EMC prevalidation tests

[hamdi.tn]

Hi all,
totally noob question i have a DSA832E spectrum analyzer and need to make some EMC measurement and validate a board before doing the final one in the chamber. i have no idea  what should i do and  consider to get some realistic results.

Can anyone please explain to me the basic setup that i should put in place.
lot of thanks.

[AndersJ]

if you "have no idea" then give up on this.
EMC is too difficult if you have no background experience and no test equipment.
A spectrum analyzer is a good start, but you will need more equipment.

[crazyguy]

EMC is a very boring topic, you need to read so many documents first

If you need to do in-house EMC testing, read all these harmonized standards

EMC Directive 2014/30/EU
http://ec.europa.eu/growth/single-market/european-standards/harmonised-standards_en
https://ec.europa.eu/growth/single-market/european-standards/harmonised-standards/electromagnetic-compatibility_en

Testing for EMC Compliance: Approaches and Techniques 1st Edition
https://www.amazon.com/Testing-EMC-Compliance-Approaches-Techniques/dp/047143308X/ref=sr_1_1?ie=UTF8&qid=1489317626&sr=8-1&keywords=EMC+testing

[billfernandez]

Might this be helpful?

http://www.emcfastpass.com/emc-testing-beginners-guide/

[nctnico]

The first thing to realise is that you can't make accurate measurements on your bench where it comes to measuring radiated emissions. What you can do is check the amount of noise coming from cables and see if there isn't much noise on long PCB traces. You also have to realise that radiated emissions aren't the only test. Radiated immunity and ESD can the though to pass as well.

In my experience it is important to design for EMC compliancy.

[Jester]

I agree with nctnico, most likely if you did not design to pass you will most likely fail unless you only have really low frequency circuitry.

Typically you get a pre-scan so you know what problems frequencies you have,  and then use your test equipment to get a baseline at the problem frequencies and then tweak to get them down x dB and then re-test.

If you have cable connected equipment you will also need to do conducted emissions testing. To do that you will need a LISN, there are some good threads on this site describing LISN characteristics and how to build one.

[AndyC_772]

Your process is back-to-front. You can't perform useful tests using un-calibrated equipment in your own lab first; you'll learn nothing useful. You could easily miss problems, or alternatively, spend a lot of time and effort trying to eliminate sources of noise that don't actually fail the statutory tests.

What you can do, is book a day's testing in a proper lab. Carry out some testing there, which will tell you what frequencies your equipment is actually emitting, and how far over the limit you are.

Then, you can go back to your own lab and look for sources that are generating the specific frequencies you know are a problem, and you can perform qualitative (but not quantitative) before/after comparisons when you make changes. This might give you some idea as to whether or not any fixes you apply are effective.

[T3sl4co1l]

Your process is back-to-front. You can't perform useful tests using un-calibrated equipment in your own lab first; you'll learn nothing useful.

I disagree: you'll see if there are gross sources, and you can identify what they come from.

No, you won't determine precisely what's critical, because the standard is a sharp cutoff, and even a "standard" setup will vary a few dB depending on how it's laid out.  But being able to tell if something is 20dB above or below, that's still very useful.

The easiest to perform is conducted emissions.  They only check external cables, like power and data lines.  You can build your own networks, or buy them, and set up a simple test.  You can also use the networks to couple noise into the circuit, to test susceptibility.

Radiated testing is more difficult: you need big antennas, some way to sweep the antenna (altitude above ground) and DUT (rotation), and some way to isolate ambient noise sources (usually a shielded test chamber, or open air testing).  This is one situation where, unless you want to spring for the expense of a representative setup, you should do qualitative inspection instead: check ambient electric or magnetic fields with probes, or cable currents with a wideband clamp probe.

You can also do ESD testing with a crappy piezo clicker device.  Should be possible to adjust its pulse response by adding high voltage capacitors and resistors, calibrating it by zapping a high voltage attenuator (Minicircuits has some I think).

I want to create a cheap EFT generator so you can do that, as well.  Still in research phase.

Tim

[nctnico]

IMHO what is generally lacking from application notes and example diagrams are the components needed for EMC compliance. Too many are hiding behind the phrase 'it depends' while there are simple recipes you can follow. For example: series beads and TVS diodes on ethernet lines, a common mode choke and (again) TVS diodes on USB lines, a common mode filter on a DC power input. Now everyone has to reinvent the wheel for themselves.

[T3sl4co1l]

IMHO what is generally lacking from application notes and example diagrams are the components needed for EMC compliance. Too many are hiding behind the phrase 'it depends' while there are simple recipes you can follow. For example: series beads and TVS diodes on ethernet lines, a common mode choke and (again) TVS diodes on USB lines, a common mode filter on a DC power input. Now everyone has to reinvent the wheel for themselves.

But... those are all terrible examples?

Well, the last one, CMCs on power supplies, is fair game.  But whether they work or not, depends on the rest of the circuit.

The biggest downside is that EMC is more holistic than usual circuit analysis.

You have to be willing to see your circuit as a much smaller simplification.

Riddle: when are mains hot and mains neutral shorted together and effectively identical?

Answer: when you put a fuckoff huge capacitor between them, so the impedance from hot to neutral is very small at higher AC frequencies.  At HF, you can reduce the mains input to two connections: mains and ground.  Now the common mode circuit is much more obvious: it's a series inductor and some parallel capacitors.  And the capacitors better not be parallel from a noise source, but shunting to ground.

Using ground planes or supply rails as Faraday cages provides similar reference and optimization in other circuit areas.

The most often-repeated mistake is adding impedance where it doesn't matter, or where it's worse.  Example: a cable in free space, acts like a large ratio coax transmission line, so has a high transmission line impedance.  Adding a ferrite bead in series, even a large one (~300 ohms?), provides only a modest reduction in noise (~6dB).  (This is, however, an ideal amount to dampen resonances on the line, which can have a big impact on susceptibility, and on emissions at "unlucky" frequencies.)  Whereas if the far end, or both ends, of the cable were grounded, you get an impedance divider and much more reduction (>20dB easily).  Another example: adding a ferrite bead to a ground connection (an oft repeated mistake on USB connections, even in appnotes!). 

Tim

[hamdi.tn]

Hi guys, thanks for all the replays but i think i miss-expressed myself, i already done design to pass EMC testing and i did testing in proper well equipped certification labs, and my design did pass,but never did it my self, by that i never did some pre-validation tests in my lab. so am not sure what it take and "how to" use the spectrum analyzer properly to get a "it may pass in final" kind of results. I did some modification on my pcb and i need to do some testing before doing the real thing.

I know i have the spectrum analyzer but what i kind of antenna i need to use, can i manage to get some results with some DIY antenna, some half-wave length antenna maybe or any setup that get me close to the real deal.

Typically you get a pre-scan so you know what problems frequencies you have,  and then use your test equipment to get a baseline at the problem frequencies and then tweak to get them down x dB and then re-test.

that's exactly what i need to do

thanks all 

[crazyguy]

Hi guys, thanks for all the replays but i think i miss-expressed myself, i already done design to pass EMC testing and i did testing in proper well equipped certification labs, and my design did pass,but never did it my self, by that i never did some pre-validation tests in my lab. so am not sure what it take and "how to" use the spectrum analyzer properly to get a "it may pass in final" kind of results. I did some modification on my pcb and i need to do some testing before doing the real thing.

I know i have the spectrum analyzer but what i kind of antenna i need to use, can i manage to get some results with some DIY antenna, some half-wave length antenna maybe or any setup that get me close to the real deal.

Typically you get a pre-scan so you know what problems frequencies you have,  and then use your test equipment to get a baseline at the problem frequencies and then tweak to get them down x dB and then re-test.

that's exactly what i need to do

thanks all 

If you have no idea what you going to do.
You have to familiar with the standards, before you go to lab set-up your test. Different Products/Applications have different international standards. You cannot skip this part.
Everything you want to know are described in the Standards. Spend some time review the standards first. Again, it is a very boring job.

billfernandez already posted a good link for the beginner
http://www.emcfastpass.com/emc-testing-beginners-guide/