Design analysis of random superimposed coding methods for data storage

摘要

The various possibilities for the use of random superimposed coding methods popular in edge-notched, punched and feature card systems are described. The number of false drops—unwanted items obtained by false matches of codes—obtained in any search can be derived from the system's “false dropping fraction”, defined as the ratio of the false drops to the total unwanted items in the collection for a search on the average single coded descriptor. Formulae for the false dropping fraction are derived for several popular system designs. Figures and tables are given for the calculated values in the most popular scheme (using constant length codewords superimposed into a single field) for various codelengths, code-field sizes and indexing depths (both for fixed and Poisson variable distribution of indexing depth). The results, which are verified by application to an actual example, show that such systems often perform less well than previously expected, especially if Poisson distribution of indexing depth is assumed.