SIGNAL PROCESSING AND DATA CODING FOR DIGITAL HOLOGRAPHIC DATA STORAGE SYSTEM

Abstract

Holographic data storage has been investigated extensively due to its high storage capacity, fast sustained transfer rate and parallel recording and access potentials. However, the research and demonstration of the digital holographic data storage appeared only in recent years. The information capacity per data page achieved so far was small. It is limited by a lot of problems, such as problems relating to the materials and peripheral components, as well as the shortage of a powerful two-dimensional signal processing and data coding method.

This thesis makes contributions towards a high-density, reliable digital holographic data storage system. Firstly, the fundamental concepts of holographic data storage are revisited. They include the existing multiplexing methods, holographic recording materials and the corresponding fixing methods, etc. Among them, the nonlinear photorefractive effect and the angular selectivity are investigated in detail. Based on the good understanding of the holographic storage channel, a channel model is developed and used to guide the design of the coding and detection methods. From the channel model, it is also found that a proper defocus distance can be chosen in the defocus method instead of using the random phase mask to achieve the same signal to noise ratio (SNR). Secondly, in order to implement a digital data storage system, a novel random array exclusive OR modulation (REOM) coding technique is developed. The channel model is used to simulate and verify that it gives a better performance compared to other existing modulation coding schemes. Moreover, a q-ary balanced code is designed. Using the proposed q-ary balanced code, the REOM can be easily extended to the multilevel data page storage case. Furthermore, a modified Reed­ Solomon (RS) error corrective code (ECC) is designed for the digital holographic data storage system. Its performance is analyzed in detail. Thirdly, in the experimental setup, a novel neural network aided alignment technique is proposed and developed to address the data registration problem. Finally, the REOM coding technique is tested experimentally, and a multi-page digital data storage with a raw BER in the order of 10-3 is demonstrated. There are 10,404 bits per data page in the demonstration and no error is observed after the error corrective decoding.