EMC equipment: Coupling-Decoupling Networks CDN DIY, EN 61000-4-6

[piotekp90]

Hi,
The price of EMC-related instruments can cause headaches. Therefore, why not try to build some EMC equipment yourself, such as coupling-decoupling networks, especially if you have a vector network analyzer. After reading the information in EN 61000-4-6 standard, I decided to make the CDN M2, M3, S24, and appropriate calibration adapters myself. I used Wurth ferrite cores 74272251, 74271221, 7427153, 74270044, Vacuumschmelze core T60006-L2040 and Kemet PME 271 Y2 series capacitors to make these. The CDNs are based on the GAINTA BS 39MF aluminum enclosure. On the EUT side, the front of the case was cut out and the 3D print was screwed to it. For attaching the cores, a base was printed which further reduces the parasitic capacitance between the housing and the wires inside the CDN. The measured impedance values are within the range provided by the standards in 150 kHz - 230 MHz range The HP 8753E vector network analyzer was used to measure the impedance, along with adapters suitable for the type of CDN.

30W broadband RF amplifier in the 100 kHz - 230 MHz band under construction - I will share the results once I have developed the final version;D

There is a CDN RJ45 waiting in the queue to be built, if anyone has pictures of the inside of this type of network then please contact me or share pictures 

[piotekp90]

CDN M2 and S24 USB C

[Vtech]

WOW! Great build! 
Exactly what I was looking for. Maybe a bit too professional for me .

I started by bodging together a 50W RF amplifier. Mine is a piece of crap but should do the job after calibration. It is based on some ancient RACAL military amplifier overdriven to the extreme and a Chinese 2W amp as a preamplifier.


This thing can work from around 7MHz up to 80MHz. I don't need more. Problems only occur above 10MHz and I don't have to test above 80MHz.
I already managed to fry a scope probe with it (guess what, scope probes can't handle the RF power )
It is quite non-linear and generates a lot of harmonics but should be enough for recreating the problem found in the proper EMC lab.

Now I need to bodge together a 6dB/50W attenuator (easy) and this M3 CDN which is some devil's invention (the hard part). My CDN probably won't have a case and will be heavily inspired by the photo found in the EMC Standards guide (linked below).

You've probably already seen this guide but I'll leave it here for others: https://www.emcstandards.co.uk/files/61000-4-6_immunity_to_conducted_rfi_2_1.pdf
There is also this useful article: http://www.elmac.co.uk/PNP_61000-4-6-a.pdf

A company called TEKBOX from Vietnam has a lot of cheap EMC equipment. It is really nicely designed and manufactured but a lot cheaper than professional stuff - good enough for pre-complience. I use their LISN for emissions and it's great.
They have a useful manual for their CDNs: https://www.tekbox.com/product/tbcdn-m3-coupling-decoupling-networks-m3/

[nctnico]

CDN M2 and S24 USB C

Thanks for sharing! 

[wilhe_jo]

Hmmm... Quite a lot of ferrites... I have a quite simple "Kit" for my CDNs... Just 2 cores of cable for 80MHZ... 2 or 3 random 100MHZ clap ones get it to 230MHZ.

I get you the part numbers.. in the end the core component is an epcos smps core

73

[piotekp90]

For measurements of the amplifier it is worth making a simple RF sampler which is described here: http://roland.cordesses.free.fr/rfcoupler.html, it works well up to 300-400 MHz or use a directional coupler. Then there is no risk of damaging the probe

The reason for the large number of ferrites is to make sure that the CDNs have the right impedance no matter what is connected on the AE side (short or open). Each ferrite is for a slightly different range, for example, the shorting on the AE side had a big impact on the impedance below 1MHz so I added this large Wurth ferrite which provides adequate decoupling 

[wilhe_jo]

As promised, my go-to for quick CDNs (I do pre-compliance testing and "fault-finding" work, so I regularly need CDNs with specific connectors that you can't really get off-the-shelf):

some (>7) turns on Epcos B66397G0000X187 (that's already the needed 250µH!)
+ some (>3) turns on Würth 74270097 (that helps with 5-80MHz)
( + 2-3 clap-ons or another 74270097,  if I really need to go above 100MHz).


Works 99% of the time.
if the cable gets thicker than maybe 8mm, I need to "split" the big (abt 250µH) inductor formed by the epcos core.

Most difficult part is getting the capacitors right... usually those foil ones have nasty resonances.
For the resistors, I tend to go for TO220 thick film resistors... those give me 0-inductance every time.
With other types, Murphy always gets me and I pick the one with a nasty impedance the first try

I attached a photo of my "kit" (in the back you can also see my lisn; it uses the same concept)...

It's basically a candemboss enclosure and some PCB-panels.
Whenever I need a special connector, I 3d print a matching plastic part, glue that into the hole and add the cabeling (so no drilling or other mechanical work needed).
No photo from the inside... I use cable straps because of space constraints and that doesn't really look that neat

All-in-all its maybe a 2h job to have a new CDN (including calibration and letting the glue set).

Ground connection is simply done with cheap copper tape used to protect plants from snails (so it's the cheap non-conductive-sticky-tape-stuff).
That is screwed 2 times to the metal chassis of the CDN. So-far that's pretty reliable (I tried soldering it as well, but that's not needed in my experience).
I also tested metal screws instead of the rubber feet.
For my liking, the copper-tape works better... maybe because the rubber feed press the tape more evenly to my ground plane than the stiff metal feed do (they may unnoticeably rock because of the 4 point contact).

I have quite some experience with commercial ones....
After some time, they all need extra weight on them to get good ground connection (otherwise you fail the calibration >100MHz).
So far, my copper strips seem to do a better job... but their contact surface need to "mature" a bit to give me a real comparison.... right from the supplier the commercial ones also work without any additional downward forces.

BTW: if you look carefully, you see these hex-standoffs.
I use these for calibration. I have made some PCB that holds a BNC right at the center of the hole (or the mains connector for the LISN).
With these, making individual cal-adapters is done in minutes as well

I was thinking of selling these as kits (cores, enclosures, panels + hardware to fix the cores).
For now, I'm my own customer because I couldn't see real demand...

After building "some" (around 10), I find the enclosure is a little tight.
So the next batch of panels will be for the next bigger enclosure and maybe, I revisit 3d printed core-holders since Prusa now has V0 rated PETG that I really want to try for one-off enclosures for prototypes

73

[piotekp90]

As promised, my go-to for quick CDNs (I do pre-compliance testing and "fault-finding" work, so I regularly need CDNs with specific connectors that you can't really get off-the-shelf):

some (>7) turns on Epcos B66397G0000X187 (that's already the needed 250µH!)
+ some (>3) turns on Würth 74270097 (that helps with 5-80MHz)
( + 2-3 clap-ons or another 74270097,  if I really need to go above 100MHz).


Works 99% of the time.
if the cable gets thicker than maybe 8mm, I need to "split" the big (abt 250µH) inductor formed by the epcos core.

Most difficult part is getting the capacitors right... usually those foil ones have nasty resonances.
For the resistors, I tend to go for TO220 thick film resistors... those give me 0-inductance every time.
With other types, Murphy always gets me and I pick the one with a nasty impedance the first try

I attached a photo of my "kit" (in the back you can also see my lisn; it uses the same concept)...

It's basically a candemboss enclosure and some PCB-panels.
Whenever I need a special connector, I 3d print a matching plastic part, glue that into the hole and add the cabeling (so no drilling or other mechanical work needed).
No photo from the inside... I use cable straps because of space constraints and that doesn't really look that neat

All-in-all its maybe a 2h job to have a new CDN (including calibration and letting the glue set).

Ground connection is simply done with cheap copper tape used to protect plants from snails (so it's the cheap non-conductive-sticky-tape-stuff).
That is screwed 2 times to the metal chassis of the CDN. So-far that's pretty reliable (I tried soldering it as well, but that's not needed in my experience).
I also tested metal screws instead of the rubber feet.
For my liking, the copper-tape works better... maybe because the rubber feed press the tape more evenly to my ground plane than the stiff metal feed do (they may unnoticeably rock because of the 4 point contact).

I have quite some experience with commercial ones....
After some time, they all need extra weight on them to get good ground connection (otherwise you fail the calibration >100MHz).
So far, my copper strips seem to do a better job... but their contact surface need to "mature" a bit to give me a real comparison.... right from the supplier the commercial ones also work without any additional downward forces.

BTW: if you look carefully, you see these hex-standoffs.
I use these for calibration. I have made some PCB that holds a BNC right at the center of the hole (or the mains connector for the LISN).
With these, making individual cal-adapters is done in minutes as well

I was thinking of selling these as kits (cores, enclosures, panels + hardware to fix the cores).
For now, I'm my own customer because I couldn't see real demand...

After building "some" (around 10), I find the enclosure is a little tight.
So the next batch of panels will be for the next bigger enclosure and maybe, I revisit 3d printed core-holders since Prusa now has V0 rated PETG that I really want to try for one-off enclosures for prototypes

73

Very nice job ! and clever solution with PCB front panel 
Good idea with the tape used for earthing, I hadn't thought of it as easy to ensure a proper connection.

Which resistor are you using?
Did you build the CDN T8 RJ45?



I am wondering how to make L2 coil in such a way that it does not adversely affect the Ethernet interface....
I'm thinking of using any of these CMCs in this role: https://www.mouser.pl/ProductDetail/Wurth-Elektronik/744290683?qs=Zt3UNFD9mQhnz1qxepZ48w%3D%3D , https://www.mouser.pl/ProductDetail/EPCOS-TDK/B82720H0015A035?qs=pGJ4H8VyKtUtaQl7VCQQSw%3D%3D , https://www.mouser.pl/ProductDetail/Bourns/SRF0905A-652Y?qs=xGZga0pI15Pqz3g3oimQVg%3D%3D , https://www.mouser.pl/ProductDetail/KEMET/SU9V-R01180?qs=iaWy59%2Fd4KAQ3WHbQAv1lw%3D%3D . How do you think any of these CMCs could be suitable for this?

[KlausKragelund]

Very nice.

I build a CDN-T2 a couple of months ago, did not want to spend 700 USD on a new one.

Build it from the standard, and along the way found a good guide on what to watch out for:

http://www.elmac.co.uk/PNP_61000-4-6-a.pdf

Later I bought a "professional" CDN, CON AF2E. Pictures attached. Extremely simple

[seuberte]

As promised, my go-to for quick CDNs (I do pre-compliance testing and "fault-finding" work, so I regularly need CDNs with specific connectors that you can't really get off-the-shelf):

some (>7) turns on Epcos B66397G0000X187 (that's already the needed 250µH!)
+ some (>3) turns on Würth 74270097 (that helps with 5-80MHz)
( + 2-3 clap-ons or another 74270097,  if I really need to go above 100MHz).


Works 99% of the time.
if the cable gets thicker than maybe 8mm, I need to "split" the big (abt 250µH) inductor formed by the epcos core.

Most difficult part is getting the capacitors right... usually those foil ones have nasty resonances.
For the resistors, I tend to go for TO220 thick film resistors... those give me 0-inductance every time.
With other types, Murphy always gets me and I pick the one with a nasty impedance the first try

I attached a photo of my "kit" (in the back you can also see my lisn; it uses the same concept)...

It's basically a candemboss enclosure and some PCB-panels.
Whenever I need a special connector, I 3d print a matching plastic part, glue that into the hole and add the cabeling (so no drilling or other mechanical work needed).
No photo from the inside... I use cable straps because of space constraints and that doesn't really look that neat

All-in-all its maybe a 2h job to have a new CDN (including calibration and letting the glue set).

Ground connection is simply done with cheap copper tape used to protect plants from snails (so it's the cheap non-conductive-sticky-tape-stuff).
That is screwed 2 times to the metal chassis of the CDN. So-far that's pretty reliable (I tried soldering it as well, but that's not needed in my experience).
I also tested metal screws instead of the rubber feet.
For my liking, the copper-tape works better... maybe because the rubber feed press the tape more evenly to my ground plane than the stiff metal feed do (they may unnoticeably rock because of the 4 point contact).

I have quite some experience with commercial ones....
After some time, they all need extra weight on them to get good ground connection (otherwise you fail the calibration >100MHz).
So far, my copper strips seem to do a better job... but their contact surface need to "mature" a bit to give me a real comparison.... right from the supplier the commercial ones also work without any additional downward forces.

BTW: if you look carefully, you see these hex-standoffs.
I use these for calibration. I have made some PCB that holds a BNC right at the center of the hole (or the mains connector for the LISN).
With these, making individual cal-adapters is done in minutes as well

I was thinking of selling these as kits (cores, enclosures, panels + hardware to fix the cores).
For now, I'm my own customer because I couldn't see real demand...

After building "some" (around 10), I find the enclosure is a little tight.
So the next batch of panels will be for the next bigger enclosure and maybe, I revisit 3d printed core-holders since Prusa now has V0 rated PETG that I really want to try for one-off enclosures for prototypes

73

Any thoughts about open sourcing the designs? Could always manufacture them, but if some fellow engineers want to save a buck or have use it as a learning experience, would be great to be able to DIY. 

[seuberte]

Just wanted to clarify, the RF in is connected to the USB shield (through the 100 R), otherwise the cable is just wrapped through the cores?

I'm looking at the standard and your M2 diagram. Do the lines around the inductors imply something particular, like the shield and the lines are inductively coupled through the same iron core? Is that just another symbol for a common mode choke?

[Vtech]

Your schematic is for CDN-S1 designed for shielded cables (like coax). For those cables, only the shield is coupled through the 100 Ohm resistor to the signal input (signal line will be coupled as well through signal-shield capacitance).

I think this design can be used for shielded usb cable (specially something like usb 3 with many data lines). For simple 4-wire usb 2.0 you could couple the signal through 400 Ohm resistor and 10nF cap for each wire. You can extend this approach for any number of signals. For 1 signal the coupling resistor should be 100 Ohm, for 2 signals 200 Ohm, for 3 signals 300 etc. You have to keep the parallel resistance of all the coupling resistors at 100 Ohm (together with 50 Ohm generator resistance they form the required 150 Ohm coupling resistance/impedance).

The lines around the inductors indicate coupled inductors (as you figured out). In fact, this is a single core that the cable is passed through (it doesn't have to be only 1 core but all the signal lines and shields are wrapped around all the ferries together/parallel). The inductor part is for decoupling the auxiliary equipment side (AE) from the device under test side (DUT). The standard requires maintaining 150 Ohm impedance from DUT side to ground (with signal input terminated with 50 Ohm) in the tested frequency range (150kHz to 80Mhz or 230MHz). This requirement has to be fulfilled for AE side shorted to ground (PE). See the graph in the first post. This is the reason for using multiple ferrites to keep the AE-DUT impedance high in the required frequency range. This is the hard part.

[wilhe_jo]

Which resistor are you using?

I use "random" TO220 thick film resistors... so far, none of them had issues up to 230MHz...

Did you build the CDN T8 RJ45?

I did a CDN T4 ... the problem is that you need to decouple common-mode and differential mode.

So you need the cores to get the CM-Impedance and then you need some additional inductors to get the signal decoupled from your interference.

These SMD CM-chokes for data-lines were ok for me... just be carefull with the current-rating.
3V test-level should be no issue, 10V, however, are tricky to find.

In the standard are suggested inductance-values... just stack some (tiny, small, large, huge) to get into the desired range. (IIRC, I used 3 different values).
Remember, if you use, these CM-chokes wrong, they get useful for diff-mode signals

Any thoughts about open sourcing the designs?

Yes... I even considered kind of a building kit (ie. panels and cores; you'll need to print the plastic mount for the connectors, get some cable, and add your resistors,...)
So far no real demand... hence, the documentation postponed
This might be some chicken/egg problem, I guess...

73

[piotekp90]

As I wrote earlier, I made an amplifier to power the CDN  It operates in the 100kHz - 230MHz range (and even 300MHz at lower power) and achieves 20W and even 35W for some frequencies. The distortion level is better than -18dBc. MRF101A/B transistors are used as output stages. The driver for these is a BLP15H9S100GZ fed from a GALI84+ preamplifier.

The gain is approximately 47dB +/- 1.5dB. Power consumption is 150W, 24V power supply.
What remained was to clean up the PCB, build in an attenuator to improve the S11 and protect against an overly high input signal, improve the bias of the transistors to drift less with temperature, add overcurrent protection.

[Kuba_eM]

Hi, I am asking the question on behalf of a friend who does not have an EEVblog account. Perhaps the information will also be useful to other forum users. He is interested in creating a test network according to your description. Several different ferrite cores were used there, but one (Vacuumschmelze core T60006-L2040) has many types defined by the further digits of the catalog number. These types differ significantly in inductance. Is there a chance to obtain information about the full order number of this Vacuumschmelze core T60006-L2040? The next question concerns the amplifier: is there a chance to share the diagram? My friend is building his own EMC LAB, hence his interest in the subject.

[basilic]

Hi,

We (my company) have purchased equipment for our EMC tests. We are trying to build our CDNs (for the first tests, an M2 and an AF2). We are using the tracking generator of our SSA3075X, the EMCView Pro software, the TBMDA-CDN25 amplifier, and the Tekbox calibration kit.

After building the M2 CDN, we attempted the calibration to achieve 10V rms at the output. Following some poor resistor choices (too inductive) and after reading up on this topic, we are now able to reach the level and reproduce the failures on a product.

For consistency, we want to verify the various elements of the CDN. While I understand how to measure RFin/EUt and RFin/EA attenuation, I do not understand the method for measuring the common mode impedance.

Could you please clarify this point for me?