Case of the Failing Switching Regulator IC

When I worked as the electronics design engineer on a medical device prototype, I designed an interface board as part of a larger project. The goal was to enable signal breakout from a Cyclone 3 FPGA development board to other development boards.

In addition to signal breakout, each board had different power supply requirements. To accommodate all of the requirements, I included a DC-DC switching regulator in the design to generate 5V DC, 3.3V DC, and 1.8V DC. I chose the LT3507 based on its small form factor and its highly integrated monolithic multiple switchers. Thereby, only a small number of external components were required.

The switcher was initially tested and characterized using a 12V DC power supply input to the interface board. I tested it multiple times. I also did support testing on the other development boards and the rest of the hardware.

After about two weeks, we decided to integrate all the hardware with the software and test the system. Given that the input supply required for another power control board in the system was +24V DC, and the LT3507 can accept a maximum input supply of +36V DC, we decided to split the input +24V DC to the interface board, as well. That consolidated the input power to a single source. So far, so good -- everything was falling in place.

Upon moving the interface board from one test setup to another, we had multiple occasions of power cycle. During one of the test setups, when some signals were being prepped to be probed, the 3.3V DC rail was shorted to ground due to a stray wire, and there was excessive current draw for a few moments. On further power cycles, the switcher worked intermittently, then finally stopped working altogether.

On further investigation, we determined that the switcher had indeed blown during the signal probing, and we replaced it. We tested the second switcher intensively for an hour or so at a supply voltage of 12V DC from a linear bench top supply. Once it was characterized satisfactorily, we moved back to the complete test setup, where we powered it with the +24V DC input from the linear bench top supply.

On the first power cycle, the switcher didn't seem to generate any of the rails, in spite of the fact that it had previously worked on the primary test setup for an hour or so. On further investigation, we determined that the second switcher was blown, as well. That was just hours after the first one was replaced.