Machine simulation models allow the use of embedded numerical computation techniques. The numerical model and computational accuracy is usually gauged by comparison with experimental measurements. In this paper a large hydro generator is modeled and the sensitivity of various model parameters is investigated by comparing the simulated results with experimental measurements from four units of the same design. The machine open circuit voltage was found to be very sensitive to the airgap and stator B-H curve. Moreover, for 2D simulations the effective length is also important because it should account for differences in the stator and rotor lengths, in addition to the presence of radial air ducts. A 20% variation in operation airgap between two machines of the same design induced over 18% difference in open circuit voltage. Reduction in permeability of the soft magnetic materials resulted in agreement between the numerical and measured results at saturation. Consequently for large machines, the B-H curves from the Epstein measurements are insufficient and should be adjusted accordingly.