Author Topic: EMC testing, problems with MeanWell power supply  (Read 4487 times)

0 Members and 1 Guest are viewing this topic.

Offline honeybadgerTopic starter

  • Contributor
  • Posts: 42
  • Country: cz
EMC testing, problems with MeanWell power supply
« on: March 16, 2019, 01:12:34 am »
Dear EEVblog forum,
I am really desperate  |O

I designed, build prototypes of a new product for my colleagues. It is rather simple, Arduino Nano (will probably be replaced for just Mega328 in the future), 4x20 LCD, linear voltage regulators, some relays, OPAMPs and bunch of passives. It has a pack of 8x2,2mF  capacitors. And it is powered with 12V/1A DC power supply - MEAN WELL GS12E12-P1I.

Now it is being tested for EMC.

I was not present for the test but it failed to survive the 2kV pulses injection to the 230V AC. (we will sell it as a complete unit with the power supply so we were told it needs to be tested with the power supply)
My colleague (not an electrical engineer) was there and the only info I have is - it was dead and it was not able to recover by itself - unplugging of power supply was necessary.

I added the watchdog (I know it should be there from the beginning) and added a massive ferite of shame to the 12V power cable.
Again I was not present but it failed again - this time at least I have a video of that. It indicates that Arduino is working but the LCD is dead and the rest is almost dead - the circuitry (mainly relays) which is powered directly with 12V is not working.

Now I have two questions:
Is it possible for the hd44780 LCD to freeze without self recovery with such pulses?

Is it possible that MEAN WELL GS12E12-P1I power supply enters some kind of half dead state in which it does not provide 12V but maybe only 5V (5V-1,2V drop = 3,8V...  enough for Arduino to be alive, probably not enough for the LCD, not enough for 12V relays)? I mean it already has all the certifications. Or is it some kind of undocumented safety feature?

From what I saw it looks the power supply is the problem.
Next time I will try different power supply and will measure voltage after the test but I need to be prepared.
« Last Edit: March 16, 2019, 01:14:30 am by honeybadger »
 

Online nctnico

  • Super Contributor
  • ***
  • Posts: 28429
  • Country: nl
    • NCT Developments
Re: EMC testing, problems with MeanWell power supply
« Reply #1 on: March 16, 2019, 01:31:04 am »
Post a picture of your contraption. Likely the wiring is picking up the pulse and distributing it where you don't want to have it. Did you create your own board or slammed several boards together?
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline honeybadgerTopic starter

  • Contributor
  • Posts: 42
  • Country: cz
Re: EMC testing, problems with MeanWell power supply
« Reply #2 on: March 16, 2019, 01:47:17 am »
I would not say it is a contraption  ;D . It is a product ready to be sold when we get the required papers. There is a proper PCB. From symptoms I am 90% sure it is the power supply (I will be 100% sure next week).

I am not allowed to post pictures of the product. But there are no long power traces - no it is definitely not picking anything. It either goes through input power circuitry (polyfuse, TVS diode - I guess TVS diode is too slow for such pulses, 2x2,2mF caps dedicated for 5V rail, LM1117-5.0 and few other caps according to LM1117 datasheet). Or it messes up the power supply. That's why I am asking if someone know if this is some kind of undocumented safety feature.

Is it possible that MEAN WELL GS12E12-P1I power supply enters some kind of half dead state in which it does not provide 12V but maybe only 5V ?
 

Offline IconicPCB

  • Super Contributor
  • ***
  • Posts: 1564
  • Country: au
Re: EMC testing, problems with MeanWell power supply
« Reply #3 on: March 16, 2019, 01:53:15 am »
Ahhhh...
Yes, the Lazarus state readily broken by unplugging the corpse from power and breathing new current of life into it.
You have the hardware,,, be there the next time it fails and make the measurements.
 

Offline Tomorokoshi

  • Super Contributor
  • ***
  • Posts: 1212
  • Country: us
Re: EMC testing, problems with MeanWell power supply
« Reply #4 on: March 16, 2019, 02:06:36 am »
2kV pulses onto AC... Specifically, is this the EFT test? Which test standards and methods? It should be on the data report you got from the lab.

Who knows about the Mean Well supply itself. It's possible it could be running at perhaps half voltage after being hit. The switcher could be going into some kind of protective mode.

As for the LCD, they can absolutely get blanked out. If the LCD controller is only initialized once, then after the interference it won't process the commands and data properly. LCD interfaces for those types of displays are a marvel of cost reduction. No redundancy anywhere. Interference could be through power glitches, corrupted data packets, etc. A couple options are to attempt to read back status registers to see if the controller is still running properly, and to have either an occasional re-initialization or some kind of user-initiated initialization without dropping power. However, that won't help if your supply drops out.

A small plug-in supply may very well pass a fair amount of transient through when hit with 2kV. A ferrite won't do much; these really aren't high-frequency RF transients getting through. You would need some good filtering at the input of your board, and ideally the Mean Well would be able to filter it out.

4uF on 5V could get jolted by one of these tests.

Take apart one of the Mean Well supplies. See if you can identify the switcher, any protective components, etc.

Note that for some EMC tests, you get to somewhat declare what the symptoms of failure are.
 

Offline tkamiya

  • Super Contributor
  • ***
  • Posts: 2178
  • Country: us
Re: EMC testing, problems with MeanWell power supply
« Reply #5 on: March 16, 2019, 03:07:24 am »
I've done something long time ago.  For something like this, an engineer has to be present during testing.  That kind of pulse will bypass filtering and re-radiate inside the case, then gets picked up from all kind of places.  In our case, R/W line of Z80 board has to be tweeked with small capacitors.  Also, cabling had to be re-routed. 

It's not something you can guess from non-technical third party's statement.
 

Offline ejeffrey

  • Super Contributor
  • ***
  • Posts: 4032
  • Country: us
Re: EMC testing, problems with MeanWell power supply
« Reply #6 on: March 16, 2019, 03:38:11 am »
Yeah if you want to fix this you really need to be able to reproduce it in the lab where you can take measurements on it.  Otherwise declare that in the event of a surge, the device is allowed to lock up and may require a power cycle. If the device isn't safety or mission critical, that should be good enough. 
 

Offline honeybadgerTopic starter

  • Contributor
  • Posts: 42
  • Country: cz
Re: EMC testing, problems with MeanWell power supply
« Reply #7 on: March 16, 2019, 12:56:26 pm »
Hi, thank you all for your reply.

2kV pulses onto AC... Specifically, is this the EFT test? Which test standards and methods? It should be on the data report you got from the lab.
...
Take apart one of the Mean Well supplies. See if you can identify the switcher, any protective components, etc.
...
A small plug-in supply may very well pass a fair amount of transient through when hit with 2kV. A ferrite won't do much;
...
4uF on 5V could get jolted by one of these tests.

It is a part of declaration of conformity testing for CE. I don't have any protocol from the lab yet, it is a relatively small local lab with an older enthusiast who has a lot of patience with us. We were told it needs to recover by itself after the pulse to pass the test.

I will open the PSU to see what switcher is inside.

What kind of filtering do you recommend? I read about TransZorbs how good and fast they are - are they enough or is is a marketing BS? There are two of them - 14,6V on 12V input and 6,7V after the 5V linear vreg.
Should I add MOVs?
I still need to prepare for the eventuality it is not the PSU fault  >:(

Also the caps are not 4uF but 4400uF (4,4mF).


So, you put 2kV EFT waveform to the PSU, and it knocked out your electronics, not the PSU, is that correct?
That's the thing. I thing is it knocked the PSU in the first place. The PSU after the "knock" is not able to provide full 12V without a power cycle -> thus our device is also dead. I didn't believe this was possible. And there is no proper datasheet for this PSU.


« Last Edit: March 16, 2019, 12:59:10 pm by honeybadger »
 

Offline Gyro

  • Super Contributor
  • ***
  • Posts: 10172
  • Country: gb
Re: EMC testing, problems with MeanWell power supply
« Reply #8 on: March 16, 2019, 01:59:03 pm »
Ahhhh...
Yes, the Lazarus state readily broken by unplugging the corpse from power and breathing new current of life into it.
You have the hardware,,, be there the next time it fails and make the measurements.

I can't emphasise that one enough. There is absolutely no point in having someone non-technical accompany the equipment to the test house, you should be there, no excuses!

Most decent test houses will allow you to play after a test failure and the test techs are usually helpful in suggesting possible fixes - they watch a lot of gear go past. They usually have ferrites and copper tape handy too!

Go along with your own provisioned EMC test toolkit - debug equipment, soldering gear, your own copper tape, selection of caps, ferrites etc.

It is time very well spent, you get to witness the testing, see what effect on emissions etc. that small changes make. They have equipment that you will never have - these are invaluable learnings for getting your next product through quicker and easier.

This can make the difference between getting it through in one session versus having to book additional expensive test slots.
« Last Edit: March 16, 2019, 02:03:12 pm by Gyro »
Best Regards, Chris
 
The following users thanked this post: nctnico

Offline ebastler

  • Super Contributor
  • ***
  • Posts: 7375
  • Country: de
Re: EMC testing, problems with MeanWell power supply
« Reply #9 on: March 16, 2019, 02:46:25 pm »
I thing is it knocked the PSU in the first place. The PSU after the "knock" is not able to provide full 12V without a power cycle -> thus our device is also dead. I didn't believe this was possible.

Frankly, I don't think this is possible (or at least plausible) either. And you seem to be basing your assumption on rather circumstantial evidence, namely: "My circuit does no longer work after the 2kV spike, so the PSU must no longer be supplying the appropriate power to it".  ???

I think the more likely explanation is that the PSU does pass a spike to its 12V output, then goes back to supplying 12V as it should. Your regulator for the 5V rail may provide additional filtering, so the 5V devices don't experience a problem; but the devices running off 12V do.
 

Offline chris_leyson

  • Super Contributor
  • ***
  • Posts: 1549
  • Country: wales
Re: EMC testing, problems with MeanWell power supply
« Reply #10 on: March 16, 2019, 02:57:33 pm »
Quote
I've done something long time ago.  For something like this, an engineer has to be present during testing.  That kind of pulse will bypass filtering and re-radiate inside the case, then gets picked up from all kind of places.  In our case, R/W line of Z80 board has to be tweeked with small capacitors.  Also, cabling had to be re-routed.

It's not something you can guess from non-technical third party's statement.

I tend to agree with tkamiya, you've got a very fast pulse, several kV with a 50ns risetime, on any combination of L, N and Earth. I've recently had to test a device at 4kV and it still had to function flawlessly, not an easy task by any means. The product also had a small 15W flyback supply built in.  Secondary side regulated controllers worked flawlessly, not even a tiny glitch on the output whereas a few primary side regulated controllers either stopped working during the EFT pulse, went into a shutdown mode for 200 to 300ms and one let the magic smoke out.
Sanken STR4A100 series flyback controllers get my thumbs up.

As for the rest of the electronics you have to increase the test voltage until something fails functionally and then fix the problem with additional decoupling caps either on the supply or signal lines, then rinse and repeat. Good luck with your investigation.
 
The following users thanked this post: vintagemaniac

Offline honeybadgerTopic starter

  • Contributor
  • Posts: 42
  • Country: cz
Re: EMC testing, problems with MeanWell power supply
« Reply #11 on: March 16, 2019, 02:59:15 pm »
There is absolutely no point in having someone non-technical accompany the equipment to the test house, you should be there, no excuses!

My colleague is quiet eager so he will try it with a different PSU next week. If this fails I will be there in person next time - I am not able to clone myself.


Now assuming it is not a PSU fault. And he transients go through. What can I do? What should I prepare?

As I wrote, there are TVS diodes (TransZorb) already - are they fast enough for these surges?
I can add another ferite to the power cable.
Should I add MOV in parralel to the TVS diode?
I guess adding few 1uF MLCC to the Arduino and LCD wll not hurt (there is a few 100nF already)

Damn, I hope it is the PSU!



"My circuit does no longer work after the 2kV spike, so the PSU must no longer be supplying the appropriate power to it".  ???
...
 Your regulator for the 5V rail may provide additional filtering, so the 5V devices don't experience a problem; but the devices running off 12V do.

The only evidence I have is the video from my colleague. There is an NPN transistor switching a relay (12V) and a LED (with a resistor in paralel with the relay coil). After the 2kV spike the LED blinks but the relay does not click - this is an evidence for me - there must not be enough voltage for the relay to click.
The relay is powered directly from 12V rail - there is no VREG there.
There are no "smart" devices (which can lock on) on the 12V rail - only relays, NPNs, LEDs and passives.
 

Offline wolfp

  • Contributor
  • Posts: 35
  • Country: de
Re: EMC testing, problems with MeanWell power supply
« Reply #12 on: March 16, 2019, 03:10:21 pm »
What kind of pulses are you talking of?
Surge acc. EN 61000-4-5 or Burst acc. EN 61000-4-4 ?

The first one is a single shot 1.2/50 usek with rather high energy. To handle this you need transient-absorbers or MOVs. A surge sometimes destroys your components.
The second one is more comparable to a high frequency signal 5/50 nsec, it is radiated from the wires and if it succeeds to get into your circuit you can find it anywhere. Because of its high frequency and low energy it can be blocked by ferrites or small capacitors. A Burst normally disturbs but not destroys.

Wolfgang
 

Offline ebastler

  • Super Contributor
  • ***
  • Posts: 7375
  • Country: de
Re: EMC testing, problems with MeanWell power supply
« Reply #13 on: March 16, 2019, 03:10:49 pm »
There is an NPN transistor switching a relay (12V) and a LED (with a resistor in paralel with the relay coil). After the 2kV spike the LED blinks but the relay does not click - this is an evidence for me - there must not be enough voltage for the relay to click.
The relay is powered directly from 12V rail - there is no VREG there.
There are no "smart" devices (which can lock on) on the 12V rail - only relays, NPNs, LEDs and passives.

Thanks; you had not described it quite like that before. Having the relay and the LED in parallel, and only the LED operating, might indeed suggest some persisting problem with the voltage supply.

Meanwell provides quite decent datasheets and compliance testing information on this power supply:
https://www.meanwell-web.com/en-gb/ac-dc-wall-mount-adaptor-output-12vdc-at-1a-input-gs12e12--p1i

Their test report summary suggests that EFT testing has only been performed with a 1kV surge (which they claim to be the requirement for "light industry" environments):
https://www.meanwell-web.com/content/files/pdfs/productPdfs/MW/GS12E/GS12x24-rpt.pdf
« Last Edit: March 16, 2019, 03:12:29 pm by ebastler »
 

Offline honeybadgerTopic starter

  • Contributor
  • Posts: 42
  • Country: cz
Re: EMC testing, problems with MeanWell power supply
« Reply #14 on: March 16, 2019, 03:16:19 pm »
WOW, than you for these links for certifications! I was not able to find these and my local Mean Well supplier does not know anything.
 

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7678
  • Country: ca
Re: EMC testing, problems with MeanWell power supply
« Reply #15 on: March 16, 2019, 08:02:53 pm »
It's not the power supply, Mean Well GS12 is using TYN279PN SMPS IC which has auto-restart every 2.5 seconds after a perceived fault. I don't see this PSU going to lockout and needing a reboot, or doing anything strange after EFT.

Without some pictures of the product, it is hard to help with EMC troubleshooting. You can black out some parts of the photos, or use an EMC consultant.

For EMC testing, you have to be in the lab and not let techs or people who don't know the design, putting it through testing. Your company will soon learn why. Usually it is an idiot manager or arrogant engineer that thinks EMC tests are done this way, without the designer there. It costs a lot of money to keep repeating EMC tests and guessing what the problem is.
 

Offline Gyro

  • Super Contributor
  • ***
  • Posts: 10172
  • Country: gb
Re: EMC testing, problems with MeanWell power supply
« Reply #16 on: March 16, 2019, 08:18:37 pm »
My colleague is quiet eager so he will try it with a different PSU next week. If this fails I will be there in person next time - I am not able to clone myself.

I will not comment further (it's not really possible with the lack of solid information anyway), other to say that non technical 'eagerness' is no qualification / justification for the task. If you were there, you wouldn't be wasting a greater amount of time quizzing your colleague, talking about it, second guessing, sourcing other PSUs etc.  There is no need to clone, just prioritize. There are few activities for an engineer that have a larger payback per hour spent. I can't understand your plan, given the results so far, but good luck with it anyway. :)
« Last Edit: March 16, 2019, 08:50:57 pm by Gyro »
Best Regards, Chris
 

Online nctnico

  • Super Contributor
  • ***
  • Posts: 28429
  • Country: nl
    • NCT Developments
Re: EMC testing, problems with MeanWell power supply
« Reply #17 on: March 16, 2019, 09:01:34 pm »
I fully agree with Gyro. If you can't be there then there is no use to do another test. Time at test houses is expensive and I can't imagine anything else being a better use of money other than being there.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline honeybadgerTopic starter

  • Contributor
  • Posts: 42
  • Country: cz
Re: EMC testing, problems with MeanWell power supply
« Reply #18 on: March 16, 2019, 09:53:49 pm »
blueskull: I think the problem is someone decided we need to test it for 2kV as "Industrial" and the PSU is tested only for 1kV.

There is the power input side. It is as simple as it can be. V_SENSE goes through voltage divider -> filtration cap -> comparator -> Arduino. Arduino survives the blast so this should not be a problem.
LCD_PWR is only for the backlight LEDs.

D5 and D6 are TVS diodes.

« Last Edit: March 16, 2019, 09:57:47 pm by honeybadger »
 

Offline honeybadgerTopic starter

  • Contributor
  • Posts: 42
  • Country: cz
Re: EMC testing, problems with MeanWell power supply
« Reply #19 on: March 18, 2019, 10:29:53 am »
Good News Everyone! It was the power supply in combination with 2kV test which was not necessary.
 

Offline Kasper

  • Frequent Contributor
  • **
  • Posts: 791
  • Country: ca
Re: EMC testing, problems with MeanWell power supply
« Reply #20 on: March 18, 2019, 03:27:35 pm »
Thanks for the update. What was happening to the power supply after 2kV?
 

Offline honeybadgerTopic starter

  • Contributor
  • Posts: 42
  • Country: cz
Re: EMC testing, problems with MeanWell power supply
« Reply #21 on: March 19, 2019, 10:12:45 am »
This time there was audible clicking coming from the PSU while tested at 2kV. I am not sure if there is a spark gap inside or destroyed MOV which act like a spark gap now... I don't know. I will try to open this PSU in near future. Lab assistant was scared of that PSU might get damaged even more and then take our device with it - he stopped the test. To be honest I am not 100% sure about this but it all seems that with 2kV pulses the PSU somehow locks and its not able to recover with its watch dog timer. With 1kV there is no problem. Not even single PSU power outage.
 

Offline station240

  • Supporter
  • ****
  • Posts: 967
  • Country: au
Re: EMC testing, problems with MeanWell power supply
« Reply #22 on: March 19, 2019, 11:31:12 am »
Could be something as simple as a failed class Y cap, or bridge rectifier diode.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf