Better solver for simscape power electronics simulation

Hello,

I have an LLC simulation file in which I need to do power loss breakdown analysis.

I built my model with realistic components(parasitic capacitance, resistance etc.).

My problem is about selecting the correct solver configuration.

At first I was using local solver "Backward Euler" with 2.5 ns sample time. However, I realized that there are something wrong when I am try to find the power loss across an inductor. The power loss calculation ( average of I(t)* V(t) over a switching period ) is not correct.

Then I try to use "Trapezodial Rule" as the solver type. With this the type simulation time has increased a lot but the power loss calculation becomes correct. But the transient of the current and voltage curves distorded.

At last, I used ODE 23t with max step size 2.5 ns, which results significantly better than "Backward Euler" but worse than "Trapezodial Rule" results. But the transient of currents and voltages are better than "Trapezodial Rule". Also simulation run time is shortest in this solver.

So I am aiming to with ODE 23t but is there a way to enhance the accuracy of ODE 23t solver?

I already tried to change "Relative tolerance", "Solver reset method" but not really upgrade.

Any help would help a lot.

Thanks in advance

Answer 1

[Dr Narayanaswamy P R Iyer]

In case of uncertainty as to which solver to be used, select AUTO solver selection mode from the drop down menu.

Answer 2

[Joel Van Sickel]

You can set your absolute tolerance to be the same as your relative tolerance, and uncheck the option for absolute tolerance. Typically setting both of these to 1e-4 or 1e-5 should be more than enough to get good accuracy. You can also switch to DAESSC for it be very slightly more accurate but also slower. There shouldn't be a problem in how you are getting the inductor losses regardless of solver method. Can you provide screenshots or share a model of how you are determining that the losses are incorrect?