Storage Channels with Write Errors: Two-dimensional Magnetic Recording and Advanced Memory Stystems

For emerging data storage systems, allowing errors during writing and correcting them by customized coding and signal processing during reading may provide higher customer densities than the conventional approach of trying to completely avoid write errors. Two high-density storage channels with write errors are investigated in this thesis; two-dimensional magnetic recording (TDMR) and advanced memory systems.;The main contents of this thesis are (1) channel model for channels with write errors; (2) investigation of information-theoretic limits of channels with write errors; and (3) coding schemes with side information for TDMR and advanced memory systems. Channels with write errors are modeled by state-dependent channel models with side information, where the side information about write errors can be used during encoding or decoding or both. Information-theoretic study shows that potential performance improvements can be achieved by taking into account write errors as side information during coding. Several coding schemes are proposed for TDMR and advanced memory systems, where the encoder can use complete or partial side information about the channel states associated with write errors. Numerical evaluations show that the proposed coding schemes using side information outperform conventional schemes by taking advantage of write errors in these emerging storage channels.;The contributions of this thesis to the field of coding and signal processing for emerging data storage systems can be summarized as follows. Channel capacity bounds are derived (1) for TDMR channels with write errors based on a random Voronoi grains media model with idealized write and readback and (2) for memory channels with write errors where error correcting code (ECC)-provided partial side information is available at the encoder. Then, for TDMR, (3) additive encoding with low-density parity-check (LDPC) codes using complete or partial side information is proposed for ideal write and readback TDMR. In addition, (4) coding and signal processing for channels with write errors and other read channel impairments are investigated with position and timing uncertainty effect. For memory systems, (5) a scrubbing scheme using side information is proposed for radiation-tolerant memory systems and (6) an iterative encoding scheme with side information based on the cross-entropy method is proposed for advanced, memory systems. Also, (7) an endurance coding scheme is proposed for flash memory systems, which can enhance the lifetime of flash memory systems by reducing the amount of charge travelling the tunneling barrier in flash memory cells. These coding schemes are evaluated numerically using MATLAB and C, and simulation results show significant improvements in customer bit storage density with the proposed coding schemes.