Why I need to use output rate transitions for an atomic subsystem to speed up simulations? Which is the correct use of rate transitions?

Hello,

I'm currently using an atomic subsystem block combined with a simulink proteced model.

The atomic block is my attitude controller, while the protected model is the system attitude dynamics. I cannot modify the protected model.

The attitude controller is modeled with an atomic block, since I need to simulate the real frequency for wich the attitude control is running (which is 1 Hz). I use input rate transitions, as also suggested by simulink debugger, to adjust the rate of input signals, since the I'm using 1 Hz for the attitude controller and 20 Hz to propagate the system dynamics (computed by the protected model). Solver setup are fixed step ODE4 at 1/20 s.

If I start the simulation with this configuration, the simulation is extremely slow: for each simulink refresh, the simulation is propagated of 1.4 s each time.

Then, I added rate transition also for atomic block output signals, with a sample rate of 20 Hz, the same as the simulation sample time. According to this configuration, the simulation runs much faster: about 54 s for each simulink refresh.

The problem is that I obtain very different simulation results, depending by which configuration I'm using. Since the attitude control law is based on an LQR, I must use different weighting matrix Q dependig by the rate transition configuration I'm using in order to obtain good simulation results.

How it is possible?

Which is the correct use of input and/or output rate transitions for an atomic (and non-atomic) subsystems?

Than you all!

Matteo