The last time Hackerfall tried to access this page, it returned a not found error. A cached version of the page is below, or clickhereto continue anyway

EMC, come and see! – HELENTRONICA


For the past month or two, I’ve been dealing with the EMC issues on our projects. I’ve become our team’s EMC guy due to my high-frequency background and because the most of the team members are either IC or FPGA design people. And because no-one wants to deal with tedious things like compliance, standards and regulations. I, on the other hand, have found the topic entertaining. Here’s what I’ve learned over the time.

What we do have and what we should have

So, we’re building this high-power device. Let’s just call it a “device” since my project manager, product manager, quality manager, CTO and CEO won’t be happy if I share more about it. It works on 3-phase power supply (yeah, I live dangerously) and delivers up to 20 kW of DC power (room gets hot quickly). You can imagine it contains whole bunch of switching elements and creates electrical noise enough to kill a weaker pacemaker carrying person.

In order to prevent such devices from ursurping our environment with noises, some smart people have come up with a “few” regulations, a long time ago. These regulations are tweaked by government agencies, such as FCC in United States or IEC and CENELEC in Europe. The regulations are disclosed in form of a standard that gets changed more often than not, so we now have standards like FCC Part 15 and MIL-STD-464 in USA and CISPR 22 in EU.

For instance, if you’re about to sell electronic teddy bear, you better make sure it doesn’t disturb a WLAN router next to it.

Understanding these standards is everything but simple, and you need very smart and experienced person to tell you how to comply with them. That person is, at the moment, not me, but all the colleagues from various EMC labs that I frequented during my assignment. To sum up the long and painstaking story of EMC regulations: in order to make your electronics electromagnetically compatiblekeep two things in mind: a) it should’t disturb other things in its surrounding, and b) it shouldn’t get disturbed by other things in its surrounding. For the first part, we have to deal with things like emissions, both radiated and conducted. For example, the standard called EN 50011 Class Asays that your device shouldn’t create more than 40 dBV/m of electrical field at distance of 10 meters. If it does, you’re not allowed to sell it within the given country, which makes your product good for trash can. For the second part, we deal with device’s immunites, both electromagnetic fieldimmunity and electrostatic surge immunity. That means that your equipment shouldn’t die and/or malfunction if there’s some EM field or ESD source next to it. The immunity is often not a legal requirement, but surely is a benefit if you sell the product that won’t burn-out when customer opens it. We felt confident about our device’s robustness and immunity, but the emissions were completely other thing. So, I got the task to check how dangerous our gadget really is.


The aforementioned compliance certificates are, technically said, super-duper expensive. They are issued by special laboratories where all EMC tests are being carried out. The less time you spend there, the better. Ideally, you go there only once, they test your device, and, after everything works fine, give you a certificate and you’re good to go. The problem is when your device is not compliant – then you have to go to the lab over and over again, tweaking your device every time you go. Usually, they charge you couple of hundreds of per hour and you have to be there to test every single modification of your design, be it an improved shielding, new PCBs, new cables etc. The costs can pile up really high.

Then again, these labs are usually very busy and won’t take you in immediately. If you are not a “familiar customer” you might wait for weeks before you’re scheduled. Another thing to keep in mind is infrastructure and equipment quality. Some labs aren’t able to test 3-phase power system devices. Some cannot test devices that consumer more than 1 kW. Finally, the most important thing about the EMC lab is the staff: are they eager to help, are they going to answer all your questions and be available for all your inquires?Are they willing to meet your irregular testing conditions (because, almost every device out there will be irregular in some aspects)? Can you make a nice offer if you’re about to collaborate for the next couple of months?

Friendliness and costs aside, what decent EMC lab needs to have in order to carry measurements out?

Anechoic chamber

If you’re going to measure radiated emissions in an open area, your antenna will pick up all the RF garbage present in the ether. That’s why the RE measurements are carried out in special rooms, called anechoic chambers, that prevent any electromagnetic field from coming in. They’re truly the closest we get to the ideal Faraday’s cage. The thick walls, floor and ceilings are covered with large steel plates on top of which there are layers of ferrite material. This way, the isolation from both electric and magnetic fields is ensured. Looking from the inside, the ferrite serves as an RF absorber, so no reflections occur when the device is on and radiating. On top of that, the walls are covered with those spongy pyramid-like wedges, the kind you can see in music recording studios.


Inside the chamber, there is a wideband antenna, which must be able pick up noise from down to 9 kHz all the way to 40 GHz, if you’re doing serious EMC certification. The antenna is usually two parts antenna, one part beingbiconical antenna, and the second part being log-periodic antenna. Both of these antennas are wideband, though the biconical one is used for the lower frequency range and log-periodical one for higher. Because EMC requires measurements of both polarizations, and on any height, the antenna must be mechanically movable. That’s why there is pneumatic rotor, as well as a height adjustment.

EMC antenna is actually two small antenna’s pretending to be a big adult antenna

The antenna is usually placed 3 meters away from DUT. For many EMC interesting frequencies, this distance is not enough for a proper wave propagation, meaning that antenna operates in the near-field. Meaning that all that elegant antenna theory of signal reception falls apart. Noone can do field calculations in the near field and survive to tell.

EMI Test receiver

EMI test receiver doesn’t have joyful demodulation and signal analysis options like classical spectrum analyzer, but it provides insanely fast frequency sampling .

EMI test receiver is nothing more than a spectrum analyzer, similar to the one I’ve been playing with before. But there are major differences. When measuring noises, we’re not sure where exactly in frequency range the measured signal lies. For that reason, receiver’s filter must slowly sweep full frequency range with a great precision. Imagine, a typical FCC standard specifies 120 kHz filter bandwidth, and radiated emissions are supposed to be measured within the range 30 MHz – 1 GHz. A “classical” spectrum analyzer would take typically 500 measurement points in one go, and for a given IF bandwidth we need point separation of max 60 kHz. It means that we get 500×60 kHz = 30 MHz for a maximum span. To cover the full standard range, we’d need to change frequency span 33 times! EMI test receiver must be able to measure much more points than ordinary spectrum analyzer, and do it in reasonable time.

Second difference is detector: EMC standards require a special type of detector, called quasi-peak,along with average and RMS detectors. QP is cool because it scales down the recurrence of given frequency. For example, if 100 MHz peak noise occurs only every couple of milliseconds and is not constantly there, QP detector will take it into account.

Third important difference is possibility of inclusion of something calledtransducer factor. The antenna (transducer) picks up an electric field (measured in V/m), while the receiver measures antenna’s output voltage. The problem is that the antenna does not pick electric field equally wellat every frequency. That’s why the antenna’s producer must specify so-called antenna factor in dB/m that is added to receiver’s measured value in dBV to obtain the actual field value indBV/m. In the end, the value of the measured electric field is calculated as:

Finally, what would EMI measurement be if there is no nice graphics? That’s why test receivers allow to draw the line of the measurement limit specified by the standard.

By the way, the test receiver is used not only with the antenna as an input but also as conducted emissions receptor. More about it a bit later. Much more differences between classical spectrum analyzer and EMI test receiverare laid out in this R&S pamphlet.


Line impedance stabilization network, LISN for friends, is probably most overpriced piece of electronics I’ve ever seen. Why? It costs a couple of thousands even though it’s nothing more than a couple of coils and capacitors. Yes, they’re special high voltage caps and high current air coils, but still, the damn thing costs like a small car! Anyways, the role of the LISN is that high frequency currents that originate from DUT and propagatetowards the mains grid see stable and frequency independent impedance, that is 50.


LISN has mains input, mains output and RF output. Between two mains ports there is LC filter that a) diverts high frequency currents from the mains to ground via C1 and b) blocks high frequency currents from the DUT via L1 and C2. Given that R2 is high resistance, the HF noise is propagating towards the EMI receiver. In 3-phase LISNs, there is LC network for every hot wire in the power cable.

Typical LISN structure for one hot line. LC filter is always referenced to the mains ground. Mains current pass to DUT unaffected, while HF noise is directed to EMI receiver.

LISN is a crucial equipment for conducted emissions measurements, what brings us to a discussion of importance of conducted emissions limits. You’d think that CE is limited to prevent the disturbance of other devices connected on the same grid, which is correct. But the disturbance mechanism is often misunderstood by many people, including myself. It is not that hf-current interfere over the wires to other devices on the grid, it is radiation they produce when passing through the power grid installations. It is a huge cabling system across the building and a terrible radiator. To put it as Henry Ott did, CE requirements are really RE requirements in disguise.

One important thing regarding LISNs that I learned on the hard way – LISNs have big leakage currents. Just think of it: C1 is typically 1 F, which means that in 220V/50Hz grid, there will be 70 mA of leakage current. The problem is that the most modern power grids have additional circuit breaker in the mains board, called residual current switch in EU or ground fault circuit interrupter in North America, that monitors current flowing to the GND and trips when the current is higher than 30 mA. That means that typical old fashion LISN can’t be used directly in the lab. Either purchase better LISN or, if you feel brave, get some high power isolated transformer.

No EMC lab?

Before I got on the quest of looking for an EMC lab for our needs, I did some preliminary measurements in our company’s storage hall. I think it’s clear that these are not valid EMC measurements, but, they gave me a feeling of how far off we are from real compliance. Of course, assumption is that I have at least proper antenna and spectrum analyzer. Using the memory storage and math functionality of the analyzer we can subtract background noise from the measured value and get some quasi-calibration that way.I managed to identify major sources of noise from the DUT and suggest some improvements before going to EMC lab. This kind of unofficial and non-compliant measurements sometimes can save you a couple of hundred bucks.

If you don’t have a real EMC lab, then everything is an EMC lab. While antenna picks up most of the RF content around you, the EMI test receiver can perform magical “diff” function and get you some insight in EMI your DUT is generating.

Emissions measurements

Once in the lab, I got this sensation that convicts must feel when they enter the courtroom. High above, in the big hall, there was the EMC antenna, looking at me like a judge, threateningly. Lab technicians, jury, seemed ready to lough at my device and point the design flaws in a blink of an eye. Gravy silence in the anechoic chamber. I’ve put my device, a defendant, on the turntable in the middle of the room and waited R&S test receiver to pass my verdict.

3-phase switching supply facing its arch enemies– EMC standards!

A second later, measurement is done. Bottom two images show how far off Class ACISPR 11 standards are we. No need to enlarge the picture, you can see that it was up to 30 dB above the limit. Interestingly, image on the right shows a bit better results. It’s the measurement output of the horizontally polarized antenna. Accidentally, power supply cable hangs vertically from the device and consequentially, the horizontally polarized antenna sees much less of the noise. Unfortunately, people of EMC only care about worse case, and we definitely do not pass the test.


No better luck there was either with conducted emissions. Bellow you can see the device in its CE examination setup: on the table, next to the huge grounded metal plate, connected to the 3-phase LISN and R&S test receiver. Notice how cable is folded in meander shape. Standards.

Conducted emissions measurement setup. There you see a 3-phase LISN and an old fashioned test receiver. Keep an eye on the huge metal wall on the side! Spectrum analyzer gave its verdict – you shall not pass CE requirements!

Identify, locate, eliminate

It is completely normal to get negative EMC results on the first date… er, measurement in the lab. However, instead of crying in the room corner, time is to pull up sleeves and solve the problem. There are two ways to solve EMI problems – either and tack the problems in their root and change your design(more effective way) or try to damp the emissions by all means. The latter is usually not very practical way, because the only way to get rid of emissions is to seal everything in a big metal box.

If you’re novice, like me, maybe the good start is to learn something about EMI origins and their prevention. I consulted with a couple of books from any “top 10 EMC books” list and took a sneak peek into design files of the device. I’ve learned a good deal from the books and can’t recommend them enough. Definitely a must-read for every electronics engineer out there.

First, I tried to identify the source of the noise. One, the most obvious one, was 100 MHz clock. My problem was that this clock was routed as a differential pair. In theory, differential pair shouldn’t radiate much, right?

Well, the book says it can. There is actually a nice little analysis of the E-field produced by the current in the loop, and differential pair is, actually, a loop. E-field will be proportional to the area of the loop, current and the frequency. Even though the differential pair is routed tightly together, (spacing was set to manufacturer’s minimum, 15 mil), there still was a certain area enclosed by the traces. I crunched the numbers and found out that area is 7 times bigger that it should be, if we want to satisfy the EMI requirements. That was basically the only noise source that I could identify – there were, of course, the 100 MHz harmonics, but most of the frequency garbage was spread spectrum noise that I had no clue how to diagnose.

This left me with the other option for EMI reduction: shield everything. But, if I put whole device into the completely box, it’s going to be useless. I’d need holes for cables, buttons, screens. Also, the ventilation holes. The problem is, that any opening, or slot, on the shield reduces its effectiveness. The slot behaves as antenna, with the same radiation pattern as a wire of the same length.

Home made gaskets from coax cable, attached to the housing using the Cu tape.

The existing housing we had, was full of these slots on the edges. To cover them up, I’ve taken steel plates, bent them in L-shape and placed over the edges. I drilled holes and tightened steel plates to the housing. But, this is not enough. Microscopically, where steel plate and housing meet, there is no perfect contact. There is always some gap that will let some radiation leak. Usually, you’d cover these with gaskets. But, I had to do it quickly, so I improvised. I found an old, thin coax cable. I stripped off the outer isolation while preserving the braided shield. I milled the paint away from the housing and attached the cable on it. The cable should come between the housing and the steel plate. Pressing it strongly from the outside, the cable flattened enough to make screwing possible. Home-brew gasket was born.

Once the housing was solved, I remade the all the cables that enter the device. I’ve ordered triple axis shielded cables for 3-phase power input and double axis shielded for load output. The problem now were connections. Coax cable pigtails do not connect elegantly to the improvised housing. I could either order custom-made connectors and pay for them or…use a great load of copper tape. That little thing is incredible. Flexible, adhesive yet conductive. The best friend of EMC engineer!

L-profile steel plates covering EMI leaks. EMC is just like plumbering.

Cu tape and line filter. The best friends of EMC engineer.

Shielding and gap covering should help with radiation emissions. However, conducted emissions, still remain unaddressed. Well, standard solution on that matter is a neat little thing called line filter. It’s not that we didn’t already have the line filter in our device, it’s that it didn’t work properly, and was not positioned at the right place. For line filters, it is crucial that it is placed immediately after the power cord. So, I had to put one externally. And of course, seal it with copper tape.

I packed my device into the carton box and hit the road to EMC lab again.

Conducted emissions first. It seems that little line filter did its magic. Noises were reduced significantly, below the limit. Seems like little victory on that side.

Radiated emissions next. Incredibly, but impromptu shield I have made has fixed things a lot. Actually, only frequency component above the limit was that 100 MHz clock I mentioned earlier.

I couldn’t change the board, nor routing, so I went for the second best thing. I’ve put 50 resistance at the start of the clock line, hoping that it will reduce the current from the TTL driver and not disturb normal operation. Small wonder, it didn’t work. The FPGA, for which this clock was intended didn’t respond, as the voltage level dropped. So, I had to tryout multiple resistor values to find the one that reduces EMI but not the voltage level for FPGA. Luckily, the “home-made” EMC lab I mentioned before turned out to be very helpful for this task.

Again in the lab, I discovered a new helping tool – a ferrite ring. One of the technicians was thoughtful enough to get me one off-the-shelf to put around some of my cables. Apparently, if the cables are not shielded properly, ferrite might help absorb some of radiation.

Final measurements, turned out to be complete success:

Completely under the limit!


So, I guess we can sell our device now. But at what cost? If the EMC is not taken into consideration from the beginning of the design, it will be a devil’s job to make it work in the end. Chances of being able to fix EMI when your product is finished will drop closely to zero, while the costs of doing so will skyrocket. My investigations showed serious flaws in both mechanical and PCB layout design. The only thing left, if the company doesn’t want to go on the marked with my hand-made shield and bunches of Cu tape, is to completely redesign the shield and PCB inside. Complex and expensive task.

Moral of the story: kids, think of EMC from the beginning.

Like this:

Like Loading...

Continue reading on