A variable metric descent method developed by Davidon, Fletcher and Powell (the DFP method) is widely used to optimize (minimize) objective functions. Like other descent methods, it suffers from poor convergence if the objective function has multiple minima. If an initial guess is not carefully chosen, it reaches a local minimum. We present a combinatorial optimization technique that uses the simulated annealing (SA) algorithm in conjunction with the DFP method. The combinatorial optimization technique avoids local minima and ultimately reaches the global minimum. This technique is applied to optimization of the PR-IV equalizer using analog filters and is shown to be superior to the DFP method. Finally, this combinatorial technique is used in optimization of a PR-IV equalizer for a Lorentzian recording channel, and it is found that the PR-IV equalizer composed of a 6-pole. 6-zero amplitude equalizer and a 6-pole, 6-zero phase equalizer is adequate to shape the playback signal to the desired PR-IV format at densities of interest (between 1 and 3).
|頁（從 - 到）
|Proceedings of the National Science Council, Republic of China, Part A: Physical Science and Engineering
|已出版 - 11月 1995