At high recording densities, recorded transitions interact magnetically upon writing in such a manner that they are partially erased and shifted from their nominal positions. Such phenomena represent nonlinear distortion and can be the dominant impediment to reliable detection. There exist in the literature a number of techniques for mitigating nonlinear distortion, including maximum likelihood sequence detection (MLSD) [1], Volterra equalization [2], neural network equalization [3], and nonlinear cancellers [4]. While transition-shift nonlinearities may be effectively reduced via write precompensation, this is not the case for partial erasure. One approach is to eliminate transition shift via write precompensation and then apply an MLSD detector to a trellis which incorporates a simple partial erasure model together with PR4 equalization [1]. In this paper, we show how the MLSD approach may extended to include also the transition-shift nonlinearity, without increasing the number of states (11 states).
展开▼
