Problem with output capacitors of the DC/DC converter with TPS54260

1) What could be the reason for the explosion of the capacitor?

Identifying the cause of the failure may require some additional information.  The most common cause of failure in Tantalum capacitors is over-heating.  Two common causes of overheating are:  Excessive RMS current or Reverse Bias. 

2) Is there an error in our circuit or layout?

The first thing I would check would be to see if one of the tantalum capacitors has been installed backwards.  Tantalum capacitors are polarized.  They operate as a capacitor when DC biased in one direction, but offer a low-impedance when DC biased in the reverse direction.  If one has been installed backwards, it would be appear as a short across the ouptut voltage, carry a lot of DC current and eventually fail.

The second potential cause would be excessive RMS current in the capacitor.  With 80mOhms of ESR, RMS current can heat up the capacitor fairly quickly.  For this, you'll need to look at the expected output ripple from your design.  for 48V to 12V the inductor experiences -12V for 3/4 of the switching period.  The peak to peak ripple current will be 9V-s / (L * Fsw)  where L is the inductance in Henry and Fsw is the switching frequency in Hz (or uH and MHz if that makes the numbers easier for you)

The RMS current of this triangle waveform is 1/sqrt(12) of the Peak to Peak value  (0.288).  If this was the cause, I would expect a rapid failure of both capacitors, however if the first capacitor to fail is failling as a short (very common for solid tantilum capacitors) it may be protecting the second device from failure.

Unfortunately the AVX datasheet isn't providing a RMS current rating, but using a similar KEMET device as a design guide, we have an estimated RMS current rating of 1.36Arms.  With 2 such capacitors in parallel that should allow upto 9.4A pk-pk in the inductor, which is very high for a 1.3A output, so that's unlikely but still worth verifying.

3) Did we use the wrong type output capacitors? If yes, it must happen with all converters, right?

With proper compensation (COMP to GND Resistor and Capacitors) there's no reason the TPS54260 shouldn't be able to support a tantelum capacitor, however you may want to revisit the compensation design to ensure you don't have a stability issue.  While the ripple current is quite low, a low-frequency instability that tries to force a low-frequency sine wave on the ouptut capacitors could easily force more than 2.6A of RMS current on the ouptut.    Make sure your loop design is accounting for the ESR zero of the tantelum capacitor around 43kHz, without a compensation pole to counter this zero, it could result in instability.

4) Did you hear about such failures, that the output capacitor begins to smoke?

Yes.  I've had tantelum capacitors bubble, smoke, discolor, explode and even burst into tiny flames on the board.  As I discussed before, this is typically a result of over-current and happens most often due to reverse voltage bias on the capacitor.

5) Unfortunately, we are under extreme time pressure, because we have to deliver the systems. Which measures do you suggest?

First, check the assembly and make sure the capacitors are installed correctly.

Second, check the design inductor ripple current, inductor saturation rating and compensation stability.

6) How is the probability that the other modules fail later by the customer? How do you assess the situation?

If the problem is an assembly issue with some capacitors installed backwards, they offer no risk to field failures because they will all fail on first power-up.  If the issue is a marginal design due to instability, excessive ripple current or inductor saturatrion, they'll offer a high risk to field units as devices heat-up and age over time.

Please let me know with a reply to this message if you identify the issue, or if you need more advice.  To help more, I may need details of your design.