How to speed up the inverter simulation based on blocks of matlab

I have a 3 phase inverter simulation based on block ready from matlab (IGBT) .

The simulation is so slow, After runing the solver profiler it lead me to the most block crossing 0 which is a relational operation block that is inside sub blocks responsible for creating sawthoot signal and comparing it with a reference to generate a voltage for inverter (PWM)

My questions: knowing that the block is a matlab block , how it is possible that it takes too much time ?

Question 2 : Some people replacing inverter simulation by using controlled voltage source instead !

which take less time, is this technic effective ?

Photo attached

[Raghava S N]

Hi @YASSER, could you please share the model you are working with? That will help in better understanding the issue.

[YASSER]

Dear @Raghava S N all simulink pre-built blocks of PWM are zero crossing problem

Matwork has provided a faster equivalent block in a new version (2020B)

i do not want to upgrade my version 2020a

So i have to find a solution

here is a photo

Answer 1

[Joel Van Sickel]

hello Yasser, we need the full model to help with this problem. It could be solver settings, it could be the model, or it is possible that your model is actually simulating well. If you simulate the switching of an inverter, you must have a zero crossing for every time the PWM changes. This is true regardless of software or approach if you want to model the actual switching. So the zero crossings just tell us how many switching cycles that you are simulating. If you want to simulate a lot of switching cycles, it is going to take a long time, that is the nature of power electronics models. If you want to really speed up the simulation, you can use average models instead, and NOT simulate the switching behavior. This is a valid approach for when you are doing the control design or system level simulation. It is not effective for detailed design or power quality analysis. If you want to try and speed things up, you might try using a fixed time step that is your switching cycle time divided by 100. If that doesn't speed things up, you are better off with a variable step model.