6
\$\begingroup\$

Edit, changed circuit diagram to make explicit the signal source is the load

I'm designing a medical device and I'm in process of determining the line filter I'm going to need. I'm not especially concerned about incoming noise but rather the outgoing conducted noise. The device in question is somewhat chunky (1kW) so the conducted emissions are non-trivial.

Now on medical devices you're not allowed to have the common garden Y-filter, medical device line filter looks like this:

Medical CMC

I've created differential and common noise models for the input filter using Würth electronics WE-CMB series 1mH 10A CMC, I'm using 220nF for the X2 cap here but that's subject to change. 53R in the differential model is the 1kW load I've got. 1meg resistor in the output is to discharge the plug when disconnected to comply with regulations.

The common mode variant tracks pretty well what Würth says it should do so I presume that's fine.

Spice Circuit

My problem is that the differential filtration is too good, I'm suspicious of it, around -40dB at 1MHz. Can anyone point out something obvious I'm doing wrong with the differential model?

The 50R star mains line model I've seen in couple of different variations, one example uses 68R but the test equipment is 50R, so..

There's some really unfortunate peaking going on around 135kHz but there's not much point addressing it if my model is out of whack.

enter image description here

\$\endgroup\$

3 Answers 3

4
\$\begingroup\$

Can anyone point out something obvious I'm doing wrong with the differential model?

No, but you may want to put in some parasitics to make the model closer to the real world.

One area that I could see a big difference is on V1, and maybe providing at least 100mΩ of resistance between the terminals of V1 and the choke. Wire's also have inductance, so there will be a few nH of series inductance there also.

Another difference is there might be a few pf's of capacitance between the terminals of the choke but I think they would form poles higher than 10MHz with other components.

Any other real world parasitics you may want to estimate and put in (it would be interesting to see Wurths test fixture.)

\$\endgroup\$
3
  • \$\begingroup\$ Make that up to 100 uH for weak domestic grids. Some are awful, yes, though I'd expect a medical environment to be more stiff. \$\endgroup\$ Feb 4, 2022 at 9:55
  • \$\begingroup\$ The Würth CMC model should have parasitics baked in, the internal wiring from the control PCB to load will indeed have some added inductance and resistance. My concern here was/is that my model is fundamentally wrong as far as the connections go. It seems you agree that the basic circuit scheme is correct, if only needs parasitics added in, that's already very helpful, I shouldn't be off by 30dB. I believe I'll add ~100uH in series with the CMC L and N wires, that'll take care of peaking and suppress differential noise even more plus pushes the peaking further below 150kHz \$\endgroup\$
    – Barleyman
    Feb 4, 2022 at 14:09
  • \$\begingroup\$ I'm talking PCB and source parasitics, also the cap might need some some series inductance \$\endgroup\$
    – Voltage Spike
    Feb 4, 2022 at 14:40
2
\$\begingroup\$

Your expectations and results are normal and will lead you towards a more optimal solution after you define the requirements.

There are 3 terminal Murata capacitors used for EMI filtering and recent patents to enhance the suppress EMI defined by the unified theories of Maxwell/Heaviside along with the many books by Henry W. Ott, on “Noise Reduction Techniques or EMI Noise in the last half-century (< which are well worth getting.) Many books exist, some on www.archive.org

  • Common mode chokes have well-defined characteristics yet interact with differential caps and orientation of other components nearby.

  • It is normal to expect a peak and notch along with a LPF differential or CM filters.

  • CM filters can be cascaded to extend the frequency range beyond 2 or 3 decades.

  • Using combinations of CM chokes and differential filters are selected to meet the stringent requirements for ingress and egress.

  • As you stated the ground leakage filter currents are more strictly limited in medical power supply qualifications.

  • Each unique reactive component adds an order to the RLC filter.

  • Wires also give parasitic inductance (~8nH/cm) and interwinding capacitance and add to the Order of any RLC filter.

  • a well-defined spec and detailed question is better-suited to a well-defined solution.

There are also two types of CM chokes bifilar circular wound and parallel sections which tradeoff interwinding capacitance for leakage and balance. REF video Wurth

Also in high power, Thermal resistance is almost equal to the half the cube root of effective volume but more precisely defined in research papers. This can be important to prevent thermal runaway in surge situations.

\$\endgroup\$
10
  • \$\begingroup\$ "A well-defined spec and detailed question is better-suited to a well-defined solution." - The question was if my spice-circuit/model is broken, not how to design a filter. The detailed solution is what I'm getting paid for but to refine parameters it would be helpful to know if my simulation is broken outright or a reasonable approximation. I can also just default to intuitive/experience-based solution and see how it does in pre-certification. Actual specifications, are, obviously, class II medical device according to IEC 60601-1 and Group 1, Class B device according to CISPR 11. \$\endgroup\$
    – Barleyman
    Feb 4, 2022 at 13:44
  • \$\begingroup\$ @Barleyman. OK but I said his "results are normal" albeit perhaps not optimal. Is that clear now? His graph has no reference level to specs so I gave general purpose solutions... although one can imagine situations where >15 dB gain at 130 kHz could be a problem due to high Q RLC responses \$\endgroup\$ Feb 4, 2022 at 13:50
  • \$\begingroup\$ Oh yes, peaking is on my radar as I mention there, but since I suspected the simulation might be incorrect, there's not much point to tweak the circuit values. Now that you've confirmed as well that the concept is valid (thanks), I can tweak that. I think the easiest way will be to add series inductance to the CMC, adding a damping RC circuit may have implications and additional X2 film caps eat space. With 100uHx2 + 100nF would work out to no meaningful peaking and 30dB suppression at 150kHz.. \$\endgroup\$
    – Barleyman
    Feb 4, 2022 at 14:21
  • 1
    \$\begingroup\$ I added answer about the damping circuitry for what it's worth. AC circuitry is a gift that just keeps on giving but for 1050W load e.g., a 24V DC supply just doesn't work for reasonable values of "work". \$\endgroup\$
    – Barleyman
    Feb 15, 2022 at 16:32
  • 1
    \$\begingroup\$ FYI, the finished filter works fine in anger and passes EMI no problem. I went with a RC damper in parallel to the filter cap instead of RL parallel to the CMC as inductors are less universal than caps and tend to cost more, at least in this application. So it seems the simulation is indeed a reasonable approximation. Würth was/is resistant to assuming CMC inductances are actually according to spec for this kind of filter tuning but in practice it works fine. The exact application uses half-wave blanking to control output power in case you're interested. Symmetric pulse train is the key. \$\endgroup\$
    – Barleyman
    Jun 28, 2023 at 13:01
0
\$\begingroup\$

There's that nasty spike in the differential attenuation seen in the graph. In order to dampen it, you need either parallel RC circuit to the X2 cap or a parallel RL circuit to the main inductance. Texas SNVA538 walks through how to that, in essence the damping capacitor needs (ideally) to be ~4x larger than the filter cap, although you can reduce that if you don't mind less-than-ideal performance and/or losses.

RL damping circuit damping inductor does not have to be large but on the other hand, the performance is not as good so it's bit of a give and take. RL circuit is also not subject to a large turn-on spike unlike the RC circuit.

https://www.ti.com/lit/snva538

enter image description here

One thing to bear in mind is that the RC circuit resistor R18 will be subject to a very high switch-on power spike indeed although intuitively a low-voltage designer would dismiss that. The spike can be up to 10kW during EMI surge test depending on actual circuitry, no ordinary film resistor will survive that. You need either ceramic or carbon composition resistor (CCR) for the job. "Everyone" uses the Tyco CCR series resistors for the job but they're pricey and with the component shortages, in short supply. Ohmite Little Devil and Stackpole RC series resistors do the same job as alternatives.

700V pulse is selected based on 275VAC varistor clamping performance.

enter image description here

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.