Thank you Ryan1 for your prompt reply.
More analysis of the capacitor that exploded on one of our MCP23017s shows it was the 47uF 25V rated tantalum capacitor associated with the 5V step-down converter on the front left of the board. This board did not have any other I2C loads connected to its output I2C connector. There were no solenoids connected to any of the channels. This was just a lab test to test the timing of the controls to each channel. All channels operated normally until the failure of the cap,
At the time of the failure the board was in operation for many days with input voltage from a Li-ion battery that cycled from about 16V to a maximum of about 18 volts at the time of failure.
We need a solution now to operate our many MCP23017s reliably. Do we really need to limit the input voltage to 15V ? This was not clear when we first purchased these boards. What if we limit the board’s input to 16V using a well controlled and filtered regulator? Would there be any advantage to adding loads to the output I2C 5V line. Other boards that have many NCD I2C sensors attached have not failed in the last 2 years.
Considering a custom design - Ideally, we would like to have this board in a version that can work with a Li-ion battery that cycles from 17.5V to 21V Max.
It would be nice if we could get
- an input voltage up to 24V using a similar buck converter to that shown on TI’s LM2675 data sheet with fixed resistor feedback (page 19)? Or
- a version of this board without its limited 5V regulator or if we could bypass it completely? If so we would use a separate 5V buck regulator from 3rd party.