Stressed out over capacitor failure - EDN

2022-07-15 21:52:26 By : Ms. Sally Dong

While working at a company that makes automotive embedded control systems, my colleagues and I set out to create a new PCB layout for one of our products. The product worked fine, but we needed to simplify the PCB assembly and increase production speed.

We totally changed the PCB layout and performed validation tests on the new PCB. All the results were satisfactory and encouraging, so we sent the PCB into volume production.

A few days after the unit had begun being installed in vehicles, we received a complaint from the production line that none of the functions of one unit were working. We studied the unit in our laboratory and found that one multilayer ceramic (MLC) capacitor in the circuit was burned. We soon started receiving more complaints about the same failure. It was easy to imagine a resistor burning in this fashion, but it was surprising to find such a failure in a capacitor.

We first suspected a problem with the manufacturing quality of the capacitor. We consulted with the capacitor manufacturer, who had been checking only a sample of his products, and he agreed to check each and every capacitor at the time of its manufacture. But even after he took this precaution, the rate of capacitor failures remained the same.

Do you have a memorable experience solving an engineering problem at work or in your spare time? Tell us your Tale.

We further analyzed the issue by studying the new layout. The MLC capacitor was being used for ESD protection at one of the digital inputs, which was how it had been used in the original layout. We checked whether the capacitor had the proper voltage rating, and it did. We also observed that several capacitors with the same specifications were used on different inputs of the same board for the same purpose (ESD protection). But only the capacitor at this particular location was getting damaged. Finally, we switched to using the previous PCB and not a single failure occurred.

Further study of the situation revealed that when we redesigned the PCB layout, we had shifted the placement of the capacitor. It was now placed near one of the surface-mount holes of the PCB, and all the failures were found after the PCB was mounted in the housing.

We read several application notes about MLC capacitors and learned that because of their brittle nature, multilayer ceramic capacitors are more susceptible to excesses of mechanical stress than other components used in surface mounting. In our case, when we tighten the mounting screws, the PCB bends slightly. Excessive bending of the board can create mechanical cracks within the ceramic capacitor. Over time, moisture penetrates the crack and can cause a reduction in insulation resistance that is accelerated by humidity and temperature, and this generates a conductive path. As a result, it gets shorted and because of high current flow through the capacitor, the capacitor gets burned. Although these mechanical cracks may not lead to capacitor failure during the final assembly test, failures may occur once the product is in the field, when they are expensive and time consuming to correct.

For confirmation of our theory, we removed one capacitor from a board that was working satisfactorily but that we suspected had undergone stress during the mounting process. We sent the capacitor for cross-sectional analysis and discovered it had a crack that could lead to damage in the field. Now, we have modified the PCB layout again to shift the placement of that capacitor away from the mounting hole. After this modification, all the critical tests related to the mechanical stress were successfully carried out and no such failure was found in the product.

Hardware designers for automotive embedded control systems pay a great deal of attention to electrical overstress, but they also need to take into account mechanical overstress on small components like resistors and capacitors. Although many international standards provide guidelines for protecting components against overstress, similar standards for mechanical overstress are conspicuous by their absence.

Shilpa P. Jadhav works on embedded hardware design at Tata Motors in Pune, India.

Hi It is nice to know that it was not just me that had ‘bad’ capacitors. We have had a few failures on a 100uF X5R ceramic capacitor. The end to end resistance which is noramlly better than 4M Ohms, fell to a few hundred Ohms. The resistance was none linear, almost like an intermittent short. A search of the internet seemed to suggest mechanical impact could cause the failure. Our capacitor lives on a PCB inside a small metal box. A brand new unit came back because it ‘just did not work’. Which was true. Also the box would not sit flat on the desk, it was bent. And the BNC connector on the front panel was folded over. It had received a massive impact. The capacitor was conducting. This was repaced and the unit worked. I examined the dead capacitor with a small microscope, but it looked good! Nothing to see at all. Tim Orr

Very nice to see such a detailed description of the cause and effect of a less-than-typical failure. I was expecting to see another account of counterfeit electrolytics the likes of which caused all 38 of our large format monitors to fail under warranty, with no admission by the manufacturer there was a problem. And it is also good to see credit given to this part’s supplier/manufacturer for doing their part in diagnosing, or at least eliminating, one possible cause. Like Mr. Jadhav, I am not satisfied with a problem resolution until there are not only no additional failures, but also a thorough understanding of the reason the failure occurred in the first place. Many employers do not support full Root Cause Failure Analysis, but in my opinion, it is the most cost effective response to a problem.

Interesting to me that you were using a capacitor for ESD protection. I’ve not seen that before; usually a transient voltage suppressor (TVS) or multiple diodes is/are used. We have in some cases used a TVS to ground with a ferrite bead in series to our chip’s input pin.

The detailed failure analysis here was very interesting, although I suspect moisture alone does not cause enough fault current to overheat the capacitor. It may be that in the presence of moisture and voltage in the cracks, the metalization migrates into those cracks and essentially plates a short circuit across them Early in my career when I worked for a well known voltmeter manufacturer in the Everett Washington area, we had the same problem with the first generation MLC capacitors. They got the nickname “blue bombs” (because of their color) and were pretty much banished from our designs. I’m not sure anyone ever figured out exactly why they would occasionally, and apparently spontaneously and randomly, explode or catch fire.

Another trick to help cure this is to put a narrow slit through the PCB next to the capacitor to relieve the flex. Some ceramics are also slightly piezoelectric so it could help with that issue as well.

The slit PCB trick is slick. The MLCC “gotcha” that bit me was using an exceedingly large one, surface mount 2512 size. The units worked fine if assembled at an outside vendor, but if assembled or reworked in-house by our techs, the 2KV MLCCs had about a 10% failure rate in a month. Turns out differential heating by soldering irons on the two ends, even if two irons were used at the same time, was enough to cause thermal stress cracks in the ceramic. And this was with flexible end caps that should have been OK. The manufacturer (once we read every footnote in the spec sheet – a process I recommend to all) even mentions this risk. Sigh. No reworking the large size SMC MLCC. We ended up going to axial lead caps for this application (soldered to the SMC pads), and are considering using a lot more parallel 1206s for the next layout.

Stress cracking is a common failure mode of MLCC’s. (Which I also found out the hard way!!!) While a visual inspection of the device usually won’t indicate anything is wrong, if you go after it with an x-ray inspection system, the internal cracks will immediatley become apparent. “Back in the day” these were uncommon machines, but with the advent of BGA’s, any board house worth their salt is gonna have one of these machines on-hand.

“Further study of the situation revealed that when we redesigned the PCB layout, we had shifted the placement of the capacitor. It was now placed near one of the surface-mount holes of the PCB, and all the failures were found after the PCB was mounted in the housing.” A surface-mount hole? I seem to recall Wile E. Coyote designing some schemes using some sourced from the ACME Corporation, but I thought they were long out of production, due his notable lack of success.

Ceramic capacitor manufacturers have been thinning the dielectric the past 15 years. A good graph for this is Figure 4 in “How Hot is Too Hot”, a link to which can be found below. http://dfrsolutions.blogspot.com/2012/03/how-hot-is-too-hot.html

TVS is used on the ‘power supply line’ for the protection of circuit from high voltage transients. For ‘signal input pins’ we can use the capacitor for the protection against transients while ferrite bead in series of chip’s supply/input pin helps to supress the high frequency noise, which further helps to improve the EMC of the electronic circuit.

Years ago a ceramic capacitor expert told me that you can accelerate the failure of a cracked capacitor by applying vinager. He personally liked red wine vinager. Appearantly, if ther is a crack, it will get in and cause an immdiate short thus revealing that the crack was there. I have never tried this myself. I suspect also that, once cracked and exposed to moisture, dendrites might form. More about dendrites in this article: http://mathews-engineering.com/ArticlesByTomMathews/CleanPCBs_2.pdf

From the “Topper”: ( A Dilbert reference for those unfamiliar) A 2512 is an exceedingly large capacitor? Why heck, that’s nothing compared to my design using 7565 capacitors from NovaCap. I hand soldered those puppies using an iron and solder not unlike laying a bead using a MIG welder. Never cracked a one. Then again, the board thickness was only .020 so the capacitors provided the bulk of the stiffness at that point.

Several MLCC manufacturers offer flexible (soft) termination options to reduce the likelihood of failures due to stress induced cracking (AVX FLEXITERM, Kemet, etc.) that I use on airborne sensor hardware designs.

Re Surface mount hole: I’ve made some of them , change a through hole pad to “top layer” on your pcb package and forget to set hole size to zero. It passes the DRC , and you can’t see the hole on screen, but when manufactured you have a plated through short circuit!!

I love stories like this…Where they find the problem was due to a mechanical vice electrical problem.

“Since my last post, I've fitted ~ 150000 10u35v X7R 1206 MLCC's on various PCB's, and getting failure rates around 1 per 3000 parts, when operated around 20-24v with a 24V transorb across the rail (and in various other places with 12v-15v), though it’s not possible to eliminate other faults causing overvoltage and also mechanical action. One of the PCB designs has a 12v-24v boost converter in a corner and the 10u35v caps and a 30vMOSFET have sporadic failures in this area, in this case the boost PWM is generated by a microcontroller, and we have modified the MCU code, installed PTC’s on the MOSFET, and modified the gate drive with a series capacitor, and a current sensing shutdown, so far two batches are OK, but will take a while to get statistics.”

“Hi Shilpa,nnThanks for your post. We were having caps shorting out on our 5V rail to ground, but they were fine when first manufactured/tested, and subsequently failed. Turns out these 0603 MLCC caps were very close to the edge of the board and near a a screw mounting hole which was causing board stress to flex across the cap. We made this conclusion based on your posting and some logical analysis. Our PCBA mfg is x-raying them to see if they can view the cracks but we are certain of it. When we measured the resistance in the shorted caps one was 0.0Ohm, another 30Ohm and another 700Ohm. After desoldering two were completely shorted to 0.0Ohm and the one that was 700 Ohm became open. This indicates the dielectric was unstable I suppose due to cracking. Thanks again.”

This is what happens when the industry insists on leadless components and brittle ROHS solder. Like tin whiskers, these problems were widely known by the industry, but the industry just made these problems “magically” (not) go away. We are going to see this more as @#$ QFN parts proliferate. Ceramic capacitors cause many other problems as well (such as “how my 1uF cap became a 0.1uF cap” because of voltage-sensitive characteristics of high K dielectrics, which are not spec’d as part of the ceramic K spec). Best to avoid them altogether and use polymer film caps, unless you are designing microwave T lines where they are small, both in size and value.

You must Sign in or Register to post a comment.