Our component sourcing organization purchases large amount of electronic components from suppliers. I believe this is through a bid process. Engineering provides the sourcing organization specification, and they purchase these parts for the plant to use in manufacturing. The issue is that recently the plant has been receiving a increasing amount of capacitors which might be defective. This is causing our products to fail. I have checked the few new incoming capacitors using fluke multimeter, and the measurements are with in specification.

It is not practical to continuously inspect these capacitors, and I have been tasked to recommend the solution. How does one technically address similar issues?

  • $\begingroup$ Are you looking for a Quality Engineering method to reduce defects below an X% probability (such as Six Sigma), or are you looking for a faster ways to identify bad capacitors? $\endgroup$
    – Mark
    Jun 24, 2015 at 17:15
  • $\begingroup$ It is not clear from your question that you have definitively determined that the capacitors are the cause of the failures. Determining this would be the first order of business. Or have you already done a root cause analysis? $\endgroup$
    – hazzey
    Jun 24, 2015 at 18:04
  • $\begingroup$ You say the capacitors "might be defective" and that's causing product failure. What other defects/errors can cause the same failure? Have those been eliminated? The first step in addressing a product failure is some kind of Root Cause Analysis to nail down what is really causing the failure before you start with quality control on components. $\endgroup$
    – DLS3141
    Jun 26, 2015 at 18:35
  • $\begingroup$ There is evidence that capacitors are one of the many problems, There are other issues beside capacitors. Like the six sigma idea too. $\endgroup$
    – user1586
    Jun 30, 2015 at 16:54

5 Answers 5


In the mechanical world, if you have a million bars, and pull on them until they yield, you'll get a wide variety of failures. These failures, if you counted them with how many failed at what strength, you usually end up with a normal distribution, though to make the math easy we usually fit it into a log-normal distribution. When dealing with a part who has a shorter life, and not a property failure (which may be the case here), you should use a weibull distribution.

Going back to the example, you'll find that a higher percentage yield at a point - the mean, and the rest scatter around the mean. Some - a small amount, will be out of tolerance because you are testing a million. 3 of those million probably would yield at a load so low that even if I used that one piece in my design to a reasonable factor of safety, my design would have failed.

Of course, in real design, I don't actually load it with a set load either - these loads vary as well. As a result, I can identify the loading as a probability as well. This can then lead to figuring out the probability of failure - the probability that the load is greater than the strength of the bar. This type of design is called Probabilistic Design - which even has a nice diagram showing this example.

Once I run my probabilities, I need to make sure my bars don't slip up. So, I run testing on every batch to see that it falls within my original distribution. I make sure using comparative probability techniques that everything I get exceeds the bare minimum the engineers need to make sure the part doesn't fail.

This principle is the foundation for the very broad field of Reliability Engineering. Using this thought that failure is inevitable - but the probability of failure can be minimized, is the basis for many Quality Control Programs, such as Six Sigma. For you it comes down to a few things with capacitors - lifetime and meeting the specified capacitance.

Here's how I would approach your capacitor problem:

  1. Enforcing your manufacturer to have an ISO quality system. This will mean that if you invest in some testing of your own to find an optimal manufacturer, they will be making the part with the same quality as when you tested it.
  2. Run some accelerated stress testing if lifetime is the problem. Perform a test using your complete assemblies and a test using just your capacitors. Find out what kind of lifetime you have - you'll need quite a few samples, but you can fit it into a Weibull or Normal Distribution. Your manufacturer may have run these tests for you - in that case, include your reliance on the manufacturer's testing in the specification.
  3. Run some capacitance, testing with the multimeter, as well some destructive (over-excess voltage), leakage, overheating, etc testing. Again, you'll need a few samples, but then you can have a probability of failure for what you are testing. Also like before, if they provide this to you, include it in the specification so purchasing won't accidentally undermine you.
  4. Run these values through the engineers to determine the probability of failure. You may find weak points you need to design around. Redesign (if needed), and certainly re-qualify your specifications. You want your specifications to ensure your final design meets the testing requirements
  5. Use control charts, run charts, and any of the other 7 Basic tools for all new parts. These lets you keep track of new parts as they come in without checking every single part, and lets you know when parts are failing prematurely - to alert or discontinue working with your vendor.

Unfortunately, Quality Engineering is a very broad topic - I've really only scratched the surface even with this very large post. But I think this will help you with your capacitor problem.

  • 1
    $\begingroup$ I'm glad it helped. There are six Sigma consultants you can hire to get you started. I only recommend that system because I've had experience with it. There are other systems, but that's the most well known. Also, Russell had some really good points when it came to actual capacitor testing, so probe him for some real electrical engineering quality advice. I'm just a mechanical engineer who happens to deal with quality problems a lot. $\endgroup$
    – Mark
    Jul 1, 2015 at 3:54
  • $\begingroup$ Broadly agreed, but I would add (a) as part of the engineering specification, /require/ ISO QA and a certificate of conformity, this moves part of the problem in-house since if this is omitted it's your purchasing peoples' responsibility. (b) Do not trust a certificate of conformity, always subject some fraction of your purchased supplies to accelerated testing and some further proportion to non-accelerated testing so that you can recognise if you're about to get a flood of returns. $\endgroup$ Nov 29, 2022 at 12:23

You do not say what sort of capacitors are involved.
Tests will vary with capacitor type.
Electrolytics can be tested for eg leakage, ESR, behaviour as they approach rated voltage, ... .
Ceramic capacitors that are "of unknown origin" may well exhibit temperature dependant variations not matched to they specifications.
[Tantalum capacitors may explode with flame smoke sound and smell, but, hey , they do that anyway, so that's no guide :-) ]

This problem is extremely easy to address by pre-emptive action - except for the part that involves getting the accountant to agree. Once they realise the real cost of using bad parts agreement should follow. Traditionally used ratios for the cost of locating a faulty part in
incoming inspection and in-production testing / final test / customer fault is
Find it early / don't let them happen at all.

Solution -

  • ONLY EVER buy known quality brand components from a known reputable sales channel.

Almost all the time you will get what you expect and if you don't you will have right of redress with the supplier. This may not extend to cost of repair but should at least cover cost of replacement. That is not enough as the cost of field replacements is often immense compared to component cost.

As a starting guide I say 'if you can buy Panasonic, do so' - it may involve price and availability - and really this is liable to apply to a range of reputable manufacturers. Note - that is NOT an "ad" for Panasonic - I have no relationship with them except as a happy customer.

Next, you do not have to buy from them but, if a brand is sold by a major reputable reseller then it is very probably of acceptable quality when used within specification. I use Digikey as a first check, but a range of otrhers are equally good. Taking aluminium electrolytic capacitors from Digikey as an example. Digikey list 12 manufacturers on that page. Most of the brands listed probably manufacture some or most of their capacitors in China. China is perfectly capable of making absolutely superb capacitors (and a;lmost any other product as well). But if you buy Chinese product via sales channels that you do not know you can trust, or if you buy brands you have not heard of, or products labelled with the logos of brands that you have heard of, but from sales channels whose integrity is unknown to you you will sooner or later find yourself " ... receiving a increasing amount of capacitors which might be defective. "


Looking for good and reliable suppliers might be one of best solutions. Vendors such as Digi-key, mouser, Newark always supply high quality electronic components. But understandably they come at a higher cost. If directly purchased from a manufacture it might be good idea to request for test data from suppliers to validate that product is meets specification. Most Suppliers have the test data readily available and if the product meets specification they will gladly share the data almost immediately. If the capacitors are purchased through an intermediary then you could request for sample data. A sample size of 30 pieces is good sample size to perform statistical analysis.

Alternatively a small sample set of components can be tested and the measured data can be analyzed using basic statistics as part of incoming inspection. If the capacitors are SMT this might become tricky because randomly sampling across the SMT reel will cause issues for the SMT pick and place machines.

Regardless after obtain this data consider performing some basic statistical analysis such as averages, standard deviation, and process capability analysis. This type of simple first order analysis will give a good indication if the incoming material is with in specification.

Below is 30 randomly generated data point to illustrate the example basic analysis. The data points are based on 10µF capacitor with +/- 10% tolerance. Some data points are marginally below or above the lower or high tolerance limits.

\begin{array}{| c | c | c | c | c | c | c | c | c | c |} \hline 9.40 & 10.00 & 10.10 & 11.30 &10.10& 8.80 &9.60 &9.20 &11.50 &8.90 \\ \hline 8.60& 11.30& 10.00 &11.10& 10.60&11.10 &10.40& 9.10& 11.50& 8.90\\ \hline 8.70& 10.70& 11.20& 10.30& 10.30& 11.50& 10.80& 9.70& 9.50& 10.80 \\ \hline \end{array}

  • Average ($\mu$): 10.17
  • Standard Deviation ($\sigma$): 0.93
  • C$_{p}$ = 0.35
  • C$_{pl}$ = 0.41
  • C$_{pu}$ = 0.29
  • C$_{pk}$ = 0.29

Data1 Data 2

Base on the example data it appears that the capacitor value distribution is very wide. Looking are such data analysis certain characteristics such as the capability of the process and incoming material can be ascertained. This type analysis can help make simple decisions about the capacitors such as whether to accept or reject the material. How can the vendor be approved/disapproved based on validation test data? post closely related to the topic and some the equations discussed in greater detail. Hope this response helps address the issue.


One has to recall, on this question, the capacitor plague problem that the computer industry experienced for a number of years. The issue with this is I'm not sure if there was ever a way to discover these effects prior to failure, and the main mitigation method was a supply-chain change.

However, it does give perspective. As Aravinthkumar (answerer here) suggests, random component testing does weed out a fairly good number of bad components, however this will only detect components that are immediately defective. It will fail to detect components that like the capacitor plague, simply have a premature lifespan, but with no other obvious issues. Also, like you yourself said, the components are usually within specification, so no red flags are raised there.

What I suggest you do is run your own product, continuously. If it's a microprocessor-based product, develop some sort of protocol much like a benchmark that is designed to stress-test the system at maximum allowable power. Cycle the power randomly (a lot of capacitors fail on initial startup, under inrush current situations). If you're in a market segment where you don't expect your devices to be run continuously (i.e. 24h/day or close to), by doing so yourself you can detect if the components meet the specification for your product lifecycle, as well as make your determination a good deal faster than it would take the end-users to come to the same conclusion.

Pushing the components a little harder might make issues come to light quicker. As it is good engineering practice to build a margin of safety into circuits (since we're talking capacitors, presumably power supplies), push the circuit harder (i.e. current draws) within the margin of safety, not so much as to cause failure but more so than you would expect from the end user.

At any rate, once you've determined that failure in your current component supply is probable (that is, higher than the manufacturer's failure rate, or your company's own arbitrary failure rate), select a different manufacturer or device line, etc.


Check some component randomly, low quality components produce heat immediately.

As you can imagine, the liquid inside the capacitor will only dry out if the capacitor isn’t perfectly sealed and/or if the capacitor is exposed to high temperatures the definition of “high temperature” for us is any temperature above the standard room temperature of 25° C or 77° F.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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