ADuM4223 as H-Bridge Driver - Strange Behavior

Hello all,

    I am currently using two ADuM4223C half-bridge drivers in a permanent-magnet synchronous motor power driver. The ADuM4223Cs are employed to control an NMOS H-bridge, as shown in the simplified schematic below. All four MOSFETs have their own isolated +9V/-4V supplies. The bus voltage ranges as high as 350VDC, and output current is about 12A maximum.

    We are experiencing some very puzzling behavior at high H-bridge output voltages and currents. The output current will occasionally drop very suddenly to zero, and remain there for a period of time ranging from 100's of microseconds to several milliseconds. See the image below. In this plot, current goes from almost 8A down to exactly 0A in one 66.67 microsecond frame. The only realistic explanation is that the ADuM4223s suddenly turn off all four MOSFETs simultaneously, completely cutting off motor phase current.

    The question is, in the absence of timing/logic errors on the input lines, why would the ADuM4223s both suddenly decide to drop their outputs to -4V? I can think of a few explanations, perhaps you can chime in with your thoughts:

1. There is some kind of glitch on the disable (DIS) line due to the extreme levels of switching noise. This DIS line is controlled by a separate latch circuit, which I can provide if needed.
2. The ADuM4223 ICs are entering undervoltage lockout. This seems unlikely as the isolated supplies are solely powering these gate drivers, and our PWM frequency is 30kHz. There should not be a great deal of power dissipation in the drivers, and in any event it shouldn't depend on the H-bridge output voltage/current.

On thing is clear, there is quite a bit of noise all over the circuit. Just attaching oscilloscope leads to certain sensitive parts of the circuit is enough to cause failures and faults!