Havn't yet plugged in everything needed for this active balancer simulation.
Reality wants more cells, but would clutter the drawing and slow LTSpice.
The theory of operation (if no grotesque errors were made) is that a brief
pulse would cause all mosfets to turn ON hard simultaneously. Isolated
windings sharing the same transformer core would AC couple all cells to
an average voltage, thus achieving balance by the forward method.
The pulse needs to be short enough not to saturate the core. And there
needs to be a core reset before another pulse can repeat that process.
Reset comes in the form of flyback, commutated by one or more of the
Whichever cell (or the capacitor on top of the stack) might measure the
lowest, gets whatever magnetic energy was leftover. The voltage of the
weakest cell may be more than a Schottky drop below average. So, we
need to allow a little more time for reset. Maybe twice as long is safe?
No problem of too much reset time, as that automatically stops when
the magnetic energy is gone.
Assuming the pulse train is alive: The fake cell on top, the capacitor,
reflects the voltage of the weakest real cell discovered by flyback.
Thus we only need watch this one place to determine when to shut
off the pack. The isolated gate driver was selected for convenient
2.5V primary side undervoltage lockout. We don't feed it any signal,
but the input hardwired ON, as only the UVLO feature is needed.
It is normal for N-Channel Mosfets to have a body diode in parallel.
That diode would allow the pack to be recharged, even in shutdown.
I wind my own pulse transformers on JBWeld cores. Sometimes cheat
and glue up blocks of ferrite, rather than work with a blob that takes
much as six hours to fully set and sticks to every mold I've ever tried.
Don't go thinking I could never find a custom 1:1:1:1:1:1:1:1:1:1:1
transformer wound for 10S, cause that's a solvable complication.
JBWeld's ferrosilicon steel filler is good stuff to 1MHz. Only a little
more permeable than air unless you add extra filler. But enough to
keep flux lines in the gap where they belong. So much distributed
gap, you could not possibly saturate JBWeld alone. I shave a slice
of hardened weld a little bigger than the air gap I might calculate,
and stack it in my ferrite glue-ups. JB is indestructible, but ferrite
blocks tend to be fragile. I'm looking at other block materials...
So again, all of the real cells (and not the capacitor) should average
together on the forward pass. On the flyback pass, only the weakest
cell (this time including the capacitor) receives whatever is leftover.
Forward for fast balance. Flyback for reset and weak cell discovery.
All untested assumptions, so don't go crazy trying to building a thing
that probably still has errors and isn't even a complete drawing yet.
Hoping someone might spot an error in my plan and clue me back
onto the correct path...
- CheckForErrors.png (30.93 KiB) Viewed 32 times