Common mode choke off the power supply PCB is better?

[ocset]

Hello,
We have failed Conducted mode emissions with our offline 150W LED driver. 

However, we added a Common Mode Choke in the mains cable (as in the attached) and then it passed conducted EMC.

We now are wondering about making room on the LED driver PCB so that we can fit the common mode choke on the PCB.
However, do you agree that if we place the very SAME common mode choke on the PCB, then we will have less chance of passing  conducted emissions, because common mode noise is high frequency and can “radiate” around the common mode choke on the PCB and get into the earth return wire like that?

(The common mode noise is doubtless coming from the 1W Buck bias supply  on this 150W Linear LED driver. The Earth wire in the mains cable does not actually get onto the PCB, but rather is screwed to the heatsink.)

[SeanB]

Put it on the board right by the input, and add a Class X capacitor across the mains input and output of the choke, and add the 2 class Y capacitors between active and neutral to the board, and add a wire to the earth point on the heatsink from the common connection of the 2 Y capacitors. That should keep you below the limit.

[fcb]

How much did you fail conducted by?  Can you post the scan.

Are you budget restrained?  Don't necessarily fit the Y caps - you need to understand what other regulations you might hit.  For example common-mode 1.2/50us 4KV spikes?

Also, how many are you making? Will putting the choke on-board help your CEM, or will it hinder them (thru-hole? SMT?) - DFM!

I've been though this process quite a few times, which is why I now have a LISN and SA (as well as an anechoic chamber!).

[CopperCone]

i think the cable radiates noise, so you want it inside of the shielded chassis, but I guess you also would want it shielded in another compartment, if possible (like having bent sheet metal to make a barrier between the choke and the rest of the electronics.

If you can afford it, obviously a mains inlet filter or a filter module is better (like delta, tdk lambda, etc)... they might not be so expensive (dc versions exist) and will offer better performance then a choke.

[T3sl4co1l]

Do precisely what you tested at the test lab.

Even if you know what you are doing, getting creative will likely result in unfortunate results.

It's a good idea to do small, incremental tests, and test many permutations.  Make the change that's the best combination of simple, cheap and dB-of-margin.

Tim

[ocset]

How much did you fail conducted by?  Can you post the scan.

Thanks, yes ..please find the scans attached...getting better as you go down, as we add more filtration each time..
The bottom one is the pass as you'll know.

I know my organisation will not mind me posting this, as  this is just basic filtering as is standard for offline power supplies.

Yes, we are extremely cost constrained. We would like the Common mode choke to be surface mount  if possible, as our driver board is all surface mount so far.
We would like preferably a height of sub 6mm if possible, but can go higher if really need be.

We'd rather not have Y capacitors if possible. Also, we'd rather not wire from the board to the earth point as it will just be an extra assembly step.

I've been though this process quite a few times, which is why I now have a LISN and SA (as well as an anechoic chamber!).

Sounds great, we thought of buying a rigol SA for £1000 and a Hameg LISN for £1000  but the price put us off.
-Plus we thought we might  struggle to get all that grounding setup done for the EMC test lab...all the copper sheeting etc, and having to earth the LISN to the ground plane by welded connection etc...all sounds expensive.
-Its a real nuisance having to pay a couple of hundred quid each time we visit the EMC place.

They dont seem to be able to offer us a common mode only scan.....even though we believe it just involves adding a ferrite splitter to the LISN?

[MattHollands]

I don't want to hijack the thread since I'm not able to offer any advice, but I was wondering exactly what those captures of the EMC testing showed.
I gather that the two solid lines, labelled EN55015 are the maximum allowed emissions at each frequency, but what exactly do the two lines show? Common mode and non-common mode? Does this only show conducted emission? Why does one line only start at  ~120kHz? Why is there a discontinuity in your blue line? Why is half of it red?

Sorry for all the questions. Just very interested as I have never seen these kind of reports before.

[T3sl4co1l]

Average and quasi-peak limits.

I don't know why the plot is different below 150kHz, but 150kHz is the FCC cutoff, while a lot of European stuff goes down much lower.  It could be the measurement method is different down there (different receiver time constant, or only average matters, etc.).  You'd have to read EN55015 to find out.

Tim

[ocset]

Thanks, please do hijack, i like to help you....yes its different below 150khz as the time window of the spectrum  analyser measurmeent is different below 150khz because these lower frequencies are deemed to be less of a problem.
Red is quasi peak limit
Blue is average limit.

Quasi peak measurement means that noisy occurrences that are very high in magnitude , but not often  in terms of frequency, are still taken proplerly account of.......its like a bomb going off every hour throughout your night.....overall, the average sound is low....but its going to jolly well ruin your night of kip....so quasi peak kind of "collars" those kind of infrequent but jolly well "loud" electrical noises....so to speak.

Yes it is conducted emc only....but it is a mixture of common mode and diff mode emissions....goodness only knows what proportion of each...theres no way of knowing lest we use a splitter to separate common mode from differential mode noise.....

its interesting you ask, because this area is steeped in industrial secrecy as are many regulatory things, because the huge corporations dont want small startups competing with them...so they make it all difficult....only the power integrations app  notes deal proplerly with emc description....is it AN15?...  or AN17 , or something similar...worth reading if you want to know a bit more.

and yes if you go to an emc test house and they do a scan for you, they wont tell you its a mixture of common mode and diff mode noise...then theyll laugh at you when you repeatedly come for re-scans having filtered out the wrong kind of emission.......they wont tell you that your massive diff mode filter doesnt even touch the common mode emissions...they will laugh and bank the extortionate fees which they charge you...........as you come back and fail the scans again and again and again.

No surprise to hear a fellow Brit not knowing what EMC scans are about....neither did i for a long time.....we Brits dont share our knowledge with each other...not like the Germans do.
Thats why UK owned companies  command less than 1% of world exports, whereas it used to be 25%.
https://massey276.wixsite.com/ukelectronics

..the UK will be a third world country within a decade.
66% of uk industry has been sold into foreign ownership
>50% of uk vital services have been flogged off overseas
eg uk owns none of its power stations or electricity network.

UK national debt is 5 times that of  Greece.

UK having sold off most of its industry will not be able to repay the national debt, and currently uk politicians are "churning" the GDP figures in order to keep the IMF from giving UK a bad credit rating.....but its all hullaballoo, and we'll drop like a stone soon.

The UK has had a negative balanace of trades for the last 33 years...ie we havent made any money for the last 33 years...just got deeper and deeper into debt....third world here we come.

I told Dave Jones about this and sent him my report, and he was shocked (he emailed me back personally) at the truth of the statistics regarding the UK and its electronics industry (or lack of it) etc..and the demise of UK.

I had sent Dave my report and SMPS course which i wanted to kind of start the rejuvination of UK industry, and wanted to pull Australia up with us...but to be honest,  Australia isnt in the mess that UK is in.

[jbb]

Sorry you had a bad time, but you might need a new test lab. Normally a test lab would offer a 'debug' scan to give you a ballpark, and (if you ask nicely) provide some coments on the patterns they see.

You can now convince your management to fork out for a LISN and spectrum analyser by saying "this equipment could have saved us a trip to the lab, which cost £XXX."

If you have a certificate now, you must keep building the power filter equipment as tested. If you don't have a certificate, you can try an SMD choke but it might fail (and cost $$$). As Teslacoil says, if you make changes you also make risk.

[ocset]

Thanks,
Do you know of cheapest conducted emc setup?.....cheapest we could find is hameg lisn for £1k, and rigol spec analyser for £1.5k
...very expensive, and we were going to just jack out of all that copper grounded sheeting stuff...and welding the lisn ground to the copper plane....very expensive....just use the bare lisn and spec analyser and hope thats ok for up to 5mhz, which is where we have mucho problemo

[CopperCone]

I typically tried to get things to pass 10dB under whatever class (usually worth arguing to boss) it was being tested against.

If someone retests the thing, or it gets integrated into another system, its nice to have some margin, plus there could be manufacturing variances (not like you are testing 100 of these in the lab, capacitance, rise time, etc of some part might change). It also gives you room for error if one of your parts you use goes obsolete and the only replacement you have is sub standard. And the engineers are not so terrified to make improvements or variants because 'oh shit its gonna fucking fail EMC if we move this trace by 0.5mm'

Also its future proofing. Literary no one wants to go back and redesign stuff because someone had a complaint. It shoots you in the foot and can hold up an entire assembly line or end up requiring drastic measures like a chassis redesign to fit some kind of giant off the shelf filter to keep things 'running smooth'.

As for common mode scan, I only ever saw scans on the conductors, you can superimpose them to see if something is common mode or not. I never saw an actual common mode scan (I think its more accurate to measure each conductor and do the math rather then trying to make some kind of impedance element to function reliably).

Plus, it basically improves product reliability. Its a protection system (most people don't realize this).

[Phoenix]

I wonder if your problem is actually not common mode, but rather differential mode. Your problem is in the 100's of kHz range - you need quite a high stray capacitance to earth to have broadband common mode problem in this range. Your MHz range (where you expect radiated and capacitive CM) is almost at the receiver noise floor to begin with.

The first step you took was to add a 220nF X2 cap (presumably across line-neutral). This is a differential filter only and got you about 20-30dB reduction! The next step was a 10mH (large value) common mode choke, which itself will have 100uH or so of differential leakage inductance.

Your circuit, you say, is a buck converter. They draw harmonic rich highly discontinuous currents - you need a lot differential filtering effort to get rid of this.

That being said EMI filters often obtain their necessary DM filtering from the leakage of CM chokes by design (or accident) - perfectly legitimate technique with its own advantages.

[skipjackrc4]

As for common mode scan, I only ever saw scans on the conductors, you can superimpose them to see if something is common mode or not. I never saw an actual common mode scan (I think its more accurate to measure each conductor and do the math rather then trying to make some kind of impedance element to function reliably).

Probably the easiest way to tell if conducted noise is CM or DM is to place a current probe around both leads running the same direction.  This measurement shows the CM spectrum.  After the CM measurement, reverse one of the leads through the probe so that you are looking at DM current flow.  It is important to note that absolute amplitude of these measurements isn't critical--we're only trying to compare the two spectra to see which contains the offending frequencies.

Another option is to use LISNs and a resistive power combiner.  This will show you common mode current, and may have better sensitivity and frequency range over a current probe.  A transformer based 180 degree hybrid coupler can then show the differential current.

[T3sl4co1l]

Probably the easiest way to tell if conducted noise is CM or DM is to place a current probe around both leads running the same direction.  This measurement shows the CM spectrum.  After the CM measurement, reverse one of the leads through the probe so that you are looking at DM current flow.  It is important to note that absolute amplitude of these measurements isn't critical--we're only trying to compare the two spectra to see which contains the offending frequencies.

Fortunately, the OP posted spectra essentially showing just this.

Tim

[ocset]

As for common mode scan, I only ever saw scans on the conductors, you can superimpose them to see if something is common mode or not. I never saw an actual common mode scan (I think its more accurate to measure each conductor and do the math rather then trying to make some kind of impedance element to function reliably).

Thanks, I thought even with separate live and neutral conducted scans, there’s no way of actually knowing from that what is the common mode spectrum only?

I wonder if your problem is actually not common mode, but rather differential mode. Your problem is in the 100's of kHz range - you need quite a high stray capacitance to earth to have broadband common mode problem in this range. Your MHz range (where you expect radiated and capacitive CM) is almost at the receiver noise floor to begin with.

Thanks, but I can assure you it is common mode, even down at 150khz….we believe the causative problem is many MHz but it “envelopes” at 150khz-400khz, hence we see it at 150khz-400khz.
We know its common mode (even at the low frequencies) because a big diff mode (only) filter barely touched it, then we put in a common mode choke (with a lesser diff “leakage” inductance) and it went right down.
I mean there is also diff mode  in the 150khz-400khz range, but mostly its common mode. Due to our entire circuit and load being on an earthed heatsink…above 400khz and yes its all common mode……our big diff mode filter didn’t touch it above 400khz.

[Phoenix]

Thanks, but I can assure you it is common mode, even down at 150khz….we believe the causative problem is many MHz but it “envelopes” at 150khz-400khz, hence we see it at 150khz-400khz.
We know its common mode (even at the low frequencies) because a big diff mode (only) filter barely touched it, then we put in a common mode choke (with a lesser diff “leakage” inductance) and it went right down.
I mean there is also diff mode  in the 150khz-400khz range, but mostly its common mode. Due to our entire circuit and load being on an earthed heatsink…above 400khz and yes its all common mode……our big diff mode filter didn’t touch it above 400khz.

I may be misunderstanding your second plot then. You say you added a 220nF X2 capacitor (i ASSUME across line-neutral) and got 30dB reduction!?

[fcb]

and yes if you go to an emc test house and they do a scan for you, they wont tell you its a mixture of common mode and diff mode noise...then theyll laugh at you when you repeatedly come for re-scans having filtered out the wrong kind of emission.......they wont tell you that your massive diff mode filter doesn't even touch the common mode emissions...they will laugh and bank the extortionate fees which they charge you...........as you come back and fail the scans again and again and again.

I have to say that's not my experience of EMC labs, they have been very helpful when it comes to hunting down the source(s) of noise.  And they are free to charge what they want, they aren't charities!  The investment in gear is extraordinary, £1M for a chamber, £130K for an EMI receiver, £7K for an antenna, £4K for a LISN - not to mention the costs of annual calibration and certification.

Also, a LISN doesn't select common mode noise - it has a switch on it that selects either L or N lines to listen to.  Worth having a look for the schematics, quite simple.  I built one in about 4 hours for £20, not as good as a proper one, but more than good enough to trace and eliminate my problem. Typically you run two scans (one L, one N) and deduce if you've got common mode noise or not.

[jbb]

Thanks,
Do you know of cheapest conducted emc setup?.....cheapest we could find is hameg lisn for £1k, and rigol spec analyser for £1.5k
...very expensive, and we were going to just jack out of all that copper grounded sheeting stuff...and welding the lisn ground to the copper plane....very expensive....just use the bare lisn and spec analyser and hope thats ok for up to 5mhz, which is where we have mucho problemo

That seems roughly right.  A low-cost spectrum analyser will not have quasi-peak detection, but that's not a deal-breaker.  Power electronics devices tend to produce fairly continuous emissions, so you can just compare Average power measurements. A certification grade spec. an. has a much better front end (more linear, lower noise, greater dynamic range), detector stage(s) and software.

Buying a LISN is probably a good plan, as it will already have passed some kind of safety review.  It's response vs frequency will probably be a bit wobbly though.  I'm not sure what the real differences are for a certification grade one (possibly hand-assembled by monks...)

I'm not sure if a ground plane would be required.  But it may not be as expensive as you fear.  For a basic system, you can use a (stainless) steel sheet, and a (short, fat) cable to the LISN for ground.  To get your equipment up to the right height for minimum money, I suggest a couple of large plastic storage boxes (empty! don't let people put stuff in them!).

The shielded room can be avoided if you take a reference measurement (with your device under test switched off) occasionally when you're testing.  I once spent days chasing after a 13.56MHz spike that turned out to be ... stray radiation from a wireless card reader on the door to the workshop.

Given the above, you can do some work 'at home.'  You'd typically get a product scanned in the certification lab, and say 'so this failed by X dB at Y kHz.'  Then you go to your 'home' lab and measure Z dB at Y kHz, and know that you want to knock it down to Z - X - 10dB margin.

If you want to get into the CM vs DM measurement, you'll need a LISN with twin outputs (i.e. 2 BNC connectors) and a set of combiners (e.g. wideband transformers). Or maybe you can find a LISN with this feature built in?  Note that the standards don't seem to care, they just measure one wire at a time (mix of CM and DM).

Another option is to use LISNs and a resistive power combiner.  This will show you common mode current, and may have better sensitivity and frequency range over a current probe.  A transformer based 180 degree hybrid coupler can then show the differential current.

Finally, try to be really consistent with your wiring.  Copy how they laid it out at the certification lab.  Tape wires down so they don't flop about ('cause you'll get different readings and can't make quality comparisons).

Good luck.

[ocset]

I may be misunderstanding your second plot then. You say you added a 220nF X2 capacitor (i ASSUME across line-neutral) and got 30dB reduction!?

Thanks, yes thats right. Yes i agree that was filtering off the diff mode bit at those frequencies.

[dmills]

Your X2 cap is making the quasi peak line drop at a very interesting 20dB per decade between about 150 and 600kHz, and that number should mean something, you may wish to think about raising the differential mode impedance downstream of this cap to put the corner frequency a bit lower (might get you a 'just' pass).

I would suggest that the earthed heatsink is probably not doing you any favours for common mode, a class Y cap or two between some carefully selected points on the PCB and the heatsink mounting screws would probably help by providing a return path for capacitive coupling.

My experience of EMC houses has been that they are usually quite willing to help, and will often suggest likely fixes, especially during a quick 'pre-compliance' test session.

Regards, Dan.

[ocset]

I would suggest that the earthed heatsink is probably not doing you any favours for common mode, a class Y cap or two between some carefully selected points on the PCB and the heatsink mounting screws would probably help by providing a return path for capacitive coupling.

Thanks, i tend to agree, but isnt it  curious that the probelm itself is caused by stray capacitive coupling to the earthed heatsink, -then the solution involves adding more capacitance (the Y caps) to the earthed heatsink.?

[Phoenix]

I would suggest that the earthed heatsink is probably not doing you any favours for common mode, a class Y cap or two between some carefully selected points on the PCB and the heatsink mounting screws would probably help by providing a return path for capacitive coupling.

Thanks, i tend to agree, but isnt it  curious that the probelm itself is caused by stray capacitive coupling to the earthed heatsink, -then the solution involves adding more capacitance (the Y caps) to the earthed heatsink.?

The problem in smps is nodes with very high dv/dt capacitively coupling larger currents to (I=C dv/dt) to ground. The only return path is through the cord. Adding a Y cap from an appropriate electrically quiet node (eg a dc bus rail) to the heatsink gives it a return path. This often requires testing between different nodes.

You also need to look out for creating additional 50/60hz leakage paths.

Do you have any Y caps as part of your CLC filter too?

Sent from my D5833 using Tapatalk

[ocset]

The problem in smps is nodes with very high dv/dt capacitively coupling larger currents to (I=C dv/dt) to ground. The only return path is through the cord. Adding a Y cap from an appropriate electrically quiet node (eg a dc bus rail) to the heatsink gives it a return path. This often requires testing between different nodes.

Thanks, so are you saying that you can prevent noise from becoming common mode noise by redirecting it (via the Y cap) back out of the earthed heatsink and having it return via either live or neutral?…….thus its no longer common mode noise since it has not returned via earth.

Do you have any Y caps as part of your CLC filter too?

Thanks, no we dont, we dont actually take the earth wire to the pcb, -only to the heatsink. We'd prefer not to use y capacitors as they are PTH and we only want SMD on the PCB.

--------------------------------------- --------------------------------------------------
During this thread about power supply common mode noise problems….
http://www.edaboard.com/showthread.php?t=368189

…there is discussion, especially by the expert Mtwieg, about common mode emissions, and even discussion (post #34) of  differential  mode noise being transformed onto common mode noise.

Post#7 of the above thread sets the scene, and states the theories about common mode emissions  that later in the thread get totally trashed.                                    

Regarding the Live and neutral conducted emissions scans being the same or different than each other,  post #34 says…

**Purely CM interference and purely DM interference would show up the exact same results in magnitude measurements on each line. When both CM and DM interference sources are present, then you will see different spectrums on L and N, however it's not possible to determine the relative contribution of CM and DM interference. In some tests, a measurement with a combiner can be used to extract just the common mode component, as discussed here.**

Post#22  of the above thread states that…
**Y capacitors force conducted EMI from the load side back into earth, away from the mains.**

…this seems to be the opposite of what has been discussed in our EEVblog thread?
------------------------------------------------ ------------------------------------------------------------------------------------

[Phoenix]

Take what people on a forum (including me) say about EMI and emissions with a healthy dose of salt!

The problem in smps is nodes with very high dv/dt capacitively coupling larger currents to (I=C dv/dt) to ground. The only return path is through the cord. Adding a Y cap from an appropriate electrically quiet node (eg a dc bus rail) to the heatsink gives it a return path. This often requires testing between different nodes.

Thanks, so are you saying that you can prevent noise from becoming common mode noise by redirecting it (via the Y cap) back out of the earthed heatsink and having it return via either live or neutral?…….thus its no longer common mode noise since it has not returned via earth.

Do you have any Y caps as part of your CLC filter too?

Thanks, no we dont, we dont actually take the earth wire to the pcb, -only to the heatsink. We'd prefer not to use y capacitors as they are PTH and we only want SMD on the PCB.

Current wants to couple from the switching nodes into the heatsink via the nice large flat transistor surfaces. This current needs to have a return path somewhere:
Your device -> Heatsink -> earth lead -> LISN (250 nF, 50 Ohm) -> active and neutral -> Your device

Adding Y caps from the heatsink to a select node short circuits the LISN path:
Your device -> Heatsink -> Y capacitor -> Your device
But also opens up the new current path:
Your device -> Y capacitor -> Heatsink -> earth lead -> LISN (250 nF, 50 Ohm) -> active and neutral -> Your device
So you need to chose your nodes carefully!

**Purely CM interference and purely DM interference would show up the exact same results in magnitude measurements on each line. When both CM and DM interference sources are present, then you will see different spectrums on L and N, however it's not possible to determine the relative contribution of CM and DM interference. In some tests, a measurement with a combiner can be used to extract just the common mode component, as discussed here.**

While true, doesn't really give any guidance as to why. For conducted emissions test there are 3 paths in/out of your device: active neutral and earth. Also note we need to obey KCL.

The way I think about it is that:
Differential mode is any current which is equal but opposite in magnitude (180deg phase shifted if you will) when comparing active and neutral. Therefore it is flowing from active to neutral.
Ia_dm = -In_dm
Common mode is any current which is equal magnitude and in phase when comparing active and neutral. Therefore the current must be flowing into earth.
Ia_cm = In_cm
Ia_cm + In_cm = Ie
Ie = 2Ia_cm = 2In_cm

So what about the case where there are different magnitudes in active and neutral e.g. Ia=0, In=2; what is that?
KCL says Ie=2
Therefore Ia_cm=1, In_cm=1
In_dm=1 and Ia_dm=-1

This is mixed mode - it contains both CM and DM components.

**Y capacitors force conducted EMI from the load side back into earth, away from the mains.**

I think this is about using Y caps in your CLC. The abstract theory is they allow the common mode currents to return back into your device without going via the LISN - effectively the same current paths as I described earlier, but different node on your device.