Author Topic: DIY mains EMI-filter (advice needed)  (Read 1642 times)

0 Members and 1 Guest are viewing this topic.

Offline 3roomlab

  • Frequent Contributor
  • **
  • Posts: 778
  • Country: 00
  • A long long time ago, in a land far far away ...
DIY mains EMI-filter (advice needed)
« on: October 30, 2015, 08:27:48 pm »
hello all, the question is

1) if a single choke is not up to current carrying capacity, if i parallel 2 chokes, am i right to assume i pass about 3dB more noise?
2) i sourced 2 chokes with relatively high inductance. how should i calculate the max permissible load? @ 50% its max current? rated max 1A = approx 100w? rated max 2A = approx 200w?
3) does X and Y capacitors need to be a calculated value based on the inductance? or more on the need of the EMI filtering cutoff point?
4) if there is high transient noise in the AC line, exceeding the max dv/dt spec of the supp-caps (ie : say 100v/us), does it mean the capacitor will fail to "capture" the noise?

any tips and hints highly appreciated :P
« Last Edit: October 30, 2015, 08:37:20 pm by 3roomlab »
[What Do You Care What Other People Think?] ... remember to brush your teeth before you read it ... *tchk tchk tchk* ...

Online T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 11468
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: DIY mains EMI-filter (advice needed)
« Reply #1 on: October 30, 2015, 09:38:52 pm »
A filter depends on characteristic impedances being met.  It's impossible to say what component values will do, without a system to test it in.  It's very particular. :-\

That said, filters are usually tested on 50 ohm systems.  This is somewhat helpful, because mains is assumed to be about 50 ohms as well (e.g., CISPR 22).  But the circuit side might be something quite different; it would be a very strange beast indeed to have 50 ohms from each line to ground inside your power supply!  Normally, a power supply is very far from a proper termination, instead having some incidental capacitance (between primary and secondary, and primary to ground), with the offending switching noise added in there.

So, understandably, filters generally won't do quite what the datasheet claims, and the actual outcome may be better or worse.  It's only broadly representative. :-\

As for your questions:
1. Parallel, uncoupled, chokes will exhibit about half the impedance.  (There may be funny back-and-forth modes with the pair, at higher frequencies, but it should be pretty straightforward otherwise.)  As for attenuation, it could be 3dB, it could be 6, it could be positive, or even negative.  I'd say it's pretty unlikely for it to be 10 or 20dB out across the band.  But if you have a lot of strong peaks and dips in the filter response, it really could be that bad!  (And for a power supply, which generates harmonics rather than flat white noise, you might not even realize where those peaks or valleys are...)

In a well-behaved system (like a 50 ohm test), it should be 6dB. :)

2. Rated current is rated current.  Check the datasheet: note that it's rated for some temperature rise (usually 40C), at rated, RMS current.  That means AC RMS, or DC average.

In this case, it's 45 or 50 or 55C rise.  Kind of toasty, but quite typical for electronic components.

You would normally use a fuse somewhat larger than this rating, consistent with the operation of the circuit.  For example, a circuit with a maximum current draw of 0.5-0.6A (50-60W at a low-line condition of 100V) might use the 0.8A choke shown, and a 1A fuse.  That also lets the choke run a little cooler. :)

3. The X capacitors depend on the leakage inductance of the choke (which is rarely rated -- occasionally, you will find a "differential impedance" or attenuation spec, which provides this information), and on how much attenuation is required.  Differential is usually not a big deal for power supplies, thanks to the main bulk capacitor.  It can be a problem for active PFC (power factor correction) circuits, which draw switching ripple directly from the AC line.  These sometimes use additional (single winding) inductors to get the required filtering!

The Y capacitors depend on the common mode impedance, attenuation, the choke used, and how much line-to-ground leakage current is tolerable.  (For example, these must be much smaller values for medical products.)

By the way, all of this can be more-or-less simulated -- but, you have to capture everything in the model, correctly.  A naive model of a power supply's grounding and isolation will completely break the deal!  Over such a frequency range, inductors usually need detailed models, too (but at least capacitors are pretty simple).

4. As long as the network remains linear, there is no "fail to capture" effect.  Everything's attenuated in proportion.

The main failure mode is self-healing of X type capacitors.  An excessive voltage (not the rate, but the peak value of a surge -- and for how long) causes the dielectric to break down.  As self-healing occurs (inevitably, during the life of the capacitor), the capacitance value drops, until eventually, so much has been healed away that the plates become disconnected from the terminals, and the capacitance value plummets.  The filtering performance can then be expected to be terrible, but performance at EOL (end of life) also isn't checked, so don't worry about it.

Regarding those chokes in particular:
FWIW, an extremely high value inductor isn't usually useful.  The physical size, and cost, are quite large for the power level, and there aren't too many places you'd need that much attenuation.  (The peak impedance is 200kohms for the 60mH part!)  The inductance is impressively high, which also suggests low frequency operation, which is handy, but unnecessary (most regulations do not measure below 150kHz).  Finally, the capacitance is also relatively high: the impedance peaks in the 100kHz range, and falls after that (due to winding capacitances).  The HF range can be extended with Y caps and a smaller choke, but that's more parts.  A single big choke, like this, may actually be worse at higher frequencies than an appropriately sized one!

I think it's the kind of part, where, if you've developed your power supply, and found that you need two, or even three, chokes to do the job, then it would be a good spot to look at, as far as cost reduction.  It's a stinkin' big choke, but if it saves three or more components, it's probably still a net win.  Further savings could be had by addressing the offending circuit itself: raise the switching frequency, change the current paths, add dampers or snubbers, add spread-spectrum functionality, etc.

By the way, if you're looking at situations where you really just don't know what to expect, you could look at chokes like these, or consider amorphous/nanocrystalline (e.g., VAC brand) chokes.  These materials give a higher impedance over a wider range, than ferrite cores do.  Of course, they're rather expensive, too...

« Last Edit: October 30, 2015, 09:47:22 pm by T3sl4co1l »
Seven Transistor Labs, LLC
Electronic Design, from Concept to Layout.
Need engineering assistance? Drop me a message!

Offline 3roomlab

  • Frequent Contributor
  • **
  • Posts: 778
  • Country: 00
  • A long long time ago, in a land far far away ...
Re: DIY mains EMI-filter (advice needed)
« Reply #2 on: October 30, 2015, 10:31:15 pm »
thanks for the details tim

i suspect there is much more LF noise in my AC line, judging from the strong spurious spikes. like operation of a large elevator motor. or maybe my used DMM is just old  :-DD (@1NPLC, Y scale in nV)
« Last Edit: October 30, 2015, 10:34:29 pm by 3roomlab »
[What Do You Care What Other People Think?] ... remember to brush your teeth before you read it ... *tchk tchk tchk* ...

Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo