Applying ferrite bead in a high impedance circuit

[floobydust]

I think ADI is wrong- putting a ferrite bead in series with a 1MEG resistor string, for monitoring mains voltage.
Murata BLM21BD152SN1 impedance peak is 1,800 ohms around 150MHz which is nothing compared to the string's resistance.
Unless it's the resistors' capacitance? Yageo RC2010 249K can't be enough to be a problem, and the anti-aliasing filter 22nF swamps that.

Am I missing something or is this is a mistake?

[Vovk_Z]

Interesting, what is the impedance of 249 k (TH?) resistor at 150 MHz?

[tooki]

Out of curiosity, could you share the original source for the circuit?

[floobydust]

ADI do the same in all of their energy metering IC eval boards.  I think they are out to lunch 
EVAL-ADE9430 pdf pg. 8 Fig. 9
EVAL-ADE7878AEBZ User Guide pdf pg. 45 Fig. 61
Also the 51pF caps seem useless with the 22nF there.

It might be related to:
"Internal RF Immunity Filter"
"Energy metering applications require the meter to be immune to external radio frequency fields of 30 Volts per Meter or 30 (V/m), from 80 MHz up to 10 GHz, according to IEC 61000-4-3. The ADE9430 has internal antialiasing filters to improve performance in this testing because it is difficult to filter these signals externally. The second-order, internal low-pass filter (LPF) has a corner frequency of 10 MHz. Note that external anti-alias filters are required to attenuate frequencies above 7 kHz, as shown in the Interfacing to Current and Voltage Sensors section."

https://wiki.analog.com/resources/eval/user-guides/ade9430

[coppercone2]

I mean they have acreddited test labs its probobly not wise to try to argue with that too much.


And it fool proofs the design against someone using the wrong kind of resistor or parts variation / stock.

What if someone wants a different power rated resistor for instance? For different form factors. This might make EMC non issue for this circuit .

[thm_w]

I would not say "wrong", maybe pointless. So if you are worried you can design in the footprints and just not populate them. If they were worried about external RF fields specifically near 10GHz there are more optimal beads for that frequency.

https://www.edaboard.com/threads/built-in-electricity-meter-with-ade7953.406460/

I mean they have acreddited test labs its probobly not wise to try to argue with that too much.
And it fool proofs the design against someone using the wrong kind of resistor or parts variation / stock.
What if someone wants a different power rated resistor for instance? For different form factors. This might make EMC non issue for this circuit .

What? If they use the wrong resistor then the thing could burn up, thats not a good excuse for designing it in..
It probably is a copy paste job, whether or not it was needed or useful who knows.

[thm_w]

Interesting, what is the impedance of 249 k (TH?) resistor at 150 MHz?

Good question:
https://jeroen.web.cern.ch/jeroen/resistor/shuntC.html
https://www.researchgate.net/figure/Frequency-Response-Characteristic-of-Four-Different-Types-of-Resistors_fig2_224110697

So maybe optimal ferrite bead would be one with a peak nearer to 1GHz?
You'd need to do some simulations of the board itself though.

[David Hess]

You know the application better than we do, so are the 249 kilohm resistors in series to raise their voltage rating or to decrease their total series capacitance?

Besides the series capacitance across the resistors to the 0.022 microfarad capacitor, there is also common mode capacitance or electrostatic coupling to the inside of the chassis.

[floobydust]

This (ADI) circuit is for monitoring mains voltage to 240VAC. The four 2010 resistors are in series to achieve the voltage/spacings and some transient withstand I think. The ground-pour around them is cut out.

I could not find RF models for 2010's but using LTSpice with 1pF shunt capacitance for plus a bit of inductance, the ferrite bead is a nothing burger to 2GHz at least, as I expect.
FB is needed on the low impedance channels (CT's, neutral connection) though.

[David Hess]

I suspect what might have happened is that a novice designer took an RF protection circuit from a low impedance design and blindly applied it to a high impedance design; that lossy inductive element is not going to do anything useful for reasons already given.  Many oscilloscopes have more inductance and loss than it represents on their high impedance inputs.

It is possible that the lossy ferrite bead does have an effect because of the common mode coupling between the input and the circuit and chassis ground.  This coupling is not represented in your model, but it could be calculated or with care directly measured.  I have had circuits where I added something similar to decrease susceptibility to 900 MHz TDMA transmissions from cell phones.

A feedthrough capacitor or network where the input passes through the chassis would be more effective, but would present its own problems for CAT ratings.

[floobydust]

Well, after much thought I did reason why they could be necessary. There is a chance ADI have their shit together. Not what I've experienced when communicating with them, it's a wall of noobs and pointing out a silicon bug or datasheet error is brutal.

I've learned while doing EMC testing that the isolated input (voltage sense, CT's) section can have terrible common-mode noise generated by the DC-DC converter powering it.
Here it is ADuM5000 iCoupler with 180Mhz (some run to 300MHz) fundamental so the harmonics it spews easily extend to 1GHz. I don't have the ability to measure 2010 resistor's with a VNA.
So what happens is the common-mode EMI moves backwards, counter intuitive - out through the high voltage divider resistors (shunt capacitance) to radiate on mains wiring.

[coppercone2]

noobs... or experts? Stonewalling you on a silicon bug is advanced PR technique 

admit to making a mistake on paper??? think of the investors!!

[mtwieg]

I don't think there's any way someone could convince me that those ferrite beads have any observable benefit for the system, even from the perspective of radiated emissions, susceptibility, etc (unless the routing of those traces is so awful that they act as a great antenna or something). It's very likely just a leftover from a legacy design, or a placebo for the designer.

[floobydust]

Here's another example of ADI ferrite bead weirdness. FB in series with the decoupling caps for the IC's 1.25VDC reference pin. For a not even 3mm trace run? 120Z 0.35Ω seems out to lunch there. Then it's weirder...
ADI specifies 4.7uF ceramic in parallel with 0.1uF ceramic. I looked and the impedance curves ain't so far off.
Either it's theoretical Spice design or some of the youngins' forgetting about them thar old tantalums. Used to be 4.7uF tant in parallel with 0.1uF ceramic, which makes sense. The IC datasheet spec's ceramic but has polarity marks, must be a tombstone from the Old West.
But paralleling both a 4.7uF+0.1uF ceramic seems nutty, and then adding a FB in series seems more nutty. A tinfoil hat might help.

120Z 0.35Ω Wurth 7427927112
0603 4.7uF 10V X5R Kemet C0603C475K8PACTU
BoM special 0402 0.1uF 25V X7R Taiyo Yuden TMK105B7104KVHF

[coppercone2]

what if its a really much smaller capacitor. like 0402 and 0805

The size requirements could get funny if there is other considerations then the smallest part (say like flex mount ceramics, ESD, voltage requirement, or automotive, reliability or whatever else they have). And thats just digikey, let alone whatever tailored requirements you might have for a company design.... whatever is the natural choice might be inappropriate for whatever mechanical or other analysis they have, its very easy for someone to say they don't want some part somewhere for some reason. Also preferred/approved vendors or design for parts cross over utilization and cost or volume requirements.


You won't go smaller then a 0603 if you narrow it down to flex mount. High reliability flex mount is 1210 for 4.7uF. For 100nF its 0201 for similar features.


but I really think someone should measure its performance on a test equipment.

I can see ADI wanting to give the most implementable solution, because half their customers don't care at all about a 100nF capacitor

[David Hess]

Here's another example of ADI ferrite bead weirdness. FB in series with the decoupling caps for the IC's 1.25VDC reference pin. For a not even 3mm trace run? 120Z 0.35Ω seems out to lunch there. Then it's weirder...
ADI specifies 4.7uF ceramic in parallel with 0.1uF ceramic. I looked and the impedance curves ain't so far off.
Either it's theoretical Spice design or some of the youngins' forgetting about them thar old tantalums. Used to be 4.7uF tant in parallel with 0.1uF ceramic, which makes sense. The IC datasheet spec's ceramic but has polarity marks, must be a tombstone from the Old West.
But paralleling both a 4.7uF+0.1uF ceramic seems nutty, and then adding a FB in series seems more nutty. A tinfoil hat might help.

That is what the calculations show, but I have done microwave layouts where 3 x 1206 0.1uF failed by a large margin, and 1206 0.1uF, 0.01uF, and 1000pF in parallel worked.

[Siwastaja]

Here's another example of ADI ferrite bead weirdness. FB in series with the decoupling caps for the IC's 1.25VDC reference pin. For a not even 3mm trace run? 120Z 0.35Ω seems out to lunch there. Then it's weirder...
ADI specifies 4.7uF ceramic in parallel with 0.1uF ceramic. I looked and the impedance curves ain't so far off.
Either it's theoretical Spice design or some of the youngins' forgetting about them thar old tantalums. Used to be 4.7uF tant in parallel with 0.1uF ceramic, which makes sense. The IC datasheet spec's ceramic but has polarity marks, must be a tombstone from the Old West.
But paralleling both a 4.7uF+0.1uF ceramic seems nutty, and then adding a FB in series seems more nutty. A tinfoil hat might help.

That is what the calculations show, but I have done microwave layouts where 3 x 1206 0.1uF failed by a large margin, and 1206 0.1uF, 0.01uF, and 1000pF in parallel worked.

The difference to general purpose data acquisition power supply however is that RF applications have a well-defined narrow peak where low impedance is specifically important, so paralleling a few values could really give you the SRF dip. The risk is that the exact frequency of that dip is not tightly controlled which is why I like to ask, did you ever qualify this against manufacturing tolerances of those capacitors, e.g. by testing with close-by different values?

[tszaboo]

Here's another example of ADI ferrite bead weirdness. FB in series with the decoupling caps for the IC's 1.25VDC reference pin. For a not even 3mm trace run? 120Z 0.35Ω seems out to lunch there. Then it's weirder...
ADI specifies 4.7uF ceramic in parallel with 0.1uF ceramic. I looked and the impedance curves ain't so far off.
Either it's theoretical Spice design or some of the youngins' forgetting about them thar old tantalums. Used to be 4.7uF tant in parallel with 0.1uF ceramic, which makes sense. The IC datasheet spec's ceramic but has polarity marks, must be a tombstone from the Old West.
But paralleling both a 4.7uF+0.1uF ceramic seems nutty, and then adding a FB in series seems more nutty. A tinfoil hat might help.

120Z 0.35Ω Wurth 7427927112
0603 4.7uF 10V X5R Kemet C0603C475K8PACTU
BoM special 0402 0.1uF 25V X7R Taiyo Yuden TMK105B7104KVHF

Not at all. The smaller ceramics will have it's self resonance at a higher frequency, and it will provide  lower impedance power across a wider frequency range. Some RF modules will define specific capacitors in the picofarad range on their power supply.
The ferrite is there to block some high frequency signal getting in. You are right that ferrite beads together with ceramics can cause an issue, they can resonate undamped and create more noise than without the ferrite. You have to simulate it. And you either have control over the noise sources, or if it comes from external sources, have a bad day.

For the original question, I don't think they do a whole lot. But also, the lack of protection is a bit concerning.

[David Hess]

That is what the calculations show, but I have done microwave layouts where 3 x 1206 0.1uF failed by a large margin, and 1206 0.1uF, 0.01uF, and 1000pF in parallel worked.

The difference to general purpose data acquisition power supply however is that RF applications have a well-defined narrow peak where low impedance is specifically important, so paralleling a few values could really give you the SRF dip. The risk is that the exact frequency of that dip is not tightly controlled which is why I like to ask, did you ever qualify this against manufacturing tolerances of those capacitors, e.g. by testing with close-by different values?

My application ran at the 2 meter, 440 MHz, and 1.2 GHz band, so it covered a wide range of frequencies.  12 x 0.1 microfarad capacitors in parallel should have worked at every frequency according to the calculations, but failed at *high* frequencies.  4 of each capacitor of 0.1uF, 0.01uF, and 1000pF worked great over the entire frequency range.  There was no lead inductance because all of the construction was part of a coaxial transmission line.  The only explanation I came up with was that the 0.1 microfarad capacitors somehow had loss at 1.2 GHz, and still significant loss at 440 MHz, which was not represented in the specifications.  All parts were C0G, but not specifically RF capacitors.  Maybe using 0.1uF RF capacitors would have made a difference.

[floobydust]

I understand staggered-value decoupling capacitors, there's many threads here about that practice, the pros and cons.

One cell modem design I worked with had 1pF, 10pF, 100pF, 1000pF etc. decade steps across the power rail which made some sense given it was up to 2A at 800MHz-2.4GHz.

Working with a vet RF engineer, I saw he relied on precise Spice sims and specified capacitors only with decent RF models S-parameters i.e. Johanson Tech..
What happened is Purchasing switched to cheaper/available parts, not knowing the design was only for that exact component. Then I asked what about PCB trace inductances, temperature, aging, voltage coefficient etc. and it becomes apparent there is too much confidence in the theoretical.

But here it's an output pin of a linear regulator.
I don't expect GHz RF there and even if the board was blasted by an RF gun to corrupt energy meter readings, as a susceptibility preventative- is ADI trying to stop RF from getting in that pin by adding impedance? How does that work? Where is the antenna?

edit: added Z curve

[uer166]

One cell modem design I worked with had 1pF, 10pF, 100pF, 1000pF etc. decade steps across the power rail which made some sense given it was up to 2A at 800MHz-2.4GHz.

That makes no sense at all, unless the package sizes and lead inductance were decades apart. 1pF would have to be basically on-die to be of any use. At 2.4GHz 1pF is ~66Ohms, so it'd doing fuck-all anyways.

[floobydust]

I can't find the modem design docs, but the datasheet gives band-specific cap values. So the LDO output has 47uF, 1,000uF, 39pF and 8.2pF then connector to the module. Which seems silly due to the trace/connector inductance.

[thm_w]

The ferrite is there to block some high frequency signal getting in. You are right that ferrite beads together with ceramics can cause an issue, they can resonate undamped and create more noise than without the ferrite. You have to simulate it. And you either have control over the noise sources, or if it comes from external sources, have a bad day.

Getting in... from where? Its an internally generated reference voltage. The traces going to the decoupling caps will be a few mm at most.

[tszaboo]

The ferrite is there to block some high frequency signal getting in. You are right that ferrite beads together with ceramics can cause an issue, they can resonate undamped and create more noise than without the ferrite. You have to simulate it. And you either have control over the noise sources, or if it comes from external sources, have a bad day.

Getting in... from where? Its an internally generated reference voltage. The traces going to the decoupling caps will be a few mm at most.

Through the power supply for example. PSRR of these at high frequencies is almost zero. And any other cables.

[thm_w]

Through the power supply for example. PSRR of these at high frequencies is almost zero. And any other cables.

So its coming through the power supply, into the IC, now its in the IC. Wouldn't it have made more sense to put the bead before the IC, instead of after it?
It has no value unless the REF was being used externally, which it isn't. And even then you'd use a much higher impedance and maybe a CLC format.

[tszaboo]

Through the power supply for example. PSRR of these at high frequencies is almost zero. And any other cables.

So its coming through the power supply, into the IC, now its in the IC. Wouldn't it have made more sense to put the bead before the IC, instead of after it?
It has no value unless the REF was being used externally, which it isn't. And even then you'd use a much higher impedance and maybe a CLC format.

From your image alone, it's not obvious what's connected where. If that's an input or output or what's going on.

[thm_w]

From your image alone, it's not obvious what's connected where. If that's an input or output or what's going on.

Not my image.
Its obvious that its a REF output on the IC buffered by two capacitors, which goes nowhere else, because there is no net label or wire on those capacitors.