In the conventional design of disturbance observer (DOB), a first order low pass filter (LPF) is used to prevent algebraic loop and satisfy causality in the inner-loop. Although a conventional DOB has a good robustness, its performance is limited by the dynamic characteristics of a first order LPF. A DOB that uses a higher order LPF is called as a higher order disturbance observer (HODOB) and can be used to improve performance of a system. However, improving performance may cause robustness problems. Therefore, there is a trade-off between robustness and performance of a DOB based robust control system, and it is directly related to order of the LPF of DOB. This paper analyzes robustness and performance of robust control systems based on DOB. The relation between performance and robustness is derived analytically, and a new design tool, which improves robustness and performance, is proposed for HODOB. A general second order plant model is analyzed, and simulation results are given to show the validity of the proposed method.