CFD Online Discussion Forums - Time dependent solution

I try to simulate a naturally ventilated building with small openings in its sides. Air blows from an inlet far away from the building, enters through one side opening and leave through the opposite opening and also I have a recirculation of air in the space above openings. Heat and mass (H20) transfer are also taken into account. After a few iterations (2000) I have a converged solution with logical values for air velocity temperature and humidity inside the building. Then I try to see what happens if I patch in the area inside the building higher (than outside air) values for humidity and temperature. I turn the solver to unsteady I patch higher values for humidity and temperature and I sellect a time step of 0.01. Each time step needs less than 20 iteration to converge. But although at the first 3000 time steps I have the same (with the steady solution) airflow pattern inside the building after this point the airflow pattern is different onlyy in the area above the opneings. That means that above openings I dont have a recirculation of air like in the steady solution and the values of air velocty are very low. Any suggestions? Thank you in advance Thomas