Design Con 2015
Home    Bloggers    Blogs    Article Archives    Messages    About Us   
Tw  |  Fb  |  In  |  Rss
William Murray

Component Count & Reliability

William Murray
SunitaT0
SunitaT0
7/29/2013 3:47:31 PM
User Rank
Master
Re : Component Count & Reliability
If the component count is decreased in design, the probability of error point is reduced. Reduces components also increase reliability. Systems engineering often includes efforts to reduce part count with the objective of cutting costs, enhancing performance, or improving reliability.

50%
50%
RedDerek
RedDerek
7/10/2013 2:04:56 PM
User Rank
Master
Re: System Reliability
Mil-Hndbk-217 Is one method for determining reliability of electronics

@WilliamM - I remember years back, data books would include the component count on the data sheet to aid in the MTBF calculations. That practice is clearly not used these days.

50%
50%
William Murray
William Murray
7/10/2013 6:28:05 AM
User Rank
Blogger
Re: How to count?
All -  Where IC's often have failures that take them below the rated # of MTBF hours are  1)  High Delta Thermal Cycles   2) ESD during Test and Maintenance  3) Thermal shock/vib.  4) Exceeding Design Ratings(Example internal hotspot due to overvolts / over-current)  5) Over Temperature of Die(Electromigration at temp, etc)

If kept within design ratings modern IC's that have been designed to pass and passed a reliablitiy qual test in a benign room temperature environment are quite reliable with #'s easily exceeding 100,000 hrs for even a really hot part like a PC, and ultra low power processors exceeding 100 million hours easily.   It is often where one cycles the part 5-10 times a day over a large delta like a delivery truck engine controller IC that one must pay close attention to everything, and numbers down in the 20K hrs are sometimes what one end's up with.     

Will

 

 

50%
50%
eafpres1
eafpres1
7/9/2013 10:14:09 PM
User Rank
Blogger
How to count?
Hi William--your post reminded me of when I learned about tolerance stack ups and how to do RSS calculations to get more realistic estimates of whether an assembly might have a problem.  Young engineers faced with their first stack up, and not being street savvy in statistics, look at all those parts, and all those tolerances, and start to worry that the assembly will never work.

Later, I learned about Monte Carlo simulations and how you could model combinations of possible distributions in an even more realistic way.  However, in both cases you still have one value for every part--the difference is in how you add it in.

Which brings me to your case, and gets to my question.  I think I'm along the same lines as DaeJ--is it correct to consider an ASIC or other IC a lumped component with a count of 1?  If the feature count is the same, what are the reliabilities of each resistor, capacitor, etc. in the integrated part?  I think that either the features in ICs are extremely reliable compared to assembling from discrete parts (where I assume some failures come from connectivity and soldering issues, as you note), or the ICs are tested directly (I know some are--they do reliabiilty testing at the IC level and publish MTTF etc.) or there is some other method, known to give more realistic estimates of combining the reliabilty of all the features.

50%
50%
William Murray
William Murray
7/9/2013 11:26:49 AM
User Rank
Blogger
Re: System Reliability
Mil-Hndbk-217 Is one method for determining reliability of electronics.   There also have been Bel-Core Standards and others.  With that said I have seen some really cooked numbers come out of Mil-Hndbk-217 compared with what one got in actual service.   Some of the reasons for this -- Mil-Hndbk-217 does not specify a thermal cycling profile for the analysis, 2) The customer may fail to specify a normal operating temperature for the equipment location for the analysis.  3) Factors like humidity, lightning(for aircaft which are often struck several times in a year), and others.   Still if one puts 50,000 parts in Mil-Hndbk-217 one will get a much worse number for reliability than if one puts in only 4500  similar parts.

50%
50%
DaeJ
DaeJ
7/9/2013 11:00:25 AM
User Rank
Master
System Reliability
It is not easy to count each component for normal operating hours, unless datasheet indicates it. I think that Chip maker defines this kind of requriement. Reliability engineer might figure out the system for long terms reliability. Secondly, all discrete components of previous system were updated to only one microcontroller. We did not measure the reliability for the updated system, eventhough the system is simplified. I think that industry has a process.

50%
50%
More Blogs from William Murray
Electromagnetic compatibility and radio frequency interference issues are increasingly more problematic. Here's how some problems occur.
SiC FETs are like conventional MOSFETs in some ways and much better in other ways. They do have a few shortcomings -- but we can live with that.
For high-rel designs, don't just read the data sheet and hope for the best. Create and implement test protocols to make sure your product keeps going.
Engineers can get in a rut when it comes to design methodology. Try thinking outside of your comfort zone to reach a solution.
With an LDO that has good PSRR specs, filtering the DC for sensitive circuitry may be unnecessary.
flash poll
educational resources
 
follow Planet Analog on Twitter
Planet Analog Twitter Feed
like us on facebook
our partners
Planet Analog
About Us     Contact Us     Help     Register     Twitter     Facebook     RSS