Repetitive control is a widely used technique for the compensation of repeatable error in systems that contain rotating mechanisms or repeat a trajectory. Generally, it includes delay chains and a low-pass filter in the positive feedback loop, which generate a periodic signal. The controller has typically been implemented in a plug-in fashion and designed heuristically with the simplest form of the filter. However, this design approach is somewhat ambiguous in the selection of controller parameters because of its influence over nonharmonic frequencies. Also, it leaves the possibility for further improvement. This paper presents an improved design method for the repetitive controller that provides minimum track misregistration (TMR) in a hard disk drive (HDD). For TMR prediction, the method identifies disturbances acting on an HDD and estimates servo performance, using the identification result. We have confirmed the identification and estimation procedure through experiments. In our method, first the basic tracking controller is designed and later the repetitive controller is designed in conjunction with a Q filter. A cost function based on Parseval's theorem, reflecting the servo performance as TMR, is defined. Then the servo performance is estimated from the identified disturbance, and the plant and designed controller's frequency response are modified as necessary by changing the parameters of the controller, whose optimization is carried out with a commercial nonlinear optimization tool. The design strategy facilitates the controller design by providing an accurate estimation for the attainable servo performance and design criteria under the optimization framework.