I had tested the new buck converter arrangement with a surplus 17 volt laptop power supply and all was well, but I wanted to also try it out with a 24 volt model. That's where the trouble began.

The instant I would plug it in, the TPS54231 (and I even tried a few TPS54331s) would blow, throwing a short across the input power. One of them even quite literally exploded, ripping the input power trace off the board.

Well, the original configuration was a reverse biased rectifier diode across the input to protect against reverse polarity, and an unpopulated polymer input filter cap, with a ceramic bypass cap for the buck converter input very close to the catch diode.

What apparently was happening was that the output of the power supply had some inductive component that led to overshoot when it was powered up. The 10 µF ceramic cap wasn't able to absorb the overshoot, so the voltage went higher than the 30V absolute max of the chip, and either the destruction threshold is not much further up from the absolute max or the overshoot was huge.

The solution is to populate that poly cap and replace the diode with a 24V TVS diode. With that, I've been able to power cycle the circuit with that suspect power supply 5 times in a row without any trouble. Before, it would have likely died the first or second try.

Fed from 24V the noise and ripple appears to be just about the same as before, and well within the spec of the 5680A.