DIY DM /CM Seperator for EMC - LISN Mate

[T3sl4co1l]

Another good way to measure it.

Yeah, two twisted pairs in parallel will do; and, it won't matter for short windings like these, but long windings like the injection transformer, may want to be aware of something: the pairs in Ethernet cable are intentionally twisted at different rates, which gives them slightly different velocities.  It may be desirable to use just one pair for all windings, take a length, cut it in half, wind those in parallel.

For very low impedances, there is a particular pattern: star quad.  Take four strands and twist them together; looking at the bundle end-on, going around the circle: connect the strands alternately to + and -.  This gives a quadrupole arrangement which gives about half the impedance again -- i.e. about 25 ohms for this type of wire!

(A pretty handy tip, when you need transformers with an impedance give or take a factor of 2 from 50 ohms -- 25 and 100 ohms are common needs.)

(For even lower, I would suggest parallel-plate transmission line, built with flex or tape.  Or use lots of pairs/quads wired in parallel.)

Tim

[TimNJ]

@Jay_Diddy_B

Due to the whacked out image attachment situation, can you attach the functional schematic (with inductors shown) one more time? Maybe just try a regular ol' attachment, and no inline image.

Thanks!

[Jay_Diddy_B]

TimNJ,

Let me try again:







Have also attached a pdf version.

Regards,
Jay_Diddy_B

[TimNJ]

@Jay_Diddy_B..

The Virginia Tech circuit, as far as I can tell, uses a 3rd transformer for impedance matching 25 ohm to 50 ohm, hence the sqrt(2): 1 turns ratio. But, in your circuit, seems you terminate the Zo = 100R transmission line transformers into 50R.

Also, I have no idea what I'm talking about.

Just wondering what's going on there.

Thanks.

[Jay_Diddy_B]

TimNJ and the group,

In this circuit:



The function of T1 is the same as in the one I built. R_int is 50 so when the spectrum analyzer is connected the common mode impedance is 25 . This is R_int // Spectrum analyzer.


T3 in this design is a common mode choke. It references the differential signal between A-B to ground. So it is a balun, balanced to unbalanced. The differential impedance between A  and B should be 100 . T3 should be built with a 100 transmission line.

T2, with a turns ratio of 1.41 : 1, has an impedance ratio of 2:1 matching the 100 to the 50 of the spectrum analyzer.

The transformer would have to be approximated to 3 : 2 or 7 : 5

In the one I built R2 and the 50 load, the spectrum analyzer are essentially in series, making a 100 differential load. There is a 6dB reduction in the signal, half the signal on R2 the other half on the spectrum analyzer. But I don't need the third transformer. There is also better symmetry.

Does that help?

Regards,
Jay_Diddy_B

[TimNJ]

Thanks for your explanation and clarification. So, the DM output signal is attenuated by 6dB, but the CM output signal is ~0dB? If that's the case, I'd almost prefer to put a 6dB pad on the CM signal.

Just thinking about how I typically approach EMI debugging...I might want the signals to at least be scaled to each other.  (Perhaps in situations where you have CM and DM overlapping.)

Thanks.

[T3sl4co1l]

Rint is 25 ohms, so if you use 50 ohms in parallel with the receiver's 50 ohms, you lose 3dB (power) to that, but 0dB voltage, and 6dB current.

Same for DM: it's down by 3dB (power), but 6dB voltage, and 0dB current.

So, depends on which mode and impedance you're calibrating to, and if you're more interested in voltage/current or power.

Tim

[TimNJ]

Rint is 25 ohms, so if you use 50 ohms in parallel with the receiver's 50 ohms, you lose 3dB (power) to that, but 0dB voltage, and 6dB current.

Same for DM: it's down by 3dB (power), but 6dB voltage, and 0dB current.

So, depends on which mode and impedance you're calibrating to, and if you're more interested in voltage/current or power.

Tim

Thanks much. I still don't have an intuitive grasp on a lot of these fundamentals...clearly. For CM, I'm a bit perplexed that the voltage at the receiver wouldn't drop by 6dB if the equivalent resistance across the receiver is 25R as opposed to 50R.

The receiver measures in dBuV since the EMI limits are specified as such. It would be nice if you could get the actual level (maybe with the exception of -0.5-1dB insertion loss).

I'd like to re-iterate that I punching above my weight, but is this vaguely what the matching circuit would need to look like? Could Coilcraft PWB (https://www.coilcraft.com/getmedia/2623c26b-542e-49f8-90ee-c7b1fe27d34b/pwb.pdf) work if using the 1:2 part?

I appreciate all of your guidance (and patience with me). 

[Jay_Diddy_B]

Hi TimNJ,

If you don't use a DM / CM Separator and you connect one of the LISN directly to the spectrum analyzer you will measure half the DM mode noise.

So you read -6dB down.

This is because the other LISN is terminated in a dummy load.

So being -6dB isn't necessarily a bad thing.

Jay_Diddy_B

[Jay_Diddy_B]

Rint is 25 ohms, so if you use 50 ohms in parallel with the receiver's 50 ohms, you lose 3dB (power) to that, but 0dB voltage, and 6dB current.

Same for DM: it's down by 3dB (power), but 6dB voltage, and 0dB current.

So, depends on which mode and impedance you're calibrating to, and if you're more interested in voltage/current or power.

Tim

Tim,

R_int is not 25

R_int in parallel with the spectrum analyzer = 25 .



The CM impedance should be impedance of two LISNs in parallel.

Similarly the DM impedance is two LISNs in series.

It is how the EMC specifications are written. The EMC specifications specify that you use two LISN and you measure them individually. The EMC specifications do not distinguish between CM and DM noise.

There are several schematics published by VPEC, this one from the patent application, I believe to be correct:




The patent has expired:




(BTW I agree that Z₀ on T2 is 100 .

I don't agree that Z₀ on T1 should be 50 . The reason I don't agree is because the signal is flowing in opposite directions in the transformer windings. The impedance of the transformer looks a shorted transmission line, half the electrical length of the winding. The transformer should be as small as possible, while achieving enough inductance. This minimizes the length of the line.)

Regards,
Jay_Diddy_B

[TimNJ]

Ah! Well now it sounds so obvious. So, the output of the separator’s DM port is 6dB from the actual level, but that’s no different than the normal measurement...so actually no problem at all.

Now it’s time to get on with building it.

[T3sl4co1l]

I was referring to this one,



(Which in the CM-DM separator, Rint is 50R || spec.)

Tim

[TimNJ]

Hi,

I'm testing my version of the separator. Does it make sense that I'm getting -6dB insertion loss from LISN1 (or LISN 2) to CM Out, when I apply a signal to LISN1 OR LISN2 (not simultaneously)? The result from LISN1 (or LISN2) to CM Out is actually the same as the result from LISN 1 (or LISN2) to DM Out. My intuition says that very little signal should arrive at CM out...but that doesn't seem to be the case. On the other hand, when testing with a resistive splitter (to inject a common-mode signal), I see -30dB at the DM Out port, and -6dB at CM out...which makes sense in my mind.

All unused BNCs capped with 50-ohm terminators.

What gives?

Thanks.

[T3sl4co1l]

Normal mode (just one port) divides evenly between CM and DM, so both should be -6dB each.

Tim

[TimNJ]

Thank you, Tim. I think I understand. If you wanted to directly test the rejection of DM at the CM Output port, would you need to drive both inputs, but one 180 degrees out of phase? Maybe with a 180 degree splitter?

[T3sl4co1l]

Exactly.

Well, depending on which one you're measuring and rejecting.  I'm not sure I understand which one you're talking about.  But driving them in-phase is common mode and rejects differential; opposite-phase is differential mode and rejects common.

Can also do it the other way, drive the DM or CM port and check the balance (amplitude and phase) on the ports.

Tim

[uski]

Hi !

So after having success with Jay_Diddy_B's LISN design, I find this topic... and I would love to try to build myself one of these "LISN mate"!

They are both transmission line transformers and have a turns ratio of 1:1 but they have different transmission line impedances.
Both transformer were wound on TDK H5C2 T16-4-8E cores. These cores may be hard to get. If there is enough interest I can try some other cores that are more readily available.

H5C2 is a high permeability (10K). The datasheet says it is suitable for low frequency (50KHz) applications. But who reads the datasheet? 

As expected this core is nowhere to be found. Is there a more readily available core I could use ?

Something like this maybe ? But I can't even find the specs of the original cores used so I don't know if this is correct either magnetically or physically
https://www.digikey.com/en/products/detail/ferroxcube/TX29-19-7-6-3E6/8021257

[Jay_Diddy_B]

Hi,

The TDK Part No H5C2 T16-4-8E

Can be decoded:

H5C2 is the material

T is the shape, toroid

16 is the outside diameter in mm

4 is the thickness mm

8 is the inside diameter mm

E is the coating, epoxy


The 3E6 material might work.
T38 is another material that might work.

I bought some T38 cores B64290L0044X038.I haven't tried them.

Regards,
Jay_Diddy_B

[TimNJ]

Exactly.

Well, depending on which one you're measuring and rejecting.  I'm not sure I understand which one you're talking about.  But driving them in-phase is common mode and rejects differential; opposite-phase is differential mode and rejects common.

Can also do it the other way, drive the DM or CM port and check the balance (amplitude and phase) on the ports.

Tim

Thanks again! I may try both ways. We’ll see.

Hi !

So after having success with Jay_Diddy_B's LISN design, I find this topic... and I would love to try to build myself one of these "LISN mate"!

They are both transmission line transformers and have a turns ratio of 1:1 but they have different transmission line impedances.
Both transformer were wound on TDK H5C2 T16-4-8E cores. These cores may be hard to get. If there is enough interest I can try some other cores that are more readily available.

H5C2 is a high permeability (10K). The datasheet says it is suitable for low frequency (50KHz) applications. But who reads the datasheet? 

As expected this core is nowhere to be found. Is there a more readily available core I could use ?

Something like this maybe ? But I can't even find the specs of the original cores used so I don't know if this is correct either magnetically or physically
https://www.digikey.com/en/products/detail/ferroxcube/TX29-19-7-6-3E6/8021257

I have built my own version with JDB’s recommended core. I’ve also purchased alternate cores from TDK and Ferroxcube. I will test them soon and leave an update here. I also have a PCB design I can share.

[bluesky1688]

Hi All,

I'm not a transformer or an EMC person but tasked to understand the FCC conducted and radiated emission test. So, any help with this task would be greatly appreciated. Fortunately I stumbled upon this excellent site with so many subject matter experts.

I first started looking for the now obsoleted LF-428 transformer with no luck then found this site. Is the TEKBOX setup cited in the following EDN article the, post LF-428, modern way to go?

 https://www.edn.com/review-tekbox-lisn-mate-is-valuable-for-evaluating-filter-circuits/

Thanks,
William

[Noy]

I also have a PCB design I can share.

Would you like to share your pcb files here? Are they kicad? Would like to build this "lisn mate" but i will use a 1590B case.

[TimNJ]

Greetings. Sorry for the delay on this. I've been so busy with many things that this got sidelined. However, I have tested my PCB design with a VNA, using the original prescribed TDK H5C2 cores, and the performance is on par with JDB's measurements. I've measured impedance, DM rejection, and CM rejection. Sorry I don't have the plots right now, but I can assure you it looked good. I will get the plots for H5C2 and the other core materials soon.

I've already designed it for Hammond 1590G. I'd recommend to just use the 'G' version. The size is a little smaller than the 'B' version with I think is a good thing.

Please see the attached zip.

The BOM is very simple:

(Qty. 1) 10 turns twisted pair 0.4mm enameled copper wire on H5C2 T16-4-8E core
(Qty. 1) 10 turns twisted pair from CAT5 cable on H5C2 T16-4-8E core
(Qty. 4) TE/AMP 5-1634513-1 (PCB mount BNC jack)
(Qty. 2) 49.9Ohm 0603 resistor (I used MCT0603PD4999DP500, 0.5%, 25ppm, 1/4W)

I used JLC black soldermask. It looks nice with the black version of Hammond 1590G.

I built mine on my R01 version PCB. Works fine, but I made the holes for the 4 posts (for each BNC shell->PCB connection) too large. It made the alignment tricky during soldering. I've fixed that in the R02 PCB file, attached.

[coppercone2]

the torroid outline with the transformer symbol nearby it reminds me of something

[Noy]

Hi, unfortunally i couldn't get the suggested: H5C2 T16-4-8E ones.
I bought the "maybe" ones (B64290L0044X038) from Mouser and hope they are useable.

I already have 1590B cases (pretty cheap on aliexpress (1,60€ for one..)) so i will use it and have to do a new board / modify yours.
I will use SMA connectors, have only a very small amount of BNC cables here ...

How critical are these:
10 turns twisted pair 0.4mm enameled copper wire ?
Are 0.4mm a "have to"? Can i use other cables?

How do you tighten the cores to the PCB? So that they aren't flapping around? Maybe 3 soldertabs and "desolder braid" to fix them onto the pcb?

I will use cherrypicked (4 wire measured) 0805 49.9 Ohm resistors (have them already here..)

[TimNJ]

Hi, unfortunally i couldn't get the suggested: H5C2 T16-4-8E ones.
I bought the "maybe" ones (B64290L0044X038) from Mouser and hope they are useable.

I already have 1590B cases (pretty cheap on aliexpress (1,60€ for one..)) so i will use it and have to do a new board / modify yours.
I will use SMA connectors, have only a very small amount of BNC cables here ...

How critical are these:
10 turns twisted pair 0.4mm enameled copper wire ?
Are 0.4mm a "have to"? Can i use other cables?

How do you tighten the cores to the PCB? So that they aren't flapping around? Maybe 3 soldertabs and "desolder braid" to fix them onto the pcb?

I will use cherrypicked (4 wire measured) 0805 49.9 Ohm resistors (have them already here..)

Regarding the 0.4mm diameter wire, hmmm, I'll have to let Jay_Diddy_B or T3sl4co1l (et. al.) answer that one definitively. They are designed as "transmission line transformers", that is, maintaining some characteristic impedance that depends on wire to wire spacing, twists per unit length, etc. If I had to guess, you can deviate a little, but I'll let them answer..

Regarding fixing the cores, I too considered implementing some sort of "proper" way to fix the cores to the PCB. I considered adding slots for zip-ties, even small hole for nylon screw + nut. But, I opted just for regular old non-corrosive RTV silicone (can also use some decent hot melt glue). The cores are light enough that I don't think it's really worth doing anything elaborate.

I would be hesitant to recommend adding something conductive/magnetic to fix the core to the PCB, as I can't be certain the effects it may have on high frequency performance.

Does your LISN have SMA connectors? Just thinking that the reason this is designed with BNC is that the vast majority of LISNs with RF output ports use BNC anyway.