MLCC capacitor failure

[tonal]

Hi all!
Hope I'm writing in the right place.

I'll get straight to the point.

In a product we have an issue with a component (a MLCC cap) in the SMPS part of the PCBA.

Although the PCBA and the component itself are tested (ICT and FVT tests) by our supplier and also in the company I work we perform functional tests in the finished product we get this kind of failure in a sampling process where the product is tested under real life conditions and we had also a couple of incidents in the field.

The cap is placed near (1mm) and parallel to the edge of the PCB. There are no depanelization points very close to the cap (the closest point were the PCBA is attached to the PCB frame is around 6cm away).
The PCB is around 150x150 and 0.8mm thick.
The depanelization of the PCBA takes place in the supplier after reflow and before through-hole parts assembly.

I was wondering whether this failure is due to a faulty part or due to bad handling/stress during depanelization.

One thought that I had was that the capacitor is mechanically stressed, cracks and after some time during operation (after a couple of hours) it fails (a resistance of a couple hundred Ohms is measured between the two terminals). There is no electrical stress applied to the cap.

Can you think of any other reasons that a ceramic multilayer cap shorts in the first hours of its operation?

 

[marce]

Nope, these are generally stress related fractures and the cap position is perfect for this sort of failure, it should be at 90 degrees to the board edge to help alleviate the problem, or look for flexi term type caps.

 

[tonal]

Nope, these are generally stress related fractures and the cap position is perfect for this sort of failure, it should be at 90 degrees to the board edge to help alleviate the problem, or look for flexi term type caps.

Thanks for your reply.

If a micro-fracture due to mechanical stress is the root cause of the MLCC capacitor's failure is it possible that the component fails after a few days or even longer or it appears shortly after normal operation?

Also are there any tests that can be performed in order to check that we have such an issue on component level apart from visual inspection with strong magnification?

Are there any characteristics of the capacitor that alter due to mechanical stress i.e. ESR?

BR,
Lampros

 

[D.A.(Tony)Stewart]

Functional vibration test is the best to find intermittent solder joints and fractured dielectrics.

Some sort of hand-held vibrator or board level shaker is part of what we call HALT testing.

over-Stress on depanelization is easy to overlook and may require more operator training or improved fixture or board stiffener.

Ceramic does not bend.

Failure sensitivity depends on orientation to de-panel warp. Stress is greatest along any edge, so it seems like bad design, bad process or bad part until further Root Cause. is known.

 

[andre_luis]

Can you think of any other reasons that a ceramic multilayer cap shorts in the first hours of its operation?

I would expect the opposite in the case of mechanic stress, such as opening the contact at terminals. If the failure is really short-circuit, don´t you think it could be due to some electric reason as for instance, underrated ?

 

[KlausST]

Hi,

I think most likely it's mechanical stress.
Although of course electrical stress could also be the cause.

Another problem is mechanical stress caused by thermal stress. This could be a fixture near the capacitor, sometimes used for (wave) soldering. Maybe a problem when the PCBs are stored during production. Or any mechanical stress during AOI, or during programming of firmware or PLDs. Check all the stations the pcb travels ... from soldering to packaging.

Hope you find the cause.

Klaus

 

[tonal]

Thank you all for your answers.

Probably someone has to visit the supplier factory to have a better look at the process (it's a looong trip and no it isn't China).

The defects till now are limited to one variant and only within two weeks of production (other variants with the same hardware design didn't have a similar failure).
Either the batch of capacitors used for these PCBAs had a quality issue or something changed in their process during a short period that caused more-than-normal mechanical stress.
If this situation is restricted to a short period and if the supplier can provide traceability regarding the component and the SNs of the PCBAs it would be very comforting.
I am waiting an answer from the supplier.

The most alarming case would be that the failure could appear after a while in the field in God only knows what percentage...

Unfortunately, the design is owned by another company site and I don't think it will change...

 

[marce]

Change the layout and move the capacitor. I did a study of stress involved with board de-panelisation a few years ago, due mainly to this problem with capacitors. The methods of de-panelisation and subsequent handling had a big effect on cap failure rates. The first link covers all the above mentioned problems, mechanical and electrical stress factors.
Start here...

**broken link removed**
**broken link removed**

 

[tonal]

Thank you macre for your help but the design is made by another department located in a different company site and it is usually almost impossible to impose a design change even though it is perfectly understandable.

 

[D.A.(Tony)Stewart]

Perhaps a patch of instant adhesive to thicker FR4 strip on underside may stiffen PCB enough to protect MLCC cap from stress during depanelization.

Ask for video of process up close.
Demonstrate the problem with edge clamp and bend to failure.
Field problems are serious and demand immediate attention.

There is a solution, just find it fast.

 

[tonal]

I am returning to this after a while, since I had the chance to examine more failed boards.

It seems that the root cause is not mechanical stress on components or anything similar.
After measuring the capacitor on the PCBA (measured a value ~3KOhm) I desoldered it, measured it again and this time no stray resistance was apparent. I also measured between the PCBA pads but the measurement was too high. I cleaned the board and soldered again the same capacitor and the power supply was again functional!

It all points to some electrochemical process that causes surface resistance to decrease so that the PSU is not functional (a compensation net is failing due to the parasitic resistance and the power supply does not start up).

I suspect formation of a conductive path formed only under voltage and after exposure to humidity (tin whiskers, dendrites, electrolysis of salts/residues??? during production process of the PCBA). Samples will be sent for chemical analysis to see if there are any residues that could cause this failure.

My question is how fast growth of dentrites/wiskers can happen? Could this happen just in a couple of hours?

 

[andre_luis]

I'm prone to believe that the problem should not be happening only due to the dentrites/wiskers themselves. but particularly due to long term exposition to high temperatures, such as would be expected at power supplies. I experienced a similar issue on IP cameras confined within small domes, what subjected them to average temperatures near to 50/60celcius degrees. This thesis can be confirmed in several referencies, like this one right below:

Soldered jointings are indispensable to high-density surface mounting of electronic equipment. Maintaining the reliability of soldered jointings is going to become a major problem that must be overcome to accurately mount miniaturized electronic parts. These soldered jointings deteriorate with exposure to long-term heat and mechanical stress, causing cracking which leads to part failure. In this paper we shall report on reliability testing to confirm the mechanism of solder cracking due to heat stress.

 

[ads-ee]

Growth rates according to research and information compiled by NASA...

Incubation (Dormancy) Period: Experimenters report the incubation period may range from days to years. This attribute of whisker growth is particularly concerning because meaningful experiments to determine the propensity for a particular process to form whiskers may need to span very long periods of time.

Growth Rate: Growth rates from 0.03 to 9 mm/yr have been reported. Growth is highly variable and is likely to be determined by a complex relationship of factors including plating chemistry, plating thickness, substrate materials, grain structure and environmental storage conditions

It's more likely that you are experiencing dendrite growth due to some process change that is leaving some residue that allows the metal migration. Is the process using a rinse instead of using a no-clean flux?

 

[FlapJack]

It is my understanding that component manufactures will analyze failures for you. Maybe someone here knows how to start the process.

 

[D.A.(Tony)Stewart]

Your failure is most certainly from bad design.

I've seen this many times in contract mfg. as Eng Mgr.

Fix the design and follow IPC stds.

With process improvements, training, or stiffener on board edge, you might be able to reduce the risk of failure until the design is fixed.

Also put in place random vibration or just a hand vibrator into your final functional test plan immediately until the design or process problem is fixed. Your yields will indicate this.

microcracked ceramics have a way of failing right after you look the other way ....or just before the warranty period expires.

 

[tonal]

Thank you all for your input.
Today we examined a sample in a lab equipped with a SEM. We couldn't notice any suspicious formation of whiskers or dendrites. We also did elemental analysis and the only thing that we found was Cl (chlorine) while observing one pad (concentration of 1.44%) and Ba (Barium) which afterwards was proved that belongs to the capacitor ceramic body (an oxide of Barium I suppose).
The sample we examined had this capacitor mechanically removed so that we could observe the area underneath the capacitor. On the surface of the other pad of the cap we found a great amount of carbon.

But this is getting even more perplexing. I was doing some tests with an affected sample. I measured the resistance across the cap and it was in the MΩ region so I thought that something else was wrong in the circuit. I powered the PCBA and it of course it wasn't functional. After powering down the device I measured the resistance across the capacitor and now I got a reading of 1K!!!
I measured it again after a minute but this time was around 6MOhm!!!. I powered again the board and checked with the thermal camera if there was any temperature rise but I couldn't detect a temperature higher than 25 Celcius degrees and the temperature would not exceed 35 degreec in the vicinity of the cap...

 

[FlapJack]

Were in the world could the carbon come from? Are you using quality name brand capacitors?

 

[BradtheRad]

The sample we examined had this capacitor mechanically removed so that we could observe the area underneath the capacitor. On the surface of the other pad of the cap we found a great amount of carbon.

This is typical of smoke deposits.

 

[tonal]

Carbon could be any organic substance I suppose.
It could even be residues from the ESD safe carbon coated carton boxes in which the supplier sends the boards.

Chlorine is more suspicious since it has corrosive action.

 

[marce]

! I wouldn't want carbon deposits on my bare boards... Not a good idea with SMD components!