Diagnosis of Nonlinear Stochastic Dynamics of Active Slider in Nanometer Spacing

Abstract

The hard disk drive (HDD) industry continues to meet the ever-growing demand for storage capacity by increasing both the number of tracks on the disk surface and the number of bits recorded per track. Characterizing the complex dynamics and sensing of the position error signal (PES) play very important roles to enhance the precision and accuracy achieved by the track-following servo mechanism. In the processing of PES signal with noisy data, diagnosis of signal and noise types is critical for filter design. Conventional filtering techniques such as the well-known median filter and mean filter cannot be expected to achieve satisfactory performance due to either the lack of the information of noise types or the presence of mixed noise in PES signal. This paper characterizes the complex nonlinear stochastic dynamics and develops a data mining approach for the noise type diagnosis and proposes a fuzzy filter design for enhancing the quality of PES signal. The experimental results demonstrate that the proposed fuzzy median-mean filter outperforms existing filters, particularly for dealing with the PES signal corrupted by Gaussian noise plus impulsive noise.