12
\$\begingroup\$

I have produced 1000 units of our new developed energy meter. Almost 20 aluminum electrolytic capacitors (2%) have failed short when we test them during the production processes. The Capacitor is a 470uF 35V (UPW1V471MPD) made by Nichicon.

The failure itself varies between total short circuit to small variable resistance. One of the caps spit out its electrolyte.

I have 3 other different aluminum electrolytic caps from the same manufacturer that didn’t fail. A year ago, I have produced 300 units (same design and part no.) and they running perfectly and didn’t find any failed capacitors.

My question has 3 parts:

  1. How this occurred?

    Is it normal to have an infant mortality of 2% with short circuit failure for aluminum electrolytic caps?

    Or, may be the cap is counterfeited and the con manufacturer didn’t do enough quality testing. (I bought them through a PCB and components supplier in Shenzhen). If they are counterfeit, how to verify that?

  2. What should I do with the rest of the batch?

    If I release the batch, would I guarantee that there will be no more short circuit failure? If it didn’t die during the manufacturing, it will not die in the near future! (U shape/bathtub graph).

    Or it is better to run my meters for a period of time (24 hours) if didn’t fail it will not fail in the future.

    Or I should go the hard way and change them all?

  3. How to avoid this in the future, because I will be going for large quantities > 100K.

    I didn’t find very useful notes about electronics parts quality and reliability other than MIL-HDBK-978B. But it is out dated and didn’t have the electrolytic caps.

Added

Operation condition:

  1. Voltage = 20V.

  2. Inrush and ripple current = the capacitor is used to store energy to supply a latch/rely only when operate. it operated about 5 times during production. latch/relay impedance 60 ohm. it draw its voltage from a capacitive power supply capable of supplying a 17mA at maximum.

  3. Temperature: about 30C.

  4. there is a guarantee that no cap have been reversed biased.

\$\endgroup\$
  • 8
    \$\begingroup\$ Purchase from a reputable supplier and contact the supposed original manufacturer to see if they have had problems with counterfeiting. They'll probably deny anything of course but the future course of actions is clear..... buy from approved dealers you can trust and, if necessary pat a little bit more to have the supplied goods traceable to source. \$\endgroup\$ – Andy aka Dec 9 '19 at 13:00
  • 6
    \$\begingroup\$ No, I don't think @Andyaka is telling you that the cap probably is counterfeit. Instead, they are suggesting how you could avoid getting counterfeit parts in the future. You may still have a design problem that you haven't uncovered, which leads to marginal reliability in the finished product. It concerns me that you haven't said anything about the operating conditions of the capacitor. \$\endgroup\$ – Elliot Alderson Dec 9 '19 at 13:17
  • 2
    \$\begingroup\$ @ScottSeidman The question is not about the design. As I stated in the question. The design already worked. The question is about a certain part and its failure mode. \$\endgroup\$ – Ashraf Almubarak Dec 9 '19 at 13:44
  • 8
    \$\begingroup\$ For those like me that aren't familiar with the lingo here, apparently "infant mortality" refers to "failures early during operation", at least according to the first sentence in the introduction of this NASA doc. It's not at all related to children dying like I was first wondering. \$\endgroup\$ – JoL Dec 9 '19 at 22:02
  • 4
    \$\begingroup\$ @JoL Agreed, as-is, I think the title is more click-bait than an industry standard term. \$\endgroup\$ – JYelton Dec 9 '19 at 22:30
27
\$\begingroup\$

That high a failure rate is unheard of for a top-quality supplier like Nichicon when properly assembled and operated conservatively. Even for no-name parts it’s not at all usual- one in 10,000 might be plausible, but that’s on the high side. Short circuit failures are very rare for aluminum electrolytics. I did once see a few in a bag of 1,000 from a Taiwan supplier that were completely missing the rubber seal so the electrolyte also had gone AWOL- that was actually funny.

You can contact Nichicon directly to confirm the parts are genuine (or not). They may be able to tell just from photos or you might have to courier samples.

You can review their application information to make sure you are not abusing the parts in some way- not only voltage but also ripple current, possible reverse voltage or reverse installation (that is one thing that will cause shorts). Poorly made counterfeits might be marked incorrectly so they are reversed even though they appear to be installed correctly. Your transformerless supply might be stressing the part upon application of power.

Also confirm that the chemicals and processes used in the PCBA and any subsequent operations such as cleaning are approved.

I would definitely pull 100% of the parts from that batch of boards and replace them with known good ones. Use good tools and skilled technicians so that reliability isn’t unduly compromised by the rework. Field failures are extremely expensive in dollars and in reputation. Give them a good visual inspection under a microscope, or at least with a magnifier, and see if you can identify differences between batches or within a batch.

As far as prevention in the future, that is a bit off-topic, but there are a few approaches to control “quality fade” and substitutions of inferior parts - one of which is third party inspections. The assembly house or supplier you used is suspect if they allowed counterfeit parts to be procured. You may have a better choice of suppliers at higher quantity levels and you can ask how they intend to guarantee genuine parts are used. The Shenzhen markets are a bit of the Wild West, so you need to take care. Nichicon will have authorized distribution channels there, but it’s also possible to procure parts of unknown history at the many retail shops in Huaqiangbei or online at Taobao etc.

\$\endgroup\$
  • 9
    \$\begingroup\$ You can contact Nichicon directly to confirm the parts are genuine (or not). YES, they probably wouldn't bother with a hobbyist, but you are a high-volume user, there is a good chance that they will take a look at a sample. \$\endgroup\$ – Mattman944 Dec 9 '19 at 23:52
9
\$\begingroup\$

There is this extensive document from Nichicon for the application of aluminium electrolytic caps.

They discuss the failure rate of aluminium caps and for the one they have pictured there the failures only appear after a test time of 6000 hours. So I would not expect an immediate failure rate of 2 % for high quality Nichicon capacitors.

They have an extensive list of advice on how to correctly design in the aluminium capacitor. One point which gets mentioned is for example that the use of a halogen containing cleaning agent might seep halogen into the capacitor which then causes different failures.

For the different kind of failure modes of the capacitors they have a paragraph on short circuits:

1) Short Circuit Short circuits in the field are very rare. A short circuit between the electrodes can be caused by vibration, shock and stress on leads. It can also be caused by application of voltage above the rated voltage, application of extreme ripple or by application of pulse current

You are saying that you have 300 boards running fine, which suggest that something fishy is going on with the new batch you received. I'd discuss this issue with the supplier and ask them if they changed anything in the manufacturing going from (I guess) pre-production runs to a complete production run. Maybe something goes wrong during the mounting of the capacitors - are the leads getting bent in a bad way, are they bent by hand?

Read all the information from Nichicon and check if any of the points might be problematic in your design. If you have an X-Ray machine accessible that might give some insight in the failed caps as well.

\$\endgroup\$
5
\$\begingroup\$

The failure rate is unacceptable.

First step is to contact the manufacturer of the defective component documenting your experience. In this case with Google this will not be a problem. Include macro photographs of the component to assist with identification of fakes. If your application is not to be copyrighted or a trade secret, and you are not concerned about information leakage then include a circuit and description of circuit in normal and abnormal operating conditions. A partial circuit diagram may be sufficient. Offer to supply component samples to assist any investigation.

Once you have identified an unacceptable failure rate with a batch of components the whole batch should be quarantined and all boards reworked. No if or buts. We I say all, I mean all. For sold items issue a product recall. Document all the steps you have taken to address the problem. A company I used to work for sold a device that had a capacitor failure which caused an office fire. I don't know the level of compensation offered but it would have more than the cost of correctly specified components. Consumer protection legislation varies and I can't address that.

Soak testing of boards is usually performed at a raised temperature for a period to promote infant mortality failures. If applicable perform the tests with marginal supply voltage(s) to weed out marginal boards.

Start reading up on Lot Quality Assurance Testing, this is not a dark art and not a new problem

\$\endgroup\$
  • \$\begingroup\$ I have contacted Nichicon, they say they wouldn't accept to receive the samples from me. They say I should contact my supplier he is the one who can send them the capacitors for testing and verification. \$\endgroup\$ – Ashraf Almubarak Dec 12 '19 at 11:01
  • \$\begingroup\$ Every manufacturer will have their own procedure for handling this situation and that is why I said "Offer to supply ..." Packaging up items and posting them to the manufacturer without a return authorisation is not appropriate. \$\endgroup\$ – PDP11 Dec 12 '19 at 19:07
4
\$\begingroup\$

Not sure if this applies to aluminum electrolytics or not, but for our high-rel applications, we have our tantalum suppliers do a surge test of the caps at the factory. We also temperature test the hardware to further reduce the probability of marginal parts getting into the final product and delivered.

Yes, all of this adds cost to the product.

\$\endgroup\$
0
\$\begingroup\$
  1. This is somewhat normal. 2% seems high, but you have a fairly small sample size so it doesn't mean much. If you're concerned about counterfeits, capacitor datasheets usually have a test setup and you can verify each specification on your own. I've seen counterfeit caps that were literally smaller caps inside of a big can. Find a reputable supplier, it is worth the extra cost.

  2. Larger companies that comply to industry standards (ISO9001 etc) all do pre-tests like you say. The amount of time you run the unit depends on the circuit and components and your specification, but you should pretest all of the boards. A yield of >90% is good. These pretests should stress your board to be extra sure.

  3. IPC will have the most up to date guidelines. Make sure you are following proper derating guidelines for the most reliability. You can send sample boards to a lab that specializes in reliability and they can predict an MTBF for you.

\$\endgroup\$
  • \$\begingroup\$ I didn't find any useful information in the datasheet. \$\endgroup\$ – Ashraf Almubarak Dec 9 '19 at 14:16
  • 14
    \$\begingroup\$ I don't think a failure rate of 2% is anywhere near normal for such a regular component. If you have, for example, 20 components on your board, the probability of your board going to rework will be more than 33%. I don't think it is acceptable for most productions. \$\endgroup\$ – dim lost faith in SE Dec 9 '19 at 16:03
  • 6
    \$\begingroup\$ @AshrafAlmubarak - "My supplier notify me when he changed some part with a Chinese one if he didn't find mine in stock." Yup. And now you know why you have a 2% failure rate. Look, if you're going to be serious about producing quality, reliable goods, the last thing you can afford to do is use unnamed Chinese brands. Sometimes you'll do OK, but sometimes you won't, and you have to way to tell what will happen, and no way to complain if you get burned. Let this be a (painful) lesson. There are some very slimy actors out there. \$\endgroup\$ – WhatRoughBeast Dec 9 '19 at 21:22
  • 3
    \$\begingroup\$ @WhatRoughBeast: I don't think that's what he was saying (though I admit it's a little hard to decipher); I read that as "In the past, my supplier has notified me of part changes, but this time it didn't happen, so I would've assumed that I got the part number I requested" \$\endgroup\$ – Aleksi Torhamo Dec 11 '19 at 10:04
  • 1
    \$\begingroup\$ @AleksiTorhamo yes that what I meant. English is not my first language. \$\endgroup\$ – Ashraf Almubarak Dec 12 '19 at 11:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.