The aim of the design method is to find a compensator that results in a ‘ basically non-interactive ’ closed-loop behaviour, for more than one model that claims to represent the same plant. The idea is to construct a nominal diagonal model, which does not necessarily belong to the given family of models, and to consider the members of this family as variations around it. If the precompensator sought is diagonal then the design can be split into a number of single-input single-output designs which are more easily carried out.