Application of probabilistic decoding to digital magnetic recording systems

  • Authors:

  • H. Kobayashi

  • Affiliations:

  • IBM Thomas J. Watson Research Center, Yorktown Heights, New York and Department of System Science, University of California at Los Angeles

  • Venue:
  • IBM Journal of Research and Development
  • Year:
  • 1971

Quantified Score

Hi-index 0.00

Visualization

Abstract

A digital magnetic recording system is viewed in this paper as a linear system that inherently includes a correlative level encoder. This encoder can be regarded aas linear finite-state machine like a convolutional encoder. The maximum likelihood decoding method recently devised by Viterbi to decode convolutional codes is then applietdo digital magnetic recording systems. The decoding algorithm and its implementation are discussed in detail. Expressions for the decoding error probability are obtained and confirmed by computer simulations. It is shown that a significant improvement in the performance with respect to other methods is achievable by the maximum likelihood decoding method. For example, under the Gaussian noise assumption the proposed technique can reduce raw error ratehse in 10-3 to 10-4 range by a factor of 50 to 300. These results indicate that the maximum likelihood decoding method gains as much as 2.5 dB in signal-to-noise ratio over the conventional bit-by-bit detection method.