Initial pressure and transverse velocity fields to initialize turbulence model
 

Surprisingly, I have certain difficulties in unsteady simulations of turbulent boundary layer over flat plate using Spalart-Allmaras model.
If I fix the velocity field and calculate eddy viscosity, the final field is realistic. Moreover, if I solve energy equation for temperature then, using constant Prandtl number and eddy viscosity obtained, heat exchange is accurate. So, probably, realization of Spalart-Allmaras model equation is OK.
Solving continuity and momentum equations was tested for laminar case and it is OK too.
But: if I calculate first eddy viscosity for the fixed velocity field and then begin solving momentum and continuity equations, the fields escape from the correct solution. The reason is that the initial pressure field must correspond to those of velocity and eddy viscosity, so that the solution is stationary.
Ok, the question ;) : How can I find this initial pressure and, perhaps, transverse velocity fields? At the moment I'm making use of solving numerically continuity dv_x/dx+dv_y/dy=0 to obtain transverse velocity and then x-momentum in the form dp/dx=... with p(0, y)=const for the pressure field. This results in pressure variations in transverse direction and program falls at the very first time step. Is there any other way to find pressure field or I should examine just the approximation of these equations?