0
\$\begingroup\$

Short version: I understand resistor initial tolerance is commonly obtained through sorting process where manufacturer only picks resistor that has resistance within the advertised initial tolerance. Does MLCC manufacturer use similar sorting process to determine initial tolerance for capacitors?

Long story: This question is related to reliability calculation I have been recently doing where RSS method is used. When I try to find the worst case capacitance change, I combine all the random terms based on manufacturer spec sheet. Let's say for example this capacitor from muRata: GCM31CL8EL225KE07L. The specsheet can be found in this link. I have made the following assumptions when doing RSS for random terms:

  1. Capacitance change due to a random event such as high temp exposure, moisture resistance, thermal shock, etc. falls within the 3-sigma range, i.e. chance of being within stated change is 99.7% under such event (I could be wrong with this too)
  2. Capacitance change due to initial tolerance is uniformly distributed. For example, if the it is a 10% cap, its one-sigma (std deviation) is 5.77% and its 3-sigma is 17.32%.

I am guessing if manufacturer determines MLCC tolerance like resistor, I should use uniform distribution. But I would like to make things easier and assume this tolerance also follow normal distribution. This means the initial tolerance corresponds to the 3-sigma probability range. In this way I can make the RSS calculation easier without worrying about the square root of 3, simply calculating the squared sum of all random terms including initial tolerance. I couldn't find related information online and hope someone can help me here. If you can explain how you would use these random terms in RSS that's also much appreciated. Thanks in advance!

\$\endgroup\$
  • \$\begingroup\$ Process controls must ensure that 100% of the parts meet spec whether it is 1% or 10% . The yield is determined from the batch and then process is adjusted to meet the target cost and bin them to different tolerances according to profits with tighter tolerances and losses from yields. They might use 3 sigma to 6 sigma or Dpk calculation and DOE to determine the cost of making the required process improvements. I have not heard of RSS being used, but what’s the benefit? They need to know the Mean Offset and Standard deviation to know in which direction to make continuous improvements. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Aug 14 at 1:02
  • \$\begingroup\$ @Tony Stewart Sunnyskyguy EE75 , thank you for your input and that's good to know... I am actually asking from user perspective. Given available data from manufacturer, we need to prove to our customers that under worst possible situation the circuit still works within spec. In this sense I need to find out how much the capacitor in our circuit can deviate from its nominal. I am not entirely sure if the way I use initial tolerance is correct or not (in RSS method)...? More specifically, I want to know if assuming initial tolerance follows uniform distribution is reasonable or not? \$\endgroup\$ – brokenegg Aug 14 at 1:26
  • \$\begingroup\$ RSS is not correct. They must meet initial spec at conditions specified with temp and voltage variations applied. Distribution will vary batch to batch. Very tight within each but may have offset. e.g. My supplier LED's test results are within +/-50 mV but many within one batch are with 5mV yet others have wide specs +500mV/-200mV \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Aug 14 at 2:10
  • \$\begingroup\$ RSS is not used when lots are 100% tested. You are thinking of something like screws, which are lot tested based on established fail rates. During my QC days I inspected military hardware, and out of a case of 1,000 grade 5 high-carbon-steel bolts, one would not have any threads on it, but it was not enough to reject the batch. \$\endgroup\$ – user105652 Aug 14 at 4:17

Your Answer

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

Browse other questions tagged or ask your own question.