Author Topic: Modifiying EMC filter in offline SMPS in order to pass conducted EMC scan  (Read 1807 times)

0 Members and 1 Guest are viewing this topic.

Offline ocsetTopic starter

  • Super Contributor
  • ***
  • !
  • Posts: 1516
  • Country: 00
Hi,
We have done an offline, 240VAC PFC’d Flyback LED driver. We have a slight failure of conducted emissions at 30MHz, as attached. In order to combat this, we wish to use two common mode chokes instead of just one like we have at the moment. That is, we wish to have a 10mH common mode choke, and a 500uH common mode choke in cascade…we believe that the smaller one will help attenuate our 30MHz problem.
The problem is that we currently only have one PCB and the scan session is on Monday. The PCB is very densely populated. So anyway, we cant layout a new PCB in time. …What about if we de-solder the existing common mode choke, and then use twisted pair “flying wires to “jump” over to a bit of stripboard, on which we will mount the two replacement common mode chokes as discussed?
Will this modified setup create so many other EMC problems that it won’t be a valid and worthwhile test?
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Another point is that we would like to experiment with different values of diff mode inductors. As such, how good (or bad) a practice would it be to have these diff mode inductors coming out on flying wires so that we can change value quickly and take a fresh conducted mode scan?
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Also, suppose we depopulated the entire EMC filter of the offline SMPS….(the section upstream of the mains rectifier bridge), and then made one using the same components, but on a piece of stripboard, and then wired this stripboard into the PCB, (so it again had a filter) how much different would the scan of the top post be, when dont with the "replacement" stripboard filter?
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
I think you can see what I am getting at here…….the situation of needing to vary the EMC filter components and then re-scan.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Having to make a whole new PCB every time an EMC filter component is changed, and then re-EMC-scanning, is not terribly practical.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Also, I have often seen offline SMPS EMC filters, with 22pF (or so) ceramic capacitors across the diff mode inductors. The purpose of these is to have a component that you can vary in order to reduce particular peaks on the conducted EMC scan. What do you make of this practice?
Also, at one company, they had done a 150W Boost BCM PFC stage, and across the boost inductor, they had a ~100pF ceramic capacitor in series with a ferrite bead. When asked about it, they simply said it failed if that was not there, so they were leaving it in. This was at a very large successful multinational TV company.
 

Online SiliconWizard

  • Super Contributor
  • ***
  • Posts: 15797
  • Country: fr
We have a slight failure of conducted emissions at 30MHz, as attached.

Isn't there one at ~70kHz as well?
 
The following users thanked this post: ocset

Online coppercone2

  • Super Contributor
  • ***
  • Posts: 11336
  • Country: us
  • $
well if you have a decent LISN and you have frequency information, perhaps you can just see what the magnitude change at your interest frequency is with your own equipment (as in direction of change) to get an idea of what is happening (despite not having accurate amplitude information). I think so long the LISN is built good, at 30MHz you won't get incorrect delta sign information.

Goes up goes down might be enough to choose components.
« Last Edit: July 27, 2019, 04:43:53 pm by coppercone2 »
 
The following users thanked this post: ocset

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
...We have a slight failure of conducted emissions at 30MHz, as attached. In order to combat this, we wish to use two common mode chokes instead of just one like we have at the moment. That is, we wish to have a 10mH common mode choke, and a 500uH common mode choke in cascade…we believe that the smaller one will help attenuate our 30MHz problem.
...

This is the bass ackwards approach: reduce the noise emission first, then tweak the filter components, if necessary. That said, it's tough to prevent stray capacitance from interfering with your inductors and stray inductance from interfering with your capacitors at 30MHz.

Snappy recovery and/or junction capacitance ringing in the secondary rectifier is the what I would look for first, followed by ringing between switch capacitance and transformer leakage inductance occurring at the instant of turn-off. Consequently, the RC damper across the secondary diode or the RCD clamp/snubber across the switch (or, better, transformer primary) should be checked more closely. It may very be that only some component values in these networks need to be adjusted, rather than resorting to a second stage of common mode filtering.



 
The following users thanked this post: ocset

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Put an RC on the switch node.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 
The following users thanked this post: ocset

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Such bump in the HF end to my untrained eye seems like HF noise bypassing EMC filter. Why there's such a HF noise in your supply in the first place?! Playing with chokes most likely will not fix it. First thing: find main EMI radiators inside your supply, then look which components may be coupling together with EMC filter components or just some mains trace, play with ferrite beads on diode legs and resistors on gate of switch transistor(s), not EMC chokes. This is reminder why everybody with intention to pass EMC shall have precompliance spectrum analyzer with LISN in-house because failed tests may cost more than that
 
The following users thanked this post: ocset

Offline ocsetTopic starter

  • Super Contributor
  • ***
  • !
  • Posts: 1516
  • Country: 00
Thanks,
There is a 60W flyback in DCM, and also a 1W flyback which operates in CCM, albeit in a pulse skipping way. The 1W flyback is controlled by a TNY286, and we have no access to the FET gate inside the TNY287. We are wondering about putting an RC snubber on the Drain-Source of the TNY286…….Well, rather, it would be the drain-ground of the TNY287 , because of course, we don’t have access to the source of the internal fet in the TNY286.
We are wondering about disabling the  TNY286 and supplying  the primary and secondary side bias circuitry via batteries….so we can see how much of the noise is coming from the 1W flyback, and how much is coming from the 60W flyback.
We have RCD clamps across the primary  of both flybacks.

Another point is that many isolated offline SMPS’s have   Y  Capacitors  across the isolation barrier on both sides of the transformer. Do you agree that this  acts to further reduce common mode emissions  compared to having just one Y capacitor on one side of the transformer?
« Last Edit: July 28, 2019, 12:31:18 am by treez »
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
There is a 60W flyback in DCM, and also a 1W flyback which operates in CCM, albeit in a pulse skipping way. The 1W flyback is controlled by a TNY286, and we have no access to the FET gate inside the TNY287. We are wondering about putting an RC snubber on the Drain-Source of the TNY286…

The 60W DCM flyback is the culprit for sure. Some of the TinySwitch ICs have reliability problems but aren't too noisy from what I recall. If nothing else, their internal switches tend to have a high Rds[on] which usually provides good damping of any resonance between the switch output capacitance the transformer leakage inductance. Also, running in CCM eliminates a lot of the potential noise sources in a flyback.

Another point is that many isolated offline SMPS’s have   Y  Capacitors  across the isolation barrier on both sides of the transformer. Do you agree that this  acts to further reduce common mode emissions  compared to having just one Y capacitor on one side of the transformer?

A Y capacitor from primary ground to secondary ground shorts out the common mode current that would otherwise be transmitted via the distributed capacitance of the transformer so, yes, it can greatly reduce the emission of common mode noise, but it might exacerbate the conduction of such, or even vice versa.

I (and Tim) already suggested the route you should be taking, which is adding/tweaking RC dampers across the switch/primary and the diode. One other thing you can try if using an RCD clamp on the primary is substituting a slow 1N4007 for the clamp diode, intentionally using the slow reverse recovery to suppress ringing at the instant of turn-off.

I'm still leaning on it being the rectifier junction capacitance ringing with transformer leakage at the instant of switch turn on, however.
 
The following users thanked this post: ocset

Offline chris_leyson

  • Super Contributor
  • ***
  • Posts: 1549
  • Country: wales
I would tend to agree with Tim and MagicSmoker, try damping the switching node. Try different RCD or R2CD snubbers, also try standard recover diodes like 1N4007 instead of fast recovery diodes. See T.I. SVNA744 app note for R2CD snubbers. http://www.ti.com/lit/an/snva744/snva744.pdf
 
The following users thanked this post: ocset

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
The problem is that we currently only have one PCB and the scan session is on Monday.

How did it go? Any results, conclusions? What's next?
 
The following users thanked this post: ocset

Online SiliconWizard

  • Super Contributor
  • ***
  • Posts: 15797
  • Country: fr
So yeah, did that work out ok?

Or is your boss very angry and decided to subcontract this design to some random chinese company? ;D
 
The following users thanked this post: ocset

Online coppercone2

  • Super Contributor
  • ***
  • Posts: 11336
  • Country: us
  • $
its in dragon temple right now being dehexed

« Last Edit: August 01, 2019, 12:11:33 am by coppercone2 »
 
The following users thanked this post: ocset

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Or is your boss very angry and decided to subcontract this design to some random chinese company? ;D
// The following users thanked this post: treez

So product failed miserably and you lost your job? Please say something besides just pushing "thank" button ;)
After all we invested our time answering. Do your part of the deal - provide feedback.
« Last Edit: August 04, 2019, 03:25:32 pm by ogden »
 
The following users thanked this post: ocset

Online coppercone2

  • Super Contributor
  • ***
  • Posts: 11336
  • Country: us
  • $
its gonna be literary hitting shelves this January (when it melts out of the plastic chassis)

custom transistors specified for 350c operation, welded, but enclosed in recycled styrene that someone gathered from all the water cooler trash bins in england
 
The following users thanked this post: ocset, InductorbackEMF

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
its gonna be literary hitting shelves this January (when it melts out of the plastic chassis)

custom transistors specified for 350c operation, welded, but enclosed in recycled styrene that someone gathered from all the water cooler trash bins in england

No offense, but not even close to funny  :-//
 
The following users thanked this post: ocset


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf