Post by John LarkinOn Mon, 6 May 2024 01:58:15 -0400, "Edward Rawde"
Post by Edward RawdePost by John LarkinOn Sun, 5 May 2024 14:42:30 -0400, "Edward Rawde"
Post by Edward RawdeDoes anyone know of a P-channel MOSFET, roughly equivalent to
XP233P1501TR-G which has a model I can use in LTSpice?
XP202A0003PR-G would be useful too.
...
Post by John LarkinPost by Edward RawdeDon't worry it won't be going into production any time soon.
That works in simulation, but the hazard of ac coupling into the fet
gate is that it has a hangup state, the fet on hard forever and
nothing limiting the current but 12 volts into the load.
I agree that there's plenty of scope to get 12V out instead of 5V.
I had the comparator with it's own 5V source for a while so it wasn't
dependent on creating its own supply rail.
I also agree that a real circuit may not behave like the simulation and may
just sit there producing 12V.
But that's a useful lesson in the difference between the simulation model
and the real thing.
A crowbar and a fuse is one way to fix the hanging fet problem.
And if it's an LED lighting circuit it's likely no-one would care. You just
have to buy a new unit.
Either that or just a suitable fuse should take care of over current as is
sometimes done to stop people paralleling too many christmas tree lights.
Post by John LarkinIt also grunts at startup, fet full on until the +5 settles. That
could cave in a wimpy +12 supply, and it may never run right. The
diode will leak more than the fet gate, so the static default is fet
on. Just remove "startup" to see the hang state.
You could power a comparator from +12, dc couple into the gate, and
slow down the startup. Something like that. Using all discrete
transistors would be fun too.
One of the goals is to see what the minimum component count/cost is.
Putting in a beefier inductor and schottky says it can do 5V at 5A no
problem as long as the fet doesn't stick on all the time.
I haven't looked at the efficiency but it looks like it's reasonably high.
I probably will do the comparator discrete to see what the minimum
comparator I can get away with is.
Post by John LarkinYou could just buy a switcher chip and copy the appnotes, as many
people do, but you wouldn't learn anything. I know a guy who started a
pretty cool company just copying eval boards, but then he started day
trading....
There's no fun in going back to the way I did most designs 20 years ago.
Post by John LarkinThe Brat assigned me to do a giant Spice sim this weekend, a dual
isolated buck switching power supply that's in trouble, so I'm almost
Spiced out.
https://www.dropbox.com/scl/fi/kzzjdfijpacf2qlch2gyq/P941C_lores.jpg?rlkey=49mv67p1c0zvu9c5v5tdftk3f&raw=1
Your circuit also looks like it takes a while to get going. And plenty of
post filtering.
I always took the view that if it can be done perfectly fine without digits
then leave the digits those who can't cope with anything other than digits.
But I have done power supplies with digital control loops before.
What I really want is a low cost 12V to LED current converter but then where
would I get the comparator supply from?
Maybe directly from the input side, or another winding on the inductor.
Post by John LarkinThe sim is all analog, but the actual control loop will be digital, in
a small FPGA. One tricky part is that the two supples can be used
independently or in parallel, and the loads can be anything, with
remote sense.
It would be interesting to put two of my circuits in parallel.
I think they would auto synchronize.
One of the issues with my circuit is what happens with no load.
68 ohm half watt fixes it in simulation but what happens in reality is
anybody's guess.
