Shielding -- Is it effective when floating?

[jerry507]

This doesn't depend on the application, it just depends on how good you need it.

A Faraday cage works because, as others have said, it completely encloses an area or volume from an electric field. Because charge is free to move around the enclosure, it can move to cancel the internal field. If you have penetrations for cables and such, it will degrade the effect with the magnitude of degradation depending on many factors such as frequency and the geometric layout.

In this respect, a fully enclosed shield such as a braid or foil wrap will form a very effective Faraday cage around the signal. What I think you're missing is that while the volume inside the shield may be at zero potential, the shield itself is not restricted. In fact, the shield will be at the same potential as the electric field it's within. The Analog Devices article tells you to ground the shield because you don't want the shield to be at a potential. Ground it so it's potential is as close as possible.

Why? Well, your shield isn't infinite. Let's reference figure 5 from the AD article. Let's say that this three segment system is a RS-485 link inside a building (something I deal with every day) and then let's break it in different ways and see what the effect would be.

If we first remove the ground reference at the left hand side, we have three connected shields. The conductors inside the shield will be effectively shielded from a noise source that is located significantly far away from either end but now that noise can travel up and down the connected portion and couple into the conductors at the far left and far right.

Now let's instead break the connections between the shields, leaving the ground reference at the far left. Now noise can be injected at the middle of the middle shield and will radiate down the center shield. It is now free to couple into the conductors at the breaks between the shields.

Now the shield does have loss and so noise signals can be attenuated simply by having to travel along the shield. This may be enough to eliminate problems if the noise is small in magnitude or the shield is very long. Also, the shield is not a perfect radiator so the coupling to the internal conductors may not be that strong. But grounding it is the best solution to attenuating that noise signal.

[CaptnYellowShirt]

I think jerry nailed it.

My hang up was the difference between voltage and electric field:



The un-grounded shield may reject an internal electric field, but it does nothing to reject an overall potential with respect to some other (remote) part of the circuit.

[SArepairman]

So, how about lossy shields? Like those RF gaskets Dave did a blog about, some of them had a few hundred ohm resistance, or more. The foamy conductive things.

When the shield is lossy, is it better if its grounded?

http://www.eevblog.com/2013/06/04/eevblog-477-mailbag/

**********this is a great episode if yoou wanna learn about shielding technology btw.

[skipjackrc4]

Jerry is absolutely right.

The shielding requirements for low-frequency magnetic and electric fields and TEM waves (RF) are somewhat different.  The following is an extension of what Jerry said.

There are two primary mechanisms when shielding against RF--reflection of the incident wave and internal absorption due to what Jerry was talking about.

I'm not sure if you are familiar with the concept of electromagnetic transmission line reflection, but if not, here is the basic idea.  If a wave is traveling through cable of one impedance (say 50 ohm, which is the most common for lab use) and it encounters another impedance (such as an imperfect connector of maybe 48 ohm), part of the wave will continue to travel through the connector, but part of it will reflect back and travel down the cable in the opposite direction.  The amount of reflection depends on the differences between the two impedances.  This is not usually noticeable at low frequencies, but every time you hook a DMM up to a circuit, you are getting reflections.  The DMM is just too slow to see them. 

An EM wave has an impedance that depends on the material that it is traveling through.  This impedance is simply the ratio between the electric and magnetic fields, just like resistance is the ratio of voltage to current.  In free space (which is basically air), the wave impedance is 377.  In a conductor, this impedance is lower.  The better the conductor, the lower the impedance.  This means that when a wave hits the surface of a conducting shield, part of it will just reflect off. 

The portion of the wave that does not reflect is then attenuated as it travels through the material.  This attenuation is influenced by material thickness and conductivity, as well as the frequency of the wave.  An increase in any one of these three parameters will improve shielding effectiveness.  However, at very high frequencies, shielding effectiveness starts to break down due to the fact that it takes the free charges in the conductor a finite amount of time to respond to the high frequency incident wave.  If the incident wave is changing faster than the charge carriers can reorient themselves to oppose it, then the attenuation will not be as effective.

If you are interested in learning more, Dr. Clayton Paul has an excellent book called Introduction to Electromagnetic Compatibility that describes this process in more detail.  I believe it is at least partially available on Google Books.

[jerry507]

I read the article a bit more while munching on my lunch and realized that the article does a better explanation of the coupling at the ends. You can effectively think about the shield and the conductor as the two plates of a huge parallel plate conductor. If you're dealing with cable that actually specifies this, such as Belden 3105A (https://edeskv2.belden.com/Products/index.cfm?event=printPrev&pnum=3105A&ut=english), then this effect becomes pretty obvious.

As an aside, I have a friend who worked at a company who did a lot of cellular networking stuff. They had a Faraday cage with a mini cell tower inside that they would use to test their devices in isolation from the regular "live" networks. They had a lot of problems with the < 2 sq in slit that they used to run serial cables and coax through. Given that the models have a typical sensitivity of -110dbm, they would sometimes lock onto the outside networks instead of the intended one. It took a significant amount of fiddly work to seal that gap around the cables. Even when they did use metal fingers to extend the shield, the signals would couple in on the serial cables. Ferrites were the key to that problem.

[SArepairman]

I read the article a bit more while munching on my lunch and realized that the article does a better explanation of the coupling at the ends. You can effectively think about the shield and the conductor as the two plates of a huge parallel plate conductor. If you're dealing with cable that actually specifies this, such as Belden 3105A (https://edeskv2.belden.com/Products/index.cfm?event=printPrev&pnum=3105A&ut=english), then this effect becomes pretty obvious.

As an aside, I have a friend who worked at a company who did a lot of cellular networking stuff. They had a Faraday cage with a mini cell tower inside that they would use to test their devices in isolation from the regular "live" networks. They had a lot of problems with the < 2 sq in slit that they used to run serial cables and coax through. Given that the models have a typical sensitivity of -110dbm, they would sometimes lock onto the outside networks instead of the intended one. It took a significant amount of fiddly work to seal that gap around the cables. Even when they did use metal fingers to extend the shield, the signals would couple in on the serial cables. Ferrites were the key to that problem.

Did they stuff the opening with ferrite foam?


Would a box made of a lossy material like ferrite be a better equipment housing then a box made of something like aluminum? could it be worse in certain cases?

[skipjackrc4]

I read the article a bit more while munching on my lunch and realized that the article does a better explanation of the coupling at the ends. You can effectively think about the shield and the conductor as the two plates of a huge parallel plate conductor. If you're dealing with cable that actually specifies this, such as Belden 3105A (https://edeskv2.belden.com/Products/index.cfm?event=printPrev&pnum=3105A&ut=english), then this effect becomes pretty obvious.

As an aside, I have a friend who worked at a company who did a lot of cellular networking stuff. They had a Faraday cage with a mini cell tower inside that they would use to test their devices in isolation from the regular "live" networks. They had a lot of problems with the < 2 sq in slit that they used to run serial cables and coax through. Given that the models have a typical sensitivity of -110dbm, they would sometimes lock onto the outside networks instead of the intended one. It took a significant amount of fiddly work to seal that gap around the cables. Even when they did use metal fingers to extend the shield, the signals would couple in on the serial cables. Ferrites were the key to that problem.

Did they stuff the opening with ferrite foam?


Would a box made of a lossy material like ferrite be a better equipment housing then a box made of something like aluminum? could it be worse in certain cases?

The best way to shield enclosure openings is to run the cables through a waveguide structure.  Below a certain cutoff frequency, waveguides exhibit strong attenuation that is proportional to their length.  If you are trying to shield everything below 10GHz for example, you would use a waveguide with a cutoff of 25GHz (just throwing a number out there--the actual cutoff frequency depends on the length and desired attenuation).  It doesn't have to be an expensive commercial waveguide--copper pipe that is securely connected to the enclosure will work nicely.  Take a look at your microwave oven--each of those holes in the door screen act as little waveguides operating below their cutoff frequency.

As for the signal being picked up by the cables, ferrite beads are indeed the easiest way to handle it.  The lab that I work in has several shielded anechoic chambers, and all the RF and control lines are heavily ferrited (new word!).

Using ferrite for a shield material is effective against magnetic fields.  In this case, aluminum or copper are completely useless (with DC fields) because they are non-ferromagnetic.  Steel, ferrite, and mumetal/permalloy are all good for magnetic shielding, depending on the frequency of the field.  Against RF waves, you are better off with aluminum or copper for pure shielding purposes. 

Ferrite tiles and foam do have their purposes against RF, though.  Ferrite can be made so that it has a similar intrinsic impedance to free space (see my post above explaining this) which means that it absorbs RF energy without reflecting.  Ferrite tiles and foam pyramids are used in EMI anechoic chambers to prevent reflections from the walls.  A typical EMI chamber will be aluminum or copper on the outside for external shielding and have the inside walls lined with ferrite pyramids and possibly solid tiles for internal reflection damping. 

Here is some info from a ferrite absorber supplier if you are interested: http://www.djmelectronics.com/rf-absorber.html

[SArepairman]

I never thought about the impedance of free air before.

[CaptnYellowShirt]

I never thought about the impedance of free air before.

[T3sl4co1l]

^ lol

But yes, everything has impedance; and geometry shapes that.  Free air has some impedance.  A single wire in free space has a similar impedance, give or take radiation -- a 1/4 wave resonant wire has a lower impedance, but not by too much, because radiation damps the resonance.  Put another wire near it and the impedance between them becomes a function of their separation; put a solid round shield on it and you get coax; etc.

Or maybe you have a trace over a ground plane, which is a (microstrip) transmission line; suppose you wanted to short it out, so you drop a via to the ground plane.  Maybe you're bypassing plane-to-plane with a capacitor and as little trace length as possible: but the traces, capacitor and vias all have impedance.  The planes aren't even direct shorts: the impedance has an odd dependency with frequency, because the via doesn't connect to the entire plane at once, it connects first to the round zone it's touching.  The inductance of a plane goes as ln(R/r), for an outer radius R and via radius r.  You can imagine a very high frequency wave interacting with maybe the first 1/4 wavelength of the plane, while the rest acts like a short; as the wave expands over the plane, the apparent impedance drops and more and more gets reflected back at a progression of angles (the reflection is smeared out over time because it's dispersive).

In EMC testing, there's a reason they test against 50 ohms: because sooner or later, the power line is going to look like a 'random wire' antenna and radiate stuff, at which point it's going to be in the vicinity of 50 ohms.  Basically, the assumed load is one that hopefully radiates at all frequencies.

You might think your switching power supply is isolated and happy and, how can I even filter this if it's all floaty and doesn't even have an impedance?  (You can't just *make* a filter: a filter only does its job into a specified range of impedances.)  Ah, but those transformer windings have impedance, within themselves and to everything nearby!  The circuit board above a metal chassis (if you don't provide one, FCC Part 15 specifies a ground plane!) has an impedance.  So it's not as bizarre as it sounds, and in fact is a quite tractable problem.

Tim