Fixed step size

It depends on solver convergence if model can converge at 1e-3 then specifying hmax=1e-3 is sufficient to make solver to take this fixed step.

Solver always attempt smaller step sizes irrespective of hmin & hmax to find convergence during trasient simulation.

Ah, thanks for this information.

Is there a way to stop convergence iteration to fix a step size?

The integrator performs the following steps to converge to a solution:

1. Based on previous converged steps, it tries to predict the next converged solution by regression analysis of a certain order polynomial.
   
2. If the 'predicted' solution is not within the specified error tolerance it uses the Newton-Rahpson method to find a numerical solution within the integration tolerance.
   
3. If the Newton Rahpson method cannot find a converged solution within a specified number of iterations, then the solver will try to predict the solution using a smaller time step and/or a different order for the polynomial.
   

The above process is repeated until the time step size is continuously below the minimum time step size.

Hence, the only way to set a constant step size is to reduce maximum step size until all the time steps become constant and this is dependent on the problem at hand.

Also note, the maximum solver time step is different than the output time step.  'Print Interval' controls the output time step and 'h_max' controls the maximum solver time step.