I lowered Rgoff from 100 ohms to 33 ohms and voila, the gates shut down faster, the bridge is still shorting, but the amount of shorting is now minimal!!!
I tested it under realistic conditions, such as OCV of 85 volts, the ripple was completely acceptable, with the inductor in series. See below.
This prototype is almost the same as Prototype 1 (read about it here), except that I lowered resistance of the off resistor. That has the effect of turning the bridge's sides off a little faster. Therefore, it would be expected to spend less time shorted, which happens when all sides are on. The old Rgoff was 100 ohm, the new Rgoff is 33 ohm.
The results you see on the scope are results with an air core inductor in series with the supply. The air core inductor is simply a coil of
12 gauge Home Depot wire, about 300 feet.Some results:
The value of current of cross conduction went down almost to zero (but not exactly zero). That means that there is still cross conduction, which I want to happen, but very little. The amplitude of ripple under power (when it carries current) is about the same as the bridge voltage The amplitude of ripple without power is less than the bridge voltage. I measured DC voltage and AC voltage with a voltmeter. The AC voltage is always below the DC voltage. I am still not using a resistor in the snubber.
The bridge makes annoying audio noises, no doubt due to magnetic forces acting on wires.
(Click on the thumbnail images to enlarge)