Multitrack Detection With a Two-Dimensional Soft-Transition Assisted Multitask Neural Network for Heat-Assisted Interlaced Magnetic Recording

Heat-assisted interlaced magnetic recording (HIMR) further increases recording density compared to heat-assisted magnetic recording due to the interlaced track layout. However, the smaller circular thermal profile of the top low-temperature-written tracks results in severe transition curvatures, which causes nonlinear distortions of the readback signal and degrades the bit error rate (BER) performance. Moreover, the increased recording density causes severe two-dimensional (2-D) intersymbol interference (ISI) along both down track and cross track directions. To mitigate the effect of nonlinear distortion and 2-D ISI in an HIMR system, we model a 2-D soft-transition, information-assisted, multitask neural network and modified Bahl–Cocke–Jelinek–Raviv (BCJR) detector (2DST-MTNN-MB) algorithm to detect three tracks simultaneously. The readback signal and 2-D soft-transition information are fed into the multitask neural network to obtain equalized signals and soft bit estimates of three tracks simultaneously. Then, the signal of the current track and soft estimates of the side tracks are embedded into the branch metrics of the modified BCJR detector for data detection, and the low-density parity check decoder is cascaded for error correction. The simulation results show that the 2DST-MTNN-MB algorithm provides 5.0 dB signal-to-noise ratio gains with reduced computation complexity compared with a single-track, single-task neural network and one-dimensional BCJR detector algorithm for the low-temperature-written track at channel bit density of 3.51 Tb/in2, thereby narrowing the gap of BER performance between high- and low-temperature-written tracks.