Welp. Version2 is done. And it's pretty damn close to working. But... not quite.
See it here.
*Note that I removed R5 and R7 as well.
Everything is switching cleanly. Nice square waves input. Nice square waves output. Mosfets that should be OFF are, in fact, staying off. Nothing getting remotely warm.
Only problem: my high-side V-gate-to-source is only reaching ~3.8V. My mosfets need 4.5+ to turn on (they are logic type). I find my mosfet total gate capacitance to be somewhere around 7000pF (31nC at 4.5V). The driver datasheet says Cboot should be > Qgate/Vdroop. So that's 31e-9C / 0.1V = 3.1e-7 or 0.31uF. It also says Cboot should not exceed 1.2uF. I am using 1.0uF, so I'd say that should be plenty.
Highside: Gate-Source - Voltage Readout
HighSide: Gate-Source - Freq Readout
LowSide: Gate-Source(Gnd) - Voltage Readout
LowSide: Gate-Source(Gnd) - Freq Readout
If I watch Cboot with the highside Mosfet held low, I get a steady 4.5V. Why would this not go up to the full 5V supply of the driver? Some kind of internal drop in their level-switcher-doodad?
If I watch Cboot with the highside Mosfet PWM active, I get a steady 3.8V. With the scope set to AC-coupling and 10mV scale, I find only 25mV transients matching the repetition of the PWM (109hz). edit: (The transients are a bit weird, though, because the voltage increases very slowly over the span of PWM=OFF and then drops rapidly at the GateCharge Event. Almost like the driver has a large internal resistance in it's Cboot-charging-thingy. (But I need to setup a 2nd probe so I can trigger off the PWM rising edge to really confirm that my timing guess is right).
To eliminate the possiblity of Cboot being too small, I laid another 1uF directly on top of it (paralleling to 2uF) and I get identical results.
I then tried laying a 22uF on top of it. With identical results.
I tried removing R5/R7 (the Gate-Source 10k resistors on high side) to eliminate them as leaking too much current. Operation remained identical.
Increasing PWM frequency makes things worse.
Changing Duty Cycle has negligible effect. Even a 1/255 duty cycle results in only 3.8Vgs.
Load is still a simple resistor.
Ideas? Next time. P-Channel highsides