Signal Processing for Bit-Patterned Media and Heat-Assisted Magnetic Recording Systems with Media Noise

• Main Author: Ng, Yibin
• Format: Dissertation
• Language: English
• Published: ProQuest Dissertations & Theses 01-01-2012
• Subjects: Applied Mathematics; Computer science; Electrical engineering
• ISBN: 9781267644022; 1267644028

To mitigate the effects of media noise, we propose an analytical approach to jointly design the two-dimensional (2D) or one-dimensional (1D) equalizer with one-dimensional (1D) channel PR target for BPMR systems with media noise, according to the minimum mean square error (MMSE) criterion with monic constraint. Simulation results show that the proposed 2D/1D equalizer with 1D target based approach achieves significant performance gain over the system designed without considering media noise. We further investigate the performance of staggered array against regular array islands for BPMR with bit-aspect ratios (BAR) of 1 and 2, and with different amount of media noise, inter-symbol interference (ISI) and inter-track interference (ITI). At recording density of 4 Tb/in 2, we find that staggered array islands with BAR of 2 offer better bit error rate (BER) performance and better tolerance to media noise. We also investigate equalizer design with and without jitter noise considerations for heat-assisted magnetic recording (HAMR) channels. Instead of an analytical approach, we use a numerical approach in this case. Similar to the case for BPMR, equalizers designed with jitter noise perform better than equalizers designed without. This performance gain becomes more significant as the jitter level increases. We propose a novel hi-directional pattern-dependent noise prediction (BiPDNP) detector to improve the performance of the HAMR channel under high jitter noise conditions. The BiPDNP detector utilizes backward linear prediction in the noise prediction process, as well as the conventional forward linear prediction. At BER of 10–3, simulation results show that BiPDNP outperforms conventional PDNP by 0.5-1.5 dB in HAMR channels. These simulations are performed at areal density of 4 Tb/in2 with 10%, 15%, 20%, and 30% microtrack jitter levels. We also investigate the performance of BiPDNP in BPMR, where it is found that no significant gains are achieved. We present a theoretical analysis listing the circumstances under which forward PDNP will not make the same errors as backward PDNP and vice versa. Based on this analysis, we find that using BiPDNP with one target scheme, we are able to achieve the same performance with lesser complexity compared to conventional PDNP. Using the BiPDNP detector, we investigate the performance of HAMR channels with fluctuations in coercivity, coercivity gradient, and peak temperature. Whereas fluctuations in coercivity and coercivity gradient cause performance degradation, we find that peak temperature fluctuations are particularly detrimental to the performance of the HAMR system. Subsequently, we focus on ECC schemes to mitigate the effects of media noise. We first investigate the performance of HAMR with LDPC code, where superior gains are reported for coded case over the uncoded case. Next, we propose modifications to the BCJR detector to incorporate BiPDNP. Simulation results show that at BER of 10–4, BiPDNP with LDPC outperforms conventional forward PDNP by 0.4-1 dB. These simulations are performed at areal density of 4 Tb/in2 with 15-30% microtrack jitter. We also compare simulation results with our proposed equalizer designed with jitter noise and BCJR with BiPDNP detector against the equalizer designed without jitter noise and BCJR without PDNP scheme. Here, the performance gains are up to 17.4 dB in the presence of 30% microtrack jitter. Lastly, we report the results of an investigation of BPMR channels containing insertion/deletion errors that are introduced because of the write synchronization problem. We describe a simple channel model for insertion/deletion errors in BPM, as a result of write clock frequency offset. Based on this channel model, we propose a new ECC scheme to correct insertion/deletion errors - picket-shift. Simulation results show that picket-shift performs significantly better than a no-ECC scheme in the presence of insertion/deletion errors. (Abstract shortened by UMI.)