We're using a buck switcher, LM2574HV-5.0 in SOIC package on a custom PCB. The load is less than 200mA and the input voltage can vary, but most testi ng as been at higher input voltages, around 55V. The IC has a max. input v oltage of 60V and absolute max of 63V so this should be fine.
We were careful about component selection and when possible, went with the recommended parts in the datasheet. We started seeing failures where the i nternal switch shorts closed, which absolutely fries the load. The load is a microcontroller circuit, with the part causing the harshest load conditi on being a small relay.
In an effort to track this down, we connected several boards to a power sup ply and a device that cycles power on and off. The only units that have ev er failed are potted. The un-potted units never fail when the input voltag e is kept within spec.
This is bizarre to me. The potting epoxy we use is very thermally conducti ve, a very good dielectric, and has been used for years in other products. The best I can come up with is the chip overheats, as it does get quite wa rm with such a high input voltage. But, surely open air is worse than the potting when considering cooling effects. Even so, the chip supposedly has thermal and overload protection.
I have dug out the potting and removed the faulty chip in a few units. The output pin isn't hard-shorted to Vin, but a diode check shows about 0.3V b oth ways which is much different than what a new chip measures.
Any ideas? We also have power MOSFETs on that same board and when potted, are able to dissipate a good deal of power without failure. However, the F ETs are not carrying any current when the switcher fails.