Is your load inductive?
View attachment 300966
ADDED:
Only full bridge is resistive to HV spikes.
Bidirectional power switch does not.
Even first transition from ON to OFF may destroy FETs.
EDIT:
Simulation is reconstructed.

I restimulate it, the waveform has a high voltage spike when the bidirectional switch turns off. Do you think I should use a higher breakdown voltage MOSFET? or do you think the layout problem makes it self-turn-on? According to the experiment waveform, when the bidirectional switch turns off, there is no high voltage or current spike. Do you guys have any ideal about it?

 

FETs already are broken, therefore you can not see spikes.
Spikes are absorbed by weak FETs.
Use snubber R1, C1 (see second simulation).

 

Hi guys, is there any way to turn on/off a GPIO signal at the dead-time period of PWM? I used the ePWM event trigger of TI DSP 28335, but it seems it doesn't work well.