Line filter, proper LTspice simulation

[flowwolf]

TL;DR Summary

Proper simulation of mains line filter with feedthrough capacitors in LTspice

Dear forumers,

I have two high performance AC feedthrough capacitors (47nF, 100A, class Y2, DC-60Hz) which I want to either combine with a general purpose EMC line filter, or build a custom one to filter the mains from of about 100khz to as high as possible (for shielding purposes).

I can see in documents (like this one https://pdf.directindustry.com/pdf/premo/emc-filters/50493-247797.html) that the typical configuration for EMC filters are: X capacitor, discharge resistor, common-mode choke, 2xY capacitors.

I gave it a try in ltspice. A few schematics I saw on the internet simply placed spice's ground between the two Y capacitors. But to my knowledge, neutral is essentially on the same potential as PE, which would mean that another symbol must be placed on the neutral line in ltspice. And there is also cable impedance.

It would look like this:

or maybe this ?:

I gave approximate parasitic values for capacitors and inductors (.5 ohm and 20pF for L3 & L4, 1m ohm and 1nH for C1), but the simulation gives very strong resonances this way (~76khz, 270mhz) so I'm not sure if I'm doing it the correct way.

So is the schematic and the simulation correct this way?

(L1 and L2 meant to be cable impedance, I assumed ~ 0.5ohm, 1nH, 1nF)By resonances, I mean if e.g. a 76khz signal were present on the mains line, it would cause excessive heating on L3 and L4. Is this correct or am I missing something here?

In another thread someone said that for line filters "The filters should not all be simple LC or RF choke filters, there must also be resistive snubbers, AC capacitor coupled, to absorb the reflected noise, or you will end up with all sorts of resonances." Source https://www.physicsforums.com/threa...feedthrough-filter-for-a-faraday-cage.956487/So I tried to place resistors in series with the inductors in another schematic, this greatly reduced those resonances (and voltage to 215-220vac) but probably still not what that person meant:

Or are these assumptions incorrect because published data is usually given for both common mode and differential mode and what I check here is the incoming voltage not noise?Another question is the place of the feedthrough capacitors. I saw in at least 2 documents that EMC filters can be combined with feedthrough capacitors in which case they don't simply replace the regular Y capacitors but place the feedthroughs after the regular y capacitors.
So would it be a bad practice to place the feedthrough right before the other EMC filter components (like in the 2nd spice schematics)?
The feedthrough on the neutral line would be shorted this way?Regards,
Akos

 

[berkeman]

I would not ground Neutral in your simulation. At the most, you could ground it behind some model of the distributed impedance of the AC Mains (the connection is back at the breaker panel, which is usually far enough away that it will not affect the HF noise in the general case).

Summary:: Proper simulation of mains line filter with feedthrough capacitors in LTspice

Or are these assumptions incorrect because published data is usually given for both common mode and differential mode and what I check here is the incoming voltage not noise?

You need to consider both CM and DM noise. Are you mainly concerned with keeping your noise out of the AC Mains (for conducted emissions compliance), or filtering incoming noise?

 

[flowwolf]

But if L1, L2 and L6 are considered as the impedance of the mains wire (Neutral, Active and PE) [correction]:

Actually, I don't know how the mains wires are connected at/after/before the breaker panel so I'm going to look for some information.

You need to consider both CM and DM noise

OK

I would like to filter incoming noise.

 

[berkeman]

I would like to filter incoming noise.

EMI line filters are meant for suppressing power supply noise to keep it off of the AC Mains feed, not for filtering incoming noise. What frequency bands do you want to filter entering your product?

 

[flowwolf]

I have a shielding enclosure and I want to keep interference out but I still want to bring the mains voltage into the enclosure. EMI line filters are advertised/described e.g. as "eliminate EM noise from both internal and external sources at the connector interface" and "external noises will not be conducted into the device through the leads."
or "if the antenna wire is taken through a hole in the mesh (shield), the radio starts working. This shows that if any conductor is passed (unfiltered) through a shield, the shielding is violated"

The interference is of unknown frequency.

 

[berkeman]

Can you link to the datasheet for the filter that has that information in its feature list? Thanks.

Generally, AC Mains line filters are to help products pass conducted emissions testing. And mainly they are to lower the CM and DM noise from the switching power supply conducted back out of the product onto the power lines. You can see that they are asymmetric with respect to CM and DM filtering diretionality in their normal configuration.

They do offer a little filtering for incoming noise, but mainly in the frequency range of switching power supply noise (say up to a few hundred kHz or a MHz). If you have higher frequency noise that you want to keep out of the shielded enclosure, there are other filters that you can add into improve the rejection bandwidth.

If you haven't done so already, look into how shielded rooms and enclosures handle various types of feedthroughs. Here is the hit list from a quick Google search of such websites:

https://www.google.com/search?client=firefox-b-1-d&q=lindgren+shielded+enclosures

 

[DaveE]

First, my appologies for not reviewing your simulations/modelling in detail. But here are some more general comments about EMC input filter design:

Summary:: Proper simulation of mains line filter with feedthrough capacitors in LTspice
neutral is essentially on the same potential as PE

Yes, at 50/60 Hz. This is irrelevant for typical EMC designs, which only care about higher frequencies (>100 KHz). Neutral is treated identically to the hot conductor. All measurements are AC coupled to reject LF stuff. EMC filters are nearly always symmetrical in this regard.

In the real world conducted EMC on the lines is highly dependent on the details of what and where you measure the current as well as the impedance of the source. Unfortunately the source is the building wiring which is highly variable. There is lots of parasitic inductance and capacitance which will dominate the impedance model at high frequencies.

Long ago, EMC types measured typical installations and developed a standard to use for measurement methods which includes a standardized source impedance network. This network is a model for the source and is not part of the equipment design. It includes measurement ports to standardize that pert too. This network is called a "Line Impedance Stabilization Network" (LISN, in practice). There are some differences for the different test specs, mostly Military vs. commercial. In practice you can use the standard CISPR 22 model for commercial stuff.

As much as I normally hate "designing for the test", it's unavoidable in the EMC world. There is just too much variation without the standardization. So, you need to include this in your model as the source impedance. You can search google for LISN schematics (here's one). In practice we would always buy the pertinent standards (they are hard to find for free, unless you know someone who bought one to steal from) and use that as the basis for our testing. You will probably want some of the EN 61000 series, it's the most common.

For lab measurements you should either pay to use someone's EMC lab or buy your own equipment. It isn't worth building it yourself if you are new to this field, there are implementation details, like calibration, that are significant.

And finally, a more general comment about filter design. You can seldom design a good filter without a good model for the circuit in goes into, specifically the source and load impedances. Whether you like it or not these are part of your filter.

 

[DaveE]

Another comment about low pass filter design. Almost always you want to design multi-stage filters so that the schematic alternates between series inductance (high series impedance) and shunt capacitance (low shunt impedance). So you wouldn't normally connect two pi sections (or two tee sections) together; instead you would alternate Pi - Tee - Pi - Tee etc.

Most EMC filters have shunt capacitors at the input and output because they are small and cheap compared to the series inductors. This means that adding additional capacitors at the output is only a little helpful. You aren't really adding a cascading filter section as much as you are lowering the cut-off frequency of the existing filter.

This doesn't mean you shouldn't try it, it could help a little. Normally adding a series inductor and a shunt capacitor will work much better, but may not be practical. Here's where you really want to look at the schematic of what's there to decide what to add. It is pointless to add a 1nF cap in parallel with a 10 nF cap, for example.

 

[anorlunda]

For the purpose you cited, the search term for products should be "power conditioner" not "filter". There are many on the market.

I don't vouch for any product on this page but you could use the page as a jumping off point.

https://ehomerecordingstudio.com/power-conditioner-reviews/

 

[DaveE]

OK, one more comment about adding to EMI filters. I know you didn't suggest this, but for completeness:

Be careful about adding Y capacitors (the position in the circuit, not the component rating). They will increase leakage current to ground which may violate that safety requirement, depending of course on your specific details.

 

[flowwolf]

Thanks for the responses, I'll check LISN and the power conditioner in more detail. The filter would feed a SMPS of a computer or maybe a UPS.

I've checked the description of several shielding enclosures and the other types of feedthrough (usb, ethernet etc.) are irrelevant at the moment.
For power, as far as I've seen, feedthrough filters are used, but those are quite expensive. I thought I can combine a feedthrough capacitor with an EMC filter.

datasheet:
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=2ahUKEwiblIXZ7bHoAhXHlYsKHZWHCQcQFjAAegQIARAB&url=https://www.schaffner.com/products/download/product/datasheet/fn-751x-ac-feedthrough-capacitor/&usg=AOvVaw0e1NkR1_JbCn9V1XtPSzwX

 

[berkeman]

I've checked the description of several shielding enclosures and the other types of feedthrough (usb, ethernet etc.) are irrelevant at the moment.
For power, as far as I've seen, feedthrough filters are used, but those are quite expensive. I though that I can combine a feedthrough capacitor with an EMC filter.

How much power do you need to feed into your shielded enclosure? If you don't need AC Mains level voltages, you can do a lot more filtering on low voltage (SELV) feedthroughs.

Can you share a bit more about what you are trying to do? What is inside the shielded enclosure?

 

[flowwolf]

I need 100-150W.
I was thinking about bringing low voltages (5V, 12V etc.) into the enclosure but it would be a bit complicated to filter all the outputs of a SMPS (AFAI could imagine). I want to power a PC inside the enclosure and I'm planning to do RF/microwave measurements both inside and outside the enclosure.

 

[Baluncore]

1. The PC SMPS is so noisy that noise from the mains entering the PC will not be a problem. It is the escape of RF from the computer, and from the SMPS, back out through the mains wiring that will be the problem.

2. The N–PE link is irrelevant to RF modelling. There is no need to model that link or the mains voltage. You are modelling RF so excitation should not be by 230V AC. Note that; if it was 230V AC then you would need to use 230V * √2 = 325V, since LTspice defines AC as peak voltage, not RMS.

3. The common-mode choke L3,4 effectively isolates the two sides of the supply. In LTspice the PE and N can be grounded to the same ground as the chassis.

4. The mains plug and 3 wire cable can be modeled as a 2 or 3 wire twisted pair transmission line. For RF purposes, a 2 wire line will appear to be about 120 ohms. I would model the mains supply simply as a 3 wire line terminated in a star of three 68 ohm resistors. RF radiation from the twisted wire transmission line is then proportional to the RF current through the line, or the RF voltage on the line.

5. The model schematic can be drawn and modeled backwards with standard AC 1 stimulus at PC on the LHS and RF radiation to mains on the RHS. You will then see sensible dB values for RF noise.

6. It is a mistake to try to do too much with one spice model. If you break the circuit into sensible chunks it will be easier to design and model. The mains filter is one module, maybe as far as the mains rectifier diodes where transient noise is injected. The mains power lead is an antenna and cannot be easily modeled with LTspice.

LTspice examples attached, remove the .txt extension.

 

[flowwolf]

Thanks for the points and the model Baluncore.

LTspice defines AC as peak voltage, not RMS.

- I missed that

This seems very complicated. Should I still include a LISN between the modeled mains wire and the filter?

There are many things I don't understand. Does your simplification (68R resistors) have something to do with LISN? In a document I saw, they simplified the LISN to 50ohm resistors.
https://incompliancemag.com/article/conducted-emissions-measurements-voltage-method/

Because at RF the impedance is high, is that the reason why those measurements were reported in 50/50ohm?

Y bypass capacitors suppress common mode noise which is defined as "This perturbation is present between the safety Earth and the conduction lines"
Why or how the N–PE link is irrelevant in this case?

In your model Baluncore, can I or should I add some modeling of an SMPS?

If EMC filters tend to suppress noise from e.g. a SMPS, then what method should I use to block interference that could come through the mains line? My objective is to bring mains voltage into a shielding enclosure. Many documents describe that leading a wire through a hole in a shielding can make the shielding useless because the piece of wire acts as an antenna. Aren't feedthrough filters/capacitors made for this purpose?Sorry for the questions, but many things to read and understand..

 

[Baluncore]

There are many things I don't understand. Does your simplification (68R resistors) have something to do with LISN?

At some point your model will have a bulkhead or external interface. If that is the mains cable, then you need to model it as a transmission line of unknown length. That is the reason why I use a star or delta of resistors in the spice model. Notice my MainsAC1.asc mixup where I used a delta of 68R when a star of 68R would be a more realistic model for the three wire cable and 3 pin plug.

In a document I saw, they simplified the LISN to 50ohm resistors. … … … Because at RF the impedance is high, is that the reason why those measurements were reported in 50/50ohm?

The coaxial cable used for RF instruments has been standardised to a 50R characteristic impedance. That is why they use 50R, to make it easy to connect RF instruments with internal 50R loads to the physical LISN. The single 50R represents the coaxial cable, connector and instrument being used.
(Line Impedance Stabilization Network; LISN, = Artificial Mains Network, AMN.)

Y bypass capacitors suppress common mode noise which is defined as "This perturbation is present between the safety Earth and the conduction lines"

A and N are treated as a two wire power transmission line. Differential RF noise between the lines is attenuated by the Cx and Ls ladder, but the common mode RF noise then occurs between the PE and the A–N line as a pair. Identical Cy is used on each of the pair to symmetrically attenuate the CM noise.

Why or how the N–PE link is irrelevant in this case?

The distance in wavelengths to the PE–N link is unknown. Since the link is a short circuit, it will reflect most incident RF energy. There is line attenuation in both directions, and mismatched junctions in the cabling. The link is only important for electrostatic purposes, to protect the cable insulation. It's distant presence has little effect on the RF environment.

In your model Baluncore, can I or should I add some modeling of an SMPS?

That would get too complex too quickly. You do not know the spectral radiation characteristics of the equipment that will be operated, so keep it simple and attenuate the whole of the band needed.

If EMC filters tend to suppress noise from e.g. a SMPS, then what method should I use to block interference that could come through the mains line? My objective is to bring mains voltage into a shielding enclosure.

A screened enclosure or room should have a separated double conductive wall. Where the power passes through you will need a low-pass ladder filter built into a series of cavities in a conductive box. The Protective Earth at the mains end is connected to the outer wall and one end of the box, the other end of the box to the inner wall and PE internal. It can be built with Cy feed-through caps between the cavities, inductors in the the cavities, with some Cx. Unwanted energy must be reflected by an inductor, and/or absorbed in snubbers. Use snubbers resistors with some of the caps.

Many documents describe that leading a wire through a hole in a shielding can make the shielding useless because the piece of wire acts as an antenna. Aren't feedthrough filters/capacitors made for this purpose?

Not only does a wire through a hole cause a problem, but a hole or slot in the screen supports circulating boundary currents that effectively couple the two sides of the wall. The effect has a first maximum when the hole or slot reaches a length of λ/4. Keep all holes smaller than about λ/20.

The values used for the example mains filter you gave, are good for the attenuation of 1 MHz and above, but certainly not good for removing SMPS noise at 350 kHz. You will need to cascade several filters covering a wider bandwidth from about 300 Hz up.

Sorry for the questions, but many things to read and understand..

Ask again if I missed something...