One subject that confuses specifiers of
power supplies is comparing MTBF numbers from different manufacturers, who
often use different standards for calculating the number of hours. There are many well written articles going in
to great detail available on the internet on the calculation of MTBF, but this
blog article will attempt to simplify things.
MTBF (Mean Time Between Failures) is the mean
time between successive failures, and only really applies to a part that
will be repaired and returned to service.
So if the up-time of the power supply was a year in each case below,
then the MTBF would be ½ x (1 year + 1 year).
A low cost power supply will probably not
be repaired and if it is under warranty, it will normally be replaced. In this case, the numbers to look for would
be MTTF (Mean Time To Failure), but that figure is not widely stated. Usually life testing of a large number (to
cut the test time down) of power supplies is used to calculate that.
The MTBF number is often thought to be the
minimum (guaranteed) time before a failure; that is certainly not the case! Reliability “R” is based on the probability
that a piece of equipment, in our case a power supply, will operate
satisfactorily for a given time period “t” (based on specified conditions – for
example ambient temperature and output load).
For random failures, the probability that a
power supply will survive to its calculated MTBF is just 37%, no matter what
the MTBF number is:
R(t)
= e –t / MTBF = e-1 = 0.368 (when t = the MTBF number)
To complicate things further, a variety of
methods are used to calculate MTBF.
Prediction Standard
|
Applications
|
Disadvantages
|
MIL-HDBK-217F
|
Provides failure rate data and stress models for
parts count and parts stress predictions. It also provides models for many environments
ranging from ground benign to projectile launch
|
Hasn’t been updated since 1995, gives higher failure
rates of commercial parts than is seen in actual product life
|
Telcordia SR332
|
Gives three prediction methods based on parts count,
lab testing and field life
|
Narrow ambient temperature range
|
RCR-9102
|
Produced by JEITA - Japan Electronics and
Information Technology Industries Association.
In each update component failure rates (FIT) have
been changed, particularly fans
|
Issued in 1994, based on MIL-HDBK-217F
|
RCR-9102A
|
Issued in 2000, based on MIL-HDBK-217F (Notice 2)
Includes SMT parts & pcbs
|
|
RCR-9102B
|
Issued in 2006
|
Usually for commercial power supplies, the
figures are calculated at 25oC, ground benign or fixed
Taking TDK-Lambda’s RWS150-B series as an
example, the calculated numbers are as follows:
RCR-9102 444,089 hours
RCR-9102B 218,172 hours
Telcordia 2,235,743 hours Ta=25℃
Telcordia 1,063,230 hours Ta=40℃
It can be seen from the above numbers, that
there is a 10-fold difference
between RCR-9102B and Telcordia, and more than a 2 fold difference between RCR-9102 and RCR-9102B. Several customers have asked why our newer
products calculated using the JEITA method appeared to be less reliable than
older products, but did not know the significant impact of the updated, harsher
standard.
Engineers should be more concerned about electrolytic
capacitor and fan life (if used) as these are the typical failure modes. Many manufacturers are showing expected
capacitor lifetimes in their reliability reports. Below are the plots for the RWS150B, which
was designed for long capacitor life. As
a note, some manufacturers show similar plots, but state in small print that
the convection cooled power supplies had external forced air cooling applied.
