Source-channel coding for digital magnetic recording

The authors explore the possibility of dedicating a disk to store the binary representation of analog data samples and using source-channel codes both to compress the data and to protect it from the increased error rate that results from higher bit density. To determine what gains are possible using this approach, a system was simulated consisting of an IID Gaussian data source, a tree structured vector quantizer for source-channel coding, a channel model similar to that of C.A. French and J.K. Wolf (1988), and two kinds of detector/demodulators. For IID Gaussian data, for which only modest compression is possible, it was found that gains on the order of three to four in data density were achievable with this approach.<
>