I'm performing a random response analysis with FEMAP 11.1 and NX Nastran on a model with a very high number of modes (about 850 from 0 to 2000 Hz). The PSD spectrum is in 20-2000 Hz range. The modal frequency method gives me a result with a modal critical damping of .03, although I had to use only the modes up to 2000 Hz and to remove a lot of points from the frequency table to reduce the computation effort and the need of RAM. I also tried the direct frequency method, that should avoid approximations, with damping in the format param,G,.06. The results should be equal to the modal frequency method, but seem to be completely wrong in both acceleration field and absolute values. Could anyone help me?