Simplification of neural network equalizer for perpendicular magnetic recording
Hisashi Osawa, Toshimasa Shimizu, Takumi Nakaoka, Yoshihiro Okamoto, Hidetoshi Saito, Hiroaki MURAOKA, Yoshihisa Nakamura · Electronics and Communications in Japan (Part II Electronics) · 2006
In this paper, we examine neural network equalization in a perpendicular magnetic recording channel with jitter medium noise and MR nonlinear distortion. First, we propose simplifying a neural network equalizer by using a hybrid genetic algorithm. Next, we determine the bit error rate of a PR2ML method provided with the simplified neural network equalizer and compare it to those of a conventional neural network equalizer and a transversal filter equalizer. The results were improvements of about 0.8 and 2.0 dB in the SNR of the PR2ML method using the simplified neural network equalizer compared to the conventional neural network equalizer and the transversal filter equalizer, respectively. © 2006 Wiley Periodicals, Inc. Electron Comm Jpn Pt 2, 89(2): 19–27, 2006; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjb.20197