Full-surface data storage.

摘要

Data-storage systems in today's market use a one-dimensional paradigm to store data. Usually, a storage medium spins past a head that reads and writes data; consequently, signals are viewed as values over a one-dimensional index set. This serial approach to data storage has motivated the formulation and solution of many interesting problems in modulation and coding that allow for greater storage densities.; New technologies that read and write bits as two-dimensional signals offer significant advantages over serial systems. These full-surface systems require substantially less power and have smaller access times. Data is stored in the pattern of holes in an optical mask. The optical mask is backlit and the resulting light pattern projected onto a charge-coupled device which produces a two-dimensional image of the data pattern. The emergence of this full-surface technology motivates the study of algorithms and techniques that can improve data density by retrieving data from noisy, full-surface, intersymbol-interference (ISI) channels.; First, serial techniques are generalized to full-surface applications and the resulting systems analyzed for bit error rate (BER) performance under varying data density and noise levels. A multitrack Viterbi algorithm (MVA) performs well; however, the computational complexity limits the achievable data density. Equalization methods are introduced to increase achievable density. Iterative techniques based on greedy optimization are also studied. Though requiring relatively large computation, these iterative techniques have manageable memory requirements and can be used to implement soft-output algorithms. Finally, methods are developed that link maximum-likelihood (ML) data demodulation with ML image processing.