Comparative performance between drop-out detection and viterbi reliability metric erasure flagging
W.E. Ryan · 1992. Digests of Intermag. International Magnetics Conference · 1992
It is well known that reIiable byte erasure flags can significantly enhance the decoded performance of an interleaved Reed-Solomon (RS) code on a digital magnetic recording channel. This paper compares the improvement afforded by a drop-out detector erasure flagger to the improvement provided by an erasure flagger based on a Class IV partial response (PR4) Viterbi detector. It is shown that, in addition to being less complex, the drop-out detector is the superior erasure flagger when decoded performance is burst-error limited and burst errors are a result of signal level drops. Although the proposed Viterbi erasure flagger is derived heuristically, it performs as well as the optimum method discussed in the references and, hence, the drop-out detector can be claimed to be superior to all such Viterbi flaggers under the assumptions of this paper. The focus is on interleaved Reed-Solomon codes over GF(288), that is, codes whose codewords consist of 8-bit symbols. Hence, we would like to flag byte errors rather than bit errors (although a bit error flagger trivially extends to a byte error flagger). Furthermore, we assume that the recorder decoded performance is limited by burst errors due to drop-outs, tenting, etc., rather than by additive white Gaussian noise as in the references. This is the environment typical of a magnetic tape recorder.