Takao Sugawara

ECC-less LDPC coding for magnetic recording channels

Summary form only given. Low-density parity check (LDPC) codes are receiving considerable attention as a next-generation coding technique for magnetic recording channels, because they show a 5 dB gain over the conventional PRML method. The gain, however, reduces to 1-1.5 dB if the code is combined with an error correcting code (ECC). After the LDPC code, the ECC is not effective in the sense that its gain is less than the rate loss. In fact, we found that ECC-less LDPC codes perform better than LDPC codes with ECC. However, it is still difficult to avoid use of the ECC since the iterative decoder is unable to correct erasures caused by thermal asperity (TA) or media defects. This paper presents a method of erasure compensation for the iterative decoder. The method enables the use of ECC-less LDPC codes, which leads to better performance than using LDPC codes with ECC.

Viterbi detector including PRML and EPRML

The authors develop an integrated-Viterbi detector which can handle partial-response maximum-likelihood (PRML) and extended PRML (EPRML). Class 4 partial response (PR4) and extended PR4 (EPR4) equalization can be accepted according to the recording density. The EPRML performance is superior to that of PRML for a normalized recording density above 1.8, but inferior at lower densities. Experimental results agree with theoretical values and prove that the performance of this Viterbi detector is superior, by 8 dB, to that of conventional peak detectors when the normalized recording density is 2.7. >

A nonlinear model for magnetic recording channel with an MR head

We propose a model for a magnetic recording channel to estimate the influences of the nonlinear distortion of an MR head on the performance of the PRML system, using a superposition of a magnetic field and a transfer curve obtained by a micro-magnetic model. A readback signal was simulated using a nonlinear model. The simulated signal was compared with an actual readback signal in the analysis of the isolated pulse, frequency response, and extracted dipulse response. We found that the nonlinear model represents actual MR head output satisfactorily, and that the performance of the PR4ML system was affected by the nonlinear distortion of the MR head for a 5% loss of the amplitude margin when the asymmetry was within 10%.

Efficiency of short LDPC codes combined with long Reed-Solomon codes for magnetic recording channels

This paper proposes the use of low-density parity check (LDPC) codes with short block lengths such as 1-K bits for magnetic recording. In general, there is a degradation in bit error rate as the block length of the LDPC codes decreases. Yet, we show that short LDPC codes do not suffer from degraded sector-error-rate performance if the sector size is 32 K bits and if long Reed-Solomon codes over GF(2/sup 12/) are used. This is due to the unique error distribution of iterative decoding. Shorter codes significantly reduce the amount of buffer memory in the iterative decoder and make hardware implementation more feasible.

An experimental study of signal equalization for thin film heads

Signal processing for a thin film head, especially the influence of negative side-lobes, are discussed. Numerical analysis shows that a (1,7,2,3) code is superior to other run-legth-limited codes for well-balanced margins in a system featuring high density and a high transfer rate. Two techniques of reducing negative side-lobes were presented. One is a new equalizer, using pulse slimming and elimination of negative side-lobes. The other is an asymmetrical pole-face head. Experimental and theoretical results clarify that the two techniques are extremly effective in increasing the level margin.

A method for suppressing error propagation in (1,7) MDFE detectors

MDFE (multilevel decision feedback equalization) type detectors have been proposed for high recording densities and high data rates in magnetic recording. However, error propagation is a serious problem in all DFE approaches. In this paper, we present some solutions to the error propagation problem using existing signals with less complexity in MDFE and M3DFE detectors. We discuss the effectiveness of these suppression techniques with respect to simulation results and experimental results in the MDFE read channel. The results show that the methods can considerably suppress error propagation.