A procedure to determine the standard controllability form of a matricial pair (A, B), using orthogonal similarity transformations, is proposed. All rank computations are performed via singular value decompositions. The procedure can be used to test controllability of the pair (A, B) or in the model reduction of linear systems.