Magnetic recording is the most widely used technology for reliably storing digital information. A stored bit is physically realized as a collection of tiny magnetic grains magnetized to one of two stable polarities. This project relies on a new paradigm of two-dimensional magnetic recording, which exploits a small number of grains to store one bit as well as two-dimensional coding of user data combined with signal detection to ensure reliable storage and retrieval of information.
The project addresses the main challenge in two-dimensional magnetic recording which arises due to the fact that grains are irregularly shaped and randomly positioned on the media surface, and hence, the signal read back is dominated by noise due to this random granularity of the medium. The project develops novel data recovery algorithms capable of compensating such high noise levels. These algorithms are based on advanced information theory and coding theory concepts, namely, constrained codes, probabilistic graphical models, and iterative detection and decoding. A two-dimensional constraint is viewed as a colored tiling of a plane, which affords using the rich theory of planar gas models from statistical mechanics and the theory of domino and lozenge tilings from combinatorics. On the detection front, the project develops generalized belief propagation and linear programming algorithms for dealing with the large number of loops in two-dimensional magnetic recording graphical models.
In light of the tremendous growth of the amount of digital data created, processed and stored, it is vitally important to ensure the continued rapid increases in capacity of magnetic hard disk drives. The project is concerned with a new technology that has the potential to increase hard drive capacity by an order of magnitude. Signal processing methods for improving storage capacity developed in this project have a crucial role in providing a foundation for development of future storage systems. They provide not only a new standard of performance in the area of data storage, but have a direct impact on the performance capabilities of a new generation of computers, data networks and services provided on the internet.
