ESD problem on SMPS

I'm using an Aimtec AMEL20-515DMAZ in a product (http://www.aimtec.com/amel20-515dma...tion-encapsulated-dc-dc-isolated-power-supply)

Without going into too much unnecessary detail right now, some ESD events cause the supply to drop its 5V output for as much as 100ms. It always recovers, but that is bad for microcontroller circuits. I have simplified the circuit as much as possible to demonstrate it is the power supply and not the other circuitry per se:

The power supply has a 3 conductor AC cord directly soldered to its input terminals. The earth ground wire is left disconnected, as there is no earthconnection on the power supply. The other end of the cord is plugged intoa properly wired outlet.

The 5V output is connected to a 25 ohm resistor, with an LED and current limit resistor connected across the output to monitor the presence of power supply output. By the way, the low tech monitoring is on purpose to avoid introducing any other factors into the test.

I hit the free earth ground wire with an ESD gun and the power supply dropsout. If I use a completely separate earth ground wire, and place the end of it near the power supply and hit it with ESD, no problem. However, thisthinking doesn't translate when I try the device the power supply is used in. The device is in a metal case connected to earth ground, and this damnpower supply likes to drop out when the case is hit with ESD. Doesn't matter if the case ground is provided with a separate wire or not. I've triedconnecting the earth ground to the power supply output common, but that does no good neither (BTW Aimtec suggested that. They weren't much help otherwise. The PS is rated such that it's allowed to malfunction, but must recover). A new power supply and different cord was tested, along with other outlets. And this isn't a cheap PS either; right around $30 in singles.

I'm no EMC expert, but myself and a few colleagues have tried their hand atsolving this.

And yes, we have thought about putting a big cap at the output. The problem is it would need to be about 30,000 uF.

Any ideas, suggestions, criticisms appreciated.

 

Try to figure out if the input or output is the cause.
I would try wrapping the L&N input wires around a ferrite bead, then
testing again.
The same with the output, perhaps just a bead on the +5V.
Have you grounded the Common?

My thinking is the fast transients for the 6kv (or8kv) ESD test is being
conducted into the supply.
You can get the EMC lab reports from Aimtech, they should show how the
device was setup to be tested.
After all this has all the approvals even 610000-4-2.

Cheers

 

This looks like the ESD event capacitively couples over from the
otherwise not connected PE wire into the line and neutral wires in your
power cord. The same might happen with the metal enclosure, provided
there is enough capacitance between it and the switcher module plus
power cable.

Without knowing the innards it's anyone's guess what happens inside. A
possible scenario: The ESD spike goes into the direction where the
switcher in there "thinks" there was a dip in the input voltage and then
goes through an undervoltage lockout scenario, followed by a re-start.
Other than TVS'es and/or filtering on the outside there probably isn't
much for you to do about it, other than finding a better switcher module.

You could try a common mode choke at the input, right at the switcher
module.

 

No luck yet. I tried a big clip-on ferrite around the L and N wires but didn't see much change. Will try more tomorrow. Also found a similar unit from CUI that is cheaper and might not have the problem.

 

No surprise there. A clip-on ferrite is for EMI, it does next to nothing
against ESD. You could try a CM choke but only if it has several
millihenries. Has to sit right next to the switcher module.

 

Or even a Cx EMI line cap may help. A clip on bead wouldn’t do much,
but 2 or 3 turns thru a bead may do more.

Cheers

 

Or even a Cx EMI line cap may help. A clip on bead wouldn’t do much,
but 2 or 3 turns thru a bead may do more.

Cheers

 

No, because AFAIK there is no metal. And the OP wrote, quote "I've tried
connecting the earth ground to the power supply output common, but that
does no good neither (BTW Aimtec suggested that". This is not much of a
surprise because the hiccups usually happen on the primary side. Either
the IC goes through an UVLO sequence in the unlikely event that there is
an IC in there at all, or the more likely transistor oscillator trips
over itself.

 

Which is not connected to anything at the switcher, meaning it can only
be coupled into line&neutral.

As I said, I'd first try a common mode choke. TVS across the input may
be another option but before I'd go that far I'd look for a better
manufacturer.

 

Ok, maybe a solution. I tried ferrite beads on the input, but no luck. However I found a very similar supply made by CUI; same footprint, but only 12V out, not 5V+12V. However this isn't a big a problem to solve. It's cheaper too, and seems to be stable under the given conditions.

 

There is actually an IC in there. I took one apart because it was damaged by another test. It's a TOP256.

 

I like their LNK-series deivce but not too fond of the TOP-series.
Anyhow, those TOP256 have a line surge shutdown protection. So if there
is an ESD surge that protection feature could be tripped and then it
goes through the re-start sequence. There is also an output OVP
protection and the effect is very similar.

 

maybe there's a current sense resistor in the ground line?

 

The topswitch feedback pin needs local decoupling with a ceramic part,
not usually included in the application literature, to prevent UVLO
events that are associated with transients on the switch drain -
whether self-generated or otherwise.

If this local decoupling is not present, that's a long-term fix,
probably outside your development budget or time schedule.

RL