Resumen de Model detection with application to probe based data storage

Sayantan Ghosal, Govind Saraswat, Murti Salapaka

• Abstract Detection of a model from a set of models that best describes the behavior of a system is of primary importance in many applications. In this article two discriminating signals are derived from measurements for a plant that switches between two model behaviors, where the transfer functions from inputs to the two signals are identical when one model is effective while they are negative of each other when the other model is effective. The developed detection algorithm called the innovation squared mismatch is utilized for the read operation of probe based data storage. The innovations squared mismatch method offers better detection performance with significantly less computational complexity compared to prevalently used maximum a posteriori probability based methods in current data storage systems. The article further proposes employing maximum likelihood sequence detection based methods where the plant behavior can switch from one model to another at high rates and the transients from a previous behavior affect the current behavior causing inter-symbol interference. Exhaustive simulation and experimental results corroborate the efficiency of the proposed methods.