I encountered an unexpected boost-converter failure mode.
My beehive monitor runs from 5V, 0.25A USB power, or from a 3.7V Li-Ion cell. A pulsed 25mA LED current source runs off 14V, using a compact sot-23-6 TPS61040 boost converter, which makes its boost switching cycles only when needed. It takes 125mA average from a 3.7V supply during the pulse. E.g., with 250mA peak sawtooth charging currents. Here we have cycles of the '61040's 400mA peak-current setpoint.
For the 50ms between LED pulsing, I disabled the converter. After enabling for a new burst, the converter would quickly re-establish +14.0 volts. After spending time measuring dark levels, LED pulsing is started. But the TPS61040 failed to keep up, and the 14V bus could decay to about 11V, damaging LED output, before it finally recovered and enforced 14V.
The cause? The TPS61040 has a soft-start function. This is desirable, given limited current availability from USB power. It sets the ramp current to 100mA for 256 cycles, and 200mA for 256 more cycles. But the '61040 doesn't need a full 512 cycles to restore its output to 14V, which means when it was finally expected to deliver full power, it could not do so.
One solution is to stop disabling the converter, and leave it on continuously. Thankfully it only needs 28uA to run. But this means a PCB roach wire. There might be a software solution. Early operating regimes didn't have the problem. Making the change allows full operating mode freedom.