Component Failure Rates/ reliability

[johndon2000]

Has anyone ever carried out statistical analysis of component failure rates/ reliability?

For example, tantalum caps would have a high failure rate so would be high up on the list... it would be good to have a list of components in order least reliable to reliable...

It would also be good to have similar list for device package types (e.g. PLCC, DIPs...) or any other failure analaysis info!

Thanks

[AndyC_772]

Manufacturers do this all the time, and you can look up the results on their web sites. You just need to go digging a bit, that's all.

It's impossible to generalise because the failure rate of any type of component depends on how it's being used - what voltage is across it, what's its ambient temperature and self-heating, how is it mechanically supported and mounted to the board, and so on. You can't just say that "1% of this type of component fail every 10 years", though there are mathematical models which try to predict the MTBF for a given part given some details of how it's being used.

Typically if I'm asked to assess the reliability of a product I start by looking at fans and other items with moving parts, then batteries, then capacitors being used in power supplies. If a product contains any of these items, then they're likely to govern the MTBF of the complete product, and the MTBF can be predicted with a reasonable degree of accuracy. If it doesn't, then the product is likely to last a great deal longer, and the uncertainty increasees enormously.

Often a product that's designed without 'unreliable' components will fail prematurely because of manufacturing problems (eg. bad soldering, thermal cycling), or because of some external influence which has caused it some type of stress that it can't handle - for example, ESD strikes when a cable is connected.

[Rufus]

Has anyone ever carried out statistical analysis of component failure rates/ reliability?

Search MIL-HDBK-217F

[KJDS]

Has anyone ever carried out statistical analysis of component failure rates/ reliability?

Search MIL-HDBK-217F

but whatever you do, ignore the FIT values in it, they are so out of date that most modern computers would die before they'd completed their boot cycle.

[WBB]

Well that certainly never happens. 

[ignator]

Has anyone ever carried out statistical analysis of component failure rates/ reliability?

Search MIL-HDBK-217F

The reliability engineers would ALL say when using this MIL STD method, "What number do you want?".  It's a method, but only government projects used it (specified in contract). 
They had their own failure rates (big company, in biz since the 30's, and I only had to do one MTBF calculation in my career, the new guy, and arrogant flight test engineers, didn't trust REL guys).

What is the OP trying to compute, MTBF, MTBUR.  Is the unit fan cooled, what operating temp range, what vibration levels, thermal shock, etc.  Failure rate is a function of environment, you can end up with a real unreliable estimate.

If just failure, then you don't care how it fails, short, open, intermittent.  This you need for a FMEA, which each failure has it's own numbers.