Multihead multitrack detection with reduced-state sequence estimation in shingled magnetic recording

Intertrack interference (ITI) is one of the most severe impairments in shingled magnetic recording (SMR). The use of Multihead Multitrack (MHMT) detection can best handle this impairment, but increases complexity. Consider a symmetric two-head two-track (2H2T) system given by r^a(D)=x-^a(D)h(D)+_εx^b(D)h(D)+n^a(D) and r^b(D)=εx^a(D)h(D)+x^b(D)h(D)+n^b(D), where x^a(D), x^b(D) and r^a(D), r (D) are the input sequences and received sequences, respectively. The intersymbol interference (ISI) is characterized by h(D) and the ITI is specified by parameter ε. Both the noise sequences n^a(D) and n (D) have variance σ². A similar model is also used in [1] and [2]. The maximum likelihood (ML) 2H2T detector simultaneously decodes two tracks by searching a joint trellis, thus operates at complexity O(4^υ), where ^υ is the channel memory. In this work we design detectors that use fewer trellis states while retaining good performance. Reduced-State Sequence Estimaton (RSSE) was proposed in [3] to reduce the detection complexity when the system has a large input signal set and/or a long channel response. We propose to apply RSSE to the 2H2T model by carefully designing the set partitioning principles. Simulation results show that in many cases RSSE achieves nearly ML performance with 50% fewer states.