Hey everybody, I want to model an EMI filter that we're experimenting with so that I can play with some values and see how the filter response would change. Most of the filter schematic is very straightforward, but I still have a lot to learn about inductors and also about LTSpice. I've done a Spice directive defining a relationship between the two inductors, but I'm not at all sure that I've done it right. Can you guys confirm that I've set up the simulation correctly, particularly regarding the common choke? Thanks!

In the schematic below, CFM1001S is a SMPS which is being powered off of the AC mains. The filter is to prevent EMI from conducting from the CFM1001S back into the mains power (so the mains power hooks up to the two lines on the left.)

 

@Alec_t , @ericgibbs , @crutschow
Any of you guys have advice on setting up the simulation of this common choke component? I've seen all of you do really impressive things in LTSpice, so I was hoping you might be able to help me sort this out.

Is the polarity of the inductor relationship set correctly? Am I right in thinking that an 11mH common choke is 11mH on each winding, or is that some sort of total which must be divided between windings? Are there other obvious errors in my setup?

Naturally I don't expect a perfectly accurate simulation, but if I could at least get a ballpark idea of what this filter can/should accomplish, that would be great!

Thanks!

 

I think you need to reverse L1 or L2 to get the phasing correct.
Why use 4 1Meg resistors in a sim?

 

I think you need to reverse L1 or L2 to get the phasing correct.
Why use 4 1Meg resistors in a sim?

Ok, I'll try reversing the phase of one of the inductors.

As for the resistors, I confess I hadn't even thought about it - I just copied the schematic because that's exactly what we had on the two sample pieces we got.

Thanks!

 

Depending how accurate the simulation needs to be, the default inductor model may not be sufficient. The 'Help' details other models (but I've never used them, so can't help there).

 

Yes, the phasing of one of the inductors needs to be reversed.

 

I'm not sure but I think that the inductors are lossy similar to ferrite beads. Can anyone confirm this?

If they are lossy then needs to be simulated. I am not sure how to do that. Again, anyone have thoughts on that?

 

Yes, the phasing of one of the inductors needs to be reversed.

Maybe I've misunderstood how to reverse the phasing. If I spin the component around, it acts like a high pass filter with a really high cutoff frequency. The filter should be a low pass filter, shouldn't it?

With the original phasing it's a low pass, albeit with much better predicted noise reduction than what we're seeing in practice.

I originally wrote this question simply as an LTSpice question, but I might as well give it some context: We failed an EMI test, isolated the problem to noise from the SMPS we use, and were advised to add an EMI filter between the mains and the SMPS. We've now tried two different filters and both have still failed the test - one was designed for us by the SMPS manufacturer. In actual testing, depending on which test lab you believe, we're seeing flat response up to 400kHz, 10-45dB noise reduction at 700kHz, 12-40dB at 1MHz, 20-35dB at 2Mhz, 23-35dB at 3MHz, and something that looks like a resonant peak between 4-5MHz... above that the filter has little effect. It's an odd response curve to my eyes. The supplier who made the filter for us did EMI testing, but their test results completely disagree with the test results from the international test lab we're trying to get certification through. It's all very frustrating. Here's a rough compilation of all the various lab test results:

We need to get the EMI from our machine (the two blue lines, as tested at two different labs) below the orange line. According to Cincon's lab test results, their filter should accomplish this with miles to spare. According to the test results from the other lab, neither of our filters is effective enough at lower frequencies.

For the moment, I'd really like to get a reasonable simulation of the filter that I can play with (doesn't have to be great, but less than 90dB of error would be nice!) In the meantime, if any of you guys have opinions on what these test results mean, and why they would be so different, I'd be happy to hear those too.

 

Your circuit in the first post #1 is a differential mode filter. And in yor diagram in post #8 you have a common mode filter. But I all can see a "differential mode signal" source Vnoise no common mode signal.

 

Your circuit in the first post #1 is a differential mode filter. And in yor diagram in post #8 you have a common mode filter. But I all can see a "differential mode signal" source Vnoise no common mode signal.

So I need to change the way I'm injecting simulated noise in so that I can see common mode noise and how the filter affects it? That makes sense. I'll give that a try after lunch and see how it goes. May be back with more questions after that attempt. Thanks!

 

Oh, boy. So I took the advice from @Jony130 and rearranged my noise source to create common mode noise instead of differential noise. This got the phase relationship and filter behavior to make sense, but I was still seeing around -100dB filter response at 2MHz, while the real world response is only maybe -20dB there.

So then I thought to add a resistor to simulate the source impedance of the noise source, and then another between neutral and ground to simulate the imperfect nature of the neutral/ground relationship. Any change to either of these values has a major impact on the simulated filter response... all of which shows me that I'm simply not ready to simulate this in any meaningful way. I haven't the faintest idea what a reasonable estimate of source impedance on noise from an SMPS is, nor what the resistance/impedance characteristics of the neutral leg are. All I could do is make wild guesses. If all these things significantly impact filter performance, then my simulation (based on guesses) would be meaningless.

I've always known that some things are just much better dealt with in the real world than with simulations, but this is the most dramatic example I've experienced so far. Back to the physical world and empirical testing for now...

 

Can you show us a picture of how the filter is wired in your system. How it is wired can make a huge difference.
How have you connected the filter compared to how Cincon connected it?

 

Typically you are interested in the conducted current into the power line so you measure that, not voltage.
Is that what you are measuring?
Are you trying to meet some EMI spec?

 

Typically you are interested in the conducted current into the power line so you measure that, not voltage.
Is that what you are measuring?
Are you trying to meet some EMI spec?

 

Typically you are interested in the conducted current into the power line so you measure that, not voltage.
Is that what you are measuring?
Are you trying to meet some EMI spec?

The EMI spec we're trying to meet is EN55022 Class B.

The test result graphs all come in units of dBV or dBmV, which is why I've been thinking in terms of voltage not current. As with most of this, I'm learning on the fly, so I wouldn't be surprised if I've got it wrong, but that was my thought process.

 

You need to get a copy of EN55022 and see how they do the test.

This something I found on the measurement conditions:

Testing is performed in an Open Area Test Site. Equipment is arranged on the turntable as described in section 3.1.2. Each individual current-carrying power lead shall be individually connected through a 50Ω/50µH Line Impedance Stabilization Network (LISN). A 2 meter x 2 meter vertical coupling plane is placed 40 cm to the rear of the EUT. The EUT is set into operation such that all parts of the system are exercised, while the RF voltages across the 50 Ω measuring port of the LISN are recorded. The test is repeated for each current-carrying power line of the EUT.

So the voltage is measured using a LISN which I believe converts the current to a voltage.

 

You need to get a copy of EN55022 and see how they do the test.

This something I found on the measurement conditions:

Testing is performed in an Open Area Test Site. Equipment is arranged on the turntable as described in section 3.1.2. Each individual current-carrying power lead shall be individually connected through a 50Ω/50µH Line Impedance Stabilization Network (LISN). A 2 meter x 2 meter vertical coupling plane is placed 40 cm to the rear of the EUT. The EUT is set into operation such that all parts of the system are exercised, while the RF voltages across the 50 Ω measuring port of the LISN are recorded. The test is repeated for each current-carrying power line of the EUT.

So the voltage is measured using a LISN which I believe converts the current to a voltage.

That's good stuff. Thanks for sharing! I tried to search for things like that, but everything I found seemed to indicate that the documention was one of those things you have to pay tons of money for (like all the UL books that cost a small fortune and go obsolete every few years.)

Were you just searching the web better than me, or did you have some better sources?

As for finding out how they do the test, I'll be seeing it first hand in a little over a month. We've scheduled a test session with a local test lab so that we can see how our machine performs ourselves and try potential solutions on the fly. We'll almost certainly be bringing the EE who designed our board up for the trip so that he can do real-time troubleshooting, brainstorming, and experimentation.

I've been trying to learn all I can, and I would've liked to have gotten a useful simulation running in order to predict alternate filter possibilities, but ultimately we're counting on the EE to figure this out. Nevertheless, I want to understand as much of this as I can, and the more I learn before we get to the test lab, the better I'll be able to understand what's happening while I'm there

 

Doh! I apologize - I may have provided bad information on the test spec. On both of our first two failed test runs, we were provided a test report which showed our EMI levels vs. the maximum limits, but it didn't say anywhere what the standard was. The graph perfectly matched the EN55022 Class B standard, so I got it in my head that that's what we were dealing with.

On our most recent failure, they finally have us a standard to shoot for, saying it was EN55014. I've found precious little on this standard, although I eventually found an image which seems to confirm that the limits are exactly the same:

I don't know if the test methods will be exactly the same. I'll try to remember from here on out to use the right numbers, now that we've finally been told what they are!

 

Here's where I got the info I posted.

 

Wouldn't you need some capacitors between each phase and ground?
I'm not sure what you need exactly, but when I had some EMI problems, I used something from these guys (Corcom). They have circuit diagrams for their products, which had capacitors between the phases like you did, but also had capacitors between each phase and ground after the common mode choke.

Also - if you were going to copy one of the Corcom designs, on theirs, you would want your capacitor that ties the phases together on the other side of the common mode choke, and the capacitors to ground on the side of the choke that your phase-to-phase capacitor is now ...