Staggered 50uH and 250uH inductor design for LISN

[CopperCone]

So i am going to make a LISN for mains power. I looked at the tekbox design and it seems strait folward.

My only design constraint right now is 2.65mm wire, which i have alot of.

I see that the tekbox design splits a 250uH inductor into four sections each in parallel with what looks to be 68 ohm resistors.

I can try to emulate their design from pictures but i am wondering why they have the resistors in parallel with the inductors.

Also to extend the highfrequency performance isaw that tesla suggested the inductors be staggered to imitate a conical torroid. I am interested in these performance gains.

I know how to calculate an inductor but i am not sure what geometry to go for, like the ratio of lenght to width, and i am not sure how to stagger the four sections.

Can someone suggest some initial dimensions and step ratiobetween these inductors and some general rational for these designs?

Or i would be for making a conical inductor that fufils the design requirement.

[BNElecEng]

I don't have the answer to your question but I did come across an interesting read on a DIY LISN.

http://www.feng.pucrs.br/~fdosreis/ftp/publicacoes/Conferencias/IECON/IECON2003/lepuc6elio.PDF

[T3sl4co1l]

I think you'll find it's the other way around, the 250uH are on the spools (more layers = more inductance in a smaller package), while the 50uH is sectioned, in single layers.

The spool winding has deeper resonances, and at lower frequencies, making it unsuitable for the inductor facing the EUT; but where it is, it doesn't need much impedance (compare to the 5 ohm resistor it's effectively in parallel with), so it's a fine way to save space.

Estimate the dimensions and enter them here:
http://hamwaves.com/antennas/inductance.html
This is the best calculator I know of anywhere, for single layer solenoids.  It is more accurate than anything you can physically measure, and includes frequency dependent effects.

The inductance of the chain will be slightly higher than the sum of each section in series, due to mutual inductance.  There's probably k ~ 0.1 between sections, giving a similar (~10%) increase above that figure.

Dimensions?  L/D ~ 1-4 is good.  Longer means higher impedance peaks, but less efficient (inductance per wire length, which peaks around 1:1).

Play with dimensions, see what fits.  Also see the self resonant frequency, or the estimate at least.  (If you're getting warnings and no output, try increasing pitch or decreasing wire size a bit.  That's usually what does it.  The lower Q won't affect the calculation much.

As for winding form, they probably used phenolic tube or something like that.  Easily found at McMaster Carr or the like (...if you don't mind paying for it).  Cardboard (say, paper towel tube) can be used, but it's not dimensionally stable and your winding will come loose.  The tube or winding can be coated in glue or varnish to help with this.

Tim

[CopperCone]

Does anyone know about dampening the coil sections? I made 4 12uH sections on a single core. I can adjust their distance to tune the overall inductance via mutual inductance.

I see in the commercial one they have 68ohm resistors in parallel with the coils.

Wont these resistors significantly degrade the highfrequency responce? Don't you need parallel inductors that block high frequencies? Does anyone know what the standard says about this?

[David Hess]

The parallel resistors lower the Q of the inductors to prevent resonance peaks.  This is commonly done in decoupling networks.

[floobydust]

I have only these pics of 50uH chokes for a 277VAC 15A LISN. They are quite big inductors 3" dia and have 4 damping resistors 100R every 10 turns or so, same as prev. post mentioned:
Building a Low Cost Line Impedance Stabilization Network for EMI Tests

There's a thread on eevblog where ferrite core inductors are used, makes me cringe. Ferrite core is proven usable to 30MHz by JDiddy

[CopperCone]

You don't happen to have a schematic do you?

Unfortunately I wound my inductors in 4 sections on a PVC pipe, I was hoping to be able to tune the two to have equal inductance by moving the sections around slightly. I have more like 20 windings per section with a smaller core to make 12.5uH measured of course.

I am interested in the undocumented things these guys have, like what appears to be RF feedthroughs and small inductors, etc. Why the diodes? Limiters?

Also, what exactly happens if you use ferrite cores? It only is for 30MHz (don't worry, I wound giant air coil inductors on PVC pipe ,and it took me way too long )

Could you possibly make a schematic for us with your equipment?
I found this one in addition to the tek box one (it explains more)
http://www.ets-lindgren.com/sites/etsauthor/ProductsManuals/LISNs/3816-2%20LISN%20399198%20C.pdf

I believe the principle difference is the first stage, where they went for a parallel RC dampener, rather then a serial one. I am not sure what the trade off is, I believe one requires larger cap values, and the other one has a poorer high frequency response.

[floobydust]

Not to highjack your inductor design question, I drew a rough schematic, it has a some unknowns and I never did figure out small inductor color codes.
Notice the distributed resistor/capacitor arrays, staggered inductor damping resistors - I believe it's all to lower parasitics. The 10dB attenuator seems to have a BPF.

It's good you went with air-core inductors, I can't see typical SMPS ferrite cores having the bandwidth and self-resonant frequency out of the way. How many amps?

I'm surprised the ETS Lindgren unit is so basic compared to this Com Power unit which has clamp diodes to protect your spectrum analyzer's front-end. Who would leave those out!?   

For coil forms, threaded ABS plastic pipe nipples would work, the wire just lays in the groove.

I find I really need CM/DM capability in a mains LISN, something you might consider adding.

edit: updated schematic
edit2: corrected R8 value in Compower schematic

[CopperCone]

How many watts should those 5Ohm resistors be? Non inductive type?

The 30K is exposed to line voltage directly, so thats easy enough, but whats reasonable for the ones in series with the cap?

say, 330 volts, so its like 3.4 watts, so 5W resistors here.

[floobydust]

It's too bad we don't have a 'home brew' design out there, for all us poor people.

Nearly all power resistors are spiral-wound, have steel end-caps so parasitic inductance is a problem at 30MHz.
I'm not sure what resistors are suitable, you'd have to model to see the effects of their inductance here.
This resistor should be made up of several in parallel, like 5 of 25R 1W as a guess.
Sees large inrush currents if switched in at peak line. Carbon comp are probably ideal but hard to find.
Ohmite WH/WN non-inductive.pdf WW rated 1nH at 1MHz.

5R resistor in series with 8uF:
120VAC is 1.1Wpk and average 0.55W
240VAC is 4.4Wpk and average 2.25W (5Vpk)

This older design had huge wirewound parts: https://www.eevblog.com/forum/testgear/lisn-50uh-solar-electronics-9252-50-teardown/

I would check the frequency response of the LISN when you are finished.
I still don't like the Lindgren 0.47uF cap from mains direct to the LISN output, that would kill a spectrum analyzer input IMHO.

[T3sl4co1l]

Note that the 5 ohm resistor's inductance appears in series with the 50uH choke, so if its inductance is a suitably small fraction of the total, it doesn't matter.  Namely, under 0.5uH would be more than fine.

Another way to put it: a resistor with an F = R / (2*pi*L) cutoff frequency over 1.6MHz.

I think I'd be shocked to see a 5 ohm with even that much.  Typical wirewound vitreous resistors are in that range, and smaller parts are much better.

A similar argument applies to higher value resistors.  Besides the 30k resistors not being in the signal path (I think they're intended only to discharge the capacitors when off), they're much more likely to have an overall capacitive characteristic.

The general rule is this: resistors over 200 ohms are capacitive, while resistors under 50 ohms are inductive.  The exact crossover point varies with construction (a wirewound of several kohms may still be inductive at a few MHz, before capacitance takes over), but the trend remains.

Tim

[CopperCone]

Floobydust in your last picture there are capacitors in a circular configuaration.

Do you know if this is one of the higher values in the schematic or is it the ? Value one?

I assume they are lower capacitance caps meant to make some kind of rf feedthrough?

[floobydust]

The Compower LISN ring structure of capacitors and resistors I believe mainly is to lessen parasitics, such as the capacitor's inductance.
Using a single 0.1uF cap vs 10 of 0.01uF cap in parallel, the self-resonant frequency of the resulting capacitor array is moved above our 30MHz top.

But that ETS-Lindgren LISN uses (single?) 0.47uF caps, so the SRF is not a problem... or one manufacturer here is a better design.

TDK/Epcos EMI supression X, Y caps

[T3sl4co1l]

Note that series resonance isn't very important, as it causes high transmittance -- in contrast, the lack of resonance causes insertion loss!

The impedance is also very low, a small fraction of the system impedance (50 ohm).

An even more subtle point to make: the high frequency asymptote is not inductance, but the low frequency equivalent of a transmission line, consisting of the component leads and body.  The characteristic impedance depends on how close these parts are to ground, and their relative sizes; it seems unlikely that you'd be able to arrange a ground so that the impedances are all 50 ohms.  In a typical ground plane situation, you would expect the leads to have a modestly high impedance (roundabouts 100 ohms) and the body, probably something a lot lower, so you get a lowpass filter effect.  There can also be a mode with the body resonant on top of the leads, which would transmit a lot of energy into space (if not shielded), resulting in a notch in the stop band.

In the pursuit of that, though, it might pay to remove the ground plane beneath the body, to compensate for its size.  The advantage would be extending the pass band flatness, by maybe up to an octave?

Tim

[Jay_Diddy_B]

Hi,
I am the designer of the LISN that was described in this message:

https://www.eevblog.com/forum/projects/5uh-lisn-for-spectrum-analyzer-emcemi-work/msg404662/#msg404662



This LISN uses 5 pieces of a Wurth 744314110 inductor.

This was chosen for its high self resonant frequency.



The LTspice model reveals the equivalent circuit of the inductor:



These inductors can be placed in a simple 5uH LISN:



The simulation shows that input impedance matches the requirement of the CISPR specification. There is no impact on the performance from using this inductors.



The transmission is also 'textbook':





There is nothing wrong with using cored inductors in a LISN.

The important feature is that the impedance of the inductor is high, compared to 50 Ohms over the frequency band of interest.

Here is the inductor in a test circuit to measure its impedance:



And the results from that simulation:



You can see that even above the self-resonant frequency the inductor still has high impedance.


I also designed the Line voltage LISN shown in this message:

https://www.eevblog.com/forum/projects/5uh-lisn-for-spectrum-analyzer-emcemi-work/msg641108/#msg641108

It was designed using similar ideas to those described above.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi,

The reason of the ring of capacitors in this picture:



Is probably the result of somebody copying somebody work without understanding it.
This is the main coupling capacitor in the LISN. One end is connected the line voltage, the other end is connected to a Spectrum analyzer.

This capacitor should have a very high safety rating, a Y class capacitor. It is not easy to get high values of Y capacitors so several Y capacitors may need to be connected in parallel.

The requirement is that this capacitor is low impedance, compared to 50 Ohms across the frequency range of interest.

SPICE is a very tool for engineering LISNs.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi,
I have looked at the schematic for the Com Power LISN 115A contributed by floobydust in reply 7.

This is my interpretation of the 10dB limiting circuit.



This is the frequency response of the schematic shown above.




I didn't make any attempt to use the same component designations.

I have attached the LTspice model.

I believe the switch bypass the filter/attenuator/limiter. This arrangement gives the 10dB attenuation.

Regards,
Jay_Diddy_B

[charliedelta]

Just follow the CISPR design which includes the  470 ohm carbon resistor across every few turns. Download CISPR 16 standard and it has all the design and construction details in the appendix.

[Jay_Diddy_B]

Hi,
I am going to continue my analysis of the Com Power LISN 115A by looking at the ring-o-caps:

r

Humor Old Technology



Humor\

The believe is that is a countermeasure for parasitic resonances in the coupling capacitor. The impedance curves for typical capacitors is like this:



A model can be built which includes parasitic elements:



And this produces the results that match the capacitor datasheet. I have included the model for 6x 0.01uF capacitors in parallel. This assumes that the capacitors can be connected in parallel without any additional inductance (most unlikely).




The capacitor models can now be inserted into a LISN model. All other components in the LISN model are ideal. The first model looks at input impedance:




When you look at the results, there is very little difference between the three models.



Voltage Source - Zero Impedance

In this test the transmission of the three LISN is measured with a zero impedance. This low impedance is similar to the actual use case. In most case the emissions from the unit under test are low impedance.





The results are:



Apart from the obvious difference caused by 0.1uF versus 0.06uF at low frequencies, there is no significant advantage to using the 6 capacitors in parallel approach, to reduce the effect of the self-resonance.

There may be a benefit form a safety view point. You can get better grade of capacitors in low values (X versus Y)



Test with 50 Source

The test can be repeated with 50 source. This represent the test condition, for testing the LISN with a signal generator or VNA.










This analysis suggests that there is very little or no benefit to the ring-o-caps.

Regards,
Jay_Diddy_B

[floobydust]

Jay_Diddy_B thank you for all your work, especially your analysis of the ComPower LISN.

I've only seen ring-arrays in high power RF transmitters, so the designer might have carried that experience over.

We disagree on one point- the use of power ferrite-core vs air-core inductors. It would be great to know SMT inductors that could instead be used for pre-compliance work, in a mains LISN.

The inductor core material is important and something I don't see covered.
Spice simple inductor models assume air-core, with some winding capacitance to model self-resonance, so I take simulation results as all rosy. Rising ESR due to core losses I think should be taken into account. I don't have a network analyzer to measure these parts, so I rely first on datasheet information.

If we need 50uH (and I have not seen the 250uH parts, assuming they are not critical and a line filter would suffice) then 5-10uH parts under consideration.
Your EEVBlog mains LISN with 5 of Wurth 7443331000 10uH 9A SRF 35MHz, not sure of the core material.

Wurth parts do not have frequency response curves but WE-HCC ferrite, Wurth says suitable to 5MHz.
WE-WCC iron powder, Wurth says suitable to 5-100MHz (but I'd say 40MHz), 4.7uH is max. offered.
Wurth "superflux" material alloy powder Wurth training module for the WE flatwire series states "switching frequency range: up to 10MHz".
Coilcraft SMT flatwound example SER805x datasheet has flat frequency response curves ending at 10MHz but look promising.
Fair-Rite 61 Material have to do math for gapped/rod.

I keep looking at complex permeability vs frequency. Can we ignore?

[CopperCone]

Well, just to review, the air coils I made came out to the following dimensions:

250uH coil = 2.65mm wire, 33 turns (2x), approximately 6 inches wide and 5 inches tall (made on a bobbin of wood and pvc pipe).

50uH coils - 13 inches long, 2 inches wide

Its going to be rather large.

using magnetic core materials would be nice yea

[Jay_Diddy_B]

Jay_Diddy_B thank you for all your work, especially your analysis of the ComPower LISN.

I've only seen ring-arrays in high power RF transmitters, so the designer might have carried that experience over.

We disagree on one point- the use of power ferrite-core vs air-core inductors. It would be great to know SMT inductors that could instead be used for pre-compliance work, in a mains LISN.

The inductor core material is important and something I don't see covered.
Spice simple inductor models assume air-core, with some winding capacitance to model self-resonance, so I take simulation results as all rosy. Rising ESR due to core losses I think should be taken into account. I don't have a network analyzer to measure these parts, so I rely first on datasheet information.

If we need 50uH (and I have not seen the 250uH parts, assuming they are not critical and a line filter would suffice) then 5-10uH parts under consideration.
Your EEVBlog mains LISN with 5 of Wurth 7443331000 10uH 9A SRF 35MHz, not sure of the core material.

Wurth parts do not have frequency response curves but WE-HCC ferrite, Wurth says suitable to 5MHz.
WE-WCC iron powder, Wurth says suitable to 5-100MHz (but I'd say 40MHz), 4.7uH is max. offered.
Wurth "superflux" material alloy powder Wurth training module for the WE flatwire series states "switching frequency range: up to 10MHz".
Coilcraft SMT flatwound example SER805x datasheet has flat frequency response curves ending at 10MHz but look promising.
Fair-Rite 61 Material have to do math for gapped/rod.

I keep looking at complex permeability vs frequency. Can we ignore?

O.K.

First thank you for kind comments regarding the Com Power LISN.

I was planning on doing some more analysis on LISN today.

There are 5uH LISNs and 50 uH LISNs. We can talk about the 250uH part later.

The inductors in my Line voltage LISN are Wurth HCC (Ferrite). I don't know how to model the change in permeability with frequency in LTspice. I can model different values of inductance.

The requirement is that the impedance of inductance is large, say 300 (for a 20% error in input impedance) at all the frequencies that you are interested in.

At 1 MHz need 50uH

At 10 MHz need 5uH

At 50 MHz need 1uH

so the permeability of the material could change and you still meet the minimum impedance requirements.

Similar if the inductance changes with current, it is not a big deal.

This set of model show the effect on impedance of changing the inductance.




And the results




Remember, the emissions are low Impedance, if they were high impedance, they would very easy to deal with. So an error in input impedance doesn't mean a (large) error when measuring emissions.

The 250uH inductor has very little impact on the LISN performance. you can model this inductor. If you look at input impedance, transmission with a 0 source and transmission with a 50 source you will see it has very little impact.

The place were it does impact is the transmission from the mains port to the output. The 250uH acts like an input filter. In my LISN I replaced the 250uH with a commercial line filter.


The air cored inductors have some issues. The self-resonant frequencies aren't very high. The other issue is coupling. I have seen reports were the inductor measures 58uH in free space, but 50uH when installed in the box. This means the flux from the inductor is coupling the metalwork or other things. Single layer solenoids with space between the turns will be better than multi-layer coils made with magnet wire.


Regards,

Jay_Diddy_B

[T3sl4co1l]

We disagree on one point- the use of power ferrite-core vs air-core inductors. It would be great to know SMT inductors that could instead be used for pre-compliance work, in a mains LISN.

The inductor core material is important and something I don't see covered.
Spice simple inductor models assume air-core, with some winding capacitance to model self-resonance, so I take simulation results as all rosy. Rising ESR due to core losses I think should be taken into account. I don't have a network analyzer to measure these parts, so I rely first on datasheet information.

Whose simple models? -- Well, if it's a 2nd order approximation, it would be an inductor, with some DCR and EPC, and no saturation.  (I don't know of anyone who's published saturable models, FYI.)

The best models I know of are from Coilcraft, and can be run in SPICE with little or no modification.  Of course, they're specific to what parts they have data on, but the model can be adjusted pretty easily to fit other Z(F) data.

And you can generate such data for an air core solenoid using: hamwaves.com/antennas/inductance.html

Higher order AC data isn't important here (ACR, Q, behavior above resonance), so we don't need to be any more precise than 2nd order, anyway.  Not a big deal.

The biggest problem is saturation: if you're testing a 1kW SMPS without PFC, peak currents will easily be 2-4 times the PF=1 (sine wave) peak current.  You don't want to miss EMI during those peaks, which are likely the most important, after all, because the FWB diodes are conducting the EMI straight into the LISN during that part of the mains waveform.  Reduction in inductance raises the LF cutoff, so you could potentially miss the switching fundamental ripple by some dB.

Well, a few dB still isn't very important for precompliance purposes, so YMMV.

Fair-Rite 61 Material have to do math for gapped/rod.

I keep looking at complex permeability vs frequency. Can we ignore?

Careful!  The effective permeability of a rod is much lower than for the material itself; and mu'' of air is zero.  Result: even if the material itself is lossy (#61 only at VHF+!), the resulting inductor has better Q than you expect from the plot.

Tim

[Jay_Diddy_B]

Hi,

Time to break out the network analyzer. These measurements were made with an HP 8714C VNA. The same circuit was modelled in LTspice so that modelling and the actual components can be compared.

Test Circuit



The model shows the test circuit. I did not have the 10uH inductor in stock so I used the 6.8uH from the same series. The inductor is Wurth 744 332 0680. Here is a picture of the packaging, from Digikey:



Construction

A small piece of copper clad circuit board was used. A BNC connector was attached to one end and two 100 resistors were connected in parallel to make 50 .



There is a very small amount of inductance in the ground connection.

Resistive Test



The VNA shows that this is a resistor   


Add the inductor

The inductor was soldered in parallel with the resistor.



And the impedance was measured again:



This looks very much like the result from the LTspice model:




I zoom in to the 10MHz, the VNA gave me:




And the LTspice model gives me:





Remember, I am testing just one 6.8uH from the same family as the five 10uH parts used in my LISN.

The VNA test shows very little deviation from the ideal results from the LTspice modelling.

Regards,

Jay_Diddy_B

[Jay_Diddy_B]

Hi,

The previous LTspice model used an ideal inductor. If I use the model which includes the parasitic components:



I get the following result:




Which is even closer to the measured result:

.


The model is accurate.

Regards,

Jay_Diddy_B