In this particular case I think part of the issue is the very high turn-on delay. With a switching period of 50 us the turn-on delay was set to 5usec or 1/10th the period. Perhaps this is a data entry error? Indeed this can influence the sampling behavior leading to a DC offset. You are no longer sampling at the midpoint of the waveform, but offset by a factor introduced by the dead-time. The figure below illustrates the concept.
> How will the RTbox implement my c-script control? Will it do it in a faster step? How this Enable Synchronization affects my control procedure?
With the PWM synchronized to the model step, the C-Script will execute according to the sample time you specify. You will receive a warning if you have an invalid sample time. For a C-Script with a discrete periodic sample time, it will work so long as it is an integer multiple of the specified discretization step size. The PWM output will be updated at the next specified update period (carrier min, carrier max, carrier min/max) after the computations have completed within the model step.
Without synchronization, the model execution will drift relative to the PWM carrier. You will end up with a variable delay between the model step and the next update period. That is why for most rapid control prototyping applications synchronization is enabled. Within the synchronized model, it is possible to further adjust the synchronization behavior by shifting the PWM carrier and ADC sampling time.