DIY small EMC test chamber / box - antenna?

[max_torque]

I'm doing some basic EMC pre compliance testing on a small (~10cm square) control unit.  I've got a suitably sized die cast metal box which i plan to use as an ad-hoc test "chamber", but i'm wondering what antenna i should use to to  collect the radiated emissions from the EUT and route them into my spectrum analyser?

Anyone built something similar with good results?  (i guess i'm not looking at a "calibrated" solution, just an antenna with enough band width (say in VHF range) so i can see the problem frequencies and attempt to remove or attenuate those source frequencies)

The enclosure i have is this one:


and is 250mmx250mmx100mm

I've tried a 2.4Ghz antenna i had lying around and it works, but it probably is a long way from optimum in the frequency band i want to test in.

[Neilm]

Personally, I would not put it in the box - you will probably get too many reflections and as EMC emission testing is done at 3m or 10m in that box you don't know what are far field and what are near field effects.

There is nothing to say you have to be in a chamber for radiated emissions. The standard test site is an Open Air Test Site (OATS) - literally they are fields. The test house does a background scan to see what is there, then turns on the Device Under Test (DUT). Then they compare the two readings. Nothing to stop you doing something similar. If you do, try to make sure things that will radiate (phones etc) are turned off to reduce the background noise.

[MagicSmoker]

Yeah, can't use a box to do any kind of EMC testing... unless it is at least a few 10's of meters on a side and covered with lossy ferrite tiles!

Seriously - you won't get any - and I mean *any* - useful data from this.

[T3sl4co1l]

You can still get informative results by placing bias tees against the box.  That is, grounding the circuit (at its normal ground locations) to the box, and passing all connections through the box and measuring their signal outputs (providing DC/power and signals as needed).

Bias tees are just single ended LISNs, after all.

To measure common mode, you'll need a 0 degree power combiner with as many inputs as you need to average over.

Tim

[MagicSmoker]

You can still get informative results by placing bias tees against the box....

But that's a conducted emissions test, not radiated, and the results will still be of questionable value unless the actual enclosure is used, rather than some random metal box (that said, the box will have less effect on conducted emissions, so not a total waste).

[T3sl4co1l]

Yes, that would be conducted only (if at a higher bandwidth than usual).

Radiated measurements depend critically on the board placement, orientation, shielding, and number, length and orientation of cables attached.

You can still see if it's just blasting shit out all over.  Which is what I mean by "informational".

If you're blowing the limit by 20dB (assuming uV/m radiated == uV conducted, which is as accurate a ballpark as any), you can be sure to fail radiated.  If you're out by that much, you have a lot of homework to do, and your PC test will have been useful, saving you a premature trip to the lab.

If you're close to the limit, or below, it's worth going to the lab to test it for real.  You might not pass, but you at least will have a better chance of passing with only minor fixes: component substitutions, ferrite beads.

Tim

[MagicSmoker]

...
You can still see if it's just blasting shit out all over.  Which is what I mean by "informational".
...

Fair enough, though I think a set of handheld E and H probes for the scope is much more useful as you can usually pinpoint the offending components or traces with just a bit of practice and no real special equipment besides a decent scope (a spectrum analyzer would be better, but that's the sort of thing I only use once a year, maybe, so I'm not too proficient with it, unlike my scopes, which I use pretty much every day).

[max_torque]

Some more background info:

The EUT was never designed to meet any EMC standard. although it was designed with "best practice" it was also designed to be lowest cost and simplest to manufacture.  The original use did not require any EMC compliance. (think plastic enclosure, 2 layer pcb, minimal i/o filtering etc)
As is the way of the world, things change, and there is now a requirement to use this device, with the least changes possible, in a MIL spec environment that meets the full MIL spec EMC limits!  Yes, yes, i know how stupid that sounds, but bear with me here!

So, the device was tested (in a certified test chamber to MIL limits) and, no surprises, failed miserably! 

What i am trying to do here is to give the customer my best estimate of what it would take to make this device (or more accurately a version of it) meet the MIL specs, with the minimum changes possible.

So, what i plan to do is to test the device in stages, ie enable / disable various bits of the h/w on the pcb (like the dc:dc, or the CAN buses etc) and to enable/disable various bit of code and processor resource.  The idea is to cheaply, and quickly get an idea of the relative contribution of each component to the overall emissions level.

Of course, the results aren't going to be accurate in any way, but should give an idea of where we need to focus our changes!


So, using the metal box (with all it's reflections etc), mounting the EUT in it in a fixed position, relative to the antenna, i can run simple tests (like run the processor off a 9v battery and not the DC:DC etc) and get an idea of the relative contributions.  I can also try "bodges" to the pcb to attempt to attenuate / modifiy the RF energy being emitted.  The primary aim will be REPEATABILITY, and not accuracy.

Ultimately, the device WILL have to be relatively extensively re-designed, but if i can get at least a road-map for what needs to change and what brings the biggest improvements, then we will have saved a huge amount of time and effort before we get to the point of an official re-test

And whilst the idea of standing in a field to do these tests sounds nice, the British Summer (ie RAIN!) makes than somewhat non-practical (be nicer in Sydney ;-)


So, back to the original question, what antenna for 30Mhz to 300Mhz, that is cheap, and fits in a small box?  I'm happy to make/bodgeup something if that is the best option? 

[Karel]

We use this: https://www.tekbox.net/test-equipment/open-tem-cells

Works surprisingly well for small boards and it's affordable: https://www.amazon.com/Tekbox-TBTC1-Cell-Pre-Compliance-Testing/dp/B00S5YDU9Q/

[T3sl4co1l]

There's no concept of "antenna" when the distance is near field.  The box has to be several wavelengths long, and the antenna a half wave or more, at the lowest frequency, to be meaningful.

At best, you'll simply be doing capacitive or inductive probing, depending on the relative orientation of "antenna", board, box and actual source.

Which is why doing it with near field probes is as good of an inspection method (but, again, not representative of real lab conditions).

The TEM cell ^ is probably the nearest, most representative substitute.  It's also somewhat self shielding.

Tim

[Jay_Diddy_B]

Hi,

I built my own OPEN TEM cell using similar construction techniques to the Tekbox design.

Here is a picture:



It is made from copper clad FR4. The large pieces are 12x 18 inches, 300 x 450 mm.

The filters on the bottom were found to be useless.

Here is a picture of the spectrum analyzer with it in use:



I have a switch power supply board in the cell. The two traces are with the board on and with the board off. You can see the FM radio band at around 100MHz is visible, but I am getting useful results everywhere else.

I also built a 30- 300 MHz bi-conical antenna. The dimensions were described in some of the earlier revisions of MIL 461. I found this to be useless, too much background EMC.



Regards,

Jay_Diddy_B

[hs3]

That TEM cell looks nice. Would you have any other measurements in addition to the large pieces?

I was looking into building one for myself last year and also got some large copper clad FR4 pieces to cut at the time but then didn't get to actually cutting those into pieces yet. I checked the drawings I made at the time and noticed that I had also used 300 x 450 mm as the sizes for the large pieces. For the other dimensions there was more uncertainty what to use I think. So it could be helpful to know the dimensions used for an existing one that someone has already built.