Connect with us

calculate MTBF

Discussion in 'Electronic Design' started by Raveninghorde, Jan 21, 2010.

Scroll to continue with content
  1. How do you calculate MTBF?
     
  2. Rich Webb

    Rich Webb Guest

    For a component? For an entire system? 217 or Bellcore/Telcordia or ...

    The Wikipedia article isn't a bad place to start. If you're tasked with
    doing a full-up parts-stress reliability prediction analysis, good luck!
     
  3. Thanks

    The National site is good and I found the info for the their parts.

    Microchip no luck, IR no luck. So where do you normally find the
    information?

    I suppose one also assumes perfect ESD procedures, and perfect lead
    free soldering.
     
  4. For a lithium ion battery charger.

    My brief post was a cry of despair. I ship these by the hundred. Now a
    BIG US company wants them and I get asked for a bucket load of
    information.
     
  5. Rich Webb

    Rich Webb Guest

    You're new at this so, even if you're not doing a MIL-STD prediction,
    I'd strongly recommend downloading MIL-HDBK-217 from the online site
    <https://assist.daps.dla.mil/quicksearch/> (search on MIL-HDBK-217 in
    the Document ID field) and skimming though sections 3 and 4. The
    Bellcore process is similar.

    John's basic equation is correct. You can use the tables in 217 to
    estimate the reliability of items for which the manufacturer does not
    provide the info.

    There are provisions to consider connections (e.g., "Quantity of Hand
    Soldered PTHs [plated through holes]"). It's a hell of a lot of work to
    do a full analysis.
     
  6. Rich Webb

    Rich Webb Guest

    Their root certificate is back in the .mil domain rather than the
    "normal" root authorities that are built into most current browsers.
    There may be some reluctance on the part of browsers or it could be a
    statutory limitation.
     
  7. Artemus

    Artemus Guest

    It's been a long time since I did any of that stuff but IIRC the MIL
    calcs used 125°C for the high temp and Bellcore used 85°C. This
    would account for the better numbers.
    Art
     
  8. Given the customer I'll go with 217. I'm trying to avoid spending a
    week doing this so I'm looking at other solutions.

    I've seen a few websites that offer online calculation ($500) or send
    the BOM and results in 48 hours ($1500). I've also seen some programs
    that calculate MTBF.

    Anyone tried these options?
     
  9. Nope. They would prefer the BOM and do the calculation themselves.
    Since I won't give the BOM it's down to me sort it out.
     
  10. Rich Webb

    Rich Webb Guest

    When I got roped into doing one for a piece of shipboard equipment
    (nobody else would even admit to knowing how to spell "RM&A"), the
    company had a copy of Relex that I used for the crunching. Packages like
    that do save time, since they have the tables "built in" and one just
    needs to select the appropriate factors for each component. It can also
    make auditing a bit easier, as well as printing pretty reports.

    You might also want to start with the Appx A "Parts Count" method which
    is less detailed but does give you a ballpark figure (but one that tends
    to be more conservative).

    The intended use of either method is to establish an "apples to apples"
    comparison baseline, sort of like the EPA estimated MPG ratings. You
    don't expect to get exactly that MPG but you can have some reasonable
    expectations if one vehicle is rated 18 and another 32.

    The other big use is to help out the logistics chain in estimating the
    initial spare parts buy. The reliability prediction is paired with a
    FMECA (Failure Modes, Effects, and Criticality Analysis) to help decide
    which, how many, and where spares need to be stocked. Once the equipment
    has a field history, of course, the actual failure rates and spares
    usage drive the numbers.
     
  11. JosephKK

    JosephKK Guest

    Basically it is the _statistical_ point where the root sum square failurerate
    of the individual components becomes 50% probability.
     
  12. Rich Webb

    Rich Webb Guest

    Note that for the often used exponential distribution (applicable for a
    constant failure rate), we will then have the reliability at time t as
    R(t) = exp(-t/MTBF). So the probability that any one component actually
    survives to its MTBF is only about 37%.
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-