Reduced Complexity Path Selection Networks for M-Algorithm Read Channel Detectors

Abstract

The M-algorithm is a reduced complexity, breadth-first, trellis search alternative to the Viterbi algorithm. It is particularly useful when the number of trellis states is so large as to render Viterbi detection impractical. When dealing with complex or long target responses, values of M ges 16 may be required in order to approach MLD performance. Performing path selection for M ges 16 using previously reported methods can be computationally complex and prohibitive to practical hardware implementation. This paper details reduced complexity path selection networks for use with large values of M. A strategy for the construction of reduced complexity networks is presented and can be used to create networks for a variety of values of M and for other applications where M-algorithm detection can be employed. The use of these networks will allow for practical realisable hardware implementations of M-algorithm detectors for M ges 16.