Thanks. My problem has 9.300 sdp variables and a bunch of constraints. If I minimize the problem down to 1.100 sdp variables the solver can find a solution, but with the more specific case it just returns NaN's. I am also quite sure, that the bigger problem has a solution, as I only change the dimension to include more information, which therefore is closer to the original problem, which has a solution.