have to do something funny and nonintuitve then something is wrong with the circuit or its really really low noise.

I suspect that the biggest problems might be complicated IC's that have fast internal clocks or multiple frequencies for other reasons to very high levels, but I don't know much about fast IC's. A older PIC will have emissions into the GHz range for instance, so if its a really big enclosure maybe you can have unexpected behavior from reflections /radiation. I would imagine a good solution to be to put a ferrite pad ontop of such devices that pollute in such ways, if they have a problem with the shield, or to generally dampen the whole thing.

I think you want to keep the shield as tight as possible and then use ferrite to minimize waveguide behavior?

Someone put this into better language please.

For instance  Capacitors also tend to be have a ground referenced shield (so they are directional). It makes me think that its fairly OK to just put something in a shield box and assume it will generally work better. There are also shielded transistors:
https://www.radiomuseum.org/tubes/tube_af114.html
http://pdf.datasheetcatalog.com/datasheets/90/74138_DS.pdf

In spectrum analyzers I have seen shielded transistors on a double ground plane backed PCB inside of another can (for a logarithmic amplifier)

I want to know why that guy got worse performance from his PLUTOSDR when he shielded it though. I want to know how that shield was grounded.

[chrisl]

I don't really understand the question, especially the bit about solder mask opening in the context of traces? Surely your shielding can is going to touch the ground plane and stay well away from the traces... Can you post any pictures?

Hello,

I want if there are any layout considerations I have to follow for placing RF shield on PCB

And what is the optimum clearance between components and shield ?

Thanks in advance

What type of shield are you using? One that solders down to the board or bolted down?
I usually leave 5 to 10 mil solder mask opening on the each side of the shield wall. Then I add extra 10 mil of spacing for the component placement.
If you are using a tall solder down version of shield you will have to add more spacing for the pick-n-place machine.

[mc172]

Jesus Coppercone, are you on speed or something?

The only way to know how much clearance you need between the can and the components is to experiment. Build a prototype can out of random pieces of copper or brass sheet and do some before and after measurements.
The closer you put the can to the components and traces, the more shunt capacitance you add, so you don't want it too close. I can't give you a number.

As for design considerations, you will need to solder the can to the ground plane in as many places as possible. At what interval it is soldered to the ground plane depends on the operating frequency, but it's a good idea to keep the distance between solder joints as short as possible.
For tracks exiting and entering the can, the clearance between the can and the track can be important. You might have to create a small slot to get the tracks in and out. Don't rest the can on the track, especially if running high power, but you might get away with it if you have to for low power because the length of the low impedance section and amount of added shunt capacitance will be quite low. But if your circuit is critically matched or just on the edge of being acceptable with no can, you might end up wrecking it.

Just try it.

[ogden]

I want to know why that guy got worse performance from his PLUTOSDR when he shielded it though. I want to know how that shield was grounded.

RF "shield" works both ways. By adding shield that guy 1) shielded external interferers - as intended 2) "canned" internal interferers - created new leakage path(s) between internal parts of radio.

Either put each individual part of radio (LNA, VCO, mixer/downconverter, digital_part) into individual shield or use proper absorbing foam inlay. For single board SDR's obviously there's only "foam way".

[CopperCone]

But the shield has a really low impedance, wouldent that only be true if it was floating or had significant impedance? I just see it being a ground, so it should not allow leakage between circuit elements, only leakage to ground?

[CopperCone]

Can you design some kind of simple sensitive circuit/experiment that I can test this with?

I want to see and take measurements on a system that is negatively effected by a overall shield and requires shield segregation or nested shields OR I guess a circuit that requires two nested shields or a circuit that can't be reasonably shielded (i.e. requires ridiculous shield dimensions)

so I can take measurements with the nested shield removed? I have plenty of copper scrap to make shields with and quite a bit of parts.


at the lowest frequency where its possible to see this effect so its easier to build and I don't second guess myself a bunch thinking about construction quality. i.e. does not require 15GHz

[mc172]

But the shield has a really low impedance, wouldent that only be true if it was floating or had significant impedance? I just see it being a ground, so it should not allow leakage between circuit elements, only leakage to ground?

It doesn't have really low impedance at all frequencies, though. It has to be made, usually very cheaply, and fitted to the board. Naturally you have cracks, slits, holes, everywhere and not a constant solder joint all the way around. Thus it's a waveguide, slot antenna, tuned resonator, shunt inductance, shunt capacitance, and etc. and combinations of some or very few of those at all different frequencies. In reality the screening effectiveness of what looks like a decent shielding can, can be all over the place for broadband devices.

By shielding a large circuit "block" (say a local oscillator and a low pass filter in one can) you might see worse performance, especially if high isolation is required. The can might promote better coupling between (hypothetically) the LO and LPF output, undermining the LPF. I'm making this scenario up off the top of my head, not sure why you would want to do that in reality.

When performance degrades, it could also be a result of the addition of the fairly evenly distributed shunt capacitance that is added with shallow shielding cans.

A fairly simple way of messing about with this is to make a high pass or low pass filter (your choice - low pass might be better because people normally have an upper frequency limit on their test gear that is approached sooner than the low frequency one) with around 30 dB of rejection on a piece of copper clad board, and stick a screening can around the lot. Use air coils you can tune.
What you'll hopefully see is that if you bring the screening can closer to your air inductors, you get less rejection, and knackered return loss. The return loss is the shunt capacitance and absorption, the rejection is the coupling.

...But you might not see any coupling effects. Impossible to say!

[CopperCone]

I will run some experiments when I clean up my area. Do you think any particular passive lumped filter topology might be effected the most?

[mc172]

Just to keep it simple, try a LPF with only series inductors and shunt capacitors, no resonators and start and end with a series, not a shunt, element. So let series elements equal n, and shunt elements n-1. Perhaps 4 series and 3 shunt. You can then add shunt elements at each end to get the match if required.

The topologies that will be affected the most will be ones with LC resonant lumps everywhere, but lining the thing up will probably have you reaching for the Scotch.

Use this program to get realistic values:

http://www.iowahills.com/9RFFiltersPage.html

[mc172]

BTW - the higher the frequency the better but something realistic might be a few hundred MHz. What gear have you got?

[ogden]

Can you design some kind of simple sensitive circuit/experiment that I can test this with?

Best way to learn about radio wave propagation outside and inside cavities - using 2.4GHz transceiver modules (like CC2500 or NRF24). For really close distances better use module designed for external antenna but don't connect any or just desolder on-board antenna coupling cap. Obviously set low TX power. Get three, not two modules

[CopperCone]

BTW - the higher the frequency the better but something realistic might be a few hundred MHz. What gear have you got?

26.5ghz spectrum analyzer, 1.5GHz spectrum analyzer, 300MHz HP vna, buncha function generators that go to 10MHz, 100MHz pulse generator, 100MHz oscilloscope,18GHz sweep/signal generator, some tone/function generator stuff under 1 MHz, buncha different signal generators from 10-50MHz, power supplies, whole bunch of microwave amplfiiers/filters/diplexers/other shit for microwave and a bunch of low noise/precision classic analog stuff that everyone on this forum seems to have that is not really fast.

I also have a 100-500MHz 50W amplifier and some more modest traveling wave tube stuff between 1-20W

I do have a few directional couplers good to high frequencies so I can run this experiment in  the high GHz but I doubt my construction will be up to par, I did think to do  this with a PCB filter, HP 11691D

my interest kind of goes all over the place so I accumulated alot of shit that I have not gotten my moneys worth from yet.

I also have some counters good to like 2GHz if something related to detuning a oscillator can be used for this experiment