Abstract:
To increase the areal density (AD) of data storage systems, the bit-patterned media recording (BPMR) is considered a promising candidate for next-generation data storage ...Show MoreMetadata
Abstract:
To increase the areal density (AD) of data storage systems, the bit-patterned media recording (BPMR) is considered a promising candidate for next-generation data storage systems. In BPMR, magnetic islands are used to store the user data. Therefore, these magnetic islands are positioned close to each other to improve AD. However, this results in a degradation of the bit-error-rate (BER) performance of BPMR owing to 2-D interference. The 2-D interference includes inter-symbol interference (ISI) and inter-track interference (ITI) from cross- and down-track directions, respectively. To address the 2-D interference, the ISI/ITI estimator can be used to remove ISI/ITI from the received signal. Therefore, this study proposes a method to enhance the accuracy of estimating 2-D interference using a long short-term memory (LSTM) for BPMR systems. LSTM is developed from the recurrent neural network (NN) to address the challenge of gradient vanishing. Simultaneously, LSTM can be used to estimate the relationship between the samples in the training process. Therefore, LSTM can predict the interference from the adjacent bits in BPMR systems. In the proposed model, the LSTM was used to predict the ISI and ITI (interferences from adjacent bits) in the BPMR channel. Thereafter, ISI and ITI were removed from the received signal, which is the output of the 2-D equalizer. Consequently, the equalized signal retained only the 1-D interference that can be easily detected by the 1-D Viterbi algorithm (VA). In the simulation, the results indicated that the proposed model improved the BER performance of the BPMR system compared with conventional methods.
Published in: IEEE Transactions on Magnetics ( Volume: 60, Issue: 9, September 2024)