IEEE Power Electronics Magazine - June 2021 - 99

rated at, say, 175 °C I would design
such that the junction temperature
never exceeds about 115−120°C
even under worst case. Am I too
conservative? Perhaps, and in very
cost sensit ive designs, I have
been known to stretch my derating
guidel ines to meet the cl ient's
needs but I always have to grit my
teeth to do so.
But achieving high reliability goes
beyond what I as the circuit designer
can do. Achieving high reliability
also means continuous diligence in
monitoring the supply chain and the
manufacturing processes. Essentially,
high reliability and high quality go
hand in hand. It is so easy to wreck
both by an imprudent purchasing
decision. In times, when components
can be in short supply, such as the
ceramic capacitor shortage of the
past few years and now shortages of
key semiconductors, there can be
tremendous pressure on the purchasing
group to " find those parts somewhere
and get them in here now! " in
order to keep production lines going.
But, when buyers turn to the grey
market and fly-by-night parts brokers,
they are inviting a quality and
reliability disaster.
Things can go wrong even with
the best processes and checks. When
I was managing a dc-dc converter
design group, we had some high
power full brick converters being
sold to a customer that was putting
them in high end computing systems.
All was going well until suddenly
about three to six months after we
shipped the converters started failing
in the field. We had done a proper
design and verified it through very
tough stress testing. The purchasing
group was only buying from either
the original manufacturer or an
authorized distributor. The manufacturing
process was under control.
Working with our suppliers on
the failure analysis we found that in
a high power semiconductor there
was a metal migration problem.
After a few months of voltage and
temperature stress metal would
migrate across the die and cause a
failure. This was a well-known and
highly regarded semiconductor manufacturer.
But they had a slip in their
process. The cost to them, to my
company, and to the end customer
was enormous as we all scrambled
to get good parts, build new dc-dc
converters, and replace all suspect
units in the field. So even when
everyone is doing all the right things
" stuff happens. "
There is one area of power electronics
reliability research that I
hope turns out to be as good as the
hype. I see more papers on trying to
monitor the condition of power electronics
and attempt to predict failure
before it happens. These methods
tend to use some combination of
actual monitoring combined with
" digital twins " and algorithms that
track changes over times. Many of
these papers feature artificial intelligence
(AI) and machine learning
(ML). I am a bit skeptical of the AI
and ML aspects but do think there is
value in condition monitoring for
some applications. For example, in
higher power utility interface applications
like solar farms and wind turbines,
monitoring the power electronics
and scheduling a service call to
replace an ailing converter could be
economically worthwhile. An example
might be monitoring the baseplate
temperature of a power module
over time. If it is noticed that for the
same operating and environmental
conditions the baseplate temperature
is increasing by about one degree per
week that probably indicates that a
semiconductor device is starting to
degrade. Replacing the module or an
entire converter before there is a
complete failure would probably save
a lot of money. Is the cost of monitoring
for a multi-megawatt solar in -
verter or wind turbine economically
worthwhile? Probably. Is such monitoring
worthwhile for the thousands
of power supplies in a data center
that is already highly redundant?
Probably not.
I will say that I am now glad to see
university research on power electronics
reliability even though I suspect
industry is actually way ahead
but unable to publish their results
due to business confidentiality needs.
Anything that can help move us
ahead is welcome. As power electronics
becomes more and more central
to our lives (I haven't even mentioned
the role of power electronics in transportation!),
reliability, service life,
and availability will become increasingly
important. While we can never
achieve perfect reliability and availability,
we can strive to make our
semiconductor based power electronics
as reliable and long lived as the
copper and iron based electronics
they will be replacing.
About the Author
Robert V. White (bob.white@ieee
.org) has more than 30 years of industry
experience as a power electronics
engineer. He has worked in product
design, systems and applications engineering,
and technology development.
He has been an active volunteer with
the IEEE Power Electronics Society,
serving several years on the Administrative
Committee, two terms as technical
vice president, and as a Chapter
chair. He earned a B.S.E.E. degree
from the Massachusetts Institute of
Technology and an M.S.E.E. degree
from Worcester Polytechnic Institute.
He is currently pursuing a Ph.D.
degree in power electronics at the
University of Colorado, Boulder. Presently,
he is the chief engineer of
Embedded Power Labs, a power electronics
consulting company in Highlands
Ranch, CO 80130, USA. He is a
Fellow of the IEEE.
June 2021 z IEEE POWER ELECTRONICS MAGAZINE 99
IEEE Power Electronics Magazine - June 2021

Table of Contents for the Digital Edition of IEEE Power Electronics Magazine - June 2021
