Simulink does messes up step sizes (and doesn't do what I tell it)

Hi all,

I want to simulate a variable-step size system. The model includes a coupler to a circuit simulator and hence transmits the current step size to the other simulator on each step.

When I run the simulation, I get errors like these on the other side:

ERROR (SIMLNK-57): [Internal] An invalid message was received from the AMS simulator: NextTimeHit must be in future, NextTimeHit=3.875e-07, lastTime=3.875e-07, now=3.875e-07

It seems that simulink makes too small time steps, probably below numerical precision so they arrive as identical numbers on the other side. Alright, so I set:

Min step size: 1000e-12

This should prevent Simulink from making too small steps. But the problem persists!

Now I just remove the coupler and observe the difference in time steps (I export "Clock" to workspace):

plot(diff(simout_time));

This is the result:

Clearly, there are steps which are below the numerical precision (3.388e-21) although I told Simulink that the minimum step should be 1e-9 !!

How can this be fixed?

PS: Taking a fixed-step solvers works as expected but this is not what I want.

[Joachim Kahr]

Has anyone found a solution to this problem?

I face the same problem and are not sure if this problem is only simulink related or if the communication over the socket leads to this problems.

KR,

Joachim

edit: setting the sample time of the coupler to a fixed step size is kind of a workaround as the min step size of the simulator does not affect the step size inherited from the coupler. anyway, this is a tradeoff between simulation time and accuracy....