In digital magnetic recording, bit densities is limited by peak-shift or S/N ratio. In either case, limitations are caused by bit-interferense and discrimination of fixed threshold. Compensation of many bits interferense by viterbi detection may be difficult due to complex algorithm. However, neural network can recognize patterns having complex algorithms without programming. Therfore, if readback waveform is recognized by neural network as several bit patterns after learning of readback waveforms. It may be possible to obtain the higher bit density over the limitation of usual method, because learning of waveform include interferense algorithm and compensation effects. Results of software simulation are discussed.
View full abstract