Computer-based simulation of data storage channels and an investigation of adaptive replay signal equalisatio

This thesis describes research into the simulation and optimisation of replay channels for magnetic recording systems. Two major activities are described, the first being the simulation of a peak-detection channel for recovering stored data and the second being the adaptive equalisation of a partial-response channel. The peak-detection channel simulation combines computer simulations for both the electronic circuits and the magnetic components of the recording channel. Mathematical models of the read processes are implemented that relate the head and media parameters to the readout signal. A replay equalisation scheme suitable for high-density recording is presented and its performance evaluated with the models developed. The simulation results show that the performance predicted compares favourably with results obtained from practical measurements. Partial-response equalisation provides a means for processing signals containing intersymbol interference introduced as a result of increasing recording density in storage systems. For high-density applications a new adaptive equalisation technique is proposed to obtain improved data integrity of the replayed signals. The self-adaptive partial-response equalisation process described is based on evolutionary principles, similar to those found in nature, and they are embodied in a novel genetic algorithm. Equaliser parameters are adjusted by the algorithm to achieve the required channel response. The performance of the genetic-based optimisation algorithm has been tuned to the requirements of replay equalisers. An evaluation of the genetic algorithm has been undertaken with equalisers using transversal filters having various numbers of taps in environments with and without replay signal noise. Equaliser simulations and analysis of results have confirmed the superior performance and robustness of the new optimisation algorithm when compared with gradient-based techniques. A rapid convergence rate has been achieved with less susceptibility to finding localised minima. Robustness of performance of the optimisation algorithm in the presence of noise presents the possibility of working with the poor signal- to-noise ratios expected in future high-density storage systems.