radau1 solver fails on a simple model

[anonymous]

the solver "radau1" fails on the following simple model:

model Test1
 parameter Integer n = 1;
 parameter Real x0[n] = {1};
 Real x[n](start=x0);
 parameter Real a = 1;
equation
  der(x) = a*x;
end Test1;

For me the solver runs forever without any solution.

[Martin Sjölund]

radau3 works fine. radau1 does get stuck though...

[Vitalij Ruge]

No wonder.
Probably, we should rename radau1 in implicit Euler.
(The eigenvalue from the ode is 1. So the solver is not inside the stability region.)

[Lennart Ochel]

Well, it is okay that radau1 cannot handle this model. But the solver should not get into an infinity loop. The simulation should stop at some point and print some meaningful message.

[Lennart Ochel]

But there is an error. If I change a to -1 radau1 still gets into an infinity loop.

[Vitalij Ruge]

don't forget the stepSize ;) for stability region (z = lambda*h).
I think it's "only" stability of the solver.

Infinity loop should be fixed now.

[Lennart Ochel]

Sure, the step size needs to be taken into account. But for the stability of radau1 (stage 1 as well as stage 2) it does not matter in the case of negative real eigenvalues. It should be fine for each step size.

[Vitalij Ruge]

You are right.
I have changed the linear solver in Kinsol.
Now, it's should work.

[Lennart Ochel]

Then, the ticket can be closed. (fixed in r16258)