DCDC output capacitor blown up.

hi

in my design, there is a 13A/5V DCDC step-down converter. Other DCDC converters use this 5V as input, and also there is a lot of direct load on it.
The capacitor is 330uF/10V tantalium SMD.

So, this capacitor blown up and made short circuit, when my colleague disconnected and connected back the input power connector to the system, when it was already hot (around 80-90°C) after few hours of usage.

so, what was the cause? how to prevent this to happen in the product?
use bigger Irms capacitor? (there is already in the next version)

This is from memory but I think that Tant's should be derated about 3 to 4 times the voltage rating.In other words you should use a 15V to 20V cap for your 5V output. I remember reading this in a Kemet application manual I think. For an 8v application I'm using 25V SMD.

They also may need a current limit resistor in series for start-up inrush.

Of course observe polarity.

Measure your RMS ripple current worst case.

I just noticed You are using only 1 cap for a 13A load current? What is your rms ripple? For 13A output it must be high.


Kemet has tons of info go here.

i have also 2 pieces of x5r ceramic 22uF caps in parallel.

I'm researching caps for an SMPS I'm building.I did come across this which may be releavant to your problem. From AVX.

"Question: I have read that manufacturers recommend
a series resistance of 0.1 ohm per working volt. You suggest
we use 1 ohm per volt in a low impedance circuit. Why?

Answer: We are talking about two very different sets of circuit
conditions for those recommendations. The 0.1 ohm per
volt recommendation is for steady-state conditions. This level
of resistance is used as a basis for the series resistance variable
in a 1%/1000 hours 60% confidence level reference.
This is what steady-state life tests are based on. The 1 ohm
per volt is recommended for dynamic conditions which
include current in-rush applications such as inputs to power
supply circuits. In many power supply topologies where the
di / dt through the capacitor(s) is limited, (such as most
implementations of buck (current mode), forward converter,
and flyback), the requirement for series resistance is
decreased."

You say that the cap blows during a warm start? Electrolytic ESR decreases with temp. Maybe the ESR was to low to limit the inrush to the cap?

The whole AN is here.


I've chosen polymer aluminums for my output caps. I've used them before a little pricey but excellent performance. Have you considered them rather then Tant's.

These

it blown up at hot start.
i was tinking in tantalium only.
you know, the ESR has maximum limits because of the output ripple voltage, so it can not be high. it has to be very low.

In another appnote from kemet, they say the ESR should be low, to decrease the power dissipation, caused by ripple current.

so, is it a contradiction? is it possible to balance these things:
-Vripple needs low ESR
-Pdiss needs low ESR
-inrush current needs HIGH ESR

Yes I know there does seem to be some contradictory info regarding appropriate uses (derating, esr etc) regarding tantalum usage. Hence the title of one of there AN’s “2004-10 Article Ripple Current Confusion 10/04 (116.3K PDF)”.

I personally only use tantalum when specifically recommended by manufactures of the components I use (LDO’s, vcc bypass). Using them as output filtering of an SMPS wouldn’t be an option for me because of the additional components that may be required, that being a resistor. They also have a higher failure rate I believe. Using the resistor wouldn’t increase the temp of your cap the power is dissipated in the resistor. That being said obviously this would contribute to ripple.

Have you at least tried a low value R in series with the cap to at least eliminate it as a possible cause? This wouldn’t be a practical solution but at least you’d know if that is what is causing the failure. If this eliminates the failure then you will be shopping for a different cap.

Are you using a quality name brand cap? i.e. Panasonic nichon etc. Learnt the hard way cheap caps aren’t worth the savings.

in the proto, there was a simple capacitor, bought from Farnell.
In the second version, I asked the production company to buy a very good capacitor (higher voltage, low esr, higher rms current.)

the blowup happened at the software testing guys, after 2-3 months of continously working on the board. so not at the first moment.

There is a discussion similar to this in Usenet. You can accesses it also using Google groups.

The thread is here in Google groups. (sci . electronics . design) I shortened the url.


There are a lot of experienced people there if you post good details I’m sure someone had a similar problem, and could offer suggestions.

To me it does sound like it’s failing due to high inrush, if as you say it only happens on a warm start.

Mind you, who knows what a bunch of software geeks are doing to it.