This paper addresses the problem of deterministic worst--case identification of non Schur plants in an open-loop setting. Thus, the proposed approach does not require prior knowledge of a controller that stabilizes the unknown plant. The main result of the paper provides an LMI based approach to obtain a nominal model of the plant as well as worst--case bounds of the identification error. The proposed method is interpolatory since it generates a model of the plant in the consistency set, and in the limit case of complete information converges in the l₂-induced topology.