EMI/EMC Conductive Emission Testing

One common question I get from my customers is "What is conducted emission test?". In simple words: it is about measuring the emissions on cables attached to your product. There are quite some differences between the different types of these so-called ports (ie. the "type" connected cable - eg. a.c mains, ethernet, antennas,...). In this article we have a look at the basic ideas behind these measurements and briefly see what could be done to reduce the emissions.

Measuring principle

As stated above, our goal is the evaluate the conducted noise emissions of our equipment-under-test (EUT) wrt. EMC/EMI. In this article, we're focusing on the principles and leave standards and limits out for another article.

In the next figure, we see a pretty minimal, but still common, example of a test setup:

We have our product (the EUT or equipment-under-test), a mains supply (this can also be some ac/dc adapter), and some kind of load. For obvious reasons, the supply should be pretty clean. Otherwise, the emissions from the source may contribute to the measured noise. The load is either some other product, or just some resistors.

The 2 boxes missing, the LISN and the CDN, are our probing devices. LISN stands for Line-Impedance-Stablilization-Network and CDN is Coupling/Decoupling-Network. There are many more possibilites, but for the purpose of this article, we're using these options.

With our setup, we now can power our EUT, connect something to it and measure the noise on the input and the output. Now what's next?

Understanding the LISN and CDN

The next question we need to answer is what a LISN or CDN is used for. It's most important taks is to present a defined impedance to the emissions of your product. Another thing to mention is the isolation from the rest of your measuring setup. This also explains, why there are so many different LISNs and CDNs. It's not just the different connectors, it's also about their internals.

From the figure above we can now get 2 crucial differences between a CDN and a LISN. A LISN, in general, defines the differential-mode (that's what is "between" line/L and neutral/N) and common-mode (beween the combination of L and N and the ground). Contrary to this, the CDN is affecting the common-mode only. This can be seen from the coupled inductor for the + and - lines in the CDN where-as the LISN has an seperate inductor for L and N.

In addition to this, you have very different impedances! For our measurements, we want our spectrum analyzer (SA in the figure) to have an input impedance of 50Ω. This means the LISN will have this 50Ω at least at higher frequencies (the coupling capacitor gets effectively a short at about 1MHz). So L and N to earth have 50Ω - common-mode signals will see 25Ω (L and N are effectively in parallel) and for the differential mode, you get 100Ω!

Now for the CDN, we see, that there's some voodoo going on. We have again some coupling capacitors, then some resistors combining the singals to get the common-mode and, again, our SA representing 50Ω. Standards usually call out for 150Ω impedance CDNs. So these 2 resistors need to have 200Ω.

As a tip, if you don't have a CDN, just get some common-mode-chokes, clap-on ferrites, or similar to get really high common-mode impedance. Then add caps (maybe 10nF) and resistors (they should give in parallel 100Ω), if you connect this to a SA, your golden already - if you try your luck with an oscillocope, either add 50Ω resistor and measure with your 1MΩ input, or switch your scope to 50Ω input.

Testing procedure

The testing procedure is quite easy. Exercise your EUT the way it is descibed in the relevant standards (as a basic guideline, just maximize the emissions) and measure each cable with an appropriate measuring technique (there are some more than the LISN and the CDN, but that's for another article). But there is one caveat we need to talk about: the physical arrangement.

In the figure above, you see a setup which is different to the ones you see on other resources. I went down the rabbit hole of EMC standards quite often for my business. So I came up with some setup which is quite conform to standads and suits my needs for pre-compliance measurements way better.

Outside the US, CISPR standards are the go-to for EMC testing. The measurement techniques are laid out in the CISPR 16-2-x series. Especially CISPR 16-2-1 is a must-read, in my opinion.

The requirement for table-top-equipment (ie. products you normally use on a table) is, that you place one side 40cm from a ground plane to give the setup a proper reference. Other conductive surfaces should be further away. Now for the particular interesting part: CISPR 16-2-1 allows you to have everything on a ground plane and rise the EUT 40cm with some spacer. In the current edition this is in clause 7.4.1. I'm stating this, because I'm pretty sure someone will argue against this.

Now, the only thing I'm deviating from this for my pre-compliance work is, that the ground plane. For a fully standard conform setup, it should be at least 2m by 2m. Obviously, my table is smaller. Having quite some experience in this field, so that I find, for the equipment I test on a regular basis, this does not matter too much.

The basics on reducing conducted emissions

Now you did some testing (either with your precompliance setup or at some external lab) and you failed the test. What's nex? Well, I propose to you to go up to the beginning of this article and look at the schematics of the LISN and CDN.

I mentioned the differences in the impedances. What I have not told you so far is, that the LISN has 6dB of DM-to-CM conversion factor. This means, that only half of the differential-mode-signal will show up at the measurement port while the CM will be present in full.

The first important thing you should do is to determine if you product produces excessive common-mode or differential-mode noise. If it's just CM, you will need to add some CM chokes. In case you have a problem with the differential-mode, you may want to increase the inductor in eg. your PFC.

As explained earlier, the CDN is sensitive to the common-mode only (if it was perfect). So on ports where you have a CDN, you can ignore the differential-mode for the most cases. However, keep in mind what we said for the LISN in your daily work. The imperfections of a CDN can (and will!) get you. There are types of CDNs specificly made and specified for these problems.

Designing a EMC filter for conducted emissions

Now the next question is "How big do I design my EMC filter from the measurement results?". First of all, you'd be better off if you can find the source of your emissions and reduce them at the source. There are cases where you cannot do this (eg. if you use off-the-shelf modules), or you might already optimized your circuit completly without any filter and now add some to finalize your results.

In any case, this gets very difficult if you have just a single result. In case you some experiements and have a small set of data, in most cases, you can get to good results quite easily.

Again, I want to remind you of some theory: Each of your problematic frequencies could come from a different source in you product and each of these sources may have a different source impedance.

We talked about the impedances of different measuring devices already. Now we can combine this knowledge with our mesurement results. My first aproach is always adding clamp-on ferrites where I know what amount of impedance they add. These essentially add to the impedance of the emission source.

It may sound trivial, but you can think of the whole procedure as an optimization of voltage dividers. With the added impedance, you take another measurement. Convert both results to linear quantities (eg. dBµV to µV) and write down the basic equations for a resistive voltage divider.

As a first order aproximation, you can take the amplitude of the source as a constant. You now can estimate the impedance needed to meet your desired limit.

In case you have a DM problem, just add some inductors (add them to all involved lines; eg. L and N!). The principle is the same. Calculate the impedance you would need to add.

Knowing the impedances and the frequencies, you can not calculate the needed inductances and choose an appropriate choke. Sounds simple, right? Well, the devil is, as always, in the details! This simple aproach, howeve, will help you in most cases.

If you have more cables involved, and you have maybe already some filters in your product, you get loads of other problems. Resonances and problems where your disturbance currents flow somewhere between the different cable ports are just some examples for this.