Coverage of Ferrite Tiles in EMC Chamber/Shielded Room

[TimNJ]

Hi all,

At work, we have a small shielded room (about 4m x 2m x 2m) that we use to run CISPR32 conducted emissions pre-compliance testing, and we also use as a controlled environment for radiated emissions debugging.

We are trying a new-setup for CISPR25 (automotive) pre-compliance conducted emissions testing. The main difference compared to what we were doing before, is that the DUT and its associated cables are placed above a conductive metal plane. Additionally, instead of a tightly bundled cable, the cables are spread out forming a relatively large loop area.

Compared to our previous CISPR32 CE results (150KHz - 30MHz, AC input), where there are no notable resonances or setup related issues, the CISPR25 (150KHz - 108MHz, DC input) results show some pretty severe resonances/peaking at 10MHz, 20MHz, and 60MHz. We have reference data from a real test lab (for this DUT), and no such resonances appear in their plots. Granted, there are some differences in setup. For instance, due to space constraints, we arranged the cables in a serpentine, whereas the standard wants the cables arranged in a straight line. This could have a big effect. Additionally, the test setup is positioned directly against the metal chamber wall, whereas the standard wants at least 1m. These type of issues require more investigation.

While it is foolish to think we can get 100% comparable results to a lab with a (probably) $500,000 chamber, I'm curious to see if we can knock down some of the resonances without also spending $500,000. 

----

So, here's my question, which will probably neatly highlight my lack of understanding:

>>>Would there be any value in using a small number of lossy ferrite tiles in the direct vicinity of the test setup? Take a look at my attached photo as an idea. The actual coverage of the chamber would probably be 5% or less.

My idea is to try to limit reflections "at the source". Intuitively, it seems that placing tiles near the emitting source would be more effective than on a far wall, but I wonder how much coverage you really need to make a dent. I guess I want to understand if the incremental improvement with additional tiles is 'linear', or if the incremental improvement is 'diminishing'. Does that make sense? Anyone have an idea?

The usual approach is to cover the entire room in ferrite tiles, but given the price of the tiles, that's not practical. Plus, it would probably just be plain stupid to spend so much money on a setup which is not compliant with any particular standardized setup.

Thanks,
Tim

[Dulus]

Are these resonances here in the room with us now? 

Well, with nothing to absorb the RF in the air, your shielded room is a reverb chamber for sure.
Yet for the size of the room, resonance frequencies you've observed seem unlikely to be caused by the chamber resonance.
First simple test that comes to mind is to open&close the chamber door while the test is running, and observe those resonances. Any amplitude or frequency change?

My guess would be the resonances are occuring between the metal table and your circuit...

[TimNJ]

Are these resonances here in the room with us now? 

Well, with nothing to absorb the RF in the air, your shielded room is a reverb chamber for sure.
Yet for the size of the room, resonance frequencies you've observed seem unlikely to be caused by the chamber resonance.
First simple test that comes to mind is to open&close the chamber door while the test is running, and observe those resonances. Any amplitude or frequency change?

My guess would be the resonances are occuring between the metal table and your circuit...

Thank you for sharing your knowledge and intuition.

I think your comments about the resonance between metal table and circuit is definitely on the right path. In fact, for our first test of the setup, we placed the DUT directly on the metal plane (no 2" / 50mm spacer). The 10MHz, 20MHz, and 60MHz peaks were certainly much higher amplitude and we also had a severe hump at about 1-2MHz or so. After adding the spacer, the 1-2MHz hump was squashed to the noise floor, and the 10MHz, 20MHz, and 60MHz humps came down about 10dB (~about 30% reduction).

Still, compared to the reference measurement from the accredited test lab, there's still some serious peaking at the above mentioned frequencies. So, if it was purely resonance between plane and DUT, I would figure we might see those resonances on the accredited scans too??

As I mentioned, there is still some larger structural differences like serpentine cable configuration and distance from wall. I think the electrical bonding of the table to the wall is also quite poor in our setup. So, there are certainly many other potential issues.

Thanks for the idea about opening the chamber door. Will try it! What would be the range of expected resonance frequencies for the mentioned chamber size?

I am still curious to know if anyone has any intuition of using a small number of tiles, and the ideal arrangement/position...although I'm sure it will depend a lot on DUT and chamber geometry.

Thanks!

[TimNJ]

My microwave theory and knowledge of cavity resonators/waveguides is a very shaky, but ballpark estimate based on 1λ would probably indicate at least 50MHz fundamental resonance, and more like 75 or 100MHz and above.

https://learnemc.com/EXT/calculators/Cavity_Resonance_Calculator/rect-res.html

On the other hand, the cable length is in the ballpark of 3m, so based on 1/4λ, that might indicate something like 25MHz due to the cables over the ground plane. Again, I'm a total dunce with respect to most of this.

[radiolistener]

>>>Would there be any value in using a small number of lossy ferrite tiles in the direct vicinity of the test setup? Take a look at my attached photo as an idea. The actual coverage of the chamber would probably be 5% or less.

My idea is to try to limit reflections "at the source"

Note that there is near field coupling at distance less than lambda / 2, which is very large for a long wavelength like 30 MHz and below. So if you place some ferrite or metal thing at near your equipment you can make interference worse in some cases (where main roles plays near field coupling). Because that ferrite or metal thing will work like antenna for interference and your DUT will be coupled through near field with that antenna. 

In order to limit reflections you're needs to use special coating with high absorption properties and special geometry of walls to change angle of reflection and absorb it. Something like that:

[TimNJ]

>>>Would there be any value in using a small number of lossy ferrite tiles in the direct vicinity of the test setup? Take a look at my attached photo as an idea. The actual coverage of the chamber would probably be 5% or less.

My idea is to try to limit reflections "at the source"

Note that there is near field coupling at distance less than lambda / 2, which is very large for a long wavelength like 30 MHz and below. So if you place some ferrite or metal thing at near your equipment you can make interference worse in some cases (where main roles plays near field coupling). Because that ferrite or metal thing will work like antenna for interference and your DUT will be coupled through near field with that antenna. 

In order to limit reflections you're needs to use special coating with high absorption properties and special geometry of walls to change angle of reflection and absorb it. Something like that:

Thank you. I was under the impression that the ferrite tiles were especially lossy/absorptive at lower frequencies like 30-500MHz or so, and that the foam pyramids were more useful into the GHz range. Again, that’s only my perception of a few plots of return loss of different materials w.r.t. frequency. So, I figured foam absorbers would not be useful in the frequency range of interest.

[radiolistener]

Thank you. I was under the impression that the ferrite tiles were especially lossy/absorptive at lower frequencies like 30-500MHz or so

ferrite rods are very popular for magnetic antenna which works at 0.5-30 MHz with pretty high Q, so it's not lossy at low frequency. It become lossy if you put very high RF power into it, when ferrite is saturated.

that the foam pyramids were more useful into the GHz range.

Yeah, that piramides are intended probably for ten or hundrend MHz and above. It's very hard to shield low frequencies, because it requires a very thick shielding and very hard to achieve acceptable attenuation. And even almost impossible to control reflection angle, because there is almost no plane wave in the near field, and since plane wave doesn't present here it doesn't have sense to talk about wave direction vector in the near field zone... But you can use some deep underground bunker for better shielding on a low frequencies

The main problem with low frequencies is a near field magnetic interferences which is very hard to eliminate.

[Marco]

Google tells me the pyramids have to be greater than half the wavelength of the lowest frequency signal.

For lower frequencies, couldn't you use an active version of the Salisbury screen though? Instead of the outer wall at a quarter wavelength, make the outer wall segmented and relatively close to the Salisbury sheet, with each segment driven to follow the voltage of the sheet in front of it (almost any way, need to prevent oscillation, so amplification slightly less than 1 relative to outer Faraday cage).

[T3sl4co1l]

Wouldn't think the few tiles, as shown, would do all that much -- the field at the wall shouldn't be all that much, and it's not near expected current antinodes (consider the stub resonance of the table or PSU) where it can help choke off those currents.  Cavity resonance is indeed a possibility; I'd think you'd need more coverage (>half a wall??) to get the Q down.

Can test the chamber/table/etc. modes by coupling in a sig gen / network analyzer / etc. to the active port (LISN I suppose?) and hanging [passive] stuff off it to copy the geometry.  Poke around with some, Idunno, resistors between interesting surfaces, see if that changes the response?

The foam pyramids are nice at high frequencies -- the absorptive coatings aren't worth very much by themselves, but the shallow angles (encouraging surface waves and internal reflection) and bulk filler do a big part.  At lower frequencies, you can imagine the depth of foam required to match it -- perhaps several meters, depending on how dense and what filler we're talking I suppose -- but suppose instead we go with bulk ferrite tiles, well, the permeability is far higher, so too the index of refraction: if we can get the fields inside the tiles in the first place, they'll pretty much get trapped and dissipated there.  The compromise lies in choosing tiles of the right thickness and material, so that they act as 1/4 wave plates in the lowest band: interference reinforces the field within, somewhat countering the otherwise large reflection you'd get from the big step in refractive index (if it were a thicker bulk layer).

And I think the Q factor of the chamber tends not to be terribly low, i.e. a true anechoic chamber has a Q of 0, anything's absorbed no matter which way it goes.  It's more modest, to save on materials and just to kill the worst cavity modes -- but mind, I've certainly not characterized one, hopefully others can provide more insight as to design and character on that.


As for the EUT -- have you compared to free air (on the bench, maybe) behavior?  This picks up all the ambient too, but, that's easy to flag (measure the background), and, while you might not learn much in those offending bands [effectively the noise floor is too high there], maybe there's enough gaps to see what the device is doing, or at least get some idea.  And then comparing to the chamber test, maybe something that way too.

Tim

[joeqsmith]

Many years ago, we were using a few outside labs for testing when our chambers were down and I remember one of the places having a screen room (real copper mesh) with what appeared to be rotating reflectors inside.   I had asked one of their employees about their purpose and I never dug into it (no internet in those days).   From what I remember their idea was to remove any resonances.   The reflectors were fairly large compared with the cross sectional area of the chamber.   They looked something like a cross section of a biconical antenna but I think were flat.  Maybe 6' in diameter? 

Searching for such a reflector, not surprising, I found nothing. 

https://ietresearch.onlinelibrary.wiley.com/doi/full/10.1049/el.2016.3411

[joeqsmith]

Spoke with some friends about the history.  See the attached paper:

http://site.ieee.org/phoenix-emcs/files/2018/10/IEEE-Phoenix-EMC-Chapter_EM.pdf

[ConKbot]

>>>Would there be any value in using a small number of lossy ferrite tiles in the direct vicinity of the test setup? Take a look at my attached photo as an idea. The actual coverage of the chamber would probably be 5% or less.

My idea is to try to limit reflections "at the source"

Note that there is near field coupling at distance less than lambda / 2, which is very large for a long wavelength like 30 MHz and below. So if you place some ferrite or metal thing at near your equipment you can make interference worse in some cases (where main roles plays near field coupling). Because that ferrite or metal thing will work like antenna for interference and your DUT will be coupled through near field with that antenna. 

In order to limit reflections you're needs to use special coating with high absorption properties and special geometry of walls to change angle of reflection and absorb it. Something like that:

Thank you. I was under the impression that the ferrite tiles were especially lossy/absorptive at lower frequencies like 30-500MHz or so, and that the foam pyramids were more useful into the GHz range. Again, that’s only my perception of a few plots of return loss of different materials w.r.t. frequency. So, I figured foam absorbers would not be useful in the frequency range of interest.

You're right. The foam covers down to the 250-750Mhz depending on the foam size, and frequencies below that are handled by ferrite tiles in chambers designed to go that low. It's a different ferrite blend than a ferrite you'd use for an antenna.  But a lossy ferrite plate that large is basically exclusively used in EMI chambers, so you're going to be talking to an EMI chambers vendor anyway. I'd talk to them for a bit of help specifying a specific part. Make them work for that $20k for a few ferrite tiles 

[T3sl4co1l]

Also different geometry for ferrite, like you might see flat tiles behind honeycomb sections; effectively the density increases with depth, serving a similar role to the foam pyramids.

Tim

[TimNJ]

Spoke with some friends about the history.  See the attached paper:

http://site.ieee.org/phoenix-emcs/files/2018/10/IEEE-Phoenix-EMC-Chapter_EM.pdf

Wow. Thanks for those references. So far, only skimmed each for a few minutes, but for the second one seems like the idea is:

"Sweep" the reverberation/resonance modes of the chamber by incrementally rotating the "tuner" (zig-zag metal plate) to re-position the wave nodes/anti-nodes within the chamber. Run the EMI scan for each mode...and then upon comparison, the 'real' data should pop out as statistically correlated, and 'noise' due to the chamber environment should fall out as uncorrelated? Or something vaaaaguely like that?

The first link is slightly more interesting to me, in that it seems more straightforward to set up. I wonder if anyone's actually done it with success?

Conceivably, you can get a big piece of aluminum or brass, bend it in a parabolic shape over the EUT (from wall to ceiling). Then do the same on the opposite wall. Then use a small(ish) number of absorbers on/near the floor. I'd have to check the formulas they included to see if it makes any sense at the frequencies I'm interested in.

[TimNJ]

As for the EUT -- have you compared to free air (on the bench, maybe) behavior?  This picks up all the ambient too, but, that's easy to flag (measure the background), and, while you might not learn much in those offending bands [effectively the noise floor is too high there], maybe there's enough gaps to see what the device is doing, or at least get some idea.  And then comparing to the chamber test, maybe something that way too.

Tim

Thanks for the tip. Thought about this too, and think it probably makes sense to do an "open-air" (well... "lab-air") measurement instead of trying to cram everything into this chamber which doesn't really support the proper cable arrangement (according to the standard) plus all the uncertainty surrounding chamber resonances.

Given the DUT is a SMPS, any narrowband peak that shows up due to some local transmitter, radio station, etc. is pretty easy to rule out (generally speaking). For our particular workspace, the chamber happens to be nice because it is not just "general purpose" lab area, which is subject to rearranging/setup changes when you're not looking. 

Thanks for the ideas about finding resonant modes of the chamber and/or of DUT w.r.t. to metal plane. I guess I could use a VNA, inject some dBm signal into a cable harness with resembles the shape, loop area, and orientation of the DUT's harness, then measure response at LISN...I guess pretty much just replacing the DUT with a known signal source. Then could play with chamber parameters (i.e. add some parabolic reflectors, absorptive material) and/or physically move the DUT, until detected resonances are minimized. This pretty much would only apply to one particular setup, I suppose. I am not sure about impedance matching of the VNA to the "antenna" (cables), if that's necessary or not.

[TimNJ]

Google tells me the pyramids have to be greater than half the wavelength of the lowest frequency signal.

For lower frequencies, couldn't you use an active version of the Salisbury screen though? Instead of the outer wall at a quarter wavelength, make the outer wall segmented and relatively close to the Salisbury sheet, with each segment driven to follow the voltage of the sheet in front of it (almost any way, need to prevent oscillation, so amplification slightly less than 1 relative to outer Faraday cage).

Interesting. Never heard of such a thing. Something like the attached?

[Marco]

That is an active device, but it doesn't really drive the electrodes.

With the standard Salisbury screen the quarter wavelength reflector makes sure that for a single frequency all the energy in an incoming plane wave is burned up in the space cloth (the 377 Ohm square sheet). What I'm suggesting are electrodes directly behind which immediately positively reflect all the energy, by having a segmented driven guard at the same potential of the space cloth. AFAICS that would represent an open circuit reflection, instead of the quarter wavelength short circuit reflection of the standard approach. Not wavelength dependent though (for wavelengths where the segments and distance to the guard are small).

Then again, if no one ever even tried it, it might just plain not work

[T3sl4co1l]

Thanks for the ideas about finding resonant modes of the chamber and/or of DUT w.r.t. to metal plane. I guess I could use a VNA, inject some dBm signal into a cable harness with resembles the shape, loop area, and orientation of the DUT's harness, then measure response at LISN...I guess pretty much just replacing the DUT with a known signal source. Then could play with chamber parameters (i.e. add some parabolic reflectors, absorptive material) and/or physically move the DUT, until detected resonances are minimized. This pretty much would only apply to one particular setup, I suppose. I am not sure about impedance matching of the VNA to the "antenna" (cables), if that's necessary or not.

Matching not needed -- or more to say, that's the point!  Basically you can test the one-port of the LISN, or well, one line of it, into the mockup cable laid as such.  This couples to the chamber with whatever response it has, and you see peaks and dips in its impedance corresponding to whatever characteristic frequencies are thusly coupled to.  And seeing how peaks/valleys vary with orientation (cable resonances, coupling to chamber modes, etc.) and reflection/absorption/etc. hardware movement.

You won't be able to remove mismatch at low frequencies, because the overall topology is capacitive of course (a wire floating in space, or a 1/4 wave antenna if you like), but by placing absorption in key locations, you can perhaps get things flatter at the higher frequencies (30MHz+?) and thus be able to say something representative about the signals being observed.

Note that a wire in free(ish) space will be fairly high Zo, so, depends.  The LISN is made for 50 ohms, but that's more to say what your receiver impedance should be, connected to it; the EUT side however might be more like 100-150 ohms (cable elevated above GND plane?) or up to 200 or so (377/2 ohms, give or take).  You'll have peaks much higher (reflected waves returned in phase) and lower (out of phase) of course, until things are well enough damped to level it out.  So don't worry about flattening the response to 50 ohms, it's not the likely target anyway; probably you'll see peaks and dips around a [geometric] mean value, which the system will tend to converge on as you dampen out the respective modes.

Tim

[jonpaul]

alternative to screen room is free space testing

Setup a wooden or plastic table in your parking lot or lawn, at least 5 m distance to any building, vehicles etc.

Run outdoor extension cable for mains, setup lab and DUT on the table.

Voilà a low cost open space test lab....

Beats Faraday cage or screen room claustrophobia.....

Jon

[coppercone2]

I agree those rooms are unpleasant


also, buying an acre of land is cheaper then the room and actually might have a ROI etc.

[xmo]

claustrophobia???  Not in here:

[jonpaul]

Rebonjour

1. memories of my time in the screen room of similar sizes to the OP, for FM tuner development  (Sequerra) in 1970s......noisy, hot, claustrophobia,

2, free field testing is of course more suitable for conducted rather than radiated EMI.
Radiated EMI noise floor will be limited by the ambient RF environment


Jon

[TimNJ]

alternative to screen room is free space testing

Setup a wooden or plastic table in your parking lot or lawn, at least 5 m distance to any building, vehicles etc.

Run outdoor extension cable for mains, setup lab and DUT on the table.

Voilà a low cost open space test lab....

Beats Faraday cage or screen room claustrophobia.....

Jon

Yes, if the results are still not so great in "lab air" space, we'll consider going outside, but then I think we'll need to get a rain jacket for the spectrum analyzer. 

And yes, I bump my head on the light on the ceiling at least every other time I go in that room. And, with those power resistors cooking, it's a bit like a sauna, yes.

[coppice]

Yes, if the results are still not so great in "lab air" space, we'll consider going outside, but then I think we'll need to get a rain jacket for the spectrum analyzer. 

If you are going to do considerable testing in a field, go to a camping store and get a tent and accessories. Put all your equipment in there, and you'll have a nice relaxed day of testing, instead of continuous worry.