A novel contact recording slider design for HDD, which continuously contacts a disk surface was investigated. The design of the slider which has a wide air bearing surface in the front and a contact portion in the rear, enables precise division of suspension load between hydrodynamic force and contact force under certain tolerances. Therefore stable contact between the slider and the disk is accomplished with small contact force, and wear of the slider can be minimized. Experiments show the small spacing deviation at the contact portion (<5nm) with small contact force (<0.2mN).
Influence of the metal film of the micro-cantilever on the mechanical resonant frequency has been investigated theoretically and experimentally. It has been found that the resonant frequency values calculated from formula deduced using the composite cantilever model consisting of two kinds of different material films almost agree with the experimental values. The composite micro-cantilever having the resonant frequency of 1.07MHz has been also developed.
In order to provide huge files, for example, MPEG movie files or game files, from a provider to users, ultra high-speed file systems have been expected. We have made a prototype of ultra high-speed file system based on a personal computer (PC file system), which is optimized for sequential reading. Two basic technologies, "Continuous DMA" and "Multi-Layer Stripe" are proposed. The prototype system can achieve effective performance of 80Mbytes/s by means of only 9 hard drives.
The DC component suppressing method, called Guided Scrambling (GS), has been proposed, where a source bit stream within a data block is scrambled into many kinds of random bit streams as candidates for the next channel bit stream, and one of them having least DC component is selected. The convolutional GS and the GF (Galois field) multiplicative GS are known as typical techniques in some GSs. The convolutional GS has high performance in the DC component suppressing capability, but has low performance in the symbol error rate because a bit error propagates an adjacent codeword before RS decoding. On the GF multicative GS, the error rate is lower, but performance in DC component suppression is worse, especially when easing the constitution of the hardware. In this paper, we propose the GF additive GS method which generates the better channel bit stream in both characteristics. We also analyze the spectrum and the average symbol error rate of this method with computer simulation, and show its good characteristics by comparing with other GSs.
In our PRML systems, we employ a noise whitening matched filter having minimum-phase to avoid emphasizing high frequency noise and coloring noise and combine it with Viterbi detector with path feedback. First, the minimum-phase, noise-whitening matched filter which consists made of a transversal filter and a low-pass filter is described. Then, the error rate performance is obtained by computer simulation, and the performance is compared with that of the combination of a minimum-phase, noise whitening matched filter and fixed delay tree search with decision feedback (FDTS/DF). The results show that Viterbi detector with path feedback can improve the signal-to-noise ratio over that of FDTS/DF.
A TCPR system for high density two track digital magnetic recording is proposed. In our system, a reduction in the channel rate due to convolutional coding is avoidable because a two track recording method improves the rate, and the number of ACS's in our system based on the syndrome decoding is fewer than that of the conventional Viterbi decoding. The results show that the proposed system keeps good performance in high density recording and saves SNR by about 3.6dB compared with the conventional NRZ coded PR4ML system at a normalized liner density of 3 and a bit error rate of 10^<-4>.
An application of the neural network canceller to the PR4ML channel with partial erasure is studied. First, a model of the read/write channel with partial erasure and a neural network canceller are described. Then, the bit-error rate of the PR4ML channel is evaluated by computer simulation and the performance is compared with that of the conventional neural network equalizer and Nyquist equalizer. The results show that the neural network canceller exhibits better performance than the conventional neural network equalizer and Nyquist equalizer.