common-mode choke better than chip beads for EMI filter?

[floobydust]

Why would a common-mode choke be better than using multiple differential (chip beads) for low current signals, like RTD's thermistors etc. EMC filtering?

Example:
3-line CM choke TDK ACM2520-801 800Z, 150mA, 1R6
Chip bead TDK MMZ1608R102 1000Z, 400mA, 0R5

My understanding is CM chokes have lower R losses and they are used for high-currents, but chip beads have more differential mode attenuation.
I have 3-wire RTD's to filter, and a previous designer (in EMC lab) came up with: two series CM chokes, chip beads, 10nF caps, 100nH inductors, 100nF caps. This seems overboard, I have never seen so much EMC filtering coming into an A/D.
Any guidance is appreciated.

[Richard Head]

Differential mode chokes are prone to saturation due to the direct current flowing through them. This results in one having to use a much larger core  (or gapping it) than would otherwise be required.
Common mode chokes are used to filter noise that is common to both (or several) leads. A ferrite bead is a common mode choke with a single turn.
A common mode choke (aka Current compensated choke) is wound so that the flux produced from the one winding is exactly cancelled by the other winding. This allows one to use very high mu cores without the core saturating due to the DC component. The same principle is used to design many magnetic structures including Baluns for RF use, residual earth current detectors, DCCT's to name a few.

[T3sl4co1l]

How much attenuation do you need, and at what frequencies?  What kind of interference are you expecting?  (Commercial, industrial, power line switching noise, dirty equipment, RF transmitters?)

A wound CMC might be unsuitable (alone), say because you have to filter RF (radiated) in the 100MHz range and the interwinding capacitance is dominating.

You may also be considering the wrong components entirely!  If you can spare the series DC resistance (often the case for Kelvin measurements), then a plain dumb old resistor can easily have more impedance (at any frequency) than even a rather large CMC has (at only its parallel resonant frequency).

And, while a wound CMC has higher [peak] impedance than a multilayer ferrite bead, they're much larger, whereas you can put more FBs -- and bypass caps -- in the same footprint.  If you can increase the filter order, you always get better performance.

You could make a CLC or LCL filter just as well; in a controlled impedance environment, either one will filter just as well.  But when the impedances are random, like the impedances of random RF pickup on long wires, it's a good idea to have a nominal impedance, especially with some lossiness.  Which a ferrite bead will get you.

Tim

[floobydust]

Guys, thanks for your insights and wisdom. Sorry I was ill and couldn't write back.

What threw me is this schematic a designer did after many hours in an EMC lab. The circuit seemed overboard for passing immunity - two 800Z CM chokes, then 2,700Z DM chokes... note the shield is floating! evb149 I think you are bang on saying the device failing susceptibility was something else.

I have max. 500ft of shielded cable to the RTD's (16-320R) and expect mostly common-mode RF and noise in an industrial setting. That would be hum, phase-control, variable speed drives - lots of noise on the low end a few MHz. I don't have a lot of experience here with how much attenuation is required.
I thought first pass would be 1-10nF caps to earth GND, and chip beads 1,000Z. I did not see a common-mode choke useful, although some have a high leakage-inductance spec, the chip beads have higher impedance say >10MHz. Maybe I could use big inductors(100uH) to get better attenuation for low frequencies.
In software the A/D is heavily filtered which looks after hum, and the A/D has internal digital filter max. 1kHz.  http://www.ti.com/product/LMP90100

I'm not sure where to start in the design because I have to guess at the signals coming in.

[nctnico]

Where does the zener common connect to? I'd put the zeners close the the ADC's input. In this setup it is easy to couple noise on the zener common into the external wires. I definitely agree that this circuit seems like overkill.

[T3sl4co1l]

Hah, all the filtering does almost nothing because there's no impedance anchoring the other end of it all.

At least, I'm reading this left to right, as is traditional.  Which seems odd having TB3 and a dashed-line RTD3 on the right, and port symbols on the left...

Anyway, assuming that's the case...

There's C209-C211, which joins the wires all together at RF (which will quite strongly reflect any DM noise, which may have consequences elsewhere, but whatever).  No effect on CM.

R208, R209 and R242 all act in parallel (and R207 into R240, R241 in parallel), so the cable* is terminated into GNDA with about 60 ohms.  Which is a little low (150 would be more reasonable against a susceptibility test), but whatever, it won't peak strongly at least.

*Which we can treat as a shorted bundle, a single wire -- because the caps are shorting the wires together at RF.

Then there's GNDA, and C241-C243.  As long as the grounding is solid (tied to ground plane, and all other connections to cables or chassis are similarly solid at RF), the common mode currents will be moderately well sunk.

One thing is already fishy: the termination impedances are unequal.  Therefore the CM current will create unequal differential voltages.  GNDA is simply GND (assuming, again, that it's actually a reasonable ground as such), while the other signals are passed through capacitors (with their ESR and ESL).

We can expect a cutoff frequency around 200 ohms * 10nF ~= 80kHz, so attenuation will be quite good above, say, 8MHz (-40dB).  Then probably rising again in the 100MHz range, due to ESL.

L211-L213 do nothing.  Simply that.

D203 should be paired with C241-C243 so all the low impedance outside-world filtering/clamping is done in the same place.  Putting them after the inductors means a variable impedance at that point (when they clamp), which screws up the filter.  Keep the filter linear!

It's not labeled, so it's not obvious what kind of device D203 is.  If it's an array of zeners as shown, with Vz around 6V, that'll be fine as long as the return pin goes to ground at some point.  I guess these signals are biased above GNDA, so single (unidirectional) zeners "from" ground will do fine; the device could also be used bidirectionally by grounding pin 1 instead.

If lower leakage is needed (a distinct possibility!), they can be bootstrapped against clamp diodes, in which case independent circuits will be needed.

L225-L226 apparently do very little, if anything at all.  I suppose TB3-1 through TB3-3 go to an instrumentation amplifier or something. . .

Tim

[floobydust]

It's backwards, flow is right to left, TB3 connects to the RTD's out in the plant. The schematic left-side goes to op-amp buffers then the A/D converter.
The zeners are for passing ESD standard 61000-4-2, tiny ESDVAL 6V1. Zener_common goes to earth ground but AGND is floating and a bit noisy from the SMPS.
I consider this a disaster (EMC) design, took months to pass immunity tests. If I show the digital input EMC filter schematic, y'all would die laughing.

The two common-mode chokes I think are a waste. Comparing the sequence of CM choke-then chip bead verses chip bead-then CM choke, the line-line capacitance of the CM choke could help preceding chip beads attenuate better?

I would have 1-10nF caps from each incoming RTD terminal to earth GND, then 3 chip beads, and the zener's capacitance helped by 10nF's to GND to give a CLC filter. Then that going into the 200R resistors off to the op-amps and A/D.

I am not clear about filtering DM noise and what benefit that has. Chip beads would lower it.
Would it be better to draw a schematic of something I think would work, because this circuit makes my head spin trying to figure out what's going on.

[Apollyon25_]

AGND is floating and a bit noisy from the SMPS...

Oh dear.

[floobydust]

I did some research and CM chokes seem popular for about 500mA up, and there are different winding styles such as bifilar and sectional. Not always mentioned in the datasheet. If you want some leakage inductance and DM filtering vs low interwinding capacitance, it's in the details. Example, for say a DC-DC converter O/P filter:
Bourns SRF0905-500Y rated 50uH with 22uH leakage inductance.
Murata 50105C rated 1,000uH with 0.2uH leakage inductance.

T3sl4co1l helped get me on track. You have to know source CM and DM impedance, as well as what you driving.

[Marco]

Isn't a 3 wire RTD usually driven with current sources? Won't R241 screw it up?

[3roomlab]

i have a relatively related question and thought it be better to add on to a thread than to start a new 1



in the above configuration, a 3 terminal EMI filter SMD, is both on the +ve rail and then inverted for GND rail ... does it make sense to apply in such a way since +ve is referenced to GND, but the "float" quality of the GND rail is unknown. so instead of using a common mode choke, will a 3 terminal SMD inverted for GND work just as it would if it is on +ve rail ?

[T3sl4co1l]

With the big caps inbetween "stitching" them together, it doesn't make any difference that one of them is "upside down".  If they were connected opposite each other (with their capacitor pins tied), you would have a diff + CM filter.

Tim