1. Home
  2. Electronics Design
  3. EMC filter design

EMC filter design

I have been recently reading a great book entitled "EMC for product designers" on Google Books; after 2 days finally decided to check who the author is. :-O

A practical question: why does the book recommend putting the transient suppressors at the very beginning of the power train (fig. 14.34), without considering the exact structure of the filter? If the filter has separate differential chokes, as in the attached sim, the high transient impedance provided by the 500uH combined inductive inertia, even a 2kV surge lasting 10us happens not to be a big deal for a TVS (mere 38.5A peak).

Even if the filter does not contain a DM choke and DM filtering is based on the leakage inductance of a CMC choke, wouldn't it be beneficial to wind a tapped CMC choke and connect the mains to one end and the TVS/MOV and the CX capacitor to the tap in order to make the life of the suppresor easier?

Best regards, Piotr

Version 4 SHEET 1 1244 1520 WIRE 384 0 352 0 WIRE 512 0 464 0 WIRE 0 112 -160 112 WIRE 144 112 80 112 WIRE 288 112 224 112 WIRE 352 112 352 0 WIRE 352 112 288 112 WIRE 384 112 352 112 WIRE 512 112 512 0 WIRE 512 112 464 112 WIRE 624 112 512 112 WIRE 752 112 624 112 WIRE 880 112 752 112 WIRE 624 144 624 112 WIRE -160 160 -160 112 WIRE 752 192 752 112 WIRE 880 192 880 112 WIRE 352 208 352 112 WIRE 512 208 512 112 WIRE 624 240 624 208 WIRE 624 240 576 240 WIRE 624 272 624 240 WIRE -160 288 -160 240 WIRE 160 368 -160 368 WIRE 288 368 288 256 WIRE 288 368 240 368 WIRE 352 368 352 272 WIRE 352 368 288 368 WIRE 384 368 352 368 WIRE 512 368 512 272 WIRE 512 368 464 368 WIRE 624 368 624 336 WIRE 624 368 512 368 WIRE 752 368 752 272 WIRE 752 368 624 368 WIRE 880 368 880 272 WIRE 880 368 752 368 WIRE -160 400 -160 368 WIRE 576 400 576 240 WIRE 352 448 352 368 WIRE 384 448 352 448 WIRE 512 448 512 368 WIRE 512 448 464 448 WIRE -160 496 -160 480 WIRE 576 512 576 480 FLAG 576 512 0 FLAG -160 496 0 DATAFLAG 528 112 "" SYMBOL ind2 128 128 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L1 SYMATTR Value {Ldiff} SYMATTR Type ind SYMATTR SpiceLine Rser=100m SYMBOL ind2 144 384 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L2 SYMATTR Value {Ldiff} SYMATTR Type ind SYMATTR SpiceLine Rser=100m SYMBOL res 736 176 R0 SYMATTR InstName R1 SYMATTR Value 100 SYMBOL cap 336 208 R0 SYMATTR InstName C1 SYMATTR Value {CX} SYMBOL res 96 96 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R2 SYMATTR Value 0.2 SYMBOL cap 496 208 R0 SYMATTR InstName C2 SYMATTR Value {CX} SYMBOL ind2 368 128 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L3 SYMATTR Value {LCMC} SYMATTR Type ind SYMBOL ind2 368 384 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L4 SYMATTR Value {LCMC} SYMATTR Type ind SYMBOL cap 608 144 R0 SYMATTR InstName C3 SYMATTR Value {CY} SYMBOL cap 608 272 R0 SYMATTR InstName C4 SYMATTR Value {CY} SYMBOL res 560 384 R0 SYMATTR InstName R4 SYMATTR Value 1m SYMBOL TVSdiode 272 192 R0 WINDOW 3 -140 58 Left 2 SYMATTR Value 30KPA288CA SYMATTR InstName D1 SYMATTR Prefix X SYMBOL res 272 96 R0 SYMATTR InstName R3 SYMATTR Value 1m SYMBOL voltage -160 144 R0 WINDOW 3 -209 97 Left 2 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR Value SINE(0 325 50) SYMATTR InstName V2 SYMBOL voltage -160 272 R0 WINDOW 3 -315 65 Left 2 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR Value PULSE(0 2kV 5m 0 0 10u) SYMATTR InstName V3 SYMBOL voltage -160 384 R0 WINDOW 3 -335 74 Left 2 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR Value PULSE(0 200 15m 0 0 10u) SYMATTR InstName V4 SYMBOL res 480 432 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R5 SYMATTR Value 100k TEXT 80 488 Left 2 !.param LDiff=260u TEXT 80 560 Left 2 !.param LCMC=2.2m TEXT 360 536 Left 2 !K L3 L4 1 TEXT 88 624 Left 2 !.param CY=1n TEXT 88 592 Left 2 !.param CX=470n TEXT -192 720 Left 2 !.SUBCKT 30KPA288CA 1 2\n* TERMINALS: T1 T2\nDone 1 3 Dtvs\nDtwo 4 3 Dtvs\nDthr 4 5 Dtvs\nDfou 6 5 Dtvs\nDfiv 6 7 Dtvs\nDsix 8 7 Dtvs\nDsev 8 9 Dtvs\nDeig 10 9 Dtvs\nDnin 10 11 Dtvs\nDten 12 11 Dtvs\nDele 12 13 Dtvs\nDtwe 2 13 Dtvs\nRleak 1 2 288.0meg\n.MODEL Dtvs D (IS=1.0e-5 RS=0.2391 N=1.5 IBV=10m BV=54.15 CJO=4000p)\n.ENDS TEXT -496 520 Left 2 !.tran 50m

Reply to
Piotr Wyderski
Loading thread data ...

I don't think that's our TW--IIRC he denied being the author of the Circuit Designer's Companion, which Amazon thinks is by the same guy.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com
Reply to
Phil Hobbs

Your DM chokes will have limited effect when they saturate

Also best to limit the pulse first by the Tranzorb

IIRC you need to have a fuse before the Tranzorb since new rules specify that it can short which is also why UL now dictates that a CM protection to earth using Tranzorb must have 2 in series

Cheers

Klaus

Reply to
Klaus Kragelund

Sure, for some value of V*t they will saturate. But the value can be pretty high, e.g. a 290uH choke made of a 33mm HiFlux toroid will still have magnificent 100uH at 15A and about 30uH at 30A. At that point you will be where the DM-less filter already is from the very beginning, bombarding the surge protector with unimited current pulses.

MOVs also have the fail-shorted failure mode. CX capacitors too. A fuse is always a good thing to have.

Interesting. But this means you have to set the cutoff point very high, because in the case of a short the second TVS cannot avalanche under normal operating conditions. And the next big spike will short the remaining TVS too, so you need the same fusing as in the single TVS case. Not sure what UL is trying to achieve by that requirement.

Best regards, Piotr

Reply to
Piotr Wyderski

Yeah, no relation, I haven't written any published books. :^)

I will however brag that I spent the last week at the testing lab running automotive tests and my design passed with flying colors. ^_^

Placing the MOV* before the filter may be to prevent blowing out the network. On the other hand, it does offer series resistance, which makes the surge considerably easier to withstand.

Check the ratings, and if need be, test! Okay, renting a mains surge generator, or scheduling a day at the test lab, may not be the easiest, but at least you'll know.

AFAIK, they're only concerned with: withstanding surge according to the acceptance criteria; and not starting a fire.

Your simulation has a lot of ringing, which is expected because it's basically unterminated. Always have to design a filter around its impedances -- in this case you'd test it with a 50 ohm source (normal mode,

2 lines) for conducted, and 2 ohm for surge; and the load is whatever the differential and common mode impedance happens to be (which is usually a big cap differential -- FWB and input filter cap -- and a little C common mode (Y-type caps)).

Try to model the CMC impedance. They aren't inductors! They have an RC parallel component, which you can extract from the Z or attenuation curves. Example:

formatting link
formatting link
Even just a single RLC will get you in the ballpark. :)

You may find this handy by the way:

formatting link
Seems to be about the right values to reproduce the 1.5/50 voltage and 8/20 current waveforms, with the ratio of peak voltage (open load) to peak current (shorted) being 2 ohms.

*Use an MOV. TVSs don't handle nearly enough energy! Unless you get the really stonking big ones like this,
formatting link
but they're rather pricey. Well, maybe in your hi-rel projects the cost is justified.

There's other methods too, like using a clamp diode into a mostly-charged electrolytic capacitor. Series impedance is required there, for which an air-core inductor is used. You rarely see MOVs on SMPSs because they have big fat caps that handle momentary overload just fine.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website:

formatting link

Reply to
Tim Williams

Not convinced, you and your clone know far too much about this stuff... ;-P

Not surprised at all. :-)

In this case I'm much more interested in surviving the surges than passing the EMC tests. The latter is of secondary importance. Hence, I'm willing to tweak a bit the conventional circuit and do all that crazy things with high I_SAT chokes. BTW, I planned to use the HF-130125-2 core for the DM choke, but a quick experiment with a syringe shows that an air-core version of the coil wound on that is comparable both in size and in the number of turns to that powder core. It's worth to make a prototype on an ETD39 or ETD44 vertical bobbin and just see.

Yes, it is very crude due to the time constraints I had. I just wanted to see what should be expected in the large excitation case, the filtering aspect of it is pure rubbish.

One of the cases of the ringing is that the CMC coil is just too perfect, the real ferrite will be intentionally lossy. The shunt resistor R5 is there to simulate the losses and does a great job: see what happens without it.

Yes, this is the EMC testing setup, but OTOH: how do they know it is 50 ohm? That 2 ohms is also something strange. My main power line is 0.38 ohm (short-circuit resistance), so what should constitute the remaining 1.62 ohm? Transformed impedance of the power network? If you an explain it, please do, because I am puzzled. Where do the numbers come from, the ancient art of rumpology performed once by a bunch of ISO experts and repeated ever since or is there

*really* some physics behind that to justify the numbers?

But the cap is behind a PFC circuit, including a choke in series. The immediately visible cap is much smaller, O(1uF).

Great, thanks for the curves!

Gladly, but MOVs do degenerate incrementally and that freaks me out.

Holy cow, this is a big one! I was thinking about something 1/4 the ratings, but still big and expensive.

Not nearly as pricey as the hi-rel polypropylene caps in the 500uF/900V range. Acceptable.

This will be there independently.

MANY thanks for such a content-rich post, Tim!

Best regards, Piotr

Reply to
Piotr Wyderski

Hah, go figure. Well, copper savings maybe, but interesting to know it's not much of a space saver.

Well, the surge has at least the stray inductance from the nearest pole-mounted lightning arrestor to you, which is usually one every three poles (or three blocks, I forget), plus transformer LL, plus low voltage distribution stray. The first part is probably negligible because of the high ratio, so that the others dominate.

There'd be induced surge something or other for the buried wires case, which I think is more common in your part of the world than overhead wires. (If you're selling globally, nevermind that anyway.)

Again, since the waveform changes, it's still kind of disingenuous for them to call it "2 ohms", and as the network shows, something like 4.7uH is pretty close to what they use, which is equivalent to about 10-20m of wiring, say from pole pig / pad transformer to EUT. Which I guess should be about right on average.

The 50 ohms for conducted is a little more justifiable: if it were greatly different from 50 ohms, it would be coupling with signals that don't necessarily have any business being there, and aren't necessarily going anywhere (because 50 ohms is closer to the characteristic antenna impedance you expect from mains wiring).

But that's easily contradicted by noting: sure, a given mains network may have impedance peaks and valleys, where it doesn't couple to a 50 ohm source very well, but that doesn't _prohibit_ radiation at those frequencies. People have been known to use dipoles at their antiresonant frequency, and this is done with a (rather heroic) antenna tuner (i.e., the resistance might be over 500 ohms). Granted the radiation isn't even dipole mode anymore, but some monopole case most likely (namely, due to imbalance in the elements, their surroundings, or the balun), but hey... to some people, if it works, it works, riiight? In short, you can always force energy in somewhere, and there will be some part of the network that does the dissipation or radiation.

Anyway, that's just my guess; I don't know of any supporting papers justifying the standards. If nothing else, the point is to have an emission threshold safely low enough not to disturb radios in most cases (let alone other devices nearby / on the line), and to deliver enough energy somehow to get a reasonable expectation of life on a typical mains environment (surge size vs. time is probably Poissonian, so in practice you might get six months, you might get a century, but the expected case is, say, five years or something).

Cheers!

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/
Reply to
Tim Williams

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.