I am having problems with the caps failing and shorting input power. I
have found a lot of pages on Tantalum Capacitor failures, but very
little satisfactory fixes.
Here is set up: I have an 18.5 volt (7.2 amp) linear regulated power
supply. It is connected to a DC to DC power supply board with a number
of linear and switching power supplies. This power supply board is
connected to another electronic board. The overall electronics draw
about .5 amps at 18.5 volts. The caps that fail are on the input of the
DC to DC power supply board. I am using 47 uF 35 volt Tantalum Low ESR
capacitors. Actually, I have three of these caps in parallel and two
more in parallel individually shielded by a ferrite bead. All fail
How is the problem occurring? I have many other boards in a similar
set up that the caps are not failing, but in this setup the caps are
failing. I believe it is because this device has can be disconnected
between the regulated power supply and the DC-DC power supply.
Tantalum capacitors are sensitive to instantaneous voltage changes
(dv/dt). This means that they should never be connected to a
mechanical switch or in this case hot plugging the device. (See
1. Shoot and kill all service techs and clients that hot plug device.
Unfortunately, this is not an option my manager will let me pursue.
2. Exchange the Tantalum caps with Aluminum electrolytic. I see this
advice mentioned a lot, but the major problem is the life of
electrolytic. Also, the manufactures also don't recommend this, of
course it is not to their advantage... (See
solution recommended is to fix the problem and not switch to a
different component with other problems. Also, my switching power
supplies on the DC-DC power supply board recommend tantalum caps... I
don't want to change to Aluminum electrolytic.
3. Add a series resistor between regulated power supply and rest of
input of DC-DC power supply board. The idea here is to limit inrush
current. But that resistor would need to be very high wattage (18.5
volts at .5 amps)... I don't think that is a feasible option in
4. Add a MosFET in series with input to DC-DC power supply board. The
idea here is to have a RC startup delay. (This is recommended
under section "keep a
handle on inrush current.") This seems to be the only viable option.
I was wondering if there is anyone that has tried this. If so could
you please discuss results? (This article also recommends using slow
turn on for power supply. This is already incorporated as mush as
possible on the board in question.)
5. Other options? Please let me know.