Interesting failure mode after potting

[mikeselectricstuff]

1 metre x 25mm PCB, 0.8mm thick, containing LEDs, driver chips and some passives, all SMD. Tested and known working.

It gets placed component-side down into a cavity in an aluminium mould, which is pre- filled with clear epoxy, and epoxy is brushed over the rear to cover the non-component side, to form a fully encapsulated strip, about 4mm thick. There are some captive wires emerging from one end. 

After the epoxy has cured (24hrs at 25 deg.C), the strip is removed and re-tested.
On a few PCBs, one driver chip has died. Always the one nearest one end, but not the same end.
Failure mode is that a pin has low resistance to ground.
Depending on the end, this pin is either tracked to a pad very close to the end of the PCB, or connected via a 5k6 resistor to one of the flying leads.

I think I know what happenned - would be interested to see if anyone comes to the same conclusion before I explain my thoughts.

I do not believe it to be a mechanical issue.

[JuKu]

Without knowing the specifics, my guess is ESD.

[mikeselectricstuff]

Without knowing the specifics, my guess is ESD.

More specifically....? Something about the process....

[Nermash]

Without knowing the specifics, my guess is ESD.

More specifically....? Something about the process....

Brushing the epoxy over ends?

[mikeselectricstuff]

Without knowing the specifics, my guess is ESD.

More specifically....? Something about the process....

Brushing the epoxy over ends?

May be a contributory factor, but I think possibly more fundamantal than that.

[nihilism]

flux residue contamination

[Len]

More specifically....? Something about the process....

Static charge builds up, maybe because of the brushing, and when you remove the strip from the mould (picking up by one end or the other) it sparks to your finger.

[jucole]

Too much exothermic heat during the curing process?

[JuKu]

Without knowing the specifics, my guess is ESD.

More specifically....? Something about the process....

About your ESD precautions. You might have done all that should be done and handle the potting "ESD-aware" manner, in which case guessing ESD is almost an insult. If not, the method sounds like a recipe for something like in your avatar picture in smaller scale...   

[jucole]

ES charge in the brush you used?

[nctnico]

Epoxy shrinks. It's a bad potting material especially for such a large PCB. Nowadays mostly PU based potting compounds are used.

[Andy Watson]

So, you're making a one metre long parallel plate capacitor with epoxy resin dielectric!  When you separate the plates, any voltage present (due to small amounts of charge being generated somewhere in the curing process) is magnified many times to maintain the C proportional to 1/d relationship, as in a Wimshurst machine. I guess the situation is exacerbated by levering and removing one end first.

[mikeselectricstuff]

OK here's my theory.
The PCB and the mould form plates of a capacitor, the epoxy is the dielectric.
As it's liquid, the epoxy is contacting the mould extremely intimately, at the molecular level. There is also a mould-release compound, which may have some effect - I'm not a physicist or chemist but I imagine that a mould-release agent is probably using an electrostatic effect at the molecule level to repel the thing you don't want to stick.
The act of removing the now-solid piece of plastic from the mould induces a charge as electrons get ripped off at the boundary, so the PCB, which is now isolated inside the plastic, is sitting at a potentially high voltage, and due to its size, has enough capacitance to store significant energy.

Most of the time this would get discharged whne the groundplane edge, or ground wire during test gets grounded by contact,. but if the pad near the end of the PCB, which has minimal epoxy on it, or the non-ground wire touches something first, this discharges the stored charge through the IC pin.
The fact that it manages to do this via a 5K6 resistor suggests significant stored energy.

No rubbing is needed - I believe the act of peeling out of the mould is enough to create the charge.

Does this sound plausible? Any physicists here?

[Len]

OK here's my theory.
The PCB and the mould form plates of a capacitor, the epoxy is the dielectric.

You've discovered the Leyden jar!
[Edit] Or the Loughton jar?

[wiss]

I used to have a DIY photo-lab where I bought film in 30 m rolls and rolled it up myself. (If) When ripping the film of there were a visible spark. Maybe just the ripping apart operation is enough to displace enough charge... (Thinking about ripping the skin of a cat, there will be things happening)

[krivx]

Does this sound plausible? Any physicists here?

Do you have the means to take any measurements while this is happening? Is the charge something you feasibly see on an electrometer? If you think significant energy is dissipated then it would follow that it is significant enough to measure...

What's the epoxy you are using? I have found that filled epoxies (usually marketed for low thermal expansion) are more conductive, you may be able to swap out the epoxy for something that does not allow charge to build up but also doesn't short out your board.

Not a physicist, but an EE working in Physics...

e: I would guess that your release agent forms a hydrophobic layer on the mould, I think this would just act as another dielectric layer. Hard to say without knowing what's used.

[Kremmen]

I used to have a DIY photo-lab where I bought film in 30 m rolls and rolled it up myself. (If) When ripping the film of there were a visible spark. Maybe just the ripping apart operation is enough to displace enough charge... (Thinking about ripping the skin of a cat, there will be things happening)

That particular phenomenon is not a spark, instead it is triboluminescence, i.e. light induced by breakage of chemical bonds. Ordinary Scotch tape diplays this behavior. I have also processed my own films and this was an often seen thing.

[krivx]

Ordinary Scotch tape diplays this behavior.

This also: http://www.nytimes.com/2008/10/28/science/28xray.html

[mikeselectricstuff]

Do you have the means to take any measurements while this is happening? Is the charge something you feasibly see on an electrometer? If you think significant energy is dissipated then it would follow that it is significant enough to measure...

Unfortunately it's at the subcontractor's place. It would be easy enough to verify - pull it from the mould and touch the ground of the flying lead to the mould - I suspect you'd get an audible spark.
However no spark wouldn't be conclusive as it could be that the occasional nature of damage is down to either unfortunate discharge or only occasional charge - it would probably need at least some charge to start with.
Obcioulsy I've now told them to connect the ground lead to the mould when releasing.
 

[G7PSK]

Epoxy cures by an ionic exchange so there must be voltages set up within the epoxy as it cures.

[calexanian]

In my experience with potting products you are describing stress cracking. You may want to look into a softer encapsulant. there are specifically designed epoxy and polyurethane blends for electronics potting that give a little to allow the solder joints to remain intact. This situation is compounded if you have gold plated pads as well. We use a product from Aero Marine. Dow and loctite (Henkel) have a wide range of products that work quite well, some are even what I would describe as rubbery in texture for strain relief of your exiting wires. They also have a humiseal property for removing residual moisture from the components. In 25 years of using epoxy potting in sensitive electronics I am unaware of ever loosing a device to ESD from the potting process.

[ConKbot]

Ahh, the separation of an intimate bond between part and mold. Definitely a good chance for static to build up.  I remember getting 1-2" sparks off a table on hard rubber casters when I was pulling the vacuum bag off of it. The bag had been holding down a part to a form for a few hours in a 300*F oven.  take up the corners of each vacuum bag, then pull off one, then a second, carefully not getting too close to the table, and it had a nice charge built up on it.   And letting a coworker be the one to take the shock.   


 

[Dago]

Epoxy shrinks. It's a bad potting material especially for such a large PCB. Nowadays mostly PU based potting compounds are used.

I'm pretty sure one of the biggest advantages of epoxy is the fact that it does not shrink. This is the reason why epoxy is used for example high precision NC machine polymer concrete bases, the base does not change shape when it cures. There might be some differences between different epoxies though so I cannot be 100% sure this applies to all epoxies.

I think polyester based resin does shrink though.

[uwezi]

My betting would also be mechanical stress rather than an electrostatic cause.

We used to do some experiments, trying to encapsulate solar cells in clear epoxy between two sheets of 2mm thick glass. The glass would crack from the while the epoxy is hardening!  This is probably not "shrinkage" even though epoxy does shrink while curing, but rather thermal expansion and contraction during the curing: the epoxy solidifies while it is hottest and then cools down again - shrinking.

http://gertrude-old.case.edu/276/materials/polymerencyclopedia/epoxy%20resins%20(overview).pdf

http://www.dtic.mil/dtic/tr/fulltext/u2/753943.pdf

http://www.expresspolymlett.com/letolt.php?file=EPL-0000181&mi=c

If it was an electrostatic problem, then I would not expect the failure to be located at the ends, but rather in the middle - and grounding the leads while curing the epoxy should solve the problem.

I also exclude the heat generated during the curing - that should also not concentrate on the ends.

But: when one short side attaches slightly better to the mould, then the stress from expansion/shrinkage would be largest at the other short end. It would not take much stress to rip off a narrow pitch SMD component from its pads. Have a look through the epoxy with a magnifying lens.

[IanJ]

Back in the 80's/90's I used to do a lot of potting......can't remember the brand but it was 2-part black stuff that set hard with a glass like look to it.

3 pcb's stacked 7mm apart with soldered links between them. Mix of 74 & 4000 cmos.

It used to heat up quit considerably whilst curing and from time to time we end up with a faulty unit. We always put it down to mechanical stress. The faults varied.

Ian.

[uwezi]

Epoxy resins have dimensional stability, a high degree och chemical resistance, and good mechanical properties, so that they have gained wide acceptance in a variety of applications. However, various problems can arise from the volume shrinkage that often occurs during the polymerization and curing processes. ... This shrinkage creates internal stress in the resin, and may cause failure by cracking or delamination.

(Leo-Wang Chen et al., Journal of Polymer Research, Vol. 4, No. 2, pages 65-72, 1997)

[mikeselectricstuff]

It is a soft compound - feels comparable to thick polythene tubing. You can cut through it with a scalpel (though applying a soldering iron to the blade speeds things up - it hardens and crumbles to granules at higher temps.

I've now reworked 17 failed strips and am absolutely certain that ESD is the only plausible explanation.
All failed strips had either one chip , or a pair of chips which shared a single connection fail, and the failures were all low-resistances to ground (5 to 100R) on a single pin.

In every case, the dead pin was either connected to a trailing wire, on the end of the PCB where there is a pad with minimal coating, or adjacent to a large bubble hole at the edge of the strip - all prime candiates for the easiest discharge path to the sharp edge of the metal mould about 2mm away

My best guess is that the strip is peeled out of the mould starting at one end, which builds up a charge, and discharge occurs towards the other end  of the PCB, which is still in the mould.

The failures at the locations of the edge bubbles is the "smoking gun" as this is  the easiest discharge path.
 

[mikeselectricstuff]

In my experience with potting products you are describing stress cracking. You may want to look into a softer encapsulant. there are specifically designed epoxy and polyurethane blends for electronics potting that give a little to allow the solder joints to remain intact. This situation is compounded if you have gold plated pads as well. We use a product from Aero Marine. Dow and loctite (Henkel) have a wide range of products that work quite well, some are even what I would describe as rubbery in texture for strain relief of your exiting wires. They also have a humiseal property for removing residual moisture from the components. In 25 years of using epoxy potting in sensitive electronics I am unaware of ever loosing a device to ESD from the potting process.

Did you ever do 1 metre long, fairly flexible strips (0.8mm PCB) though? The "long peel" to extract would be different than most typical operations. 

[mos6502]

I agree that the peeling probably creates an electrostatic charge. Like when you peel a long piece of tape off of a roll (no, I'm not talking about the triboluminescence effect mentioned earlier) you'll notice it becomes electrostatically charged and gets attracted to the thing you least need the tape to stick to.

In fact, I think this is a universal effect that occurs when large sheets of nonconductive materials are separated. Any small charge imbalance at the beginning gets amplified the more you separate the sheets.

[SeanB]

You considered adding moisture to the area during depotting to reduce charge build up? Easy enough to do so with just a swamp box air cooler to add the moisture. Likely it is a static build up, and a humid atmosphere, or simply spraying the moulds before removal with a mist spray of water and working damp will help.

[mikeselectricstuff]

You considered adding moisture to the area during depotting to reduce charge build up? Easy enough to do so with just a swamp box air cooler to add the moisture. Likely it is a static build up, and a humid atmosphere, or simply spraying the moulds before removal with a mist spray of water and working damp will help.

..or just connecting a ground wire when removing....

high humidity when curing is probably a bad idea - not sure about Epoxy but PU really doesn't like water.
And as the charge is created by seperating an air-free joint, on something surrounded by plasic, atmosphere may not make much difference

[SeanB]

Water when cured and out of oven cold. Resin and water during curing means a mess.