one of the more involved designs I've done in a while
rather satisfying when you get to make stuff symmetric though :D
@gsuberland I’m curious, how do you learn to build complex stuff like this? I’m starting from the point of a barely remembered A level in electronics, unused for anything more complex than wiring a temperature sensor up to an ESP8266 for the last 20 odd years.
@jon it's mostly practice and learning about the building blocks in isolation, and working from reference designs as a starting point when you're working with a new IC.
while this schematic looks complicated, it's largely just a bunch of common configurations of parts stuck together, with some parasitic inductances in series to account for how it would behave on a real PCB.
@jon so for example here I've got an SFFM voltage source to emulate spread spectrum switching noise, in series with a pulse source which is configured to act like a DC supply with a ramp-up time.
then there's a pi filter (cap, inductor, cap) to filter out supply noise. this is useful to simulate here because the response of the filter may lead to oscillation as the LTC4370 tries to balance power across the two supplies.
then it's mostly the LTC4370 reference design, with some tweaks.
@jon the main changes I made were tweaking the compensation capacitors and using parallel FETs on the second stage, adding a pi filter stage between to help avoid the two sets of control loops from over-compensating and oscillating, tweaking the current sense shunt resistances slightly, and the enable circuits (NPN threshold circuit + RC filter). the output side just simulates a load.
avoiding oscillation in a two-stage LTC4370 design is remarkably challenging, in part because it's hard to tell which part of it starts the oscillation. on top of tuning the compensation circuit to the input capacitance of the FETs, I had to add a delayed enable on the output stage to keep it stable, which isn't something mentioned in the datasheet.
decoupling the two stages with a pi filter also helped a fair bit, since you don't end up with the two control loops fighting each other and reinforcing the oscillations.
interestingly they don't do this in the reference design from the appnote. I suspect they just ignore the scenario of under-damped output filters on power supplies, or hope that the dI/dt of load transients is small enough not to trigger oscillation. or maybe they just never tested it rigorously.
@gsuberland reference design? Tested rigorously? Hahahahahahahahahahahahahahahahahahahahaha!
@jpm yeah I was definitely rolling my eyes a bit while writing there haha