Power supply failures
can be frustrating, expensive, and time-consuming events. These
are highly complex circuits, with many components operating
near the edge of their envelope, by design. When they fail,
they tend to destroy most of the failure evidence with them,
and many times you can spend months on an apparently "simple"
problem to track it down to the real root cause. Some common
failure mechanisms are detailed below. These are well-known
to engineers with many years experience, but most of them are
not documented in any easily-accessible form.
Power Supply Failure Mechanisms
Don't Forget
Your Safety Glasses!
Think
Safety. Power supply circuits are potentially lethal - 400
to 600 VDC is the most dangerous voltage range to humans. At
lower voltages, not enough current flows in the body to cause
damage. At higher voltages, the current tends to flash over
the surface of the body. Never work alone on power circuits.
Even low power,
low voltage circuits can be dangerous. Apply a stiff 20 V source
to the wrong pins of an integrated circuit, and the results
can be dramatic. Always protect your eyes.
Power
Switch Overcurrent -
either the semiconductor will fail, or bond wires leading to
the die will fuse. Always implement pulse-by-pulse
overcurrent protection, whether it's a 10 W bias supply, or
a 10 kW inverter. Poor current limiting design is a very common
cause of power supply failure. Build the pulse-by-pulse current
protection into your very first breadboard, and save yourself
a lot of development time in fixing failed circuits.
FET gate
overvoltage will cause rapid failure. FETs are very rugged devices
(from the right sources - Motorola and IR make excellent parts),
they will tolerate very large drain current, and even overvoltage
on the drain for short periods. But they will not survive excessive
voltage on the gate. Make sure your gate drive circuit can never
cause excessive gate voltage, and protect with back-to-back
zener diodes if necessary.
High-side
silicon drivers are
not recommended for high power, high voltage applications. Use
a gate drive transformer - they are rugged and provide a negative
gate drive for improved noise immunity when the device is off.
Start-Up
failures are among the most common problems.
At start up, the output caps are discharged, and look like a
short circuit to the power stage. Make sure that your current-limiting
is fast enough to survive start up at maximum input line. Do
not depend on the soft-start of the PWM controller alone to
protect the switches.
FET Anti-Parallel
Diode conduction
can cause problems.
There is nothing wrong with using this diode in a circuit as
long as you do not apply reverse voltage quickly while the diode
is conducting. These diodes are very slow, and the subsequent
dv/dt rating after conduction is low. It's OK to use this diode
for conduction if that particular FET is the next one to turn
on - this sweeps out the charge in the diode junction.
A common situation
where the diode is inadvertently allowed to conduct is in a
full-bridge converter. When a conduction cycle ends, leakage
inductance will cause a ringing. If this ringing is large enough,
on its first peak, it may reach the dc rail and the FET diodes
will conduct. This is OK, these FETs are the ones scheduled
to turn on next. However, if the ringing is so undamped that
the second peak (in the opposite direction) hits the rail, the
other pair of diodes turn on and your circuit can fail at the
initiation of the next switch cycle.
To avoid this
situation, make sure the snubber provides sufficient damping.
The phase-shifted
full-bridge topology inherently controls the FET diode conduction
problem better than the hard-switched full-bridge, but it too
is not immune from this in some cases.
The current-doubler
full-bridge may run into this problem at light loads - be careful
to fully characterize the circuit during development to make
sure this won't happen.
Make sure your
Undervoltage
Lockout is working properly at start-up.
Commercial PWM ICs are usually carefully designed to prevent
unpredictable operation when insufficient Vcc is applied. However,
if you use auxiliary comparators, gate drivers, or other circuits,
you have to check them all to make sure they power up properly
in a predictable and safe manner.
Large
MLC capacitors are prone to failure at high voltage ratings
(400 V) and elevated temperature. If you need to use these caps,
make sure you test and qualify them thoroughly for all your
sources and for your particular manufacturing conditions. Also
make sure they are properly stress-relieved for the conditions
your power system will see.
Schottky
diodes usually fail because of excessive reverse voltage
(assuming they are properly cooled). Make sure you never exceed
the voltage rating under any conditions; an 80% derating worst-case
is recommended.
Watch out for
the full-bridge converter with a dc-blocking cap. If this capacitor
becomes charged for any reason, excessive voltage can be applied
on one half cycle, and the output schottky can fail very quickly.
The dc blocking
cap can also cause an unwanted resonance in the phase-shifted
bridge, and can cause the wrong FET diode to conduct. If the
control circuit is designed properly, this capacitor, which
is large and expensive, should not be needed.
Proper
Instrumentation is essential. You need a high-speed storage oscilloscope
to capture single-shot events. Power switches can fail in a
matter of tens of nanoseconds, and you need to be able to see
these failures.
Proper
grounding and probe connection of the oscilloscope is also
crucial. For high-speed events, remove the ground lead from
the scope probes, and connect to the circuit with a very small
loop. Differential probes are not recommended.
