Make your own EMI measurement probes
This story started with my pet project. One of those which are never completed for one good reason: it works and I can use it whenever I need it; so the improvements and finishing were always postponed. This project is a programmable frequency generator using a DDS (Direct digital synthesizer) AD9850 module controlled by an Arduino.
It can generate a clean and stable sine wave signal from 1 Hz up to 60 MHz just by typing the desired frequency on a keypad. A rotary encoder allows tuning the frequency around the selected value. This project was working and used for more than 6 months on a proto board until I decided to put it into a cabinet. Here it is:
Once this done, there were still a number of things to improve both on the hardware side (including the overlay!) and software side. One of them was the power supply; So far it was powered by an external power supply and I thought it would be nice to have its own power supply, and because this was the easiest part I decided to do it first… The problem was simple: I needed a power supply of either 9V or 12V with a maximum current of 1A. The best candidate (I thought) was a small SMPS that I bought some time ago, rated 12V / 2A. The size was perfect to fit into the cabinet and the maximum current more than enough. But careful, do not install a new power supply into a sensitive project without testing it first! I loaded the Power supply with a 10 Ohm resistor to do some measurements before installing it into the cabinet.
I won’t describe the measurements here as you know how to measure voltage, current etc… Here are the results:
Input Voltage: 244 V / 50 Hz
Input Current: 0.153 A
Input Power: 17.0 W
Power Factor: 0.45
Output Voltage: 12.076 V
Output Current: 1.2A
All looked OK so far; the output voltage remained stable when I disconnected and reconnected the load. The power factor was not good but quite typical for this type of power supply without compensation. At this stage I could think that this SMPS was fit for the job. However, don’t forget one of the most important lessons in electronics: The multi-meter only tells you a part of the story… Let’s have a look at the 12 V output with an oscilloscope:
Wow, with AC coupling and 50 mV/Div it doesn’t look too good but I could probably fix this with a filter at the output. However, something else was troubling me; I was listening to an FM radio station and could hear the interferences produced by the SMPS. Checking further, I could hear the SMPS over a wide range of radio frequencies, from a few hundred KHz up to over 300 MHz. That SMPS was generating more signal than my frequency generator!
This is the point where my pet project was put aside, once more…
Tired of listening to Electromagnetic Interferences (EMI), I wanted to see them! Since a while ago I had been willing to build and test a few EMI measurement probes – also called sniffers – and this wonderful SMPS seemed to be the perfect candidate to act as an EMI generator! EMI measurement probes are available on the market but quite expensive. I found a number of DIY probes design on the Internet and I was keen to test a few of those.
EMI is a very complex topic, full of even more complex mathematics! Fortunately we don’t need the math to understand the very basic principle. Here it is, a bit simplified I should admit, but enough for our purpose:
EMI (Electromagnetic Interferences) exist because of electric and magnetic fields that you probably learned about in your physics classes (remember the old, long suffering teacher!). For EMI to exist there are three situations: two different fields and the combination of both.
1) An electric field exists between two or more electrically charged particles of opposite polarity. If this field fluctuates (variation in amplitude) it will induce a current into a nearby conductor.
2) A magnetic field exists when one or more electrically charged particles are moving. If this field fluctuates (variation in amplitude) it will induce a current into a nearby conductor.
3) When variable electric and magnetic fields combine under particular conditions they create electromagnetic waves (Radio waves indeed!). Electromagnetic waves propagate over thousands of kilometers. This is the magic of radio!
When AC current circulates in a conductor you have both electric and magnetic fields around the conductor. Those fields are fluctuating at the frequency of the AC and, at a certain distance of the conductor (source), and depending on the frequency, they will combine into radio waves. The area near the source, where the two fields are not combined yet is called the near field. In this area you can measure both electric and magnetic fields separately. Beyond this area, called the far field, you can measure electromagnetic waves. The distance between the source and the far field depends on the frequency. The probes tested in this article are designed for near field measurements, so there are two different types, one type for magnetic field and one type for electric field.
The first magnetic field probe I tested is an unshielded loop. A loop of solid wire insulated with heat shrink and soldered to a BNC connector. It is not very sensitive but this can be an advantage as we can pinpoint the area of the circuit where the most of the emission comes from:
The maximum signal amplitude is obtained when the probe is exactly above the transformer, as we would expect. Moving away from this point shows the amplitude of the signal decreasing rapidly. With this probe, no signal is measured from around the input and output cables or above the load resistor. The signal shows a kind of square wave with a frequency of 227.6 KHz – which is probably the SMPS switching frequency – and strong overshoots at the rising and falling edges. The oscilloscope seems to identify a frequency of around 3 MHz, maybe the overshoot resonance frequency? Interestingly the strongest interferences on my radio were around the 3 MHz band.
This probe could be used for detecting the area of radiations on a circuit board but it will have to be re-designed and made safer with a better insulation and more distance between the BNC and the loop. This one was just OK to do testing outside the SMPS cabinet.
The next probe, the shielded loop, is a very popular design found in many articles on EMI. It is quite difficult to make, because of the gap on the coaxial shield (see picture below). This gap weakens the cable at this point making it almost impossible to make a round loop. I used a lot of heat shrink to reinforce that point and this might be why the probe is not very sensitive. On the right hand side, on the drawing below, both shield and center conductor are soldered to the shield. As bizarre as it looks, this is not a drawing error! It is indeed a loop like the previous one but it is shielded and the shield is open at the top of the loop.
We can see the result is very similar to the one with the unshielded loop. However I was able to observe a small signal around the mains cable (input) which could indicate that this probe could be slightly more sensitive. I will not use this probe until I find a better way of building it, including with better insulation.
Now it’s time to crank up the volume with something more sensitive. Using a small toroid core ferrite cut in half and winding a few turns of transformer wire (enameled) et voilà!
Using a ball pen body as insulation, it doesn’t look very pretty but, hey… just to see if we can get a more sensitive probe using the ferrite core… And look at that!
This one shows a much stronger signal (140 mVpp instead of 37 mVpp) and everywhere around the SMPS including above the input cable (mains), output wires and load resistor. We cannot differentiate the square wave anymore but the oscilloscope still sees a switching frequency of 227.4 KHz. This probe is too sensitive to pinpoint the main area of radiations in this particular situation but will be useful to measure weaker signals. I am probably going to build a few more with different types and sizes of ferrites.
Now what about the electric field probes? I found a couple of designs and the one I selected is the following:
The construction is simple, a piece of coaxial cable connected to a BNC connector at one end. The center conductor is exposed by approximately 10mm at the other end. The whole probe should be insulated. For this purpose I used an old ball pen body again.
The results are interesting: with this probe the switching frequency is clearly visible as a square wave and the overshoots minimized. The signal is strong above the transformer, as with the loop, but we measure an even stronger signal when hovering above the load resistor and over the output wires…
This is measured above the transformer:
And this is measured when the probe it over to the output wire or hovering above the load resistor:
Well, this is fun and we could try different things and different designs for hours. But now that we can “see” the EMI, what can we do about it? This is a whole science and there are books and books about it, many of them full of formulas and design rules. In our case, without changing anything in the SMPS design – not even opening the casing – we can add filters. One input filter and one output filter. Here we go:
This input filter is a huge overkill, but it was the best available in my workshop at the moment. It is a two-stage, third order filter, with a current rating of 20A! As we can see it is almost as big as the SMPS! I tried a couple of smaller filters from my junk box but no one gave as good results as this one…
The output filter is a standard LC low pass filter. The homemade inductor has an inductance of 4.06 mH and the capacitor a value of 1200 uF. This should filter any frequency above 72 Hz, the higher the frequency, the better the filtering.
With the help of this heavy artillery the output voltage is now OK. This is the output, using the same settings (50 mV/Div and 1.00 us/Div) as the first measurement:
Checking with the ferrite probe and the electric field probe there is no more signal over the input and output wires, nor over the load resistor. On the radio side, the interferences have been strongly reduced. The input and output cables were acting like antennas for the EMI signal and this signal is now blocked by the filters. Using the ferrite probe we can see the efficiency of the input filter by checking the signal before and after the filter. A strong signal is present between the SMPS and the filter and nothing can be measured on the other side of the filter.
Above the transformer we still measure a strong signal as we could expect. There is nothing we can do here without modifying the design and shielding of the SMPS.
Conclusions:
Although not calibrated, these DIY probes can be very useful to detect the presence of EMI signals and compare the situations before and after filtering or design modifications without risking the life of your oscilloscope. They can help you to pinpoint the location that is radiating EMI in a circuit. You can see the switching frequency and duty cycle of an SMPS without even opening the casing! There are many designs available on the Internet that might be worthwhile to test if you are interested in developing further. However, one world of caution: Beware of the insulation! Do not use these probes nearby High Voltage points in your circuits!
What about my project? Well this one will be fine for time being:
The good old fashion linear power supply will do the job until I find a suitable SMPS which, I am sure exists out there…
Gerald Musy
Penang, Malaysia
Please give a support by clicking on the social buttons below. Your feedback on the post is welcome. Please leave it in the comments.
P.S- If you enjoyed reading this, click here to subscribe to my blog (free subscription). That way, you’ll never miss a post. You can also forward this website link to your friends and colleagues-thanks!
You may check out his other article below:
https://jestineyong.com/quick-understanding-of-led-bulb-turning-blue/
(99)Dislikes
(1)
25 Comments
Leave a Reply
Cancel reply
skwong
October 25, 2016 at 11:28 am
Hi Gerald Musy,
Very good article on EMI, thanks for sharing. I have also tested EMI on the oscilloscope just like what you have done for a company.I found out that the oscilloscope can only display a single frequency i.e it cannot display critical interference frequencies which existed. Later on we are forced to purchase a spectrum analyzer in which the whole spectrum of interference frequencies can be observed from 150khz to 30Mhz. Most of the equipment failed EMI on the lower frequency range of 150Khz to 1Mhz. Radio is a cheap and easy method to correlate and test EMI interference signal strength.
Paris Azis
October 25, 2016 at 3:24 pm
Skwong,you are absolutely right!
Best Regards
Gerald
October 25, 2016 at 6:24 pm
Hi SK,
Thanks for your comment and support. You are correct that the spectrum analyzer is the way to go, unfortunately those are pretty expensive. The radio is good too, I use and ICOM IC-R6 that covers all the frequencies from 100 kHz up to 1.3 GHz. For the SMPs the high frequency EMI are a byproduct of the lower frequency (harmonics)so if you get rid of the low frequency EMI you will also get rid of the high frequencies as well.
Cheers,
Gerald
Albert van Bemmelen
October 25, 2016 at 2:31 pm
Super informative article on EMI Gerald! Something I could build easily too like you have showed. I wonder though why you left out building the two-stage input third order 20A filter. But I guess that probably was because it already was pre-made by a Manufacturer?
I previously was thinking about buying a 9KHz to 3GHz Spectrum Analyzer. I like the old Tektronix 2712 (1997 manufactured?). But since 5.8GHz already is used in our homes which makes 3GHz Spectrum Analyzers kind of useless already, it kept me from buying one. But if prices because of the rising higher frequencies drop, I probably will in the near future.
Gerald
October 25, 2016 at 6:41 pm
Hi Albert,
Thanks for your comments and support. Correct, the input filter is an off-the-shelf item, so much easier than building my own 🙂
Yes, spectrum analyzers are still expensive. Something I also use is a SDR radio. It costs a few dollars only and goes until 1.7 GHz. You might find one that goes higher in frequency. The main problem with the SDR is that it works with a PC and is picking up all the harmonics generated by the PC clocks... Nothing is perfect in this world!
Cheers,
Gerald
Yogesh Panchal
October 25, 2016 at 2:56 pm
Sir,
Thanks for sharing informative article.
Gerald
October 25, 2016 at 6:42 pm
Thanks for your support Yogesh
Paris Azis
October 25, 2016 at 3:29 pm
Very good presentation of this vital topic Gerald. Thanks for your effort you put in it and for sharing it as well.
It is always a pleasure for me to read your so informative articles.
Best Regards
Gerald
October 25, 2016 at 6:43 pm
Thanks for your support Paris
Robert Calk
October 25, 2016 at 8:20 pm
Thanks for the nice article, Gerald.
Gerald
October 28, 2016 at 4:01 pm
Thanks for your support Robert. Haven't seen an article from you for long time. Hope you are well and preparing a big one!
Cheers,
GM
Robert Calk
October 30, 2016 at 8:07 pm
Hi Gerald, Right now I'm working on restoring, and repairing if necessary, and if I can(lol), a 1969 Bulova Accutron 214 gold watch that was my dads. It has a 214 3-wire movement without a hack spring, 10K gold filled dagger lug(one of the lugs is bent) case, with a gold coined dial. Wasn't sure I would write an article about it since the electronics in it are pretty simple, except for the tuning fork and coils. The only modification I'll do is installing a Schottky diode to drop the voltage .2V so I can use modern 1.5 silver oxide button cells, since it was designed to use the antiquated 1.35V mercury cells. But I will write an article on it if you guys want me to; or if I don't ruin it. lol!
Robert Calk
October 30, 2016 at 8:13 pm
I forgot to mention that it has a cracked crystal, so I'll be ordering a new sapphire crystal for it, and some more watchmaker tools and stuff Thursday. I don't have the money right now.
Gerald
October 31, 2016 at 10:30 am
Hi Robert,
Of course this will be a very interesting article, something totally different. I remember the very first Bulova Accutron watches when I was a teenager. I was dreaming to own one but they were much too expensive for my CHF 30.00/month apprentice salary 🙂 They were advertising 1 min/month accuracy, totally mind blowing at that time!
It's a very brave repair that you are attempting. Good luck and looking forward to read about.
Cheers,
GM
Robert Calk
October 31, 2016 at 3:14 pm
Yeah, I was only 9 years old when this watch was made. I'll admit that the only reason I'm doing the work is to learn, and because it's in poor condition with no watchband. If it were in mint condition I wouldn't take the chance of ruining the movement, or breaking the lug to try to straighten it. It has sentimental value though, but I'm hoping that I will succeed. I'm doing as much study as I can find on the web, and in books.
Chris Massie
October 25, 2016 at 11:55 pm
A great, practical article on EMI. And thank you for the visuals. I appreciate your sharing this info Gerald. Excellent work.
All the best
Gerald
October 28, 2016 at 4:02 pm
Thanks for your support Chris
Lee
October 26, 2016 at 6:39 am
Nice informative article Gerald
Gerald
October 28, 2016 at 4:02 pm
Thanks for your support Lee
ivan (CHILE)
October 30, 2016 at 7:49 am
Thank you for sharing your knowledge.
Humberto
October 31, 2016 at 11:54 pm
Thanks Gerald for sharing a so valuable information in your article.
Gavin Harris
December 2, 2016 at 11:44 pm
First off, great article, seems to be what I am looking for to aid in my own project. I did have a question about the probes, you warn at the end about high voltage. I have a spark ignition system in a custom engineering project. Initially, I was running unshielded wire, in which the EMF or EMI spiked the voltage on a pressure transducer, I think this was the cause of the transducer ceasing to function. Since then, I've replaced all the wiring with double shielded wiring on both the spark system and the transducer, as well as encased the spark wires in a grounded metal conduit. The spark system is putting out around 10kV to 12kV. With the shielding, would this still be too high to use the probes to ensure that I've cleaned the system up enough to keep from damaging another transducer?
Raju Varghese
December 23, 2016 at 10:19 pm
Dear Gerald,
great job , thank you so much your valuable information
Ulises aguilar pazzani
September 29, 2017 at 12:10 pm
Great MR Musy thks for sharing
Varshusaran
January 21, 2018 at 11:54 pm
Hello...
This was very informative...and inspired me alot....
Thank u for sharibg your ideas...