DIY DM /CM Seperator for EMC - LISN Mate

[Jay_Diddy_B]

Hi Group,

I am going to share the analysis and design of device to separate Differential Mode, DM, and Common Mode, CM, conducted emissions from two LISNs.

If you Google this subject this device is often called a 'LISN Mate' I believe the term was invented by M.J. Nave.

These devices are available commercially, one device is the Tekbox TBLM1.

Not all of the information on the internet and the designs of some of the commercial designs are correct.

After I have analyzed some of the solutions that can be found on the web, I will present the design and construction of a device that works properly.

We will start with an analysis of the issue, so that we understand the requirements.

Analysis

I am going to start with the analysis of the LISN.

The name LISN, stands for Line Impedance Stabilization Network. It presents a defined impedance to the device under test.
Common LISNs are 5uH and 50uH.

The impedance is defined by the parallel combination of the inductor and the resistor.

LISN Model





This model can be used to measure the impedance of the LISN.



This is the impedance it is 50 decreasing at low frequencies.

2 LISNs and CM and DM Impedances

Two LISNs are used in most EMC measurements to capture CM and DM emissions.

DM Impedance

When two LISNs are used:



The result is:



It should not be a surprise, but the impedance for DM signals is 100

CM Impedance

Similarly for CM:



The resulting impedance:



The answer 25 should not be a surprise.


Target Specification

When the LISN Mate is introduced to separate DM and CM emissions these Impedances should not change.

The DM impedance should be 100
The CM impedance should be 25

To be continued …

Jay_Diddy_B

[Jay_Diddy_B]

Hi group,

Mark Nave presents his LISN Mate design in this Video:



The circuit is:



This is often implemented like this:




To be continued …

Jay_Diddy_B

[Jay_Diddy_B]

Hi Group,

I will continue with an analysis of Mark Nave's circuit.

CM Impedance

Model



Modeling Results



The results show that the CM impedance is 25
This is the correct value

DM Impedance

Model



Modeling Result



The modeling result show that the DM impedance is 25

The circuit is flawed

The desired DM impedance, as explained earlier in the thread, is 100\$\Omega\$

To be continued …

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi group,

I will continue by looking at the circuit proposed by Sanjaya Maniktala in this EDN article:

https://www.edn.com/measurements-and-limits-of-conducted-emi/



In this design the transformer is specified as 2:1. Most power supply designers would think of the turns ratio being 2:1.
RF engineers would think of this a being the impedance ratio of 2:1 or 100 to 50
This is a turns ratio of 1.412:1

Assuming the turns ratio is the impedance ratio.

DM Impedance



Modeling Results



This has the correct DM Impedance of 100

CM Impedance








The input impedance for CM signals is 50

This is flawed


The target value for CM impedance is 25

This can be fixed by placing a 50 in parallel with the CM output port.

The impedance ratio of 2:1 is challenging to achieve in practice because it requires a turns ratio of root 2, 1.412, to 1

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi Group,

The next product that I am going to look at is the Tekbox TBLM1.

A friend of mine sent me a couple of photographs of the PCB inside the TBLM1. The schematic is the same as the circuit used by Sanjaya Manitala in the EDN article. The transformer is a Mini-Circuits ADT1-6T+.

The datasheet can be found here:

https://www.minicircuits.com/WebStore/dashboard.html?model=ADT1-6T

This is a 1:1 transformer with a center tap.

Based on the -3dB point (30kHz) I am assuming an inductance of around 200uH.

There are no other components in the box.

CM Impedance








The TBLM1 has a CM Impedance of 50

This is a flaw. The CM Impedance should be 25\$\Omega\$

DM Impedance







The differential mode impedance is 50

This is a flaw. The DM impedance should be 100 

The TBLM1 has the wrong impedances for both CM and DM signals.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi Group,

So far I have shown a few circuits that don't work properly. It is now time to show one that does work properly. I found this information published by Virginia Tech CPES.

This circuit is a little more complicated. It uses two transformer and two resistors. The transformers are both 1:1 which makes them easier to wind.

CM Impedance




Results



This circuit gives the correct result which is 25


DM Impedance





This circuit gives the correct result which is 100

We have found a circuit that works !!


It give both correct impedance for DM and CM signal paths.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi Group,

Now that we have a schematic that looks promising, it is time to design and construct a prototype.

KICAD Schematic






Board Artwork



50 Coplanar waveguide over a ground plane is used to control the impedance on the board. The design is kept a symmetrical as possible.

Board Design







Mechanical Packaging

Fusion 360 was used for the mechanical design. The housing is a piece of extruded aluminium.



Regards,
Jay_Diddy_B

[prasimix]

This looks great. What about ready made RF 1:1 transformer such as Coilcraft PWB1010L_ ? It has 780 uH instead of 330 uH, but looks equal in simulation (attached)?

[floobydust]

I looked into making transformers for one back here https://www.eevblog.com/forum/rf-microwave/help-making-a-wideband-transformer but couldn't decide on a core material or stack.
Schematic taken from Electromagnetic Compatibility Engineering by Henry W. Ott pg. 707 that uses RFMD Sirenza Microdevices LF-428 excellent but obsolete and super expensive.

[Jay_Diddy_B]

Hi Prasimix, Floobydust and the group,

I have a solution for the transformers. I wound it myself. I have finished the DM/CM separator and tested it. I am getting good results.
I am writing in installments so you can follow the journey.


The journey has been modified from the journey I took, but the segments are being strung together so they are easy to follow.

This looks great. What about ready made RF 1:1 transformer such as Coilcraft PWB1010L_ ? It has 780 uH instead of 330 uH, but looks equal in simulation (attached)?

The simulation is for two 100% coupled ideal inductors. The model doesn't include parasitic capacitance, self-resonance, core loss, copper loss, coupling, skin depth permeability versus frequency etc. etc.
In the model the requirement is that the inductance is 'large' compared to the LISN inductance.

In the next segment I will talk about the transformer.

After that I will talk about the measured results.

I was looking for certain properties in the transformer beyond turns ratio and all the usual stuff. If you want to think about this, think about the 'NVT' that was discussed in this thread:

https://www.eevblog.com/forum/blog/eevblog-1104-omicron-labs-bode-100-teardown/msg1662917/#msg1662917

Regards,
Jay_Diddy_B

[hugo]

The KICAD Schematic is missing ... 

[T3sl4co1l]

Heh, the winding schedule will be more important. Though not too troublesome at these frequencies (<=30MHz ish).

Tim

[Jay_Diddy_B]

The KICAD Schematic is missing ... 

Something is happening with the forum. It is very difficult to get the right pictures to stay in the right place.


Here is the schematic:





Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi,

I am going to continue with the construction of the transformers.

There are two transformers:



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? 

Transformer T1

This has a low impedance transmission line. It was made by winding twisted pair of 0.4mm diameter enameled wire around the core.



Transformer T2

This should have a transmission line impedance of 100 .
I used a twisted pair pulled from CAT5 ethernet cable.



I measured the transmission line impedance using my Tektronix 11801 with an SD24.



The measured impedance of the transmission line was 106.5 

For this measurement the transformer was connected to the TDR with 50 and terminated with a 100 resistor.
The first part of the TDR waveform is the 50 coax. The step is the transmission line wrapped around the ferrite core.

The magnetizing inductance of these transformers at 20kHz is around 800uH.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi group,

now I am going to share the test results.

All the measurements were made with an HP8714C Network Analyzer. The data is from 300kHz to 200MHz.

All unused ports on the LISN Mate were terminated with 50 terminations.

OSL port CAL was performed for reflection measurements.
Thru CAL (Normalization) was used for transmission measurements.

The Ports are called:

LISN_1
LISN_2

These are the LISN inputs

CM
DM

These are the output ports for the Common Mode and Differential Mode signals.

LISN 1




An Ideal result would be an SWR = 1
The measured SWR is close to better than 1.1 of most the range 0 - 200MHz.



An Ideal result would be 50 at all frequencies.
The input impedance is close to 50\$\Omega\$.


To be continued ...

Jay_Diddy_B

[Jay_Diddy_B]

Hi,

LISN_2



The SWR is better than 1.1 over most of the frequency range.



The impedance is close to 50 over the frequency range.



This is the same data presented on a Smith chart. Ideally the impedance should be a dot on the 50 , the center of the chart.

To be continued ...

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi,
I am going to share the transmission measurements.

LISN_1 to the Differential Mode Output.

The nominal insertion loss is 6dB.




There is additional insertion loss of 0.4dB up to 100MHz.
At 200MHz there is 1.5dB loss beyond the 6dB nominal loss.


LISN_2 to the Differential Mode Output.



This path has insignificant insertion loss beyond the nominal 6dB to 100MHz.
The path has an additional 1.5dB of loss at 200MHz.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi group,
continuing with the Common Mode Rejection Ratio.

A resistive power splitter was placed on the on the output of the network analyzer. The outputs of the power splitter were connected to the LISN_1 and LISN_2 inputs.
Both inputs are being fed with the same signal.

The Differential Mode output was connected to the input of the network analyzer.



This is a really good result more than 30dB of CMRR over the range 0-200 MHz



Common Mode to CM Output



This should be nominal 6dB.
Almost no insertion loss to 50MHz and very acceptable results to 200MHz.

Regards,
Jay_Diddy_B

[TimNJ]

Thanks for all of your efforts. I am currently using the LISN Mate (resistive circuit) and have previously used the Sanjaya Maniktala circuit. They seemed to work reasonable well differentiating the signals. But, we only did some rudimentary testing to characterize the performance.

From my introductory e-mag courses in college, I know that poorly matched transmission lines lead to power loss, reflections, and standing waves...but typically we talked about it in the context of a fast rise time square wave signal.

What are some typical effects of mis-matched impedances in this case? Frequency shifts? Resonant peaking? Wrong amplitdues?

Thanks.

[Jay_Diddy_B]

Thanks for all of your efforts. I am currently using the LISN Mate (resistive circuit) and have previously used the Sanjaya Maniktala circuit. They seemed to work reasonable well differentiating the signals. But, we only did some rudimentary testing to characterize the performance.

From my introductory e-mag courses in college, I know that poorly matched transmission lines lead to power loss, reflections, and standing waves...but typically we talked about it in the context of a fast rise time square wave signal.

What are some typical effects of mis-matched impedances in this case? Frequency shifts? Resonant peaking? Wrong amplitdues?

Thanks.

The biggest problem is that the LISN no longer conforms to the standard that are published by the regulatory bodies.
The limits for a 5uH LISN, as found in CISPR16-1-2 are +/- 20% about the nominal value:



If you used the Tekbox TBLM1 the resistance is 25 instead of 50 .

The errors you get depend on the impedance of the noise source, the length of the cables between the LISN and DM/CM separator and the frequency.
Errors in the range of 6dB are possible.

LISNs are used in other tests, such as bulk current injection.

The real question is why not use a LISN Mate that is designed correctly?

Regards,
Jay_Diddy_B

[TimNJ]

Thanks for all of your efforts. I am currently using the LISN Mate (resistive circuit) and have previously used the Sanjaya Maniktala circuit. They seemed to work reasonable well differentiating the signals. But, we only did some rudimentary testing to characterize the performance.

From my introductory e-mag courses in college, I know that poorly matched transmission lines lead to power loss, reflections, and standing waves...but typically we talked about it in the context of a fast rise time square wave signal.

What are some typical effects of mis-matched impedances in this case? Frequency shifts? Resonant peaking? Wrong amplitdues?

Thanks.

The biggest problem is that the LISN no longer conforms to the standard that are published by the regulatory bodies.
The limits for a 5uH LISN, as found in CISPR16-1-2 are +/- 20% about the nominal value:

(Attachment Link)

If you used the Tekbox TBLM1 the resistance is 25 instead of 50 .

The errors you get depend on the impedance of the noise source, the length of the cables between the LISN and DM/CM separator and the frequency.
Errors in the range of 6dB are possible.

LISNs are used in other tests, such as bulk current injection.

The real question is why not use a LISN Mate that is designed correctly?

Regards,
Jay_Diddy_B

Thanks. Makes sense. I suppose I'm saying that once you run a "normal" EMI scan (without separator box), you will see the "true" results anyway. Maybe my EMI work is not as detailed as yours, but typically I just use a separator box to check which peaks are DM and which are CM. It seems that even the LISN mate can do this.

However, maybe I've been fooled by the existing designs and haven't even noticed. Maybe I've even wasted some time. It makes sense to get it right.

So, one question about your transformer construction. Both appears to be 10 turns, bi-filar, with the main difference (that I can see) being the gauge/stranding of the wire. What's the effect of the different construction? Different winding to core capacitance? Different inter-winding capacitance? I guess more directly, what are you trying to accomplish by changing the construction slightly?

Also, if I build this, I'll probably try jamming it in this enclosure:

https://www.pomonaelectronics.com/products/boxes/shielded-box-size-b-225-x-138-x-113-4-bnc-f-blue-enamel-cover

It might be convenient for other people to use this enclosure too since it already has 4 BNCs, solder terminals etc.

Thanks!

[Jay_Diddy_B]

Thanks. Makes sense. I suppose I'm saying that once you run a "normal" EMI scan (without separator box), you will see the "true" results anyway. Maybe my EMI work is not as detailed as yours, but typically I just use a separator box to check which peaks are DM and which are CM. It seems that even the LISN mate can do this.

However, maybe I've been fooled by the existing designs and haven't even noticed. Maybe I've even wasted some time. It makes sense to get it right.

So, one question about your transformer construction. Both appears to be 10 turns, bi-filar, with the main difference (that I can see) being the gauge/stranding of the wire. What's the effect of the different construction? Different winding to core capacitance? Different inter-winding capacitance? I guess more directly, what are you trying to accomplish by changing the construction slightly?

Also, if I build this, I'll probably try jamming it in this enclosure:

https://www.pomonaelectronics.com/products/boxes/shielded-box-size-b-225-x-138-x-113-4-bnc-f-blue-enamel-cover

It might be convenient for other people to use this enclosure too since it already has 4 BNCs, solder terminals etc.

Thanks!

Hi TimNJ and the Group,

The 'danger' of only using the LISN for DM/CM separation is that you forget and leave it in circuit for other tests. If the LISN Mate presents the right impedances it doesn't matter if you leave it in the setup all the time.

For me it was interesting to see the differences between the schematics other people had used. If I was going to build one, I might as well get it right.

To get good high frequency response, these transformer are transmission line transformers. The capacitive coupling cancels the leakage inductance.

One of the transformers was wound with wire harvested from CAT5. This has a nominal impedance of 110 Ohms. This should be fairly easily to find.
The other transformer is wound with magnet wire about 0.4mm in diameter. The two wires are twisted first. This transformer should be wound with a 50 Ohm transmission line.

If you wind them differently, it will work but the separation will not be as good.

The core is more challenging. I have had the cores that I used a long time. They may be obsolete, they are hard to buy in small quantities.
A similar sized core in EPCOS T38 material might work.  I will add some cores to my next Digikey order.

If you want to avoid the metal work the Pomona Box will be fine.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi group,
I spent most of the day designing and building a clone of the HP35676A Reflection/Transmission Test Set. This work with the HP3577A Network Analyzer. I will described the design and construction of the RT Test Set in another thread.

I have the S-parameter Test set, but this has a lower frequency limit of 100kHz.

The RT Test Set will work down to 5Hz.

DM Impedance Measurements

The target impedance is 100



The measure impedance was 97

-3dB point is

 


CM Input Impedance

The target impedance is 25




And at low frequencies



The theoretical number should 12.5

The measured impedance for DM and CM signals is very close to the values expected.
The low frequency corner is sufficiently low, so it will not impact EMC measurements.

Regards,
Jay_Diddy_B

[T3sl4co1l]

Another option for getting close to 50 ohms is a twisted pair-plus.  Twist three wires, say wire-wrap or CAT-5 stuff (peeled apart if needed), and connect two of the strands in parallel.  This gives an unbalanced E-field, which doesn't really matter for such a short winding (the effect will be poor DMRR, but at very high frequencies where it doesn't matter).

It's not easy to calculate or look up transmission line geometries like this; it's easier to get an estimate with a 2D calculator.  ATLC2 comes to mind, particularly being easy to use, just feed it a BMP of adequate size and the right colors, and it calculates the quasi-static field distribution, and the impedance, for one or two conductors with respect to ground.  (So, it can't take into account skin effect and dielectric loss, but for well-behaved materials, those should amount to small adjustments.)

Tim

[Jay_Diddy_B]

Another option for getting close to 50 ohms is a twisted pair-plus.  Twist three wires, say wire-wrap or CAT-5 stuff (peeled apart if needed), and connect two of the strands in parallel.  This gives an unbalanced E-field, which doesn't really matter for such a short winding (the effect will be poor DMRR, but at very high frequencies where it doesn't matter).

It's not easy to calculate or look up transmission line geometries like this; it's easier to get an estimate with a 2D calculator.  ATLC2 comes to mind, particularly being easy to use, just feed it a BMP of adequate size and the right colors, and it calculates the quasi-static field distribution, and the impedance, for one or two conductors with respect to ground.  (So, it can't take into account skin effect and dielectric loss, but for well-behaved materials, those should amount to small adjustments.)

Tim

Hi Tim and the group,

I did a little experiment using my HP3577A VNA and my clone of the 35676A R/T Test Set.
If you are interested in the 35676A clone have a look here:

https://www.eevblog.com/forum/testgear/cloning-the-hp35676a-reflection-transmission-test-set-for-the-3577ab-vna/msg3126330/#msg3126330

Experiment

The transmission line transformer was connected to a small 200 trim pot. The other end of the transformer was connected to the T/R test set.
SOL calibration was performed at the point where the 50 coax was connected to the transmission line.
The Smith chart was observed while adjusting the variable resistor.



The variable resistor was adjusted until the circle displayed on the Smith Chart was a small as possible.



At this point the resistance of the variable resistor is equal to the transmission line impedance, Z₀.

The dot on the Smith chart is close to 2, indicating 2x the nominal impedance, 2x 50 .

I then disconnected the RT Test Set and measured the resistance with a DMM. The resistance was 110 .
This is the same value I got using TDR and the characteristic impedance of the twisted pair in CAT5.

Two transmission lines in parallel

I then tried measuring two twisted pairs, about 1m long, connected in parallel. I twisted the twisted pairs together. I measured the Z₀ as 50 . This is probably a good result because there is a little extra capacitance from twisting the two 110 pairs together.

Two pairs from the ethernet cable will not fit on the small cores.

I can use the  technique with magnet wire to measure the impedance of the transformer.
It is probably easier to measure than modeling impedance.

Regards,
Jay_Diddy_B