Hello, I'm trying to convert a LQG-problem into the equivalent H2-problem. I'd like to do this without first removing the x-u cross terms in the penalty function. Thus I seem to be facing the problem of factorizing matrices where the factors have prescribed dimensions. For details on the matrices see the algorithm section under h2syn. Is there a standard procedure for doing this that I have overlooked? I have tried to solve the problem by applying fsolve at fun = X'*X-R, where X and R are matrices, but the problem seems to be close to semi-/in-definite because the solution fluctuates a lot with varying initial starting point.