Failed Conducted and Radiated EMC with 60W offline Flyback LED driver

[ocset]

Hello,
We recently took our 60W Offine , isolated, PFC’d Flyback LED driver (schematic attached) for Conducted and Radiated EMC scans.
The conducted EMC scan fails at a few spot frequencies  as the attached shows.     
We also failed Domestic Radiated emissions (CISPR F). –But we passed on Industrial Radiated emissions (CISPR B).

The problem is, that  we have no more room on the PCB for any more filter components. As it was, we had to use SMD Y2 capacitors  on the bottom of the board because there is no room on the PCB for radial ones.   

I believe we need a second common mode choke, but there is no room on the board. 
Are there any filter tricks we can pull off here?...unfortunately we aren’t allowed to make the board any bigger.
The ceramic Y2 caps were 3.3nF….but as you know, ceramic X7R capacitors are unfortunately only around 10% of nominal  value when operated at rated voltage. 

The 60W Flyback’s   FET and Diode had small TO220 heatsinks directly connected to  them..The FET  heatsink  is then obviously connected to the drain tab of the FET.…..however, do  you agree that this would have made the Radiated scan result worse? (…I believe we should have had an insulating pad between FET heatsink and the FET itself, do you agree?)

During the scans, we had  a 60W LED load that we “hacked”  together connected……Unfortunately, due to the “hacked nature”, much of the wiring in the LED load bank was not twisted and so there were wide area current loops. The LEDs were on bits of MCPCB, and these were not connected to Earth. The secondary 0V was just “floating”.

(We passed Mains Harmonic Emissions and Power Factor     )
 
-----------------------------------------------------------------------------------------------------------
Some datasheets of our filter components....................

VJ2220   SMD  X2  and Y2 capacitors   
https://www.vishay.com/docs/45020/vjsafetycapsx7r.pdf

ELC09D inductorshttps://www.eevblog.com/forum/Smileys/default/facepalm.gif
www.farnell.com/datasheets/2245397.pdf

Wurth 7448640415   Common Mode Choke
http://katalog.we-online.com/pbs/datasheet/7448640415.pdf

[T3sl4co1l]

Inefficient use of filtering components, and lots of peaks at critical frequencies.  Give or take what impedance the supply appears as, and what capacitance those ceramic chips are down to, at voltage (probably <1nF, you'd be better off with C0G).

I have some plots you may be interested in, but they're too big to post here.

Tim

[ocset]

Thanks, ....in the meantime, we did another quick Conducted EMC scan, this time with the attached EMC filter. This performed much worse than the first EMC filter (shown in the top post). This worse conducted EMC scan (on 8th November) is also shown attached here. (the better scan of 22nd August is also attached here, for ease of reference)
We believe that the worsening is due to the fact that we no longer have a Y capacitor from primary to secondary of the 500mW bias supply transformer. We also believe that the change to using Radial Y capacitors in the AC part of the filter, as opposed to SMD ones as used in the previous scan, has also had an impact. –This being due to the increased stray inductance of the radial Y capacitors.
Also, in the worse scan, our LED load was sitting on a much bigger heatsink. Also, in the worse scan, the Power NFET was screwed to a bigger metal heatsink. (the tab of the FET being the switching node means that this made EMC worse).

So do you agree that we need to have Y capacitors between primary and secondary of both transformers in order to improve things?

I have also attached the (better) previous scan of 22nd August here. (Also the EMC filter schematic is also atttached)

[nctnico]

I'd try and fix the problem at the source: first figure out using a small H probe where the emissions originate from and then modify one by one to get the emissions down. Filtering is not the solution! Filtering is trying to cover up the symptoms. You really need to measure and understand how the currents are flowing through your circuit and what their frequency spectrum is.

[chris_leyson]

I've just spent two weeks tracking down a stubborn band of noise at around 12MHz when doing conducted tests on a board with two sets of line filters, one filter is high current 13A or so and the other one is for a small flyback. It was the flyback supply causeing most of the problems and using a much lower inductance common mode choke helped a lot at 12MHz, I could get a reasonable margin, but then at the lower frequencies, massive fail. So it turned out to be common mode noise. The boss said "try two turns on a ring core in the earth lead" and what do you know, two turns on a small Manganese Zinc core and nearly 20dB reduction because most of the common mode ended up going down the earth return. You learn something new every day.

@nctnico, a small H probe was very useful finding one source. A high current flowing from the primary earth through the Y caps across the transformer and into the secondary earth. Removing or changing the transformer Y cap made little difference. Breaking the primary secondary earth with a 10uH inductor and that's fixed that current path but the conducted emissions go up by 10dB. It just stears more current back down the earth wire. I've got a good margin and can't change the board layout so will have to fix that problem another day.

[TimNJ]

Just throwing this out there.

For conducted emissions, most of the time, <500kHz = differential mode, >500kHz = common mode. But you don't have to guess. You can build a box like this: http://www.hottconsultants.com/techtips/CM_vs_DM%20Conducted_Emission.html

+1 for earth-choke as above, if your output is earthed.

And a lot of this comes down to your transformer construction too. Shield windings, core shields, parasitic capacitance between primary and secondary windings, etc.

[chris_leyson]

@TimNJ Thanks for for link to Henry Ott's article. I just remember I've got his book on my bookshelf  I asked the guys who run the EMC facility "can you get hold of a dual LISN ?" and they said there is no such thing, I bet there is but they haven't looked. The other option was to use two LISNs and there just happen to be two Rohde and Schwarz LISNs sitting side by side, use one for L and the other for N and a 0 degree or 180 degree combiner, or will a splitter will do because the LISN will only give you access to L or N and not both at the same time. The answer was "you're not connecting that up to my test receiver I don't wan't you to blow it up" I will just show him the article about the LISN MATE. A dual LISN would be a useful piece of kit and I can't understand why Rohde and Schwarz just don't provide the extra outputs, it wouldn't add much to the cost.

[SiliconWizard]

The 60W Flyback’s   FET and Diode had small TO220 heatsinks directly connected to  them..The FET  heatsink  is then obviously connected to the drain tab of the FET.…..however, do  you agree that this would have made the Radiated scan result worse? (…I believe we should have had an insulating pad between FET heatsink and the FET itself, do you agree?)

I think I'd have used an insulating pad here. The heatsinks will probably act as antennas.

[TimNJ]

@TimNJ Thanks for for link to Henry Ott's article. I just remember I've got his book on my bookshelf  I asked the guys who run the EMC facility "can you get hold of a dual LISN ?" and they said there is no such thing, I bet there is but they haven't looked. The other option was to use two LISNs and there just happen to be two Rohde and Schwarz LISNs sitting side by side, use one for L and the other for N and a 0 degree or 180 degree combiner, or will a splitter will do because the LISN will only give you access to L or N and not both at the same time. The answer was "you're not connecting that up to my test receiver I don't wan't you to blow it up" I will just show him the article about the LISN MATE. A dual LISN would be a useful piece of kit and I can't understand why Rohde and Schwarz just don't provide the extra outputs, it wouldn't add much to the cost.

Any time!

Also, you can use this approach to get both the DM and CM components. More complicated though. Plus, if you get very little signal through the LISN MATE, then you know it must be differential-mode noise.



https://www.edn.com/Home/PrintView?contentItemId=4169136